The Hippocampus Book

This page intentionally left blank


The Hippocampus Book

EDITED BY

Per Andersen, Richard Morris,

David Amaral, Tim Bliss, & John OKeefe

2007
Oxford University Press, Inc., publishes works that further
Oxford Universitys objective of excellence
in research, scholarship, and education.

Oxford New York

Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto

With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam

Copyright 2007 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.

198 Madison Avenue, New York, New York 10016
www.oup.com

Oxford is a registered trademark of Oxford University Press

All rights reserved. No part of this publication may be reproduced,

stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press.

The hippocampus book / Per Andersen ...[et al.], editors.

p. ; cm.
Includes bibliographical references and index.
ISBN-13: 978-0-19-510027-3
ISBN-10: 0-19-510027-1
1. Hippocampus (Brain)Physiology. I. Andersen, Per.
[DNLM: 1. Hippocampusanatomy & histology. 2.
Hippocampusphysiology. 3. Memoryphysiology. 4. Models,
Theoretical. 5. Synaptic Transmission. WL 314 H666 2007]
QP383.25.H57 2007
612.825dc22 2006007088

987654321
Printed in the United States of America
on acid-free paper
To

Kari Andersen
Hilary, Louise, and Josephine Morris
David Joseph and Jennie Amaral
Isabel Vasseur
Eileen OKeefe
This page intentionally left blank

Preface

The hippocampus is one of the most widely studied regions of card-carrying hippocampologists. Inevitably, our conversa-
the brain and is of interest to a wide spectrum of neuroscien- tion turned to our favorite brain structure and in particular to
tists, ranging from those who study its normal structure and the lack of reviews and surveys to which we could direct edg-
function to others who study its malfunction in various dis- ling hippocampologists. Eventually, Per Andersen threw out
eases and pathological conditions. Information about it has the challenge: Could the four of us write a treatise on the hip-
accrued across a considerable span of time and is spread over pocampal formation that captured the essentials, derived as
a wide variety of publications. It seems all the more remark- they were from so many branches of neuroscience? He won-
able, therefore, given the explosion of texts on every conceiv- dered whether, among us, we could cover the entire eld from
able topic in neuroscience, that there is no single, synapse to behavior by way of the cells and the nerve nets in
comprehensive source of information on the hippocampal between. The original crew had a not inconsiderable breadth
formation. It would clearly be helpful to the community of of experience from physiology, psychology, and behavioral
hippocampologists to have the essential information about science, but we quickly realized the essential need for struc-
hippocampal neurobiology gathered together in one volume. tural information as one of the foundations of such a book.
This book is our attempt at a reasonably comprehensive sur- We were happy to bring on board a neuroanatomist with great
vey of hippocampal research, as viewed through many eyes knowledge and experience of the hippocampus, and the gang
and collected with a wide variety of methods. We hope that of four became ve.
the book will be useful at several levels. For those new to the Over the next 10 years, the Editors met in numerous
arena of hippocampal research, we hope it will act as a primer delightful venues in the United States and Europe to read each
pointing to the important advances of the past decades and others work and to confer on the book. Although much writ-
suggesting important research paths to be pursued; for hip- ing was done during this phase, it gradually dawned on us that
pocampal veterans, we hope that by laying bare the many two formidable obstacles made our objective signicantly
uncertainties and open questions that remain, it will provide more difficult than we had originally anticipated. First, the
ample stimulus to see old problems with new eyes and to pur- material on the hippocampus was even more extensive than
sue them with renewed vigor. we had expected and, with the explosion of neuroscience, was
The hippocampus is a structure eminently worthy of study growing rapidly day by day. A recent PubMed search elicited
in its own right, but it has also been seen by many as a model more than 73,000 references to the target word hippocam-
structure for the study of cortical function and plasticity in pus. More daunting still, the breadth of the eld had
general. The twin aims of the book therefore are to discuss expanded with the introduction of new methods and infor-
hippocampal structure and function in its own terms and to mation, not least from the rapidly developing elds of molec-
highlight how study of the hippocampus has revealed ideas ular biology, genetics, and development. To cover these
and advances that are of wider signicance and generality for important elds adequately, we decided to ask a set of col-
neuroscience. leagues with the appropriate specialist knowledge to join us in
The inception of The Hippocampus Book occurred many the effort. We were fortunate to attract a powerful team, and
years ago, at the end of a symposium in the beautiful, stimu- the Editors are extremely grateful to these colleagues for their
lating city of Palermo in Sicily. We ended up at a street-side important contributions.
cafe, each with a glass of birra glowing in the setting We hope that The Hippocampus Book will prove useful to
Mediterranean sun. We were four old friends, all long-term its readers. We are condent that many share our affection for
viii Preface

this uniquely fascinating structure. If the book helps future ory and its involvement in a host of psychological and psychi-
neuroscientists in their planning and execution of hippocam- atric conditions, the hippocampal formation will undoubt-
pal studies, our project will have been a success. If, in addition, edly continue to be an active focus of energetic research. Our
the book facilitates new lines of thought and experimental hope is that The Hippocampus Book will act as a springboard
approaches that challenge existing ideas and concepts, our and a guide for some of that research.
reward will be even greater. With its important role in mem-
Per Andersen
Richard Morris
David Amaral
Tim Bliss
John O Keefe

Acknowledgments

Over the years, several individuals and institutions have Many dozens of our colleagues have been badgered with
helped us in various ways with the creation of this book. An questions over the years and have replied with patience
important planning session took place at The Institute for and dispatch. Per Andersen particularly thanks Theodor
Neuroscience, then at The Rockefeller University. We thank its Blackstad, Gyorgyi Buzsaki, Leif Gjerstad, Vidar Jensen,
Director Gerald Edelman and Research Director Einar Gall for Edward Jones, Bruce Piercey, Geoffrey Raisman, Eric Rinvik,
their generous support. Twice we were accommodated for Thomas Sears, and Jon Storm-Mathisen. Finally, we express
writing sessions at the Centre for Advanced Studies at the our particular gratitude to Fiona Stevens, at Oxford University
Academy of Science and Letters in Oslo, and we thank the Press, New York for keeping faith with our project through too
director of the Centre Vigdis Ystad and the President of the many fallow periods and, also at OUP to Joan Bossert and
Academy Bjarne Waaler for their hospitality. At other times Mallory Jensen for the graceful pressure they applied during
we met in La Jolla, California, Overstrand, Norfolk and in the nal months while shepherding the book to publication.
London at the Royal Society and the Novartis Institute. We Finally, we are deeply grateful for the efficient, yet gracious
enjoyed several editorial meetings in Edinburgh at the Centre manner, in which Berta Steiner at Bermedica Production
for Neuroscience. To all these institutions we express our helped us to bring the book to fruition.
gratitude.
This page intentionally left blank

Contents

Chapter 1 2.4.3 New Anatomical Techniques that Revolutionized

The Hippocampal Formation 3 Connectivity Studies 18
2.4.4 Predominantly Unidirectional Connectivity Between
Per Andersen, Richard Morris, David Amaral, Tim Bliss,
Cortical Strips 20
and John OKeefe
2.4.5 New Tracing Studies Using Axonal Transport 20
1.1 Overview 3 2.4.6 Electron Microscopy Offers New Opportunities 21
1.2 Why Study the Hippocampal Formation on its 2.4.7 Hippocampal Synapses Are Highly Plastic: Early
Own? 4 Studies of Sprouting 21
1.3 Dening the Contemporary Era 5 2.4.8 Hippocampal Neurons: Transplantable with
1.4 Organization and Content of the Book 6 Retention of Many Basic Properties 22
2.4.9 Hippocampal Cells Grow Well in Culture 22
Chapter 2
2.4.10 Development of Hippocampal Slices: From Seahorse
Historical Perspective: Proposed Functions, Biological to Workhorse 22
Characteristics, and Neurobiological Models of the 2.5 Several Neurophysiological Principles Have Been
Hippocampus 9 Discovered in Hippocampal Studies 23
Per Andersen, Richard Morris, David Amaral, Tim Bliss, 2.5.1 Identication of Excitatory and Inhibitory
and John OKeefe Synapses 23
2.1 Overview 9 2.5.2 Gray Type 2 Synapses are Inhibitory and are Located
2.2 The Dawn of Hippocampal Studies 9 on the Soma of Pyramidal and Granule Cells 23
2.2.1 A Famous Dispute About the Signicance of the 2.5.3 Gray Type 1 Synapses are Excitatory and are Located
Hippocampus 9 on Dendritic Spines 24
2.3 Early Ideas About Hippocampal Function 10 2.5.4 Long-lasting Alterations of Synaptic Efficiency After
2.3.1 The Hippocampal Formation and Olfactory Physiological Stimulation 24
Function 10 2.5.5 Hippocampal Systems: Exhibiting Several Types of
2.3.2 The Hippocampal Formation and Emotion 11 Oscillatory Behavior 25
2.3.3 The Hippocampal Formation and Attention 2.5.6 Studies of Epileptiform Behavior 25
Control 12 2.6 Development of Methodological Procedures for
2.3.4 The Hippocampal Formation and Memory 13 General Use 26
2.3.5 More Direct Evidence for Hippocampal Involvement 2.6.1 Hippocampus as a Test Bed for Microelectrode
in Memory 13 Work 26
2.3.6 The Hippocampus as a Cognitive Map 16 2.6.2 Pioneers of Intracellular Recording 26
2.3.7 Conclusions 16 2.6.3 Tetrode Development 27
2.4 Special Features of Hippocampal Anatomy 2.6.4 Field Potential Analysis 27
and Neurobiology 16 2.6.5 Histochemistry: Pioneered in the Hippocampus 30
2.4.1 Early Neuroanatomical Studies of the 2.6.6 Pharmacological Analysis of Cellular Properties 30
Hippocampus 17 2.6.7 Development of Computational Models of Neural
2.4.2 New Fiber Tracing Methods 18 Networks 31
xii Contents

2.6.8 The Hippocampal Formation: A Test Bed for 3.7.3 Functional Implications of the Transverse Topography
Several Types of Neural Dysfunction and of Connections 108
Neuropathology 31 3.7.4 Serial and Parallel Processing in the Hippocampal
References 31 Formation 109
3.8 Conclusions 109
Chapter 3 References 110
Hippocampal Neuroanatomy 37 Chapter 4
David Amaral and Pierre Lavenex Morphological Development of the Hippocampus 115
3.1 Overview 37 Michael Frotscher and Lszl Seress
3.1.1 Hippocampus: Part of a Functional Brain System 4.1 Overview 115
Called the Hippocampal Formation 37 4.2 Neurogenesis and Cell Migration 115
3.1.2 Similarities and Differences Between the Hippocampal 4.2.1 Pyramidal Neurons 116
Formation and other Cortical Areas 37 4.2.2 Granule Cells 116
3.1.3 Hippocampal Formation: With A Unique Set of 4.2.3 Local Circuit Neurons and Hilar Neurons 118
Unidirectional, Excitatory Pathways 38 4.2.4 Determinants of Neuronal Migration in the
3.1.4 Hippocampus of Humans and Animals: Same or Hippocampus 120
Different? 39 4.3 Development of Hippocampal Connections 120
3.1.5 Synopsis of the Chapter 39 4.3.1 Entorhinal Connections 121
3.2 Historical Overview of Hippocampal Nomenclature 4.3.2 Commissural Connections 123
Whats in a Name? 39 4.3.3 Septal Connections 125
3.2.1 Denition of Hippocampal Areas: Denition 4.3.4 General Principles Underlying the Formation of
of Terms 42 Synaptic Connections in the Hippocampus 126
3.2.2 Subdivision of Hippocampal Areas 42 4.4 Development of the Primate Hippocampal
3.2.3 Major Fiber Bundles of the Hippocampal Formation 127
Formation 44 4.4.1 Neurogenesis 127
3.3 Three-dimensional Organization and Major Fiber 4.4.2 Neuronal Differentiation 127
Systems of the Hippocampal Formation 44 References 128
3.3.1 Rat Hippocampal Formation 44
3.3.2 Major Fiber Systems of the Rat Hippocampal
Chapter 5
Formation 47
3.3.3 Monkey Hippocampal Formation 48 Structural and Functional Properties of Hippocampal
3.3.4 Human Hippocampal Formation 49 Neurons 133
3.4 Neuroanatomy of the Rat Hippocampal Nelson Spruston and Chris McBain
Formation 51 5.1 Overview 133
3.4.1 Dentate Gyrus 55 5.2 CA1 Pyramidal Neurons 133
3.4.2 Hippocampus 67 5.2.1 Dendritic Morphology 134
3.4.3 Subiculum 76 5.2.2 Dendritic Spines and Synapses 136
3.4.4 Presubiculum and Parasubiculum 81 5.2.3 Excitatory and Inhibitory Synaptic
3.4.5 Entorhinal Cortex 83 Inputs 137
3.5 Chemical Neuroanatomy 94 5.2.4 Axon Morphology and Synaptic Targets 138
3.5.1 Transmitters and Receptors 94 5.2.5 Resting Potential and Action Potential Firing
3.5.2 Steroids 95 Properties 139
3.6 Comparative Neuroanatomy of the Rat, Monkey, and 5.2.6 Resting Membrane Properties 141
Human Hippocampal Formation 95 5.2.7 Implications for Voltage-Clamp Experiments in CA1
3.6.1 Neuron Numbers 96 Neurons 144
3.6.2 Comparison of Rat and Monkey Hippocampal 5.2.8 Attenuation of Synaptic Potentials in CA1
Formation 96 Dendrites 144
3.6.3 Comparison of Monkey and Human Hippocampal 5.2.9 Mechanisms of Compensation for Synaptic
Formation 104 Attenuation in CA1 Dendrites 144
3.7 Principles of Hippocampal Connectivity and 5.2.10 Pyramidal Neuron Function: Passive Versus Active
Implications for Information Processing 107 Dendrites 145
3.7.1 Highly Distributed Three-Dimensional Network of 5.2.11 Dendritic Excitability and Voltage-Gated Channels in
Intrinsic Connections 107 CA1 Neurons 146
3.7.2 Functional Implications of the Septotemporal 5.2.12 Sources of CA2 Elevation in CA1 Pyramidal Neuron
Topography of Connections 107 Dendrites 148
 Contents xiii

5.2.13 Distribution of Voltage-Gated Channels in the Chapter 6

Dendrites of CA1 Neurons 149 Synaptic Function 203
5.2.14 Functional Implications of Voltage-Gated Channels
Dimitri Kullmann
in CA1 Dendrites: Synaptic Integration and
Plasticity 152 6.1 Overview 203
5.2.15 General Lessons Regarding Pyramidal Neuron 6.2 General Features of Synaptic Transmission:
Function 152 Structure 204
5.3 CA3 Pyramidal Neurons 153 6.2.1 Transmitter Release and Diffusion 205
5.3.1 Dendritic Morphology 154 6.2.2 Receptors and Receptor Activation 207
5.3.2 Dendritic Spines and Synapses 155 6.2.3 Quantal Transmission 209
5.3.3 Excitatory and Inhibitory Synaptic Inputs 155 6.2.4 Short-term Plasticity 210
5.3.4 Axon Morphology and Synaptic Targets 156 6.3 Glutamatergic Synaptic Transmission 211
5.3.5 Resting Potential and Action Potential Firing 6.3.1 AMPA Receptors 213
Properties 156 6.3.2 Kainate Receptors 215
5.3.6 Resting Membrane Properties 158 6.3.3 NMDA Receptors 215
5.3.7 Active Properties of CA3 Dendrites 158 6.3.4 Co-localization of Glutamate Receptors 216
5.4 Subicular Pyramidal Neurons 159 6.3.5 Metabotropic Glutamate Receptors 216
5.4.1 Dendritic Morphology 159 6.3.6 Receptor Targeting and Anchoring 218
5.4.2 Dendritic Spines and Synaptic Inputs 159 6.4 GABAergic Synaptic Transmission 218
5.4.3 Axon Morphology and Synaptic Targets 160 6.4.1 GABAA Receptors 218
5.4.4 Resting and Active Properties 160 6.4.2 GABAB Receptors 220
5.4.5 Mechanisms of Bursting 162 6.5 Other Neurotransmitters 220
5.4.6 Membrane Potential Oscillations 162 6.6 Special Features of Individual Hippocampal
5.5 Dentate Granule Neurons 162 Synapses 222
5.5.1 Dendritic Morphology and Spines 163 6.6.1 Small Excitatory Spine Synapses 223
5.5.2 Excitatory and Inhibitory Synaptic Inputs 163 6.6.2 Mossy Fiber Synapses 226
5.5.3 Axon Morphology and Synaptic Targets 164 6.6.3 Other Glutamatergic Synapses on
5.5.4 Resting Potential and Action Potential Firing Interneurons 228
Properties 164 6.6.4 Inhibitory Synapses 229
5.5.5 Resting Membrane Properties 165 6.7 Unresolved Issues 231
5.5.6 Active Properties of Granule Cells 166 References 231
5.6 Mossy Cells in the Hilus 166
Chapter 7
5.6.1 Dendritic Morphology and Spines 168
5.6.2 Excitatory and Inhibitory Synaptic Molecular Mechanisms of Synaptic Function in the
Inputs 168 Hippocampus: Neurotransmitter Exocytosis and
5.6.3 Axon Morphology and Synaptic Targets 168 Glutamatergic, GABAergic, and Cholinergic
5.6.4 Resting and Active Properties 168 Transmission 243
5.6.5 Other Spiny Neurons in the Hilus 168 Pavel Osten, William Wisden, and Rolf Sprengel
5.7 Pyramidal and Nonpyramidal Neurons of Entorhinal 7.1 Overview 243
Cortex 169 7.2 Neurotransmitter Exocytosis 243
5.7.1 Stellate Cells of Layer II 170 7.2.1 Introduction: Proteins Involved in Synaptic
5.7.2 Pyramidal Cells of Layer II 172 Release 243
5.7.3 Pyramidal Cells of Layer III 172 7.2.2 Reserve Pool of Synaptic Vesicles 244
5.7.4 Pyramidal Cells of Deep Layers 174 7.2.3 Synaptic Vesicle Docking and Priming at the Active
5.8 Pyramidal and Nonpyramidal Neurons of Zone: Role of the SNARE Complex, Munc18, and
Presubiculum and Parasubiculum 176 Munc13 244
5.9 Local Circuit Inhibitory Interneurons 176 7.2.4 Ca2-triggered Synaptic Release: Role of
5.9.1 Understanding Interneuron Diversity 179 Synaptotagmin 246
5.9.2 Dendritic Morphology 181 7.2.5 Ca2-triggered Synaptic Release: Role of Rab3A and
5.9.3 Dendritic Spines 183 RIM1 247
5.9.4 Excitatory and Inhibitory Synapses 184 7.2.6 NSF-mediated Disassembly of the SNARE
5.9.5 Axon Morphology and Synaptic Targets 185 Complex 248
5.9.6 Resting Membrane Properties 186 7.2.7 Synaptic Vesicle Endocytosis, Recycling, and
5.9.7 Voltage-Gated Channels in Inhibitory Relling 248
Interneurons 186 7.3 Glutamate Receptors: Structure, Function,
References 188 and Hippocampal Distribution 249
xiv Contents

7.3.1 Introduction: Ionotropic and Metabotropic 8.1.3 Subcellular Domain Specicity in Hippocampal
Receptors 249 Circuits 299
7.3.2 AMPA Receptors 251 8.1.4 Patterns of Local Circuit Connectivity 299
7.3.3 NMDA Receptors 253 8.1.5 Circuit Specic Receptor Distribution 299
7.3.4 Kainate Receptors 254 8.1.6 Convergence and Divergence 301
7.3.5 Metabotropic Glutamate Receptors 257 8.2 Dentate Gyrus 301
7.4 Trafficking of Glutamate Receptors and Hippocampal 8.2.1 Inputs to the Dentate Gyrus 301
Synaptic Plasticity 258 8.2.2 Granule Cell Projection to Area CA3 302
7.4.1 Synaptic Transport of AMPA Receptors in LTP 8.2.3 Granule Cell Interneuron Connections 303
and LTD 259 8.2.4 Interneuron Granule Cell Connections 303
7.4.2 NMDA Receptor-associated Cytoskeletal and 8.2.5 Granule Cell Local Excitatory Neuron
Signaling Proteins 261 Connections 305
7.4.3 Proteins Regulating Transport and Function of 8.2.6 Interneuron Interneuron Connections 305
mGluRs 263 8.3 Areas CA3 and CA1 305
7.5 Glutamate Receptor Mutant Mice: Genetic Analysis of 8.3.1 Inputs to CA3 and CA1 305
Hippocampal Function 263 8.3.2 Pyramidal Cell Interneuron Connections 306
7.5.1 Introduction: Building of Hippocampus-specic 8.3.3 Interneuron Pyramidal Cell Connections 307
Genetic Models 263 8.3.4 Pyramid Pyramid Local Connections 312
7.5.2 NMDA Receptor Mutant Mice 264 8.3.5 Interneuron Interneuron Connections 312
7.5.3 AMPA Receptor Mutant Mice 269 8.3.6 Gap Junction Connections 313
7.5.4 Kainate Receptor Mutant Mice 272 8.4 Summary 314
7.5.5 mGluR Mutant Mice 273 References 315
7.5.6 Synopsis of the Section 273
7.6 GABAergic Receptors: Structure, Function, and Chapter 9
Hippocampal Distribution 274 Structural Plasticity 321
7.6.1 Introduction: Synaptic and Extrasynaptic GABAergic Elizabeth Gould
Receptors Mediate Tonic and Phasic Inhibition in the
9.1 Overview 321
Hippocampus 274
9.2 Dendritic and Synaptic Plasticity in the Hippocampal
7.6.2 GABAA Receptors 274
Formation 321
7.6.3 GABAB Receptors 280
9.2.1 Naturally Occurring Structural Plasticity 321
7.7 Trafficking of GABA Receptors and Hippocampal
9.2.2 Hormones and Dendritic Architecture 322
Synaptic Function 282
9.2.3 Experience and Dendritic Architecture 322
7.7.1 Role of Gephyrin in GABAA Receptor Localization
9.2.4 Structural Plasticity Following Damage 323
282
9.2.5 Transplantation 324
7.7.2 Role of Dystrophin-associated Protein Complex in
9.3 Adult Neurogenesis 324
GABAA Receptor Function 282
9.3.1 Turnover of Dentate Gyrus Granule Cells 326
7.7.3 Plasticity of GABAA Receptor Expression at
9.3.2 Hormones and Adult Neurogenesis 326
Hippocampal Synapses 283
9.3.3 Experience and Adult Neurogenesis 328
7.8 Genetic Analysis of GABA Receptor Function in the
9.3.4 Neurogenesis Following Damage 331
Hippocampus 283
9.3.5 Unusual Features of Adult-Generated
7.8.1 GABAA Receptor Mutant Mice 283
Neurons 332
7.8.2 GABAB Receptor Mutant Mice 284
9.4 Possible Functions of New Neurons 332
7.9 Cholinergic Receptors 284
9.4.1 A Possible Role in Learning? 332
7.9.1 Introduction: Muscarinic and Nicotinic
9.4.2 A Possible Role in Endocrine Regulations? 334
Receptors 284
9.4.3 A Possible Role in the Etiology and Treatment of
7.9.2 Hippocampal Muscarinic Receptors 285
Depression? 335
7.9.3 Hippocampal Nicotinic Receptors 285
References 335
References 286
Chapter 10
Chapter 8
Synaptic Plasticity in the Hippocampus 343
Local Circuits 297 Tim Bliss, Graham Collingridge, and Richard Morris
Eberhard Buhl and Miles Whittington 10.1 Overview 343
8.1 Overview 297 10.1.1 LTP: The First Two Decades 344
8.1.1 Neuronal Classication Issues 297 10.2 Transient Activity-dependent Plasticity in
8.1.2 Input Specicity of Extrinsic Afferents 298 Hippocampal Synapses 347
 Contents xv

10.2.1 Short-term Activity-dependent Changes in Synaptic 10.4.8 Membrane Spanning Molecules Contribute to
Efficacy in the Hippocampal Formation 347 Signaling Between Presynaptic and Postsynaptic
10.2.2 Single Stimuli in Hippocampal Pathways Produce Sides of the Synapse 384
Two Transient Aftereffects: Facilitation and 10.4.9 Late LTP: Persistent Potentiation Requires Gene
Depression 347 Transcription and Protein Synthesis 386
10.2.3 Post-tetanic Potentiation Is the Sum of Two 10.4.10 Structural Remodeling and Growth of Spines Can
Exponential Components: Augmentation and Be Stimulated by Induction of LTP 393
Potentiation 350 10.5 LTP at Mossy Fiber Synapses 398
10.3 NMDA Receptor-dependent Long-term Potentiation: 10.5.1 Mossy Fiber Synapses Display Striking Short-term
Properties and Determinants 350 Plasticity 398
10.3.1 Long-term Potentiation: Tetanic Stimulation Induces 10.5.2 Basic Characteristics of NMDA Receptor-
a Persistent Increase in Synaptic Efficacy 350 independent LTP at Mossy Fiber Synapses 398
10.3.2 Time Course of LTP: Rapid Onset and Variable 10.5.3 Induction Mechanisms of Mossy Fiber LTP 398
Duration 353 10.5.4 Expression of Mossy Fiber LTP Is Presynaptic 401
10.3.3 Three Distinct Temporal Components of 10.5.5 E-LTP and L-LTP at Mossy Fiber Synapses Can Be
Potentiation: STP, Early LTP, Late LTP 353 Distinguished by the Effects of Protein Synthesis
10.3.4 Input-Specicity of LTP: Potentiation Occurs Only at Inhibitors 402
Active Synapses 354 10.5.6 Summary 403
10.3.5 Associativity: Induction of LTP Is Inuenced by 10.6 LTP Can Be Modulated by Other Neurotransmitters,
Activity at Other Synapses 355 Neuromodulators, and Effectors and by Endogenous
10.3.6 Requirement for Tight Coincidence of Presynaptic and Circadian Rhythms 403
and Postsynaptic Activity Implies a Hebbian 10.6.1 Modulation by Other Neurotransmitters and
Induction Rule 356 Neuromodulators 403
10.3.7 Molecular Basis for the Hebbian Induction Rule: 10.6.2 Cyclical Inuences Modulate Induction
Voltage Dependence of the NMDA Receptor Explains of LTP 406
Cooperativity, Input Specicity, and Associativity 10.6.3 Neurogenesis and LTP 407
357 10.7 Long-term Depression and Depotentiation:
10.3.8 Spike Timing-dependent Plasticity (STDP) 358 Properties and Mechanisms 407
10.3.9 Ca2 Signaling in LTP 359 10.7.1 Overview 407
10.3.10 Metabotropic Glutamate Receptors Contribute to 10.7.2 NMDAR-dependent LTD: Properties and
Induction of NMDA Receptor-dependent LTP 360 Characteristics 408
10.3.11 Role of GABA Receptors in the Induction of 10.7.3 NMDAR-dependent LTD: Induction
NMDAR-dependent LTP 361 Mechanisms 410
10.3.12 E-S Potentiation: A Component of LTP That 10.7.4 NMDAR-dependent LTD: Expression
Reects Enhanced Coupling Between Synaptic Drive Mechanisms 412
and Cell Firing 362 10.7.5 mGluR-dependent LTD 414
10.3.13 Metaplasticity: The Magnitude and Direction of 10.7.6 Homosynaptic Depotentiation 416
Activity-dependent Changes in Synaptic Weight Are 10.7.7 Heterosynaptic LTD and Depotentiation:
Inuenced by Prior Activity 364 Activity in One Input Can Induce LTD in
10.3.14 Synaptic Scaling and Long-term Changes in Another 417
Intrinsic Excitability 368 10.7.8 LTD and Depotentiation at Mossy Fiber CA3
10.4 NMDA Receptor-dependent LTP: Expression Pyramidal Cell Synapses 419
Mechanisms 369 10.8 Synaptic Plasticity and Inhibitory Pathways 420
10.4.1 From Induction to Expression of LTP 369 10.8.1 LTP and LTD at Glutamatergic Synapses on
10.4.2 STP Is a Transient Presynaptic Form of Plasticity Interneurons 420
369 10.8.2 LTP and LTD at GABAergic Synapses 422
10.4.3 Early LTP Involves Multiple Protein Kinase- 10.9 LTP and LTD in Development and Aging and in
dependent Mechanisms 369 Animal Models of Cognitive Dysfunction 422
10.4.4 Site of Expression of Early LTP: Experimental 10.9.1 Hippocampal Synaptic Plasticity During
Approaches 374 Development 422
10.4.5 E-LTP: Presynaptic Mechanisms of Expression 376 10.9.2 Synaptic Plasticity and the Aging
10.4.6 E-LTP: Postsynaptic Mechanisms of Hippocampus 423
Expression 377 10.9.3 Animal Models of Cognitive Decline 425
10.4.7 Retrograde Signaling System Is Required for 10.10 Functional Implications of Hippocampal Synaptic
Communication Between the Postsynaptic Site of Plasticity 427
Induction and the Presynaptic Terminal 383 10.10.1 Synaptic Plasticity and Memory Hypothesis 427
xvi Contents

10.10.2 Detectability: Is Learning Associated with the 11.4.7 Olfactory Stimulation Can Elicit Hippocampal
Induction of LTP? 429 Gamma and Beta Waves 485
10.10.3 Anterograde Alteration: Do Manipulations That 11.5 Single-cell Recording in the Hippocampal Formation
Block the Induction or Expression of Synaptic Reveals Two Major Classes of Units: Principal Cells
Plasticity Impair Learning? 432 and Theta Cells 486
10.10.4 Retrograde Alteration: Does Further Induction or 11.5.1 Distinctive Spatial Cells Complex-spike Place
Reversal of LTP Cause Forgetting? 441 Cells, Head-direction Cells, and Grid Cells Are
10.10.5 Mimicry 443 Found in Various Regions of the Hippocampal
10.10.6 Synaptic Plasticity, Learning, and Memory: The Formation 487
Story So Far 443 11.6 Theta Cells 490
References 444 11.6.1 Theta Cells Fire with a Consistent Phase Relation to
EEG Theta 490
Chapter 11 11.6.2 Pharmacology of Theta Cells 490
11.6.3 Hippocampal Theta Cells Have Behavioral
Hippocampal Neurophysiology in the Behaving
Correlates Similar to Those of the Hippocampal
Animal 475
EEG 490
John OKeefe 11.7 Complex-spike Cells and Spatial Processing 491
11.1 Overview 475 11.7.1 Place Cells Signal the Animals Location in an
11.2 Hippocampal Electroencephalogram Can Be Environment 491
Classied into Distinct Patterns, with Each Providing 11.7.2 Basic Properties of Place Fields 492
Information About an Aspect of Hippocampal 11.7.3 Place Fields are Nondirectional in Unrestricted
Function 477 Open-eld Environments but Directional When
11.2.1 Hippocampal EEG Can Be Classied into Four Types Behavior is Restricted to Routes 495
of Rhythmical and Two Types of Nonrhythmical 11.7.4 What Proportion of Complex-spike Cells Are Place
Activity 477 Cells? 497
11.2.2 Each EEG Pattern Has Distinct Behavioral 11.7.5 Frame of Reference of Place Fields 498
Correlates 477 11.7.6 Place Fields Can Be Controlled by Exteroceptive
11.3 Hippocampal Theta Activity 479 Sensory Cues 499
11.3.1 Hippocampal Theta Activity: Historical 11.7.7 Idiothetic Cues Can Control Place Fields 502
Overview 479 11.7.8 Are Place Cells Inuenced by Goals, Rewards, or
11.3.2 Hippocampal Theta Activity Is Comprised of Two Punishments? 504
Components, a-Theta, and t-Theta, Which Can Be 11.7.9 Temporal Patterns of Place Cell Firing 505
Distinguished on the Basis of Behavioral Correlates 11.7.10 Place Fields in Young and Aged Animals 507
and Pharmacology 480 11.7.11 Hippocampal Place Cell Firing Is Inuenced by
11.3.3 Both Types of Theta Activity Are Dependent on the Other Areas of the Brain 509
Medial Septal/DBB but Only t-Theta Is Dependent 11.7.12 Primate Hippocampal Units also Exhibit Spatial
on the Entorhinal Cortex 480 Responses 510
11.3.4 t-Theta Occurs During Movement 11.8 Place Cells Are Memory Cells 511
Through Space 481 11.8.1 Hippocampal Place Cells Remember the Animals
11.3.5 a-Theta Occurs During Arousal and/or Attention as Location for Several Minutes During a Spatial
well as Movement 481 Working Memory Task 512
11.3.6 Theta and Sleep 481 11.8.2 Place Field Plasticity During Unidirectional
11.3.7 Theta Activity in Nonhippocampal Areas 481 Locomotion 512
11.3.8 Does the Hippocampal EEG in Monkeys and 11.8.3 Cue Control over Hippocampal Place Cells Can
Humans Have a Theta Mode? 482 Change as a Function of Experience 513
11.3.9 Functions of Theta 482 11.8.4 Control of the Angular Orientation of Place
11.4 Non-theta EEG Patterns in the Hippocampal EEG: Cells in a Symmetrical Environment Can
LIA, SIA, Ripples, Beta, and Gamma 483 Be Altered by the Animals Experience of Cue
11.4.1 Sharp Waves, Ripples, and Single Units During Large Instability 514
Irregular Activity 483 11.8.5 Complex-spike Cell Firing and Connectivity During
11.4.2 Dentate EEG Spikes During LIA 484 Sleep Is Modulated by Prior Spatial Learning
11.4.3 Pharmacology of LIA 484 Experiences 516
11.4.4 Behavioral Correlates and Functions of LIA 484 11.8.6 NMDA Receptor Confers Mnemonic Properties on
11.4.5 Small Irregular Activity 485 Place Cell Firing 516
11.4.6 Beta/Gamma Activity in the 11.8.7 Summary of Place Cell Plasticity 517
Hippocampus 485 11.9 Head Direction Cells 517
 Contents xvii

11.9.1 Head Direction Cells are Controlled by Distal Chapter 12

Sensory Cues 518 Functional Role of the Human Hippocampus 549
11.9.2 Angular Distance Between any Given Pair of Head
Craig Stark
Direction Cells Always Remains Constant 519
11.9.3 Head Direction Cells Can also Be Controlled by 12.1 Overview 549
Idiothetic Cues 519 12.2 Patient H.M. 550
11.9.4 Head Direction Cells Are Found in Different 12.3 Methods for Studying Human Hippocampal
Anatomically Connected Brain Areas 520 Function 552
11.9.5 Dorsal Tegmental Nucleus of Gudden Provides 12.3.1 Behavioral Tasks and Terms 552
Information About the Direction and Angular 12.3.2 Behavioral Measures 553
Velocity of the Animals Head Rotation 521 12.3.3 Anoxia and Bilateral Hippocampal Lesions 554
11.10 Interactions Between Hippocampal Place Cells and 12.3.4 Depth Electrode Recordings 556
Head Direction Cells 521 12.3.5 Neuroimaging 557
11.11 Hippocampal Complex-spike Cells Have Been 12.3.6 Technical Challenge: Alignment of MTL Regions
Implicated in Nonspatial Perception and Across Participants 559
Learning 523 12.4 Dissociating Hippocampal Function 561
11.11.1 Hippocampal Cells Have Been Implicated in 12.4.1 Explicit Versus Implicit 561
the Processing of Nonspatial Sensory 12.4.2 Encoding Versus Retrieval 564
Information 523 12.4.3 Time-limited Role in Declarative Memory 565
11.11.2 Hippocampal Unit Activity May Show 12.4.4 Spatial Memory 567
Correlations with Different Aspects of Nonspatial 12.4.5 Associations, Recollections, Episodes,
Learning Tasks 524 or Sources 569
11.11.3 Hippocampal Unit Activity During Aversive 12.5 Conclusions 573
Classical Conditioning 524 References 575
11.11.4 Nictitating Membrane Conditioning in the Rabbit:
Chapter 13
Role of Theta 525
11.11.5 Single-unit Recording in the Hippocampus Theories of Hippocampal Function 581
During Nictitating Membrane Conditioning Richard Morris
of Rabbits 526 13.1 Overview 581
11.11.6 Hippocampal Unit Recording During Operant 13.2 Cognitive and Behavioral Neuroscience,
Tasks 528 Interventional Techniques, and the
11.11.7 Comparison of Hippocampal Cells During Hippocampus 582
Operant Conditioning and Place Tasks in 13.2.1 Value of Interventional Studies to Identify
Rats 532 Function 582
11.11.8 Hippocampal Units During Nonspatial Learning 13.2.2 Lesions, Functional Hypotheses, and
in Nonhuman Primates 533 Behavioral Tasks 583
11.11.9 Hippocampal Units During Nonspatial Learning 13.2.3 Contemporary Lesion Techniques: Pharmacological
in Humans 533 and Genetic Interventions 585
11.11.10 Conclusions 534 13.2.4 Biological Continuity of Hippocampal
11.12 Other Distinctive Cells in the Hippocampal Function 588
Formation and Related Areas 534 13.3. Declarative Memory Theory 591
11.12.1 Subicular Region Has Fewer Place Cells than 13.3.1 Outline of the Theory 591
the Hippocampus Proper, and Their Properties 13.3.2 Development of a Primate Model of
Differ 535 Amnesia 596
11.12.2 Presubiculum Contains Several Classes of 13.3.3 Domains of Preserved Learning Following Medial
Spatial Cell 536 Temporal Lobe Lesions in Primates 601
11.12.3 Parasubiculum 537 13.3.4 Selective Lesions of Distinct Components of the
11.12.4 Spatial Cells in the Entorhinal Cortex 537 Medial-temporal Lobe Reveal Heterogeneity of
11.12.5 Cells in the Perirhinal Cortex Code for the Function 602
Familiarity of Stimuli 537 13.3.5 Remote Memory, Retrograde Amnesia, and the
11.12.6 Cells in the Medial Septum Are Theta Time Course of Memory Consolidation in
Cells 538 Primates 606
11.12.7 Summary of Extrahippocampal Place Field 13.3.6 Critique 608
Properties 538 13.4 Hippocampus and Space: Cognitive Map Theory
11.13 Overall Conclusions 538 of Hippocampal Function 617
References 540 13.4.1 Outline of the Theory 618
xviii Contents

13.4.2 Representing Spatial Information, Locale Processing, 14.5.5 Consolidation and Cross-modal Binding of
and the Hippocampal Formation 623 Events in Memory 738
13.4.3 Using Spatial Information: Spatial Navigation and 14.5.6 Hippocampal Contributions to Various Types
the Hippocampal Formation 627 of Memory and Retrieval 739
13.4.4 Comparative Studies of Spatial Memory and the 14.6 Reconciling the Hippocampal Roles in Memory
Distinction Between Spatial and Associative and Space 740
Learning 640 14.7 Conclusions 744
13.4.5 Storage and Consolidation of Spatial References 744
Memory 650
13.4.6 Critique 654 Chapter 15
13.5 Predictable Ambiguity: Congural, Relational, Stress and the Hippocampus 751
and Contextual Theories of Hippocampal Richard Morris
Function 656 15.1 Overview 751
13.5.1 Congural Association Theory 658 15.2 Glucocorticoid Receptors and Hippocampal
13.5.2 Relational Processing Theory: Renement of the Function 753
Declarative Memory Theory 662 15.2.1 Glucocorticoid Receptors Are Present in the Animal
13.5.3 Contextual Encoding and Retrieval 668 and Human Hippocampus 753
13.5.4. Critique 676 15.2.2 There is an Inverted U-Shape Function Between
13.6 Episodic Memory, Hippocampus, and Neurobiology Level of Stress and Memory 754
of Rapid Context-specic Memory 677 15.2.3 Stress Modulates Intrinsic Hippocampal Excitability
13.6.1 Concept of Episodic Memory 677 and Activity-dependent Synaptic Plasticity Associated
13.6.2 Scene Memory as a Basis for Episodic Memory and with Learning and Memory 756
Top-down Control by the Prefrontal Cortex 679 15.3 Stress and Hippocampal Structure 759
13.6.3 What, Where, and When: Studies of Food-caching 15.3.1 Chronic Exposure to High Levels of Stress or Stress
and Sequence Learning 682 Hormones Is Associated with Structural Changes in
13.6.4 Problem of Awareness 686 the Hippocampus 759
13.6.5 Elements of a Neurobiological Theory of the Role of 15.3.2 Stress or Stress Hormones Can Impair Neurogenesis
the Hippocampus in Episodic-like Memory 687 in the Hippocampus 759
References 694 15.3.3 Fetal Programming of GC Regulation 760
15.4 Other Higher Brain Structures Implicated in Stress
Chapter 14
and Their Interaction with the Hippocampus 761
Computational Models of the Spatial and Mnemonic 15.5 How the Hippocampus Orchestrates Behavioral
Functions of the Hippocampus 715 Responses to Arousing Aversive Experiences 762
Neil Burgess References 764
14.1 Overview 715
14.2 Introduction 715 Chapter 16
14.3 Hippocampus and Spatial Representation 716 Hippocampus and Human Disease 769
14.3.1 Representing Spatial Location and Orientation: Matthew Walker, Dennis Chan, and Maria Thom
Data 716 16.1 Overview 769
14.3.2 Representing Spatial Location: Feedforward 16.2 Mesial Temporal Lobe Epilepsy and Hippocampal
Models 719 Sclerosis 770
14.3.3 Representing Spatial Location and Orientation: 16.2.1 Introduction 770
Feedback Models 721 16.2.2 Clinical Features 771
14.3.4 Modeling Phase Coding in Place Cells 727 16.2.3 Etiology 774
14.4 Hippocampus and Spatial Navigation 729 16.2.4 Pathophysiology 777
14.4.1 Spatial Navigation: Data 729 16.2.5 Conclusion 789
14.4.2 Spatial Navigation: Feedforward Models 729 16.3 Alzheimers Disease 789
14.4.3 Spatial Navigation: Feedback Models 732 16.3.1 Introduction 789
14.5 Hippocampus and Associative or Episodic 16.3.2 Clinical Features 789
Memory 733 16.3.3 Genetics 795
14.5.1 Hippocampus and Memory: Data 733 16.3.4 Pathophysiology 796
14.5.2 Marrs Hippocampo-neocortical Model of 16.3.5 Treatment Options 801
Long-term Memory 734 References 803
14.5.3 Associative Memory and the Hippocampus 734
Index 813
14.5.4 Hippocampal Representation, Context,
and Novelty 737

Contributors

David Amaral, PhD Michael Frotscher, Dr Med

The M.I.N.D. Institute Institute of Anatomy
Department of Psychiatry and Behavioral Sciences University of Freiburg
University of California-Davis Freiburg, Germany
Davis, California, United States
Elizabeth Gould, PhD
Per Andersen, MD PhD Department of Psychology
Department of Neurophysiology Princeton University
University of Oslo Princeton, New Jersey, United States
Oslo, Norway
Dimitri Kullmann, MBBS DPhil FRCP
Tim Bliss, PhD Institute of Neurology
Division of Neurophysiology University College London
MRC National Institute for Medical Research, London London, United Kingdom
London, United Kingdom
Pierre Lavenex, PhD
Eberhard Buhl, Dr Med* Institute of Physiology
School of Biomedical Sciences Department of Medicine
Leeds University University of Fribourg
Leeds, United Kingdom Fribourg, Switzerland

Neil Burgess, PhD Chris McBain, PhD

Institute of Cognitive Neuroscience Laboratory of Cellular and Synaptic Neurophysiology
University College London National Institute of Child Health and Human Development
London, United Kingdom Bethesda, Maryland, United States

Dennis Chan, PhD MRCP Richard Morris, DPhil

Dementia Research Group Laboratory for Cognitive Neuroscience
Institute of Neurology College of Medicine and Veterinary Medicine
University College London University of Edinburgh
London, United Kingdom Edinburgh, Scotland

Graham Collingridge, PhD John OKeefe, PhD

MRC Centre for Synaptic Plasticity Department of Anatomy & Developmental Biology
University of Bristol University College London
Bristol, United Kingdom London, United Kingdom

*Deceased
xx Contributors

Pavel Osten, PhD Maria Thom, MRCPath

Department of Molecular Neurobiology Department of Clinical and Experimental Epilepsy
Max-Planck-Institute for Medical Research Institute of Neurology
Heidelberg, Germany University College London
London, United Kingdom
Lszl Seress, MD PhD
Central Electron Microscopic Laboratory, Faculty of Medicine Matthew Walker, PhD MRCP
University of Pcs Department of Clinical and Experimental Epilepsy
Pcs, Hungary Institute of Neurology
University College London
Rolf Sprengel, PhD
London, United Kingdom
Department of Molecular Neurobiology
Max-Planck-Institute for Medical Research Miles Whittington, PhD
Heidelberg, Germany Department of Neuroscience
Newcastle University
Nelson Spruston, PhD
Newcastle upon Tyne, United Kingdom
Department of Neurobiology and Physiology
Institute for Neuroscience William Wisden, PhD
Northwestern University Department of Neuroscience
Evanston, Illinois, United States University of Aberdeen
Aberdeen, Scotland
Craig Stark, PhD
Department of Psychology
Johns Hopkins University
Baltimore, Maryland, United States
The Hippocampus Book
This page intentionally left blank
1 Per Andersen, Richard Morris, David Amaral, Tim Bliss, and John OKeefe

The Hippocampal Formation

rons in the brain. As a result, we now know a great deal about

1.1 Overview their development, synaptogenesis, neurotransmitter recep-
tors and ion channels, micro-circuitry, and cell biological
Buried deep within the medial temporal lobe of the human machinery. No less impressive has been the molecular analysis
brain lies a group of many millions of neurons organized into of neurotransmission in hippocampal cells; what has been
a network quite different from that found anywhere else in the learned from this has revealed general properties that seem to
nervous system. It is a structure whose bulb-like shape, pro- apply in other areas of the central nervous system. We also
truding into the lateral ventricles, has captivated anatomists know something about why and when these cells are activated
since the rst dissections took place in classical Egypt. The in the living brain. Recordings from freely moving animals as
hippocampal formation is a group of brain areas consisting of they navigate around a space with which they have become
the dentate gyrus, hippocampus, subiculum, presubiculum, familiar have shown that individual hippocampal pyramidal
parasubiculum, and entorhinal cortex. The basic layout of cells re in particular locations. This nding led to the devel-
cells and ber pathways of the hippocampal formation is opment of new behavioral tools to study the neural mecha-
much the same in all mammals (Fig. 11). nisms of memory in animals. Along the extensive dendritic
There are several reasons the hippocampus has attracted arborizations of these pyramidal cells, there can be many
the interest of scientists in the many disciplines that now thousands of miniature dendritic spines. These are the sites
characterize modern neurosciencethe hippocampus has where most of the excitatory synapses are to be found. An
something for everyone. Whether you are a psychologist important nding is that the efficiency with which these
interested in memory, a synaptic physiologist investigating excitatory synapses transmit messages can vary as a function
neuronal and synaptic plasticity, or a computational neurosci- of neural activity. The circumstances bringing about this
entist wanting to build a neural network model, the hip- synaptic plasticity are now well understood, and we are begin-
pocampus and its associated structures are an attractive set of ning to understand some of the underlying biochemical
brain structures on which to work. In parallel, clinicians con- mechanisms. Many suspect that synaptic plasticity is a key
cerned with the basis of neurological conditions such as mechanism of memory. Another exciting feature of the hip-
epilepsy or Alzheimers disease had their attention drawn to pocampal formation is that the dentate granule cell is one of
the hippocampal formation because of the pathological the rare types of nerve cell that multiplies throughout life.
processes observed to occur there and the opportunities that Adult neurogenesis is likely to yield important lessons about
scientic study of this area of the brain offers for novel thera- neuronal growth and survival. This is an area of obvious
peutics. The hippocampus has been a neural Rosetta Stone. potential importance for neuronal repair and therapeutic
The striking discovery, 50 years ago, that patient H.M. had a intervention.
relatively pure memory decit after surgical excision of the These observations have emerged from the widespread fas-
medial temporal lobe for the relief of epilepsy had a profound cination with this beautiful cortical structure that so attracted
effect on the study of memory and, through that, on our the early anatomists. As we shall see, certain structural princi-
understanding of the functions of the hippocampus itself. ples that are characteristic of the hippocampal formation have
During the last 30 years, the pyramid-shaped cells of the greatly facilitated structural, physiological, and behavioral
hippocampus have become the most intensively studied neu- analyses.

3
4 The Hippocampus Book

Figure 11. Nissl-stained coronal sections

through the rat, monkey, and human hippocam-
pal formation. Although there are obvious dif-
ferences in certain regions, the most striking
feature is the general similarity of this brain
region across phylogeny.

only some of them. There is also general agreement that some

1.2 Why Study the Hippocampal Formation types of learning and memory are propositionalthat is, they
on Its Own? relate to memories that can be described as a series of propo-
sitional statementswhereas other types, such as the process
Can one protably focus on one group of brain structures and of acquiring motor skills, are not. The hippocampus is
appear to ignore the rest of the brain? The brain is so heavily engaged in remembering information that can be described in
interconnected that it is misleading to think about the func- a propositional or declarative manner. That much is clear.
tion of one small part of it in isolation. What it does and how Beyond that, there remain many areas of disagreement and
it does it might be understood better by comparing it with debate that continue to fuel imaginative new research in
other brain areas and by thinking about how it works in con- behavioral and cognitive neuroscience. A striking feature of
junction with them to realize the seamless fabric of normal this strand of research has been the constructive convergence
brain function. We have much sympathy with this view, not of both top-down and bottom-up approaches to the study
least because the general signicance of the many exciting dis- of brain function. Work on the function of the hippocampal
coveries that have been made about the hippocampus can formation, arguably more than any other brain area, has led to
only become clear with a more inclusive approach in which it the development of current concepts about the organization
and other parts of the brain are considered together. We of memory systems in the brain.
believe, nonetheless, that there is still value in our focused Some traditionalists still hold that the various subjects that
approach. There are two primary reasons for singling out the make up the biomedical sciencessuch as psychology, physi-
hippocampal formation as a brain area worthy of a book in its ology, biochemistryeach has a self-contained level of analy-
own right. They have to do with trying to understand its func- sis that is logically independent of other levels of analysis.
tion(s) within the brain and the way in which work on hip- With this view, a psychological problem requires an answer in
pocampal tissue has revealed general principles in strictly psychological terms and a physiological problem a
neuroscience. physiological approach. The emerging view at the start of the
First, contemporary research points strongly to the hip- twenty-rst century, however, is that many years of a strictly
pocampus having a highly specic role in memory. There has psychological analysis of learning and memorybeginning
been a great deal of debate about how that role should be with the discovery of Pavlovian conditioning, the ensuing
characterized; and after the explosive period of research dur- dark phase of behaviorism, and the subsequent enlightenment
ing the last quarter century, a consensus is beginning to of the cognitive revolutionfailed to unravel the existence of
emerge. It is now generally recognized that there are multiple multiple memory systems. It was not until a different brain-
types of memory and that the hippocampus plays a part in systems approach was taken during the last part of the twen-
 The Hippocampal Formation 5

tieth century, an approach that emerged from studies of the other experiments tractable in a manner that remains difficult
memory problems faced by patient H.M. and other amnesic in other brain areas and would be wholly impracticable in
patients, that the major breakthroughs in our understanding vivo.
of the cerebral organization of memory came about. Such Beyond its key role in the development of in vitro tech-
thinking led to research that married brain and behavior in niques, the hippocampus has also been the area of choice
single-unit recording and lesion studies in awake primates for many other types of brain research. It is involved in a num-
and freely moving rodents, as well as the development of new ber of disparate neurological disorders, including epilepsy,
behavioral tests of learning and memory. There is a lesson Alzheimers disease, and cerebrovascular disease. Thus, the
here that we believe is generally applicable to other branches abnormal electrical activity that is at the root of seizures in
of functional neuroscience. Specically, the brain-systems epileptic patients is often easily detected in the hippocampus.
approach is one in which there is analytical value in juxtapos- Moreover, a hallmark feature of the neuropathology of tem-
ing the biomedical sciences with psychology and, in particu- poral lobe epilepsy is loss of neurons in several hippocampal
lar, in thinking carefully about what bits of the brain do what. elds. For example, the characteristic pathological changes of
The second justication for writing this book is that the Alzheimers disease manifest initially in the entorhinal cor-
hippocampus has preeminently been the structure in which texone of the components of the hippocampal formation
many of the general principles of modern neuroscience have and the disease spreads from there to involve the hippocampus
been studied and established. Thirty years ago, the most proper and ultimately the entire cerebral cortex. Such ndings
widely studied cell in the nervous system was the alpha have led to the development of model systems in which patho-
motoneuron of the ventral horn of the spinal cord. Today it is physiological events such as these may be studied and, hope-
the pyramidal cell of the hippocampus. One reason for this fully, alleviated by treatment.
development has been the peculiar anatomy of the hippocam- These two main reasons for writing the book therefore
pus, with all principal cells in a single layer and synaptic provide us with two intersecting themes that run through
inputs to well dened dendritic lamina. This simplied archi- many of its chapters. One perspective is functional, the other
tecture facilitated the recording of both synaptic signals and heuristic.
population discharges. Field potential recording became pos-
What does the hippocampus do?
sible because the well dened somatic and dendritic laminae
What can we learn about general principles of neuro-
allowed identication of current sources and sinks in extra-
science from studying the hippocampus?
cellular recordings made in vivo. It was through these studies
that the basic principle of unidirectional excitatory transmis- To achieve these twin aims, we have broken up our task
sion was rst described and the phenomenon of long-term into chapters that are largely organized along conventional
potentiation was discovered. Of particular importance for the disciplinary lines but where, when appropriate, both strands
study of synaptic function in the context of learning and of thought are intertwined.
memory was the fact that eld-potential recording of synaptic
responses could be made as easily in the freely moving animal
as in the hippocampal slice. The pyramidal cell also became a
popular cell to study because of the analytical potential of the 1.3 Dening the Contemporary Era
in vitro brain slice. The rst brain slices were made from neo-
cortex and piriform cortex with limited synaptic activation. Our starting point when preparing this book was to dene the
With its better identication of cells and input bers, the contemporary era of research as beginning in the 1960s and to
development of the transverse hippocampal slice revolution- regard discoveries made before that as modern and those
ized neurophysiology and neuropharmacology. Of course, made afterward as contemporary. Any division of this kind
many fundamental concepts had been worked out before- is arbitrary, but the period of the 1960s and 1970s was a water-
handin the axon of the giant squid, spinal cord, and cere- shed for both functional and mechanistic studies of the hip-
bellumbut the hippocampal slice rendered analytical pocampus. That decade saw the rst intracellular analysis
extra- and intracellular studies of mammalian cells and iden- of synaptic, antidromic, and epileptiform activities of hip-
tied synapses feasible on an unprecedented scale. Work in the pocampal neurons; the characterization and interpretation of
hippocampus has been a major contributor to our under- eld potentials signaling excitation and inhibition; and elec-
standing of the actions and mechanisms of various types of tron micrographs of synapse types and their distribution
synapse and the various classes of receptors for excitatory and along the cell bodies and their extensions. It was also at this
inhibitory amino acids, the many transmitter uptake mecha- time that several of the new tract-tracing techniques became
nisms, activity-dependent synaptic plasticity, and the deleteri- available to anatomists, replacing the classic degeneration
ous consequences of excitotoxicity for brain cells. It is the techniques hitherto used to identify regional connectivity.
two- or three-layered architecture of the hippocampus and its These advances made it possible to map the extrinsic connec-
capacity to survive in vitro for long periods of time coupled tions of the hippocampal formation at a previously unachiev-
with a strict layering of synapses in the dendritic tree that has able level of sensitivity and detail. During the same decade,
rendered the design and analysis of electrophysiological and hippocampal place cells were discovered and the phenomenon
6 The Hippocampus Book

of long-term potentiation was rst described in detail. Chapter 5 takes us forward to a detailed analysis of indi-
Methodological developments of the 1970s included the hip- vidual cells in which we present, side by side, both anatomical
pocampal slice, new tests of recognition memory for primates, and physiological ideas. To understand the principal cell types
and the open-eld watermaze, a behavioral technique to study of the hippocampal formationpyramidal and granule
learning in rodents that has since been used extensively to cellsand the several types of interneuron, one must know
analyze the role of the hippocampus in spatial navigation. what these cells look like, know about their inputs and out-
Mathematicians also started thinking about the hippocampus puts, and grapple with the biophysical principles that govern
as a neural network and so set in motion a theoretical neuro- how current injected into the cell at a dendritic synapse con-
science that has attracted increasing attention. tributes to cell ring. It is necessary to understand how
synapses at various dendritic locations differ in their synaptic
effects and how these individual effects sum to control the cell
discharge pattern. The large group of after-hyperpolarizing
1.4 Organization and Content of the Book responses is essential for cellular behavior. The important
modulatory effects of a set of controlling systems comprise a
During the discussions about when the contemporary era major topic.
actually started, we recognized the enormous contributions The intricacies of neurotransmission are considered in
made beforehand and upon which so much of modern re- Chapters 6 and 7 from physiological, pharmacological, and
search rests. Accordingly, we have devoted Chapter 2 to a his- molecular biological perspectives. Chapter 6 considers how
torical discussion of the key discoveries and concepts of the release of glutamate from presynaptic terminals in the
hippocampal neurobiology of the modern era up to about the hippocampus and its action on various glutamatergic recep-
1970s. In it, we also identify some of the key gures whose tors mediate excitatory synaptic transmission. It also discusses
work helped to usher in the contemporary era. In addition, we the essential role of inhibitory synaptic transmission at
highlight key concepts in neurobiology that were rst estab- GABAergic synapses. Because the hippocampus is one of the
lished through work on the hippocampal formation. few regions where it is possible to have experimental control
In Chapter 3, we outline the anatomy of the hippocampal over a large set of converging inputs to an individual cortical
formation, beginning with issues of nomenclature and def- cell, most of our ideas on cellular integration are derived from
inition. Here we present the three-dimensional organization studies on hippocampal tissue. This chapter describes the
of the rodent, monkey, and human hippocampal formation, intricate molecular machinery responsible for the mobiliza-
the detailed architecture of each of its component structures, tion of glutamate-containing vesicles, the exocytotic release
and the patterns of cellular interconnectivity. Straightaway, we machinery for glutamate, and its postsynaptic actions. It also
found that we disagreed about certain conceptual issues, such outlines the various ligand-binding and modulatory sites on
as the status of the lamella hypothesis that was proposed these receptors, the insights coming from the relatively new
nearly 30 years ago on the basis of electrophysiological data. work on receptor subunits afforded by molecular biology and
Modern neuroanatomical tract-tracing studies are thought the mechanisms responsible for transmitter uptake. In many
not to support this concept in its original form, although of these studies, genetically altered animals have been essential
the electrophysiological data remain on the table. After long tools. The chapter touches on the important and developing
debate, we arrived at a description of the extent to which con- topic of neuromodulatory transmitters in the hippocampus,
nectivity is primarily in the transverse plane and the extent to such as acetylcholine, norepinephrine, 5-hydroxytryptamine,
which longitudinal and other more diffuse projections are a dopamine, and peptides such as dynorphin and somatostatin.
characteristic feature of hippocampal anatomy. Similarly, The transmitters inuence hippocampal excitability and
although recognizing that the original concept of the tri- second-messenger effector mechanisms in an orchestrated
synaptic circuit may be too narrow, we continue to stress the cascade of formidable complexity. Chapter 6 also discusses
sense in which unidirectional connectivity is a distinctive fea- how the many neuromodulatory systems inuence the tradi-
ture of the hippocampus compared to the neocortex, includ- tional synaptic processes and the mechanisms by which they
ing the parallel connections made from the entorhinal cortex are mediated. These studies often are models for the general
to the dentate gyrus, and to areas CA3 and CA1 of the hip- neuronal activity seen in other parts of the nervous system.
pocampus proper. Chapter 7 discusses the molecular biology of hippocampal
Chapter 4 considers how this intricate structure develops cells with particular emphasis on synaptic function. Here,
both from the perspective of overall organization and from much of the pioneering work on glutamate receptor catego-
that of individual cells. Here the interplay moves from pure rization was made in the spinal cord, whereas the important
anatomy to genes and molecular biology. Modern studies of discovery of the voltage-dependent magnesium block of the
the development of the hippocampus are characterized by N-methyl-D-aspartate (NMDA) receptor channel was discov-
molecular studies in which the temporal expression of the ered in hippocampal tissue. Notably, the hippocampal slice
genes responsible for migration, neuronal specication, and has been of critical importance for developing the modern
axonal guidanceand for major cellular constituents such as understanding of ionotropic and metabotropic glutamate
receptor subunitsare being mapped out. receptors.
 The Hippocampal Formation 7

Having described the three-dimensional organization of and power. However, the extent to which bidirectional
the hippocampus, its cells, synapses, and their transmitters changes can be seen in the hippocampus in freely moving
and receptors, we need next to consider the local circuits that adult animals remains unclear. Chapter 10 also takes on the
these structures produce and the possible types of neuronal task of explaining the various behavioral studies that have
processing they make possible. Chapter 8 takes on this task attempted to identify the role LTP might play in learning,
and endeavors to explain how an impressive multitude of memory, or other cognitive processes. We outline studies that
inhibitory interneurons shape the number and pattern of have examined correlations between the physiological proper-
the participating cells in a near-physiological situation. These ties of synaptic plasticity and behavioral learning and the
interactions between principal cells and interneurons are par- effects of blocking it, saturating it, and erasing it. We offer no
ticularly evident during periods of rhythmical theta and denitive answer to the question of the purpose of these vari-
gamma activity. ous types of plasticity, but the tenor of our account is sympa-
The chapter on local circuits also provides an introduction thetic to a role in the encoding and storage of certain kinds of
to Chapters 9 and 10, where we confront activity-dependent memory.
neuronal and synaptic plasticity. One of the more extraordi- Chapter 11 continues the behavioral theme with a survey
nary discoveries of recent times has been that new neurons of studies in which electrical activity in structures of the hip-
can be formed in the adult brain. This is contrary to estab- pocampal formation has been recorded during various kinds
lished dogma but, interestingly, the dogma that had been chal- of behavior. The pioneering work of this kind was conducted
lenged in hippocampal studies nearly 30 years ago but was in rats and was heralded by the discovery of place cells noted
unappreciated at the time. The dentate gyrus is one of two earlier (neurons that re when rats occupy particular posi-
major sites of adult neurogenesis. In Chapter 9, we discuss the tions in space). The further discovery of a different group of
extent to which new cells and connections can be made, a eld cells that red during certain kinds of movement and in phase
of acute interest for basic and practical neurobiology. Chapter with slow-wave electroencephalographic activity was also
10 focuses on processes to explain the fact that, once formed, important, as were ndings of cells that code for the direction
hippocampal circuits are not immutable, as synapses through- in which an animals head is pointing or show grid-like pat-
out the various components of the hippocampal formation terns as an animal traverses a familiar space. These categories
show both short-term and long-term changes. The latter of cell are found in distinct areas of the hippocampal forma-
includes the intensively studied phenomenon of long-term tion and in different cell classes within a single area. Work in
potentiation (LTP). Research on this form of synaptic plastic- nonhuman primates has also identied a class of cells that are
ity has revealed a range of properties that are highly suggestive responsive to the animals view. The properties of these cells,
of a potential cellular memory mechanism. The effort to their role in mapping space, and, according to some, other
understand LTP mechanistically led to passionate debates that cognitive functions as well are gradually being unraveled.
have stimulated a host of ingenious experiments. Research on The process of discovery has been accompanied by the
synaptic plasticity, perhaps more than any other area of hip- development of a new single-cell methodology that permits
pocampal neurobiology, spans the twin themes of the book. simultaneous recording from large numbers of neurons. This
What is the purpose of LTP? Are its underlying neural mech- tetrode-recording technique enables the properties of several
anisms applicable to other areas of cortex? Research on the individual neurons in a population to be mapped simultane-
induction of this phenomenon has helped unravel one of the ously and has applications that extend well beyond hip-
most satisfyingly elegant biophysical mechanisms in the nerv- pocampal neurobiology.
ous system: the role of the NMDA receptor as a coincidence Chapter 12 begins with the discovery, almost 50 years ago,
detector. Having detected the conjunction of presynaptic that surgical damage to the hippocampal formation and
activity and postsynaptic depolarization, this receptor signals related structures on both sides of the brain causes a
the conjunction by an inux of the divalent cation calcium profound, lasting global amnesia. Once this unexpected
(Ca2), which then sets in train a cascade of biochemical consequence had been observed, neurosurgeons took care
effects. The discussion of the NMDA receptor complements not to perform bilateral medial temporal lobectomies. Such
the earlier description of its molecular basis in Chapter 7. It memory-impaired patients are therefore rare, but their very
activates signaling proteins associated with the receptor itself existence spurred the development of animal models through
and calcium-dependent kinases. It may also help trigger the which, it is hoped, the role of the hippocampus in memory
release of calcium from intracellular stores, thereby sustaining can be worked out. Modern functional imaging techniques are
the synaptically evoked Ca2 transient beyond the period that also providing new insights into the differential activation of
the NMDA receptor-associated ion channel remains open. various brain areas for different types of memory and differ-
The resulting biochemical cascade eventually results in an ent stages of memory processing. Such work has led to new
increase in synaptic strength. It should also be recognized concepts, such as the extent to which neuronal activation in a
that activity-dependent synaptic plasticity is not restricted to memory-related area reects the effort devoted to encoding or
increases in synaptic efficacydecreases can also be induced. retrieval or to the success of achieving either. Such conceptual
This bidirectional feature of synaptic plasticitythat what dissociations are almost impossible to draw on the basis of
goes up must come downadds computational exibility lesion studies alone.
8 The Hippocampus Book

Chapter 13 takes up the theme set by these human studies and patches that will eventually enable a comprehensive neu-
to lay out a number of prominent theories of hippocampal ral network account of hippocampal function to be woven
function that have been developed from work on animals. together are beginning to become apparent.
One theory concerns the role of the hippocampus in remem- Last come Chapters 15 and 16. The rst of these chapters
bering facts and events that can be consciously recalled, and picks up from a set of intriguing observations suggesting that
another has to do with its role in mapping and navigating the hippocampus, together with the amygdala and prefrontal
through space. Other theories have been developed to help cortex, play a role in regulating the hypothalamic-pituitary-
address some of the issues and problems that have arisen in adrenal axis, which is responsible for the release of stress hor-
connection with these two ideas, particularly in situations of mones. The hippocampus contains a particularly high density
predictable ambiguity. We also outline current interest in the of corticosteroid receptors of two major types, and this neu-
idea that the hippocampus plays a critical role in contextual roendocrine detection of mild or severe stress has a dramatic
encoding and retrieval, including episodic memory. This impact on hippocampal physiology and memory function,
section attempts to link behavioral studies on rapidly learned including effects on LTP. Glucocorticoid receptors act as tran-
forms of associative memory to the neurobiological observa- scription factors, and it has long been thought that the pri-
tions on circuitry and synaptic plasticity outlined earlier mary expression mechanism was an intracellular regulation of
in the book. Although one of the authors of this book is genes. However, it is now becoming clear that there are a vari-
an architect of the second theory discussed, we attempt to ety of more rapid effects of corticosteroid action in the hip-
present as objective an account of the state of the eld as pocampus. Chapter 16 considers hippocampal pathologies.
possible. This is, of course, a large subject in its own right, and we can-
Chapter 14 takes us to computational models of hip- not here do justice to a vast eld of clinical research. To
pocampal function. Some models focus on its global function, exclude it, however, would be wrong because all of us have an
but even when they work they often involve assumptions that interest in alleviating the impact of human neurological dis-
seem far removed from real neurobiology. These models are ease. The importance of these conditions is underscored by
the articial intelligence and idealized neural network models the prevalence of disease: Alzheimers disease alone affects
that can solve interesting cognitive problemsincluding nav- around 20 million people worldwide. Substantial advances
igation around the world or remembering associations in have been made in understanding the involvement of vari-
memorybut using processing units and learning algorithms ous components of the hippocampal formation in epilepsy
that may not exist among real hippocampal cells. Other mod- and Alzheimers disease, on which we focus; and this infor-
els incorporate more realistic assumptions about the hip- mation has underpinned improvements in diagnosis and
pocampal network and its neuronal components, the different treatment.
classes of cells, and their phase relationships; but their pro- This book is our tribute to the thousands of scientists who
cessing tends to be more limited. Ultimately, the value of these have contributed to our understanding of this beautiful and
models is to provide a predictive framework for understand- enigmatic part of the brain. We hope that it will help design
ing the mass of detail we have covered in other chapters of new, revealing experiments and thereby increase our insight
this book. To date, no model incorporates information realis- into its cellular, molecular, physiological, and behavioral
tically at each of the levels of the anatomical circuit, cellular processes. Should it also facilitate clinical studies, our satisfac-
architecture, and synaptic transmission. However, the pieces tion will be so much the greater.
2 Per Andersen, Richard Morris, David Amaral, Tim Bliss, and John OKeefe

Historical Perspective: Proposed Functions,

Biological Characteristics, and Neurobiological
Models of the Hippocampus

many functions, ranging from olfaction to motor function

2.1 Overview and even reason, stubbornness, and inventiveness. The
Bolognese anatomist Giulio Cesare Aranzi (circa 1564) was
This chapter has two major sections. In the rst, we provide a the rst to coin the name hippocampus, undoubtedly
selective overview of some of the historically important pro- because of its similarity to the tropical sh (Fig. 21).
posals concerning the function of the hippocampus. The cur- After the advent of microscopy, the hippocampus was per-
rent view, that the hippocampus plays a prominent role in haps even more impressive, with its characteristic, neatly reg-
memory, is dealt with in detail in Chapters 12 and 13. In the imented cellular arrangement. Although the hippocampal
second major section, we highlight the historically signicant formation is organized in a very different way from other cor-
advances in neurobiology that have been made by using the tical areas, it has still played an important role in unraveling
unique structure of the hippocampus for studies ranging some of the basic principles of cortical organization.
from dening the morphology of excitatory and inhibitory Its stunning structure, with cell populations apparently
synapses to developing computational models of memory. condensed into single layers, has attracted the attention of
This second section is itself divided into subsections that many investigators of the central nervous system. A striking
emphasize rst the biological characteristics of the hippocam- example is Camillo Golgis picture from 1886, where he used
pus and then the advantages of the hippocampus as a model his revolutionary new technique to illustrate the unique
system for neurobiological research. Many of the most recent organization of the hippocampus, here reproduced from
advances are dealt with in appropriate chapters, as we intend Golgi et al. (2001) (Fig. 22).
this chapter to be primarily a historical overview. As noted in Chapter 1, the two themes of this bookthe
overall functions of the hippocampus and how its study has
shed light on a range of general principles in neuroscience
can be highlighted by reference to the history of research on
the hippocampus. Early work inevitably focused on anatomi-
2.2 The Dawn of Hippocampal Studies cal issues, whereas later research successively brought in phys-
iological, behavioral, and other technologies in an effort to
From the very start of brain investigations, the hippocampus fathom the function of the hippocampus and through it the
has played a central role. No doubt this is due to the striking workings of the brain.
appearance of the large, bulging structure impressing itself
into the lateral ventricle and visible even to the ancient 2.2.1 A Famous Dispute About the Signicance
anatomists. Members of the Alexandrian school of medicine of the Hippocampus
were impressed by the elegant, curved structure, which when
seen with its contralateral half strongly resembles the coiled Interest in the hippocampus was particularly high during the
horns of a ram. Hence, the ancient scholars named the hip- eighteenth and nineteenth centuries, a notable example being
pocampus cornu ammonis, Latin for horn of the ram. This its curious role in the famous debate between Richard Owen
terminology survives in the acronyms for the hippocampal and Thomas Huxley at the British Association in Oxford in
subelds CA1 (cornu ammonis), CA2, and CA3. Over the cen- June 1860 following the publication of Charles Darwins book
turies, this part of the brain has been proposed as the seat of On the Origin of Species in 1859 (Gross, 1993). The superin-

9
10 The Hippocampus Book

Figure 21. Human hippocampus dissected free (left) and com- Figure 23. Hippocampus major and minor as seen in a human
pared to a specimen of Hippocampus leria (right). (Source: Courtesy brain after the dorsal cortex and corpus callosum have been
of Professor Laszlo Seress, University of Pecs.) removed.

tendent of the natural history collections of the British objected strongly, not least because of his own dissection
Museum, Richard Owen, claimed that only human brains results. Today, it is difficult to see how this particular bulge
contain a structure he called hippocampus minor, this being could have any bearing on the question of human evolution.
one essential distinguishing feature between Man and Apes. The name hippocampus minor is really a misnomer because
This structure is a relatively small bulge on the medial wall of it has nothing to do with the hippocampal formation per se
the posterior horn of the lateral ventricle (Fig. 23). Huxley but is a fold of white matter around the calcarine ssure, in
other words a visual cortical structure.
Figure 22. Section of a rabbit hippocampus stained with the origi-
nal Golgi method (1886). (Source: Golgi et al., 2001).

2.3 Early Ideas About Hippocampal Function

2.3.1 The Hippocampal Formation

and Olfactory Function

Although the hippocampus had been casually implicated in a

number of functions during the nineteenth and early twenti-
eth centuries, until the 1930s the hippocampal formation was
considered by neuroanatomists to be part of the olfactory sys-
tem. Perhaps this was because macroscopic investigations left
the impression that the hippocampus is especially large in
macrosmatic animals (with large olfactory mucosa and bulbs)
such as nocturnal insectivores and rodents. However, in these
macrosmatic animals, the absolute sizes of the olfactory parts
of the brain are quite small. In fact, it was claimed that the
ratio of the volume of the hippocampus to the olfactory bulb
is largest in humans (Rose, 1935). Indeed, as observed by
Brodal (1947):

From a functional point of view it is worth emphasiz-

ing that the development not only of the fornix, but
also of the mammillary body, the mammillo-thalamic
tract, the anterior (more particularly the antero-
ventral) nucleus of the thalamus and the cingular
gyrus runs roughly parallel with the degree of develop-
 Historical Perspective 11

ment of the hippocampus. All these structures reach pocampal formationhas subsequently proven incorrect.
their peak of development in man. More sensitive anatomical techniques have shown that in the
rodent the entorhinal cortex receives a massive direct projec-
A possible basis for the idea of an olfactory function for the tion from the olfactory bulb as well as secondary olfactory
hippocampus (an idea that was repeated in numerous text- inputs from the piriform and periamygdaloid cortices
books during the rst half of the twentieth century) is the (Shipley and Adamek, 1984). Even in the monkey, the anterior
appearance of a few early behavioral and clinical observations. portion of the entorhinal cortex is directly innervated by the
For example, David Ferrier (1876) observed movements of lateral olfactory tract (Amaral et al., 1987). Thus, the olfactory
the lip and nostrils on stimulation of the hippocampal lobe in system retains a privileged position in relation to the entorhi-
monkeys. John Hughlings Jackson and Charles Beevor (1890) nal cortex as none of the other sensory channels from which
reported a patient who had subjective olfactory sensations it receives information originates in primary or even higher
during seizures that originated in the olfactory cortex of the order unimodal sensory cortices; rather, it comes from poly-
periamygdaloid gyrus. Later, Wilder Peneld and Theodore sensory association cortices. Thus, although Brodals review
Erickson (1941) reported a patient who had olfactory sensa- was a milestone in the evaluation of hippocampal function
tions as part of his epileptic seizure and where it seems to be and is partially responsible for the currently held view that the
the hippocampus which must be the site of the discharge (p. hippocampal formation is not a major component of the
56). olfactory system, olfactory information certainly must con-
In a scholarly and comprehensive review, Alf Brodal (1947) tribute to the functions in which the hippocampus is engaged.
summarized the evidence for and against such a role for the Interestingly, olfactory learning and memory tasks are now
hippocampal formation. He noted that phylogenetic and widely used in studies of hippocampal function in rodents.
comparative neuroanatomical studies suggested that the hip-
pocampus develops in parallel with the olfactory portions of
2.3.2 The Hippocampal Formation and Emotion
the brain and that it is particularly prominent in macrosmatic
mammals such as rodents and insectivores. Moreover, even in Another inuential neuroanatomical hypothesis was pro-
the gross brain or in normal histological preparations, bers posed by James W. Papez (1937), who suggested that the hip-
from the lateral olfactory tract can easily be traced to brain pocampus was part of a circuit that provides the anatomical
regions surrounding the hippocampal formation. Even so, substrate of emotion. He was inuenced by the work of Walter
Brodal raised a number of arguments against an olfactory role B. Cannon (1929) and Philip Bard (1934), which indicated
for the hippocampus, suggesting that the association of the that the hypothalamus was essential for evocation of the
hippocampal formation with olfactory function was largely autonomic and visceral aspects of emotional behavior. He
based on circumstantial evidence. accepted the view of Cannon and Bard that emotion has two
Central to his thesis was the claim that bers arising in the component processes: emotional behavior and the cognitive
olfactory bulb did not, in fact, directly innervate any portion appreciation of emotion. In an attempt to explain certain
of the hippocampus. In particular, although he agreed that aspects of emotion, he proposed a circuit that intercon-
some olfactory bers innervated the anterior portions of the nected cortical and subcortical structures (Fig. 24). In this
parahippocampal gyrus, they did not extend back into the now famous circuit, which bears his name (Papez circuit), he
caudal portion of this gyrus where the entorhinal cortex viewed the hippocampus as a collector of sensory informa-
resides. He did not entirely dismiss an olfactory role for the tion; this information, in turn, would develop an emotive
hippocampal formation and suggested that the entorhinal state that would be transferred to the mammillary nuclei. In
cortex should be considered as concerned mainly with the addition to mediating the appropriate behavioral response to
association and integration of olfactory impulses . . . with this emotive state, the mammillary nuclei would also relay
other cortical inuences (p. 206). Among other negative evi- information to the anterior cingulate cortex via the anterior
dence, Brodal cited several studies that attested to the fact that thalamic nuclei where conscious appreciation of the emotion
the hippocampal formation was present in anosmatic and would be achieved.
microsmatic animals such as dolphins and whales (Ries and
Langworthy, 1937) and that there was substantial regional dif- The central emotive process of cortical origin may
ferentiation in microsmatic humans. Furthermore, he cited then be conceived as being built up in the hippocam-
the data of William F. Allen (1940) in which lesions of the pal formation and as being transferred to the mammil-
temporal lobe had no effect on the ability of dogs to perform lary body and thence through the anterior thalamic
olfactory discrimination tasks. Brodal (1947) concluded, No nuclei to the cortex of the gyrus cinguli. The cortex of
decisive evidence that the hippocampus is concerned in olfac- the cingular gyrus may be looked on as the receptive
tion appears to have been brought forward (p. 180). region for the experiencing of emotion as the result of
Brodals review was highly inuential in raising doubts impulses coming from the hypothalamic region, in the
concerning the role of the hippocampus in olfaction. same way as the area striata is considered the receptive
However, at least one of the pieces of evidence cited by cortex for photic excitations coming from the retina.
Brodalthat there are no olfactory projections to the hip- (Papez, 1937, pp. 725743)
12 The Hippocampus Book

objects by sight. In the emotional sphere, they were sexually

Gyrus cinguli anterior aberrant and apparently without fear of stimuli that usually
are frightening to them. It was this latter observation that
appeared to conrm the role of the hippocampus in emotion
predicted by Papez. However, it was subsequently shown that
the loss of fear and other behavioral alterations could be
Anterior attributed to damage to the amygdala and its connections
cingulum with the visual regions of the inferotemporal cortex (Mishkin,
thalamic
1954; Schreiner and Kling, 1956; Weiskrantz, 1956; Aggleton,
nucleus
1993; Hayman et al., 1998). The hippocampus was again in
Vicq search of a function.
d'Azyr's Unbeknown to Papez, and probably also to Klver and
bundle Bucy, a report appearing nearly a half century earlier by
Sanger Brown and Hans Schfer (1888) described the results
of a large bilateral temporal lobe lesion on a ne, large, active
Corpus Hippocampus rhaesus monkey. In addition to describing virtually all of the
mammillare fornix behavioral and emotional alterations of the Klver-Bucy syn-
drome, they also reported disturbances in memory. They
commented on the apparent forgetfulness of the monkey in
Figure 24. Original Papez circuit. Later versions included the
the following way: And even after having examined an object
amygdala, hypothalamus, prefrontal cortex, and some association
cortices. in this way with the utmost care and deliberation, he will, on
again coming across the same object accidentally, even a few
minutes afterwards, go through exactly the same process, as if
Again, a comment from Brodal (1981) is skeptical. After he had entirely forgotten his previous experience (p. 262).
having scrutinized the anatomical literature, he stated: There This little known observation of Brown and Schfer presaged
is no sound biological basis for selecting a few brain regions what has become the most enduring notion of hippocampal
as being involved in these complex functions (emotions), functionthat it plays an essential role in memory.
and today the theory of Papez is of historical interest only
(p. 672). 2.3.3 The Hippocampal Formation
However, we may conclude that the Papez circuit was and Attention Control
important historically because it shifted emphasis away from
the idea of olfactory functions of the hippocampus. Moreover, As early as 1938, Richard Jung and Alois Kornmller had
it was one of the early ideas about how the brain might have noted that desynchronization of the neocortical electroen-
more autonomy from the environment than suggested by the cephalogram (EEG) was temporally linked to a large-ampli-
then current stimulusresponse behaviorism. As we shall see, tude, sinusoidal wave pattern in the rabbit hippocampus at
there is only moderate evidence to support the specic idea between 4 and 7 Hz, which they named theta activity. Later
that the hippocampus is involved in emotion, although the work suggested that this particular activity was related to
idea has occasionally been championed, most notably in enhanced attention (Green and Arduini, 1954). John Green
recent years by Jeffrey Gray (1982, 2001). and W. Ross Adey (1956) found that arousal caused an inverse
Papezs speculations about the role of the hippocampus in relation between the electrocortical waves of the hippocampus
emotion appeared to receive support from the experiments of and neocortex. Endre Grastyn and colleagues (1959) pro-
Heinrich Klver and Paul Bucy, which were carried out at the posed that theta activity could be coupled to specic learning
same time (1937). Klver was a psychologist interested in the states (Fig. 213). Both hippocampal and entorhinal theta
neural locus of action of psychotropic drugs such as mesca- waves showed distinct changes during the acquisition of con-
line. Together with Bucy, a neurologist, he searched for the ditioned response (Holmes and Adey, 1960).
locus of action by removing brain tissue from monkeys and Other signs pointing to a role for the hippocampus in the
looking for changes in their behavioral response to drugs. The control of attention came from electrical stimulation of the
changes that followed temporal lobe removal were interesting anterior part of the temporal lobe, including the hippocam-
enough to warrant extensive description, and the constellation pus, which led to widespread and long-lasting desynchroniza-
of changes is now known as the Klver-Bucy syndrome. There tion of the EEG in both anesthetized and awake cats (Kaada et
are two important aspects to the syndrome: changes in visu- al., 1949; Sloan and Jasper, 1950). Such responses were taken
ally guided behavior and changes in emotional expression. as evidence of a general activation of attention. Stimulation of
The lesioned monkeys appeared to have visual agnosia but the hippocampus in awake cats produced a peculiar reaction
would compulsively follow stimuli brought into their visual with pupillary dilatation and slow turning of the head and
eld; they continually placed all objects encountered in their neck to the contralateral side, as if the animal experienced a
mouths, perhaps owing to their failure to recognize the weird sensory stimulus in the contralateral visual eld (Kaada
 Historical Perspective 13

et al., 1953). The reaction was associated with enhanced respi- Independently, Sergei S. Korsakov (1889, 1890), a Russian psy-
ration and blood pressure. Similar reactions were elicited from chiatrist, described a similar clinical picture he called psy-
the anterior cingulate gyrus, later to be associated with higher chosis polyneuritica and emphasized that the main symptoms
analysis of emotional signals. Modern imaging methods have were memory decits. Finally, Johann Bernhard Aloys von
revealed that this latter area is deeply engaged in the analysis Gudden (1896) associated the condition with pathology of the
of emotional aspects of sensory stimuli, suggesting that it is a mammillary bodies and the mediodorsal nucleus of the thal-
device for estimating potential conicts between old and new amus, both important target stations for hippocampal efferent
experiences (Bush et al., 2000; Botvinick et al., 2001; Bishop et pathways (see Chapter 3). After the rst description of the
al., 2004; Kerns et al., 2004). Wernicke-Korsakov syndrome, a similar condition with an
In summary, this evidence suggests that together with the emphasis on memory impairment was described by another
anterior cingulate cortex some portion of the hippocampus Russian neurologist, Vladimir Bekhterev (1900). He reported
takes part in a certain form of general attention control. two patients with a prominent memory decit who were
later found at autopsy to have softening of the hippocampus
2.3.4 The Hippocampal Formation and Memory and neighboring cortical areas on both sides. This study
appears to be the rst hint of a hippocampal localization for
An early effort to analyze memory functions was that by memory.
Thodule-Armand Ribot, a French philosopher and psychol-
ogist. He proposed that memory loss was a symptom of pro- 2.3.5 More Direct Evidence for Hippocampal
gressive brain disease. His Les Maladies de la Mmoire Involvement in Memory
[Diseases of Memory] (1881) constitutes an inuential early
attempt to analyze abnormalities of memory in physiological The idea that the hippocampal formation is intimately associ-
terms. He even proposed that a plausible memory mechanism ated with memory is due to observations made on brain-
could be an alteration in the activity of engaged cells in the damaged patients by William Scoville and Brenda Milner in
cortex. This seems to be the rst hypothesis to suggest a direct 1957. Scoville removed the mesial aspects of the temporal
role of nerve cells for memory functions. Strongly inuenced lobes from several patients in an attempt to relieve a variety of
by Paul Brocas reports on language localization (Broca, neurological and psychiatric conditions. The most famous
1861a,b), Ribots views on localization of function in general of these, H.M., was a severely epileptic patient whose seizures
were further supported by the clinical evidence of John were resistant to antiepileptic drug treatment. Following sur-
Hughlings Jackson (1865), which showed memory loss in gery, his seizures were reduced, but he was left with a pro-
selected lesions, and by the experimental evidence from David found global amnesia that has persisted to this day. H.M.s
Ferrier (1876), who reported on lip and nostril movements memory decit is observable as an inability to remember
upon stimulation of the hippocampal lobes in monkeys. material or episodes experienced after the operation (antero-
Richard Semon, much inuenced by Ribot but working in grade amnesia); it also includes an inability to recall informa-
Berlin (1908), was responsible for coining the term engram tion experienced for some period of time prior to the
to signify a memory trace in the form of a physical change operation (retrograde amnesia). H.M. can remember items
in the participating nerve cells. for brief periods, provided he is allowed to rehearse and is not
Remembering that the hippocampal efferent impulses distracted. Upon distraction, however, H.M. rapidly forgets.
inuence many hypothalamic nuclei, including the mammil- Thus, he never remembers for any length of time the doctors
lary bodies, the rst link between memory functions and part who test him, his way around the hospital, or the story he has
of the hippocampal system was made by three scientists been reading. He reports his conscious existence as that of
during the 1880s who sequentially and semi-independently constantly waking from a dream and everything looking
described an affliction causing amnesia. This condition was unfamiliar (see Chapter 12). The temporal characteristics of
commonly associated with heavy drinking of alcohol and was H.M.s retrograde amnesia continue to be controversial.
later to be named Wernicke-Korsakoff psychosis. Subse- Originally it was thought by Milner to extend up to 2 years
quently, this disorder was associated with thiamine deciency, prior to the operation, with memories for events earlier than
which is often a correlate of dietary lack in the presence of that remaining relatively intact. This duration of retrograde
alcoholism. In 1881, Carl Wernicke rst described three amnesia was later extended to 11 years (Sagar et al., 1985),
patients whose illnesses were characterized by paralysis of eye with the recollection of more distant events preserved. The
movements, ataxia, and mental confusion. All three patients, interpretation of the extent of retrograde amnesia is compli-
two men with alcoholism and a woman with persistent vom- cated by the fact that this length of time was approximately
iting following sulfuric acid ingestion, developed coma and equivalent to the period of the preoperative epileptic condi-
died. In all three patients at autopsy, Wernicke detected punc- tion. We shall return to a more detailed discussion of the sub-
tate hemorrhages affecting the gray matter around the third sequent work on H.M. and other amnesic patients in Chapter
and fourth ventricles and near the aqueduct of Sylvius. He 12. Suffice it to say here that his memory disabilities are due to
believed them to be inammatory and therefore named the the removal of a large part of the hippocampal formation and
disease polioencephalitis hemorrhagica superioris (Fig. 25). surrounding cortical regions (Fig. 27).
14 The Hippocampus Book

Figure 25. Cerebral localiza-

tion and memory pioneers.
Upper left: Carl Wernicke.
Upper right: Sergei Korsakoff.
Lower left: Bernard v. Gudden.
Lower right: Vladimir Bekhterev.
(Source: Courtesy of www.
alzheimermed.com.br/)

The global amnesia reported in H.M. spurred efforts to to the next) but not the delayed response task (where animals
nd an animal model of amnesia. Early studies by Jack choose a well learned response after a delay period). Even
Orbach, Brenda Milner, and Theodore Rasmussen (1960) and lengthening the delay failed to reveal a decit in the delayed
by R.E. Correll and William Scoville (1967) looked at the response task. Thus, these early attempts in the primate were
effects, in primates, of various temporal lobe lesions on mem- viewed as falling short of developing an animal model of
ory tasks such as delayed alternation and delayed response. medial temporal lobe amnesia.
Decits were seen with combined lesions of the amygdala and Work on rats also failed to nd a convincing memory
hippocampal formation in the delayed alternation task (where decit following hippocampal damage. Lesioned rats had no
animals are required to alternate their response from one trial difficulties learning simple sensory discriminations: to press
 Historical Perspective 15

levers in a Skinner box or to run down alleys to obtain food or

avoid punishment. What emerged, however, was a loose con-
stellation of decits, including changes in exploration and
habituation to novelty. For example, hippocampus-lesioned
rats placed in novel environments were hyperactive (i.e., they
moved around more than control rats and failed to reduce
their activity progressively over time). They also failed to
alternate choices on two successive runs in a simple T-maze
when no rewards were available (spontaneous alternation).
Both changes in behavior were attributed to decits in
inhibitory processes; and consequently a role for the hip-
pocampus in response inhibition or Pavlovian inhibition was
postulated by Robert Isaacson and his colleagues Robert
Douglas and Daniel Kimble (e.g., Kimble, 1963, 1968; Douglas
and Isaacson 1964; Isaacson and Kimble, 1972). Another
change was a striking decit in the ability to withhold a pre-
potent behavioral response, a nding taken as support for the
response inhibition hypothesis. Hippocampectomized rats
learned to run down an alley or press a lever for food as
quickly as normals but were decient in changing this behav-
ior when the circumstances altered (e.g., when food was with-
drawn or when an alternate response was rewarded).
Interestingly, they were consistently better at learning one par-
ticular type of avoidance task: two-way active avoidance. In
this task, the animal must learn to run back and forth between
two compartments of a shuttle box to escape punishment.
This improved learning seemed particularly anomalous in
Figure 26. Drawings of the brain of patient H.M., made on the
light of the presumed role of the hippocampus in memory.
basis of an MRI examination performed 40 years after bilateral
resection of the medial temporal lobe. Drawings are of transverse A third decit lay in complex maze learning. Whereas they
MRI sections from various anterior-posterior levels: A through the had no trouble learning to make a turn in a simple T- or Y-
amygdala; D at the posterior tail of the hippocampus. (Source: maze, they failed miserably at more complex multichoice con-
Corkin et al., 1997.) The lesion was bilateral, but its right side is gurations (Kaada et al., 1961; Kimble, 1963; Kveim et al.,
here left intact to show structures that were resected. 1964).

Figure 27. Brenda Milner (left)

pioneered the analysis of the role
of the medial temporal lobes in
memory. Donald Hebb (right)
inuenced a generation of neuro-
scientists with his theories on
memory mechanisms. (Source:
Courtesy Brenda Milner and
www.williamcalvin.com/.)
16 The Hippocampus Book

This confused state of affairs remained until around 1970

when several developments took place that together changed
the approach to the hippocampal theory. The changes, which
have continued to be inuential, involved alterations in con-
cepts about memory, the development of new tests for assess-
ing memory, and the application of single-cell recording
techniques to behaving animals.
Around this time, it was realized that there might be more
than one type of memory, only one of which involved the hip-
pocampus (Gaffan, 1974a; Hirsh, 1974; Nadel and OKeefe,
1974; Olton et al., 1978). Two related ways of classifying
human memory proved particularly inuential. The rst was
Endel Tulvings contrast between memories for episodes set in
a spatiotemporal context as distinguished from those for
semantic items such as facts and other context-independent
material (Tulving, 1972). The second was the related but dif-
ferent distinction between declarative and procedural memo-
ries, which was originally introduced by Terry Winograd
(1975) in the eld of articial intelligence, subsequently
applied to studies of amnesia by Cohen and Squire (1980),
and extensively promulgated by Larry Squire and his col-
leagues (Squire, 1988, 2004). The former include both
Figure 28. Olga Vinogradova was among the rst who recorded
episodic and semantic memory, and the latter refers to skills
single units from the hippocampus of awake animals (rabbits) while
and other stimulus-response habits. Both memory schemes
testing their behavioral responses to natural stimuli. She interpreted
have inuenced research on amnesic patients and animal her recordings to mean that the hippocampus served as a novelty
models of amnesia (see Chapters 12 and 13). detector. (Source: Courtesy of Richard Morris.)
The second important development around 1970 owed
from the realization that the behavioral tests for animal mem- lesions had with avoidance learning. They concluded that
ory available at that time were not optimal for testing hip- decits were found only when the animals had to learn to
pocampal function. David Gaffan (1974b) introduced unique avoid places, not when they were given a clear prominent
object recognition tasks. Mortimer Mishkin and Jean object or cue to avoid. Selective tests for the spatial functions of
Delacour (1975) developed a version of this task that has the hippocampus, including the Olton radial arm maze and
become the standard test for recognition memory in the mon- the Morris watermaze, were subsequently developed (see
key (see Chapter 13). Following the introduction of these Chapter 13).
more powerful testing paradigms, efforts to establish memory
systems became increasingly more fruitful. 2.3.7 Conclusions

2.3.6 The Hippocampus as a Cognitive Map We have provided a brief overview of some of the historical
perspectives concerning the function of the hippocampal for-
An important development in the analysis of hippocampal mation. Although there is widespread acceptance of the view
function was the use of implanted microelectrodes to monitor that the hippocampal formation is involved in some aspects of
single-neuron activity in the hippocampus of the awake intact memory, there may still be some surprises as to other poten-
animal (Hirano et al., 1970; Vinogradova et al., 1970; OKeefe tial functions of the hippocampal formation. As an important
and Dostrovsky, 1971; Ranck, 1973). These experimenters example, the hippocampus has been implicated in the modu-
described the relations between cellular activity and a variety lation of stress responses through inhibitory projections to the
of sensory and behavioral parameters (Fig. 28). One corre- hypothalamus. As described in Chapter 15, this putative role
latethat of the animals location in the environmentgave of the hippocampus has been implicated in a variety of nor-
rise to the cognitive map theory, which has fostered research mal and abnormal stress responses such as post-traumatic
into the spatial functions of the hippocampus (OKeefe and stress disorder.
Dostrovsky, 1971, see Chapters 11 and 13). This theory sug-
gested that the hippocampus in animals was dedicated to spa-
tial memory and allowed the animal to navigate in familiar
environments. Extension of this theory to humans envisaged 2.4 Special Features of Hippocampal
the addition of a temporal signal, allowing the hippocampus to Anatomy and Neurobiology
act as a spatiotemporal context-dependent (episodic) memory.
Lynn Nadel, John OKeefe, and Abe Black (1975) proceeded to We now turn to an overview of some of the early research on
look carefully at the problems that rats with hippocampal the hippocampal formation and the important historical
 Historical Perspective 17

gures associated with it. The outcome of some of this porting the reticular theory. He also pictured the dendrites of
research led to general principles of neuroscience (e.g., the granule cells as being continuous with blood vessels that occu-
neuron doctrine), whereas others have established the exis- pied the hippocampal ssure. Camillo Golgis most outspoken
tence of unique features of the hippocampal system (e.g., the adversary in the conict over the neuron doctrine versus the
largely unidirectional nature of the intrinsic excitatory con- reticular theory was Ramon y Cajal, whose support of the
nections). In some cases, the motivation for studying the neuron doctrine came, ironically, from preparations made
hippocampal formation was not so much to understand with Golgis method. Although he also had many examples
its function but to capitalize on some unique aspect of its from the hippocampal formation, particularly in newborn
anatomical and physiological organization to carry out an rodents, his favorite preparation was the cerebellum. After
experiment of more general interest more easily. examining thousands of preparations, he became convinced
A number of features have attracted scientists to the hip- that individual neurons did not form a syncytial network. One
pocampus for studies of general neuronal and systems proper- particularly important piece of evidence was that he observed
ties. A short list of the useful anatomical and neurobiological individual neurons that were in the process of dying. Neurons
features of the hippocampus includes the following. adjacent to these cells were entirely healthy. This was convinc-
ing evidence to Ramon y Cajal that neurons were individuals,
A single cell layer and strictly laminated inputs
not just members of interdependent groups. The debate lasted
Predominantly unidirectional connections between
a number of years with a large number of participants, and
a series of cortical regions
was notable for the vigorous exchanges between the two
Extrinsic and intrinsic bers making numerous en pas-
factions.
sage contacts with target neuronal dendrites, running
The nal resolution of this classic debate did not occur
orthogonal to the main dendritic axis
until the advent of electron microscopy, which conclusively
Synapses that are highly plastic
supported the neuron doctrine by showing the neurons as
Tissue that can be used in transplantation studies
individual entities with no connecting intercellular brils
Neurons that can be successfully grown in
(Shepherd, 1972).
culture
Acute or cultured slices surviving for prolonged periods
in vitro 2.4.1 Early Neuroanatomical
Studies of the Hippocampus
The most striking difference between hippocampal and
neocortical cortices is the aggregation of the principal cells in Hippocampal neuroanatomy beneted from the pioneering
a single layer. Another major difference between the two cor- investigations of Camillo Golgi (1886), Luigi Sala (1891),
tical types is the direction of the afferent bers. In contrast to and Karl Schaffer (1892). Santiago Ramon y Cajal (1893)
the input to neocortical areas where most afferent bers are described the stratication of the various afferent systems
radially oriented, the extrinsic and intrinsic afferent nerve and drew a distinction between cells with long and short
bers of the hippocampal formation run in a horizontal direc- axons (Fig. 29).This observation made it clear that a hip-
tion (parallel to the pial surface) and orthogonal to the apical pocampal neuron could inuence a large number of target
dendritic axis. cells and areas. Even before the formal denition of synapses
Hippocampal studies were important during the nine- by Charles Sherrington in 1897, Ramon y Cajal saw the
teenth century controversy between the neuron doctrine and functional implication of lamination. He suggested that there
the reticular theory. The neuron doctrine proposed that each was a convergence of afferent input onto a single neuron, a
neuron was an individual cell that contacted but did not notion that we take for granted today. His monumental effort,
merge with other target neurons. The reticular theory, on the including an analysis of all portions of the hippocampal for-
other hand, posited that nerve cells form a syncytium where mation in a number of animal species, allowed him to propose
one cell emits a protrusion, or bril, that continues into other a functional circuit diagram of this region. Following his
cells, thus creating an interconnected network of bers from a principle of dynamic polarizationinput to the dendrites,
large number of neurons. This controversy continued for output through the axonhe placed arrows indicating his
some time because microscopes at the time were not able to view of the direction of impulse ow through the hippocam-
resolve the neuronal membranes and the spaces (e.g., the pal formation; many of these circuit characteristics have
synaptic clefts) that separated individual neurons. stood the test of time. He emphasized the size and variabil-
Camillo Golgi, who strongly supported the reticular the- ity of the various dendritic trees and gave a detailed descrip-
ory, used observations of the hippocampal formation to bol- tion of dendritic spines and their distribution (structures
ster his arguments. Employing the reazione nera (black that were dismissed by Golgi as artifacts of his staining
reaction), as he called the staining technique, later to be called method).
the Golgi method, he repeatedly pointed to the convergence of Simultaneous with Ramon y Cajals rst studies, the
ne axonal branches from a large number of dentate granule Hungarian anatomist Karl Schaffer, who gave his name
cells to form a dense bundle in the hilus of the dentate gyrus (Schaffer collaterals) to the axons from CA3 cells that termi-
(vertical bundle in Fig. 22). Here, he believed laments from nate in CA1 (1892), found by meticulous charting of well
various axons made a large intertwined feltwork, thus sup- impregnated Golgi material not only that axons had short
18 The Hippocampus Book

Figure 29. Santiago Ramon y Cajal

and his famous drawing of the hip-
pocampus in his 1911 book Histologie
de Systme Nerveux. in two volumes.
The arrows give his interpretation of
likely impulse direction. Later func-
tional studies have vindicated his ideas
on nearly all points. Portrait from Royal
Society, London, gratefully acknowl-
edged. Permission to reproduce draw-
ing from Cajal (1911) from Istituto
Cajal is gratefully acknowledged.

branches but that some axons could be very long, connecting degenerating bers could be stained better than intact bers.
neurons in neighboring cortical elds. Nearly 40 years later Temporal myelinization gradients were also useful. Unfor-
Rafael Lorente de N (1934), building upon the work of his tunately, these methods initially met with limited success in
compatriot Ramon y Cajal, greatly extended the analysis of the hippocampus. Although the hippocampus contains sev-
the many hippocampal cell types and their axonal and den- eral ber systems with moderately thick myelinated bers,
dritic patterns. He described many detailed networks of inter- most of its axons are much thinner than in other parts of the
connected neurons (Fig. 210). On the basis of their dendritic central nervous system (Shepherd et al., 2002). These features
tree and connections, he divided the hippocampal formation are probably one reason why the rst available methods for
into a set of clearly dened divisions and coined the well experimentally establishing pathways by degenerating bers
known terms CA1 to CA4. did not give satisfactory results. For example, even after
The pioneering neuroanatomists came impressively close appropriate lesions, Vittorio Marchis original method
to modern-day thinking in their proposed schemata for func- (Marchi and Algeri, 1886) stains only a small number of
tional connectivity in the hippocampus. More often than not, degenerating bers in the alveus and mbria. In addition, the
the diagrams of Ramon y Cajal (1893, 1911) indicated the cor- stained bers are exclusively among the thicker of those
rect direction of information ow. Likewise, the detailed present.
drawings of various neurons by Lorente de N (1934)in
which he used the number of dendritic spines to estimate the 2.4.3 New Anatomical Techniques that
relative efficiency of afferents terminating at various positions Revolutionized Connectivity Studies
on the dendritesallowed him to formulate a general rule
about summation. Even so, to understand the operational The situation improved greatly when certain variants of silver
rules of the hippocampus, specic details beyond the major impregnation were applied to lesioned hippocampal path-
sources and trajectories of the main afferent and efferent ways. Admittedly, the rst silver degeneration method devel-
pathways were needed. oped by Paul Glees (1946), was disappointing in the
hippocampus despite the success it had in pathways of the
2.4.2 New Fiber Tracing Methods brain stem. However, Nautas 1950 method, soon followed by
the Nauta and Gygax variant (1954) and Fink and Heimers
In parallel with the classic neuroanatomical studies with the method (1967), were all remarkably effective.
Golgi method just described, much work was carried out to Although similar in principle to other methods, Walle
trace major afferent and efferent ber systems in the central Nautas methods turned out to be highly useful in hippocam-
nervous system. The early studies exploited the fact that pal tissue. By employing the rst variant of Nautas tech-
 Historical Perspective 19

Figure 210. Top: a rare picture of the pioneer Lorente de N, who had every reason to be
proud of his 1934 drawings shown below, signed along the margin. The diagram shows the
cell types and divisions and the terminology in general use today: fascia or gyrus dentata, CA1
to 4, subiculum, pre- and parasubiculum and the entorhinal area. Acknowledgment to Larry
Swanson for the portrait. Permission from Georg Thieme Verlag, Stuttgart to reproduce de
Ns drawing gratefully acknowledged.
20 The Hippocampus Book

niques, Theodor Blackstad (1956, 1958) (Fig. 211) showed were the basis for the so-called trisynaptic circuit (Andersen et
how commissural and ipsilateral afferents to several parts of al., 1966). With this unidirectionality, the hippocampal system
the hippocampus terminated in laminae oriented parallel to differs fundamentally from the reciprocal connectivity of
the cell layers (Fig. 211). These data not only veried the gen- nearly all neocortical areas. This principle of hippocampal
eral principles established by Ramon y Cajal on stratication neuroanatomy is discussed in detail in Chapter 3.
of afferent bers but in addition showed astonishing density When a small ber bundle of the Schaffer collaterals was
of presynaptic boutons in the innervated zone. stimulated, the amplitude of the synaptic potentials showed a
characteristic distribution over the CA1 region. Signals were
2.4.4 Predominantly Unidirectional detected in the whole transverse extension of CA1. However,
Connectivity Between Cortical Strips within the borders to CA3 and the subiculum, the largest
potentials were detected along a strip oriented nearly trans-
A key aspect of hippocampal connectivity was to emerge from versely to the longitudinal axis of the hippocampus, with
these analyses: unidirectionality. Because the Golgi technique gradually reducing amplitudes on either side, the range span-
stains a relatively small proportion of the total number of ning nearly half the length of the hippocampus. With stronger
neurons, this picture needed to be complemented by methods stimulation, the synaptic potentials generated a population
giving more quantitative data. Systematic study of each region spike (see Section 2.3), which had the same main orientation
of the hippocampal formation with suitable silver degenera- but with more restricted longitudinal distribution (Andersen
tion techniques, later supplemented by electron microscopic et al., 1971b). This general arrangement was found for all four
and anterograde and retrograde tracing methods, led to the pathways studiedperforant path, mossy bers, Schaffer col-
realization that each component of the hippocampal forma- laterals, CA1 axons in the alveusand the same orientation
tion projects to its neighboring region but generally does not was found with orthodromic as with antidromic testing. This
receive a return pathway from this target (Hjorth-Simonsen, arrangement, which was named the lamellar organization, has
1973). This was in good accord with previous physiological received considerable opposition, mostly on anatomical
studies in which activation of the perforant path led to grounds (see Chapter 3), probably because the name lamella
sequential activation of CA3 followed by CA1. A cut of CA3- invited vsualizing a set of thin slices with sharp borders, like a
to-CA1 bers removed activity caudal to the section in CA1 loaf of bread. Instead, the lamella should be seen as the main
and subiculum. On the other hand, stimulation in CA1 did orientation of a set of overlapping fan-shaped ber regions (Li
not give synaptic activation in the CA3 region. These ndings et al., 1994), with most branches and their synaptic effect
along the major orientation and gradually less inuence on
Figure 211. Theodor Blackstad, often called the father of modern either side. The efficiency of the transverse slice is an illustra-
hippocampal histology because of his ground-breaking studies of its tion of this organization, although the spread of ber orienta-
external and internal connectivity, with charting of degenerating tion allows signals to be elicited in slices cut at a less than
bers through silver staining and electron microscopy. (Source: optimal angle.
Courtesy of Jon Storm-Mathisen.)
2.4.5 New Tracing Studies
Using Axonal Transport

In addition to the connectivity of single cells, it was necessary

to determine how aggregations of neurons were connected so
early researchers could gain insight into how larger systems
could operate in concert. New anatomical techniques
exploited intra-axonal transport. Maxwell Cowan (Fig. 212)
and his colleagues introduced a new technique for ber trac-
ing based on the anterograde transport of injected radioactive
amino acids (Cowan et al., 1972) (Fig. 212). Here, they
exploited earlier information showing that radioactive pro-
teins could be transported inside axons in both anterograde
and retrograde directions (Grafstein, 1971; Lasek, 1975).
Cowan had an impressive lineage: His doctoral (PhD) studies
were carried out under the mentorship of Thomas Powell of
Oxford University (Fig. 212). Powell, Cowan, and Geoffrey
Raisman were responsible for many of the classic studies of
hippocampal and fornix connections (Raisman et al., 1965,
1966). Max Cowans laboratory, in turn, spawned a number of
hippocampal researchers, including Larry Swanson, Brent
Staneld, and David Amaral, who among them conducted a
 Historical Perspective 21

Figure 212. Thomas (Tom) Powell (top left) and W. Maxwell (Max) Cowan (top right), the
former born in Britain and the latter in South Africa. They both had illustrious careers in neu-
roscience, starting their work by tracing ber systems in the hippocampal formation. (Source:
Courtesy of Geoffrey Raisman.)

substantial portion of the modern studies on hippocampal microscopy gave an unprecedented view of the details of indi-
connectivity. The autoradiographic method had the merit of vidual cells and the contacts of neural circuits of the hip-
allowing extremely small and discrete injections to be placed, pocampus and its main target nuclei (Raisman, 1969).
with the 3H-amino acids being taken up and transported by Moreover, the electron microscope also provided a new
cells of origin, not bers of passage. One important result of method for ber tracing. When a ber system was lesioned,
this study was that it became clear that direct hypothalamic the subsequent degeneration of both bers and their associ-
projections of the hippocampal formation arose not from the ated boutons darkened within 1 to 2 days, a process that could
hippocampus proper but from subicular regions (Swanson be used for both synapse indentication and connectivity
and Cowan, 1977). However, the hippocampal formation can studies (Alksne et al., 1966).
inuence a multitude of hypothalamic nuclei through multi-
synaptic pathways. 2.4.7 Hippocampal Synapses Are Highly
In addition to the autoradiographic method, new methods Plastic: Early Studies of Sprouting
that depended on axonal transport of various proteins or
plant lectins were developed (Kristensson and Olsson, 1971; The term plasticity is used to indicate several types of change.
Gerfen and Sawchenko, 1984). In addition to the well known activity-dependent alterations
Fibers containing monoamines were also found to inner- of synaptic efficiency (e.g., long-term potentiation, described
vate the hippocampus. The rst reports described noradrena- in Chapter 10) there are numerous examples of anatomical
line (norepinephrine)-containing axons from the locus changes in cell dimensions and number and in axonal length,
coeruleus (Segal and Bloom, 1974a,b). branching and connections, and biochemical content in
response to various stimuli or injury (see Chapter 9). The hip-
2.4.6 Electron Microscopy Offers pocampal formation was the rst brain region in which
New Opportunities axonal sprouting and reactive synaptogenesis were unequivo-
cally demonstrated. Convincing experimental evidence of
The rst steps toward analysis of the number and types of adult synaptic reorganization was described by Geoffrey
synapses in the central nervous system also proted from Raisman (1969). He studied two sets of afferent bersone
work on the hippocampus. Following the pioneering work of from the hypothalamus and the other from the hippocam-
Sanford Palay and George Palade in other brain regions (Palay pusthat converged on cells of the septal nuclei. After long-
and Palade, 1955), Lionel Hamlyn (1963) carried out the rst term lesions of the mbria, there was an increased number of
electron microscopic analysis of the hippocampus. He pro- multiple synapses (boutons in contact with several dendritic
vided the rst comprehensive description of all synapses con- spines), interpreted as being the result of residual axons
tacting the various portions of the dendritic tree in both CA3 sprouting. Conversely, following removal of hypothalamic
and CA1 pyramidal cells. The introduction of electron afferents, mbrial bers were found to contact somata of sep-
22 The Hippocampus Book

tal cells, which they rarely do normally. In another report, 1981, 1984). Lesions of the commissural or perforant path
Raisman and Pauline Field (1973) exploited the fact that m- bers did not induce such effects. The grafted sympathetic
bria bers innervate a segment of the lateral septal nucleus on axons respected the lamination borders and formed typical
both sides of the midline. A unilateral mbrial lesion gave rise norepinephrine-containing boutons with target neurons.
to remarkable sprouting from the contralateral ber system Similar results were obtained when cholinergically deaffer-
such that the total number of fimbrioseptal synapses ented hippocampi received implants of acetycholine-contain-
remained unchanged. Further work by the same group using ing cells placed in a cavity formed after the original septal
transplantation of embryonic tissue for reinnervation of den- region had been removed. Again, both peripherally and cen-
ervated hippocampal areas (Raisman and Field, 1990) sparked trally located donor neurons were effective, and the reinner-
interest in the plasticity of hippocampal afferent and efferent vation by transplanted cholinergic bers displayed homotypic
ber systems. However, the authors also warned that the reor- localization. Particularly impressive was the functional recov-
ganization was not necessarily adaptive or functionally ery of spatial learning ability and the partial restitution of hip-
important. A special case was the crossed entorhinal-dentate pocampal place cell activity (Shapiro et al., 1989), which
pathway, allowing a set of control experiments to determine paralleled the regrowth of cholinergic bers (Dunnett et al.,
factors of importance for reinnervation (Steward et al., 1974). 1982).
Nearly simultaneously, sprouting of hippocampal ber sys- Following these encouraging initial observations, trans-
tems was demonstrated with another approach by Gary Lynch plantation research in the hippocampal formation ourished.
and Carl Cotman and their associates (Lynch et al., 1972; Subsequent efforts were directed at determining if cerebellar
Mosko et al., 1973; Matthews et al., 1976; Cotman et al., 1977; neurons would survive when transplanted into the hippocam-
Goldwitz and Cotman, 1978). This group rst showed that the pus and if hippocampal tissue might be electrophysiologically
acetylcholinesterase-containing septohippocampal fibers active when grafted into the cerebellum and vice versa. In the
innervating the dentate gyrus increased in intensity and dis- latter situations, the transplanted hippocampal tissue
tribution if the perforant path bers to the same region were assumed its original cellular appearance and synaptic lamina-
removed. Electron microscopy conrmed that there was an tion, and the cells showed electrophysiological properties typ-
early postlesion loss of synapses followed by a partial recovery. ical of hippocampal neurons (Hounsgaard and Yarom, 1985).
The remarkably efficient and fast regeneration is an impres-
sive example of the plastic properties of hippocampal neu- 2.4.9 Hippocampal Cells Grow Well in Culture
rons. After lesions to the perforant path, for example, the
maximal reduction of associated boutons in the molecular The hippocampus has become a favorite source of neurons in
layer of the dentate gyrus was seen after about 5 days. several forms of tissue culture. Early successes at dening the
However, after 2 weeks there was substantial recovery, and parameters for successful survival and maturation of hip-
after 3 weeks the tissue had regained normal bouton density pocampal neurons came from the work of Gary Banker in the
(Matthews et al., 1976). laboratory of Max Cowan (Banker and Cowan, 1977; Banker
and Goslin, 1991). In such cultures both neurons and glial
2.4.8 Hippocampal Neurons: Transplantable cells are readily identiable and may be manipulated or
with Retention of Many Basic Properties recorded as individual elements. A disadvantage, however, is
that the original cytoarchitecture disappears, including such
Another remarkable example of neuronal plasticity is the abil- characteristic hippocampal features as the laminated input
ity of transplanted neuronal cells to emit axonal branches that with synaptic segregation. This problem led to the develop-
grow and connect to existing neuronal target cells. The hip- ment of the organotypic slice by Beat Ghwiler and colleagues
pocampal formation was the testing ground for analysis of (Ghwiler, 1981, 1988; Ghwiler and Brown, 1985). With this
incorporation of embryonic transplants. After the work by technique, thin slices of hippocampal tissue can be grown for
Raismans group, described above, Anders Bjrklunds group several weeks with an impressive capacity for neuronal growth
(Bjrklund et al., 1975, 1976) independently investigated the and differentiation (Zimmer and Ghwiler, 1984). With care,
regenerative ability of transplanted monoaminergic tissue most neuronal phenomena seen in vivo or in acute slices can
to reinnervate normal hippocampal tissue. They showed be demonstrated. In addition, the longevity of the preparation
that transplanted tissue containing catecholamine- or acetyl- makes possible a number of experiments, including rehabili-
choline-synthesizing neurons could extend axons and rein- tation after injury and analysis of the long-lasting effects of
nervate the hippocampus in a normal fashion. When a graft of growth factors.
norepinephrine-containing neurons, whether from a periph-
eral source such as the sympathetic superior cervical gland or 2.4.10 Development of Hippocampal Slices:
from a central location such as the locus coeruleus, was From Seahorse to Workhorse
allowed to grow into a noradrenergic denervated hippocam-
pus, grafted cells were observed to send out axons that entered A major technological advance for hippocampal research, and
the hippocampal formation but only if the cholinergic indeed for the eld of neurobiology, was the development of
septo-hippocampal bers were cut (Bjrklund and Stenevi, the in vitro hippocampal slice preparation. The pioneer in this
 Historical Perspective 23

effort was the neurochemist Henry McIlwain. He set out to Role of oscillations in neuronal networks
develop a reliable and functional in vitro preparation to inves- Underlying mechanisms of epileptogenesis
tigate how various stimuli and compounds inuenced the bio-
chemistry of central nervous tissue (Li and McIlwain, 1957). 2.5.1 Identication of Excitatory
He used a variety of preparations from the central nervous sys- and Inhibitory Synapses
tem. Despite considerable success, most of the horizontally cut
neocortical slices did not allow a physiologically realistic affer- George Gray (1959) found that synapses in visual and
ent impulse pattern. More physiologically relevant data came frontal cortex could be divided into two distinct structural
when he employed slices from the piriform cortex. Here, stim- types, which he called type 1 and 2. Type 1 synapses were
ulation of the lateral olfactory tract elicited synaptic and cell more heavily stained, and the synaptic specialization was
activity in neurons of the piriform cortex (Yamamoto and asymmetrical in that the presynaptic density was less intensely
McIlwain, 1966a,b; Richards and McIlwain, 1967; McIlwain stained than the postsynaptic density. Type 1 synapses
and Snyder, 1970). Tim Bliss and Chris Richards (1971) were usually associated with dendritic spines. Type 2 synapses
showed that eld potentials similar to those found in the intact had pre- and postsynaptic densities with the same moderate
animal could be elicited in slices of the dentate gyrus cut along staining intensity, and they were found mainly in associa-
the septal-temporal axis of the hippocampus. With some tech- tion with dendritic shafts and neuronal somata. As to pos-
nical modications of the McIlwain procedure, Per Andersen, sible functional consequences of these differences, Gray (1959)
Knut Skrede, and Rolf Westgaard started to use transversally concluded that, At present there is no evidence to suggest that
sectioned hippocampal slices from guinea pigs and demon- type 1 and type 2 synapses are functionally different (p. 252).
strated that intrinsic pathways could be successfully activated When Grays colleague Lionel Hamlyn (1963) studied the
(Skrede and Westgaard, 1971). Single-cell activity could be hippocampus, he found the same two synapse types and a
recorded and showed short-term synaptic plasticity similar to similar distribution on dendritic spines and shafts. All
that in intact, anesthetized preparations (Andersen et al., synapses involving spines were of type 1, whereas synapses
1972). Rat and mouse hippocampal slices were useful as well, formed on the soma and thicker dendritic trunks or smooth
with the same type of signals as are seen in intact preparations. branches were always type 2. A simultaneous report by
A distinct advantage was the great stability of the isolated slice Theodor Blackstad and Per Flood (1963) corroborated these
preparation, which allowed long-lasting high-quality intracel- data. It was comforting that the hippocampal cortex showed
lular recordings to be made (Schwartzkroin, 1975). Important the same synaptic structures as those in neocortical areas.
also was the precision with which electrodes could be placed Conversely, this similarity made it possible that ndings in
and lesions created. The virtual lack of a blood-brain barrier is hippocampal tissue also could be generalized to other cortical
an experimenters dream. A new opportunity arose when areas. What was needed for such a purpose was a correlation
intracellular recording could be combined with iontophoretic of functional and structural data from the same tissue. Such
delivery of transmitter candidates or blockers from multiple an opportunity arose with the advent of intracellular record-
pipettes arranged to hit various parts of the dendritic tree ings from hippocampal neurons.
(Schwartzkroin and Andersen, 1975). Fortunately, acute
slices also supported well developed long-term potentia- 2.5.2 Gray Type 2 Synapses are Inhibitory
tion (Schwartzkroin and Wester, 1975), a prerequisite for and are Located on the Soma of Pyramidal
the wealth of analytical studies of this phenomenon (see and Granule Cells
Chapter 10).
In an inuential intracellular study, Eric Kandel, W. Alden
Spencer, and Floyd Brinley, Jr. (1961) reported on the ubiqui-

tous and long-lasting inhibitory postsynaptic potentials

2.5 Several Neurophysiological Principles Have
(IPSPs) following both orthodromic and antidromic activa-
Been Discovered in Hippocampal Studies
tion of cat CA2-CA3 pyramidal cells. Per Andersen, John
Eccles, and Yngve Lyning (1964a) used a combination of
Although at rst glance the hippocampal formation appears
intracellular recording of IPSPs and laminar analysis of the
to be a special part of the brain with many features not
associated eld potentials elicited by the same two activation
seen elsewhere, it shares enough general features with other
modes. They showed that the initial phase of the IPSP in both
cortical regions to make it remarkably useful for studies
CA1 and CA3 pyramidal cells and in dentate granule cells was
of a general nature. Neurobiological principles that have
associated with an extracellular positive eld potential with a
been discovered from work on hippocampal preparations
sharp maximum at the cell body layer, indicating a synaptic
include:
current leaving the somata of many principal neurons. This
Identication of excitatory and inhibitory synapses and observation could be explained if the basket cell synapses,
their localization, transmitters, and receptors which terminate at the soma and initial axon, caused activated
Discovery of long-term potentiation and long-term hyperpolarizing chloride conductances, and were in fact
depression inhibitory interneurons. The basket cell synapses were exclu-
24 The Hippocampus Book

sively Gray type 2 (Hamlyn, 1963). Andersen, Eccles, and gave rise to a localized fEPSP carrying a population spike,
Lyning (1964b) looked for cells that were not antidromically proving the excitatory nature of the connection. Following
invaded from the alvear bers and thus were not pyramidal specic surgical lesions to each of these monosynaptic excita-
cells. They discovered a subset of such neurons that dis- tory pathways, the method of Alksne et al. (1966) showed
charged repetitively with a time course corresponding to the that boutons belonging to all four pathways were
rise time of synaptically activated IPSPs found in pyramidal electron-dense, closely associated with dendritic spines, and
and dentate granule cells, respectively. The location of such Gray type 1 (Andersen et al., 1966a). Thus, the widely
cells corresponded to the basket cells described by Ramon y accepted idea that cortical (and many other) excitatory
Cajal (1911) and Lorente de N (1934). Hippocampal IPSPs synapses are Gray type 1 and are usually located on dendri-
were found to be bicucculine-sensitive by David Curtis, John tic spines emerged from these studies of the hippocampal
Felix, and Hugh McLennan (1970) and therefore were medi- formation.
ated by -aminobutyric acid ionotropic receptor type A In conclusion, within a few years during the 1960s, both
(GABAA) receptors. inhibitory and excitatory synapses of the hippocampus were
Thus, the dentate and hippocampal basket cells were the identied with respect to location, anatomical type, and func-
rst interneurons for which both the functional role and the tion. These ndings turned out to be applicable to many,
identity of the effective synapses were revealed. Subsequently, albeit not all, parts of the central nervous system.
ndings from the hippocampal formation served as a tem-
plate for similar searches in the cerebellum, thalamus, and 2.5.4 Long-lasting Alterations of Synaptic
other parts of the central nervous system. Efficiency After Physiological Stimulation
However, the basket cell inhibition was far from exclusive.
Later work in the cerebellum, hippocampus, and neocortex Since the discovery of long term potentiation (LTP), hip-
showed that many more types of inhibitory interneurons and pocampal synapses have been the most studied exemplars of
synapses were present (Eccles et al., 1966). synaptic plasticity. These investigations are dealt with in detail
in Chapter 10. Prior to the discovery of LTP, many other
2.5.3 Gray Type 1 Synapses are Excitatory processes had been studied as potential models for learning.
and are Located on Dendritic Spines In particular, post-tetanic potentiation (PTP) was a common
candidate for a number of years. Originally described by
Identication of the morphological and functional character- Martin Larrabee and Detlev Bronk in the sympathetic ganglia
istics of excitatory synapses was also rst carried out in the (1939), PTP appeared as enhanced synaptic responses follow-
hippocampus. Kandel et al. (1961) saw examples of excitatory ing a period of high frequency tetanization of afferent fibers.
postsynaptic potentials (EPSPs) in cat CA3 and CA2 pyrami- The main weakness of PTP as an adequate mechanism for
dal cells in response to subiculum or mbria stimulation, but behavioral learning was its limited duration; in the spinal
they were surprised how relatively rare such responses were cord, PTP lasted only up to 7 minutes (Lloyd, 1949). The
compared to the ubiquitous IPSPs. This relative rarity made duration was increased to 30 minutes by Alden Spencer
them difficult to analyze in detail. In rabbits and rats, eld (Spencer et al., 1966) using long-lasting tetanization of poly-
potential recording allowed better identication of synapti- synaptic spinal reexes.
cally activated cells. Limited duration was also an obstacle in the usefulness of
Identication of excitatory synapses and their localization the early studies of synaptic plasticity in the hippocampus.
were the result of a combination of extra- and intracellular Several investigators had noted the remarkable growth of
responses to various excitatory afferent pathways and a new synaptic potentials during a period of high frequency stimu-
method for marking synapses. In the perforant path-to-den- lation (Cragg and Hamlyn, 1955; Kandel and Spencer, 1961b;
tate granule cells, the rst association between a eld EPSP Gloor et al., 1964). The critical question was: Would the
(fEPSP) and an intracellular EPSP was made by Per Andersen, enhanced excitability last for a signicant period following
Birgitta Holmqvist, and Paul Voorhoeve (1966b), proving that cessation of the tetanic stimulation? Andersen (1960) noted
the fEPSPs used were monosynaptic events and that the that a period of enhanced synaptic response of commissural
synapses in question were excitatory. responses of CA3 and CA1 cells did indeed follow a short
With a similar approach, several other excitatory synapses period of 10- to 20-Hz stimulation, but the duration was only
were identied including perforant path bers activating den- up to 8 minutes and was initially regarded as an example of a
tate granule cells, mossy bers activating CA3 cells, and com- special form of PTP. It soon became clear, however, that
missural and local intrinsic bers in the stratum oriens longer-lasting enhancement could be achieved when higher
activating CA1 pyramidal cells. Using intracellular recordings stimulation frequencies and, above all, repeated tetani were
of monosynaptic EPSPs in rabbits and cats, in response to applied, as was rst reported by Lmo (1966). This important
stimulation of four independent afferent ber systems to CA1 discovery led the way to a description of truly long-term
and the dentate granule cells, the activated synapses were synaptic plasticity (hours and weeks), which is the basis for
identied as excitatory. In each system, synaptic activation studies of LTP to this day (Bliss and Lomo, 1970, 1973). For a
 Historical Perspective 25

full description of the processes and mechanisms underlying ranges, have been studied in hippocampal preparations (see
LTP, see Chapter 10. Chapter 8).

2.5.5 Hippocampal Systems: Exhibiting Several 2.5.6 Studies of Epileptiform Behavior

Types of Oscillatory Behavior
Richard Jung (1949) reported that electrical polarization (by
A few years after Hans Berger (1929) described the human long-lasting direct current) of the hippocampal formation
EEG, Alois Kornmller in Berlin began a search for possible readily gave rise to electrical after-discharges similar to those
electrical equivalents to the cytoarchitectonic parcellation of seen during epileptic seizures. By measuring the stimulus cur-
the cerebral cortex pioneered by Korbinian Brodmann (1909), rents necessary to elicit such seizures in various tissues, he
as mentioned in Section 2.1.3. concluded that compared to other cortical areas the hip-
Working with rats, Case Vanderwolf (1969) discovered that pocampal formation needed considerably less current to elicit
theta activity occurred each time the animal initiated a vol- epileptiform seizures. This observation was in good accord
untary movement, in contrast to movements initiated more with clinical experience gained from the EEG techniques
automatically or in a reex-like manner. He also identied introduced during the mid-1930s, suggesting that many
two types of theta activityatropine-sensitive and atropine- epileptic conditions originated in medial temporal structures
insensitivethe former related to attention and the latter (Fig. 2-13).
to movements (Kramis et al., 1975). Atropine is an alkaloid Further progression of an electrographic seizure is often
that blocks muscarinic cholinergic receptors. In elegant characterized by a set of depolarizing burst responses in which
analytical studies, Abe Black obtained evidence that the the extracellular responses are typically seen as a large local
higher-frequency hippocampal theta activity was related to negative wave carrying a number of high frequency spikes of
motor activity rather than to learning as such (Dalton and decreasing amplitude on its back, the so-called burst response
Black, 1968; Black, 1972; Black and Young, 1972) (see also (Ajmone-Marsan, 1961) at the onset of epileptiform dis-
Chapter 11). charges (von Euler et al., 1958; Kandel and Spencer, 1961b;
Theta activity is also of interest as a means of identifying Gloor et al., 1964). After a period with spontaneous high
hippocampal interneurons and their relation to behavior amplitude burst responses, the electrographic records dimin-
(see Chapters 8 and 11). In addition, other types of oscilla- ish and all cells cease ring because of a depolarization block.
tion at different frequencies, including beta and gamma Both in vivo and in vitro hippocampal preparations have been

Figure 213. Endre Grastyan (left), a Hungarian neuroscientist, was and named theta waves, measured the low hippocampal threshold
among the rst to investigate the role of the hippocampus in condi- for epileptic seizures, and gave the rst descriptions of visual corti-
tioned learning. (Source: Courtesy of Gyorgyi Buzsaki.) Richard cal neuronal responses to contours, contrasts, and colors. (Source:
Jung (right) had a number of discoveries to his credit. He discovered Courtesy of Volker Dietz.)
26 The Hippocampus Book

favorite tools in the search for cellular and network properties ing on the chorda tympani and lingual nerves (1926).
underlying the generation and spread of epileptiform activity Microelectrodes were developed during studies in cardiac
(see Chapter 16). Among these preparations are the epilepti- and skeletal muscle by Ralph Gerards group in Chicago
form activity provoked by kainic acid injections in CA3, (Ling and Gerard, 1949). In the spinal cord, Chandler McC.
GABA blockade by benzyl penicillin, and the effect of ouabain Brooks and John Eccles (1947) used microelectrodes to detect
intoxication. An important model for epileptiform activity is what they called a focal potential, a eld potential generated
the kindling phenomenon, described by Graham Goddard, by monosynaptic activation of motoneurons by volleys in
D.C. McIntyre, and C.K. Leech (1969), who elicited seizures by muscle afferent bers.
repeated low-strength stimulation of the amygdala. Despite The rst use of microelectrodes in the brain for extracellu-
the ease with which electrographic seizures may be recorded lar recording of single nerve cell discharges was in the hip-
from the hippocampus or the entorhinal area, these areas do pocampus by Birdsie Renshaw, Alexander Forbes and, Robert
not usually develop the kindling phenomenon. The role of the Morison (Renshaw et al., 1940). The researchers had greater
hippocampal formation in experimental and clinical epilepsy success detecting activity of single nerve cells in the hip-
is discussed in Chapter 16. pocampus than in neocortex, possibly because the latter tissue
was more depressed at the levels of pentobarbital anesthesia
used. In the hippocampus, these researchers recorded sponta-

neous and evoked discharges of what they called axon-like

2.6 Development of Methodological
spikes. Their description stated: what is evidently the dis-
Procedures for General Use
charge of the same unit - they recur repeatedly with the same
waveform (p. 90). When describing the individual elements
The hippocampal cortex has served as a test bench for the with a constant amplitude and predominantly negative polar-
development of a variety of general neuroscientic method- ity, they used what later became the classic critera for identi-
ologies. In many instances this has led to the realization that fying individual nerve cell discharges. Renshaw et al. localized
the apparently special features of hippocampal tissue is also such discharges to elements in the pyramidal layer and con-
found in other central nervous structures. For example, the cluded that the axon-like spikes were discharges of individ-
physiological properties of CA1 hippocampal pyramidal cells ual pyramidal cells.
have much in common with those of neocortical pyramidal
cells.
2.6.2 Pioneers of Intracellular Recording
Below are listed some areas where hippocampal neurobiol-
ogy has been instructive for neuroscience in general. The same experimental advantages that facilitated the analysis
First use of microelectrodes for extracellular neuronal of extracellular unit recording potentials also paved the way
studies for using the hippocampus for intracellular recording. The
Development of tetrodes for unit recording from behav- rst to succeed in this technically demanding enterprise were
ing animals Denise Albe-Fessard and Pierre Buser in 1952. They were sur-
Interpretation of eld synaptic potentials and popula- prised by the fact that nearly all penetrated cells showed a
tion spikes as tools for analysis of extracellular signals large, long depolarizing signal. Both its large amplitude and
Pioneering use of intracellular recording for central long duration made interpretation difficult. Initially, they took
nervous system neurons the response to be an EPSP. Later, it turned out to be an arti-
Isolated slices of cortical tissue for neuroscience studies factual recording of a large IPSP, which due to an electrode-
Development of histochemical methods for localization induced chloride infusion went in the depolarizing direction.
of neurotransmitters and receptor types They did not record EPSPs with their technique, however.
Transplantation studies The rst satisfactory intracellular recordings from cortical
Pharmacological studies of central neurons and neurons were made by Charles Phillips in 1956. He recorded
synapses from Betz cells in the precentral gyrus of cats and character-
Molecular biological analysis of synaptic function ized their response to both antidromic and orthodromic acti-
Formulation of computational models to explain ways vation. This too proved a technically difficult task, not least
in which neural networks can implement learning and because of the difficulty of nding monosynaptic afferent
memory bers to stimulate. Fortunately, the situation turned out to be
easier in hippocampal preparations. As described in Section
2.6.1 Hippocampus as a Test 2.3, Kandel, Spencer, and Brinley (1961) carried out the rst
Bed for Microelectrode Work comprehensive intracellular study of hippocampal pyramidal
cells. In addition to their analysis of EPSPs and IPSPs (see
Many of the standard methods used in neurobiology were above), they reported on prepotentials, which were short
developed in the peripheral or autonomous nervous systems depolarizing deections at the very start of the full action
or in muscular tissue. Recording from single nerve bers potential, interpreted as a sign of dendritically generated
was pioneered by Edgar Adrian and Yngve Zotterman work- action potentials. They also recorded major features associ-
 Historical Perspective 27

ated with the transition from normal behavior to epileptiform In the hippocampal formation, the histological arrange-
behavior in the form of slow depolarization waves and the ment is highly favorable for eld potential studies of both
occurrence of giant depolarizations with superimposed burst orthodromic and antidromic activation (see Box 21). The
discharges (burst responses) (Kandel and Spencer, 1961a,b). dense packing of the cell bodies, the roughly parallel position
This was an essential step in epilepsy research. of the apical dendrites of hippocampal neurons, and the ease
In contrast to the wealth of inhibitory synaptic data, there with which they can be synchronously activated are three
was a surprising paucity of excitatory synaptic signals. This main reasons for the striking appearance of hippocampal eld
may be related to the barbiturate anesthesia with its facili- potentials.
tating effect on IPSPs, an effect also discovered in the hipp- The main advantage of using eld potentials is that an
ocampus by Roger Nicoll and Eccles (Nicoll et al., 1975). As extracellular potential recording may give an accurate index of
mentioned above in the discussion of IPSP and EPSP identi- synaptic activity with regard to amplitude, time, and polarity.
cation (see Section 2.3.1), intracellular recording from hip- First, the parallel orientation of a large number of principal
pocampal neurons was essential for the identication and cells may generate a eld potential of considerable magnitude.
localization of inhibitory synapses to the soma and of excita- The fEPSP has maximum negativity in the region with the
tory synapses to dendritic spines. highest concentration of activated excitatory synapses.
Neighboring activated synapses add their effect largely lin-
2.6.3 Tetrode Development early, justifying the use of the fEPSP amplitude as a measure
of synaptic strength. Conversely, synchronous and local acti-
Following the recording of single-cell activity in awake, behav- vation of inhibitory bers also gives rise to large currents, but
ing animals (OKeefe and Dostrovsky, 1971; Ranck, 1973) it in the opposite direction. The usefulness of the recording pro-
became clear that hippocampal cells signaled differently from, cedure coupled with its simplicity lies at the heart of the pop-
say, visual cortical cells. The latter could be classied as sensi- ularity of eld potential studies.
tive to contrast or color and with specic and repeatable pat- Brian Cragg and Lionel Hamlyn (1955) were the rst to
terns to a standardized stimulus delivered to a restricted part record signicant elements of the hippocampal eld poten-
of the visual eldthe cells receptive eld. In contrast, hip- tials. However, because they placed their stimulation electrode
pocampal cells showed a much lower rate of discharge, more very close to the recording site, the axonal conduction dis-
variability from one trial to the next, and a receptive eld tance was very short. Consequently, their records of the synap-
related to the aberrant concept of space (see Chapter 11). tic component of the eld potentials were brief and difficult to
Subsequently, it became clear that hippocampal neurons oper- isolate from the rest of the compound signal (action poten-
ated in ensembles and deciphering the full code might receive tial in their nomenclature). Nevertheless, they were able to
simultaneous recording from a number of neighboring cells. report the rst evidence of conduction along the apical den-
This led to the development of recording techniques with dou- drites and found that the minimal latency was associated with
ble electrodes, or stereotrodes (McNaughton et al., 1983), and the dendritic position of the activating synapses. With a com-
nally quadruple electrodes, or tetrodes (OKeefe and Recce, missural input, the shape and conduction of these action
1993; Wilson and McNaughton, 1993). Such electrode assem- potentials were similar to those found with close-range stim-
blies allow simultaneous recording of a large number of cells ulation (Cragg and Hamlyn, 1957).
in the freely moving and behaving animal by taking into The large compound action potential that followed stim-
account the constant shape and form of the potential gener- ulation at close range, later called the population spike,
ated by discharges of each of the active cells as seen by each of was rst noted by Cragg and Hamlyn (1955, 1957). Although
the four electrodes. The tetrode technique has revolutionized claims to dendritic conduction had been made on the basis
the study of hippcampal neuronal activity and spatial behav- of neocortical surface stimulation, the specialized hippocam-
ior. For a further account of this eld, see Chapter 11. pal histological arrangement made interpretation of the
records much more convincing. Today, a new and much more
2.6.4 Field Potential Analysis detailed view on dendritic conduction has emerged, largely
owing to whole cell patch recordings from dendrites. Again,
After the pioneering use of microelectrodes during the 1940s, much of the evidence for dendritic properties has been
several years passed before further progress was made in stud- obtained from hippocampal preparations (Spruston et al.,
ies of the hippocampal region using cell discharge recordings. 1995; Johnston et al., 1996; Stuart et al., 1997) (see details in
During the mid-1950s, however, several groups started to use Chapter 5).
eld potentials to understand the pattern of activation or Several other groups exploited the laminated synaptic
inhibition of hippocampal systems. arrangement to identify the eld potentials associated with
In an early study of eld potentials, Lorente de N (1947) restricted dendritic synapses. To make the hippocampus a
studied the signal sequence when a nerve volley traveled useful preparation for studying synaptic activation, it was nec-
antidromically into the hypoglossus motor nucleus. He essary to ascribe a standard extracellular signal to a standard-
offered the rst theoretical explanation for the generation of ized input. Recording from various dendritic positions,
such eld potentials. Andersen (1960) described how commissural bers to CA1
 Box 21
Field potentials and current source density analysis

Field potentials are extracellular potentials recorded from groups of nerve cells in response to
synaptic or antidromic stimulation. In laminated structures such as the hippocampus, eld
potentials provide surprisingly detailed information on cellular activity.
A eld potential is generated by extracellular current owing across the tissue resistance
between the recording electrode and, in general, the ground electrode. Although measurable
extracellular voltages are generated by action potentials in a single neuron, and form the basis
of single unit recording, synaptic currents generated by single neurons are generally too small
to be detected. In the hippocampus, and other highly laminated structures such as the olfactory
cortex, the synchronous and localized currents generated by synaptic activation of a population
of pyramidal or granule cells gives rise to a characteristic and easily measured response called a
population or eld EPSP (fEPSP); with weak stimulation, the synaptic response is below
threshold for the generation of action potentials in the target cells, and a pure fEPSP is gener-
ated. With stronger stimulation, the cells discharge synchronously, giving rise to a population
spike which is superimposed on the rising phase of the fEPSP. Unlike the all-or-none action
potential generated by a single neuron, the population spike is a graded response; as stimulus
intensity increases, the number of neurons discharged becomes greater, and the population
spike becomes correspondingly larger.
In Box Fig. 21 stimulation of medial perforant path bers activates synapses in the middle
third of the molecular layer of the dentate gyrus (shaded gray). The eld responses recorded at
various positions along the soma and dendritic tree are shown on the left. Synaptically gener-
ated current ows into the dendrites in the activated region; inside the cells, current ows prox-
imally and distally away from the synaptic region, exiting where membrane area is greatest,
notably in the region of the soma. The current loop is completed extracellularly, with current
owing radially from distal and proximal sources towards the sink in the synaptically active
region. With a distal ground electrode, these synchronous extracellular currents give rise to
eld EPSPs that are negative in the region of current sinks and positive in regions of strong
current sources. Note that the current sink generated by synaptic activation is in the molecular
layer, while the passive source is in the cell body region and in distal dendrites; the converse sit-
uation obtains for the population spike; here, the active sink is in the cell body region while the
passive source is in the dendrites. The polarity of the fEPSP and the population spike are, as a
result, spatially out of phase.

Im

IL

Box Fig. 21. Field potentials. A. Field potentials generated by dentate granule cells in response
to stimulation of the medial perforant path. Activated synapses are located in the middle third
of the dendritic tree (gray band in B). In this region the eld potential (fEPSP) is negative,
reversing to a positive polarity near the cell body layer. The population spike (star) is negative
in the cell body layer, and positive in the dendritic region. (Source: Andersen et al., 1966a). C.
The direction of intracellular, extracellular and transmembrane current ow generated by
synaptic activation in an activated neuron.

28
 200

Recording Distance from Border of

-100

St.Pyramidale (m)
0

+100

200

300
5 mV
A B 10 msec

Box Fig. 22. Current source density analysis. A. Field potentials evoked by weak activation of
bers projecting to stratum oriens in area CA1, evoking a pure synaptic response without a
superimposed population spike. B. CSD analysis reveals a sink (positive) in stratum oriens and
a source (negative) in the cell body layer and proximal dendrites. Arbitrary units. (Source:
Richardson et al., 1987).
A further renement called current source density (CSD) analysis allows the precise regions
of inward and outward current to be plotted. Box Fig 22A displays the fEPSPs recorded at dif-
ferent locations in area CA1 in vitro following weak synaptic activation. For a given latency, a
depth prole can be generated, showing how the amplitude and polarity of the response
changes with location.. By Ohms law, the current owing in the longitudinal direction, parallel
to the dendrites, is proportional to the rate of change of voltage in that direction:
IL  k.V/L
where IL is the longitudinal current density at a given point L along the longitudinal axis of the
cell, V is the potential at point L, and k is the conductivity of the extracellular space. Thus the
rst spatial derivative of the depth prole yields a plot of the longitudinal current as a function
of location at the selected latency. Current owing through the membrane at any point must be
equal to the rate at which longitudinal current changes at that point, i.e.
IM  -IL/L  -k.2V/L2
The second spatial derivative of the depth prole is thus proportional to the amplitude of the
membrane current density, Im. By repeating this analysis at a succession of latencies, the time
course of current sinks and sources at each location can be calculated. A full CSD analysis,
computed from the sequence of fEPSPs evoked at different locations along the dendritic tree by
weak stimulation of Schaffer commissural bers projecting to the basal dendrites of pyramidal
cells in area CA1 is shown in Box Fig. 22B. The voltage responses are on the left, and the cur-
rent source density analysis, revealing the development of sources and sinks over time, are plot-
ted on the right.
Field EPSPs have the same time course as the synaptic current that generates them, and are
therefore phase advanced with respect to intracellular EPSPs which are delayed by the time
required for the membrane current to charge membrane capacitance (Box Fig 23). Further
discussion of eld potential theory can be found in Stevens (1966), Rall and Shepherd (1968),
Nicholson and Freeman (1975) and Johnston and Wu (1994).

+
5mV
Intra -

Extra +
1mV
-

10msec
Box Fig. 23. The extracellular eld EPSP reects intracellular events, but is phase advanced
with respect to the intracellular EPSP. Recordings were made from the dentate gyrus in vivo
(Source: Lmo, 1971).

29
30 The Hippocampus Book

and CA3 gave a local negative eld potential in exactly the cholinesterases, and aldolase to lactic, malic, and glutamic
same strata where Theodor Blackstad (1956) had found that dehydrogenases. Lowrys group also determined the lipid con-
the relevant bers terminated. Such negative eld potentials tent and type as related to cortical layers. Third, it was the rst
were taken as a sign of excitatory synaptic activity and was instance when a neurochemist explicitly exploited a special
called a eld excitatory postsynaptic potential (fEPSP). Above histological arrangement for localizing biochemical elements,
a given strength the fEPSP was interrupted by a compound initially in the hippocampus and then in the cerebellum. His
action potential, called the population spike because it was group reported impressive homogeneity in a single stratum
interpreted as the synchronous discharge of a number of and often a considerable variation among various strata. In
pyramidal cells. This interpretation was supported by the particular, he noted the much higher ATP concentration in
observation that most of the synaptically activated unitary cell the dendritic areas than in the cell body layer.
discharges fell within the time envelope of the population Using a microdissection method, Storm-Mathisen and
spike (Andersen et al., 1971a). Fonnum (1972) were able to attain quantitative data for
Thus, the extracellular eld potentials can be used as a sen- GABA and glutamate concentrations in various hippocampal
sitive, quantitative measure of the intensity of excitatory and strata. The amount of GABA was particularly high in the
inhibitory synaptic effects and the efficacy of postsynaptic pyramidal layer, in accord with the concentration of
activation. Final conrmation of the eld potential interpreta- inhibitory boutons here. However, the high amount of GABA
tion given above came with intracellular recording of dentate even in the stratum lacunosum-moleculare pointed to new
granule cells (Andersen et al., 1966b; Lmo, 1971) and the use challenges.
of isolated slices, where intracellularly recorded EPSPs were Peter Lewis and Charles Shute (1967) extended the chemi-
associated with excitatory eld potentials in CA1 pyramidal cal dissection by performing the histochemical analysis
cells (Schwarzkroin, 1975; Andersen et al., 1980). directly on sections from the hippocampus. They showed that
the distribution of acetylcholinesterase was concentrated in
2.6.5 Histochemistry: Pioneered certain lamina, an approach that heralded a new approach to
in the Hippocampus chemical analysis of the CNS. This histochemical approach
also provided the groundwork for many of the studies on
We have become used to the hippocampus as a favorite prepa- synaptic sprouting that followed during the 1970s (see
ration for histological, electrophysiological, and behavioral Chapter 9).
studies. Less well known is its essential role in the develop- The later development of histochemical and immunohis-
ment of neurochemical methodology and histochemistry. tochemical studies have largely concentrated on microscopic
Once again, it was the packing of homogeneous cells into sections. For this approach, hippocampal tissue has also been
compact cell layers and the striking stratication of afferent a favorite choice. Among these developments, the electron
bers and synapses to specic parts of the dendritic tree of microsopic identication of bouton contents with antibodies
principal cells that appealed to the exploring scientist. raised against an aggregate of amino acids and bovine globu-
Until the middle of the twentieth century, virtually all neu- lin marked an entirely new, powerful approach (Storm-
rochemical analyses were conducted on homogenized tissue Mathisen et al., 1983). Thus, the hippocampus has played an
of relatively large samples from various parts of the brain. important role in the development of histochemical tech-
This procedure was thought to be necessary because of the niques for both fresh and xed tissue.
extreme complexity of central nervous tissue. In an attempt to
perform analysis of more discrete CNS elements, Oliver 2.6.6 Pharmacological Analysis
Lowry exploited the synaptic lamination of the hippocampus of Cellular Properties
to perform a detailed neurochemical dissection of cortical tis-
sue. He noted the following: The hippocampus has also been important for the develop-
ment of ideas in the neuropharmacology of synaptic trans-
Ammons horn is a region of the cerebral cortex which
mission in the CNS. Following the demonstration of
is organized in such a manner as to invite quantitative
glutamate sensitivity of spinal cord cells by David Curtis and
histochemical study.
Jeff Watkins (1960), Tim Biscoe and Donald Straughan (1966)
Enzyme concentrations and lipid type and concentration found that hippocampal pyramidal cells were sensitive to ion-
were measured for individual strata of the hippocampal cor- tophoretic application of glutamate. Together, these studies
tex. In a pioneering series of papers (Lowry et al. 1954a,b,c; suggested the radical (at the time) possibility that glutamate
1964), Lowry and his colleagues made several fundamental was an excitatory neurotransmitter in the CNS. Support came
advances. First, they introduced a number of methodological from an ingenious study on hippocampal slices in which
improvements. By dissecting freeze-dried tissue from thin Nadler et al. (1976) measured Ca2-dependent release of
slices of hippocampus under a microscope, they were able to aspartate and glutamate. Because the release of both com-
reduce the sample size to 5 to 10 g of tissue. Second, the pounds was grossly reduced after lesions of commissural or
accuracy of the enzymatic measurements greatly improved. entorhinal bers, the release is likely to have come from bou-
The enzymes of interest spanned a large range from acid and tons of the two fiber systems, thus supporting the idea that
alkaline phosphatases through adenosine triphosphatase, aspartate or glutamate could be excitatory transmitters.
 Historical Perspective 31

Postsynaptic inhibition produced by Purkinje cells in the pocampus has been a focus of interest in epilepsy research
lateral vestibular, and deep cerebellar nuclei was rst shown to (Schwartzkroin, 1997).
be mediated by GABA by Obaka and colleagues (1967). As The hippocampal formation has proven to be particularly
mentioned above for hippocampal inhibition, David Curtis, vulnerable to various traumatic insults. Spielmeyer (1925)
John Felix, and Hugh McLellan (1970) showed that the large, noted that Sommers sector, which corresponds to the CA1
ubiquitous IPSPs in all principal cells of the hippocampus eld of the hippocampus, is a brain structure extremely vul-
could be blocked by localized application of bicuculline nerable to ischemic or hypoxic insults. This may be related to
methochloride, establishing that these IPSPs were therefore a lower level of mitochondrial oxidative enzymes in CA1
mediated by GABAA receptors. Thus, the studies of Curtis and pyramidal dendrites than in other hippocampal subregions
his colleagues extended the Obata et al. results to cortical (Davolio and Greenamyre, 1995; Kuroiwa et al., 1996). Early
inhibitory pathways. neuropathologists also noted that the hippocampal formation
is a major target for pathological changes in patients suffering
2.6.7 Development of Computational from senile dementia of the Alzheimers type (Alzheimer,
Models of Neural Networks 1909). These issues continue to challenge modern neuro-
science, as described in Chapter 16.
In a remarkable set of three papers, David Marr (1969, 1970,
1971) developed a theoretical proposal for the mode of
operation of three major components of the CNS. Arguably, REFERENCES
the triad inaugurated modern computational neuroscience.
Further description of this development is found in Chap- Adrian ED, Zotterman Y (1926) The impulses produced by sensory
ter 14. nerve-endings. Part 2. The response of a single end-organ. J
Marr started with the cerebellum, for which he proposed Physiol (Lond) 61:151171.
the main operation to be a comparison between an initial state Aggleton JP (1993) The contribution of the amygdala to normal and
of the motor system and the state after an elemental move- abnormal emotional states. Trends Neurosci 16:328333.
ment had taken place (Marr, 1969). The next article dealt with Ajmone-Marsan C (1961) Electrographic aspects of epileptic neu-
the neocortex and how the cortical matrix behaved during ronal aggregates. Epilepsia 2:2238.
perception and attention (Marr, 1970). The third paper con- Albe-Fessard D, Buser P (1952) tude de rponses neuroniques
laide de microlectrodes intrasomatiques. Rev Neurol (Paris)
tained a description of the mode of operation of the archicor-
87:455.
tex (Marr, 1971), meaning the hippocampal formation. Basing
Alksne JF, Blackstad TW, Walberg F, White LE Jr (1966) Electron
his proposal on what was then known about its neural cyto- microscopy of axon degeneration: a valuable tool in experimen-
architecture, he hypothesized that the hippocampus per- tal neuroanatomy. Ergeb Anat Entwicklungsgesch 39:332.
formed a simple memory function which allowed the animal Allen WF (1940) Effect of ablating the frontal lobes, hippocampi and
to make use of previous experience to adapt its behavior. In occipito-parieto-temporal (excepting pyriform areas) lobes on
the process, he described how elemental neurophysiological positive and negative olfactory conditioned reexes. Am J
processes could perform certain basic mathematical proce- Physiol 132:8191.
dures. For example, inhibition exerted at a dendritic level Alzheimer A (1907) ber eine eigenartige Erkrankung der
would cause relatively little enhancement of the resting mem- Hirnrinde. Allg Z Psychiatr Psych Gerichtl Med (Berl) 64:
brane potential of the neuron, a procedure equivalent to sub- 146148.
Amaral DG, Insausti R, Cowan WM (1987) The entorhinal cortex of
traction. In contrast, inhibition at the soma level, produced by
the monkey. I. Cytoarchitectonic organization. J Comp Neurol
basket cells, would represent a much stronger change of the
264:326355.
membrane potential, removing a substantial part of the Andersen P (1960) Interhippocampal impulses. II. Apical dendritic
synaptic drive, equivalent to mathematical division. activation of CA1 neurons. Acta Physiol Scand 48:178208.
He also pointed to the recurrent axon collaterals of CA3 as Andersen P, Eccles JC, Lyning Y (1964a) Location of postsynaptic
a possible neural basis for attractor networks in which the inhibitory synapses on hippocampal pyramids. J Neurophysiol
excitatory interactions between networks of pyramidal cells 27:592607.
could form the basis of content-addressable memories and Andersen P, Eccles JC, Lyning Y (1964b) Pathway of postsynaptic
support retrieval of a complete memory from partial infor- inhibition in the hippocampus. J Neurophysiol 27:608619.
mation (see Section 5 in Chapter 14). Andersen P, Blackstad TW, Lmo T (1966a) Location and identica-
tion of excitatory synapses on hippocampal pyramidal cells. Exp
Brain Res 1:236248.
2.6.8 The Hippocampal Formation:
Andersen P, Holmqvist B, Voorhoeve PE (1966b) Excitatory synapses
A Test Bed for Several Types of Neural on hippocampal apical dendrites activated by entorhinal stimu-
Dysfunction and Neuropathology lation. Acta Physiol Scand 66:461472.
Andersen P, Bliss TVP, Skrede KK (1971a) Unit analysis of hip-
It is a well established clinical observation that epileptic pocampal population spikes. Exp Brain Res 13:208221.
patients exhibit a high frequency of temporal lobe foci, often Andersen P, Bliss TVP, Skrede KK (1971b) Lamellar organization of
coupled with the neuropathological nding of hippocampal hippocampal exitatory pathways. Exp Brain Res 13:222238.
sclerosis (Sommer, 1890; Gastaut, 1956). As a result, the hip- Andersen P, Bland BH, Skrede KK, Sveen O, Westgaard RH (1972)
32 The Hippocampus Book

Single unit discharges in brain slices maintained in vitro. Acta Broca P (1861a) Sur le principe des localisations crbrales. Bull Soc
Physiol Scand 84:1A2A. Anthropol 2:190204.
Andersen P, Sundberg, SH, Sveen O, Wigstrm H (1977) Specic Broca P (1861b) Perte de la parole, ramollissement chronique et
long-lasting potentiation of synaptic transmission in hip- destruction partielle du lobe antrieur gauche [sur le sige de la
pocampal slices. Nature 266:736737. facult du langage]. Bull Soc Anthropol 2:235238.
Andersen P, Silfvenius H, Sundberg SH, Sveen O (1980) A compari- Brodal A (1947) Hippocampus and the sense of smell. Brain
son of distal and proximal dendritic synapses on CA1 pyramids 70:179222.
in guinea-pig hippocampal slices in vitro. J Physiol 307:273299. Brodal A (1981) Neurological anatomy in relation to clinical medicine.
Aranzi GC (1564) De humano foetu opusculum. Bologna: Rubrii. New York: Oxford University Press.
Banker GA, Cowan WM (1977) Rat hippocampal neurons in Brodmann K (1909) Vergleichende Lokalisationslehre der Grosshirn-
dispersed cell culture. Brain Res 126:397425. rinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues.
Banker G, Goslin K (1991) Primary dissociated cell cultures of neu- Leipzig: Johann Ambrosius Barth Verlag.
ral tissue. In: Culturing nerve cells (Banker G, Goslin K, eds), pp Brooks CMcC, Eccles JC (1947) Electrical investigations of the
171. Cambridge, MA:MIT Press. monosynaptic pathway through the spinal cord. J Neurophysiol
Bard P (1934) On emotional expression after decortication with 10:251274.
some remarks on certain theoretical views. Part 1. Psychol Rev Brown S, Schfer EA (1888) An investigation into the function of the
4:309322; Part 2: 4:424449. occipital and temporal lobes of the monkeys brain. Philos Trans
Bekhterev V (1900) Demonstration eines Gehirns mit Zerstrung der R Soc B 179:303327.
vorderen und inneren Theile der Hirnrinde beider Bush G, Luu P, Posner MI (2000) Cognitive and emotional inuences
Schlafenlappen. Neurol Zentralbl 19:990991. in anterior cingulate cortex. Trends Cogn Sci 4:215222.
Berger H (1929) ber das Elektroenkephalogramm des Menschen. Cannon WB (1929) Bodily changes in pain, hunger, fear and rage. New
Arch Psychiatr Nervenkr 87:527570. York: Appleton-Century-Crofts.
Biscoe T, Straughan D (1966) Micro-electrophoretic studies of neu- Cohen NJ, Squire LR (1980) Preserved learning and retention of
rones in the cat hippocampus. J Physiol (Lond) 183:341359. pattern-analyzing skill in amnesia: dissociation of knowing how
Bishop S, Duncan J, Brett M, Lawrence AD (2004) Prefrontal cortical and knowing that. Science 210:207210.
function and anxiety: controlling attention to threat-related Corkin S, Amaral DG, Gonzalez RG, Johnson KA, Hyman BT (1997)
stimuli. Nat Neurosci 7:184188. H.M.s medial temporal lobe lesion: ndings from magnetic res-
Bjrklund A, Stenevi U (1981) In vivo evidence for a hippocampal onance imaging. J Neurosci 17:39643979.
adrenergic neuronotrophic factor specically released on septal Correll RE, Scoville WB (1967) Signicance of delay in the perform-
deafferentation. Brain Res 229:403428. ance of monkeys with medial temporal lobe resections. Exp
Bjrklund A, Stenevi U (1984) Intracerebral neural implants: neuro- Brain Res 4:8596.
logical replacement and reconstruction of damaged circuitries. Cotman C, Gentry C, Steward O (1977) Synaptic replacement in the
Annu Rev Neurosci 7:279308. dentate gyrus after unilateral entorhinal lesion: electron micro-
Bjorklund A, Johansson B, Stenevi U, Svendgaard NA (1975) Re- scopic analysis of the extent of replacement of synapses by the
establishment of functional connections by regenerating central remaining entorhinal cortex. J Neurocytol 6:455464.
adrenergic and cholinergic axons. Nature 53:446448. Cowan WM, Gottlieb DI, Hendrickson AE, Price JL, Woolsey TA
Bjorklund A, Stenevi U, Svendgaard N (1976) Growth of trans- (1972) The autoadiographic demonstration of axonal connec-
planted monoaminergic neurones into the adult hippocampus tions in the central nervous system. Brain Res 37:2151.
along the perforant path. Nature 262:787790. Cragg B, Hamlyn LH (1955) Action potentials of the pyrami-
Black AH (1972) The operant conditioning of central nervous system dal neurons of the hippocampus of the rabbit. J Physiol (Lond)
electrical activity. In: The psychology of learning and motivation, 129:608627.
vol 6 (Bower G, ed), pp 4795. San Diego: Academic Press. Cragg B, Hamlyn LH (1957) Some commissural and septal connec-
Black AH, Young GA (1972) Electrical activity of the hippocampus tions of the hippocampus in the rabbit: a combined histological
and cortex in dogs operantly trained to move and to hold still. J and electrical study. J Physiol (Lond) 135:460485.
Comp Physiol Psychol 79:128141. Curtis DR, Watkins JC (1960) The chemical excitation of spinal neu-
Blackstad TW (1956) Commissural connections of the hippocampal rones by certain acidic amino acids. J Physiol (Lond)
region in the rat, with special reference to their mode of termi- 150:656682.
nation. J Comp Neurol 105:417537. Curtis DR, Felix D, McLellan H (1970) GABA and hippocampal inhi-
Blackstad TW (1958) On the termination of some afferents to the bition. Br J Pharmacol 40:881883.
hippocampus and fascia dentata. Acta Anat (Basel) 35:202214. Dalton AJ, Black AH (1968) Hippocampal electrical activity during
Blackstad TW, Flood PR (1963) Ultrastructure of hippocampal axo- the operant conditioning of movement and refraining from
somatic synapses. Nature 198:542543. movement. Commun Behav Biol 2:267273.
Bliss TVP, Lmo T (1970) Plasticity in a monosynaptic cortical path- Davolio C, Greenamyre JT (1995) Selective vulnerability of the CA1
way. J Physiol (Lond) 207:61P. region of hippocampus to the indirect excitotoxic effects of mal-
Bliss TVP, Lmo T (1973) Long-lasting potentiation of synaptic onic acid. Neurosci Lett 192:2932.
transmission in the dentate gyrus of the anaesthetized rabbit. J Douglas RJ, Isaacson RL (1964) Hippocampal lesions and activity.
Physiol (Lond) 232:331356. Psychon Sci 1:187188.
Bliss TVP, Richards C (1971) Some experiments with in vitro hip- Dunnett SB, Low WC, Iversen SD, Stenevi U, Bjrklund A (1982)
pocampal slices. J Physiol (Lond) 214:79P. Septal transplants restore maze learning in rats with fornix-m-
Botvinick MM, Braver TS, Barch DM, Carter CS, Cohen JD (2001) bria lesions. Brain Res 251:335348.
Conflict monitoring and cognitive control. Psychol Rev Eccles JC, Llins R, Sasaki K (1966) The inhibitory interneurones
108:624652. within the cerebellar cortex. Exp Brain Res 1:116.
 Historical Perspective 33

Ferrier D (1876) The functions of the brain. London: Smith, Elder. Hamlyn LH (1963) An electron microscope study of pyramidal neu-
Fink RP, Heimer L (1967) Two methods for selective silver impregna- rons in the Ammons horn of the rabbit. J Anat (Lond)
tion of degenerating axons and their synaptic endings in the 97:189201.
central nervous system. Brain Res 4:369374. Hayman LA, Rexer JL, Pavol MA, Strite D, Meyers CA (1998) Klver-
Gaffan D (1974a) Loss of recognition memory in rats with lesions of Bucy syndrome after selective damage of the amygdala and its
the fornix. Neuropsychologia 10:327341. cortical connections. J Neuropsychiatry Clin Neurosci
Gaffan D (1974b) Recognition impaired and association intact in the 10:354358.
memory of monkeys after transaction of the fornix. J Comp Hirano T, Best P, Olds J (1970) Units during habituation, discrimina-
Physiol Psychol 86:11001109. tion learning, and extinction. Electroencephalogr Clin
Ghwiler BH (1981) Organotypic monolayer cultures of nevous tis- Neurophysiol 28:127135.
sue. J Neurosci Methods 4:329342. Hirsh R (1974) The hippocampus and contextual retrieval of infor-
Ghwiler BH (1988) Organotypic cultures of neural tissue. Trends mation from memory: a theory. Behav Biol 12:421444.
Neurosci 11:484489. Hjorth Simonsen A (1973) Some intrinsic connections of the hip-
Ghwiler BH, Brown DA (1985) Functional innervation of cultured pocampus in the rat: an experimental analysis. J Comp Neurol
hippocampal neurons by cholinergic afferents from co-cultured 147:145161.
septal explants. Nature 313:577579. Holmes JE, Adey WR (1960) Electrical activity of the entorhinal cor-
Gastaut H (1956) Etude electroclinique des pisodes psychotiques tex during conditioned behavior. Am J Physiol 199:741744.
survenant en dehors des crises clinique chez les pileptiques. Rev Hounsgaard J, Yarom Y (1985) Intrinsic control of electroresponsive
Neurol (Paris) 94:587594. properties of transplanted mammalian brain neurons. Brain Res
Gerfen CR, Sawchenko PE (1984) An anterograde neuroanatomical 335:372376.
tracing method that shows the detailed morphology of neurons, Isaacson RL, Kimble DP (1972) Lesions of the limbic system:
their axons and terminals: immunohistochemical localization of their effects upon hypotheses and frustration. Behav Bull 7:
an axonally transported plant lectin Phaseolus vulgaris-leucoag- 767793.
glutinin. Brain Res 290:219238. Jackson JH (1865) In: Selected writings of John Hughlings Jackson
Glees P (1946) Terminal degeneration within the central nervous sys- (2 vols). (Taylor J, ed). London: Hodder & Stoughton, 1932.
tem as studied by a new silver method. J Neuropathol Exp Neurol Jackson JH, Beevor CE (1890) Case of tumor of the right temporo-
5:5459. sphenoidal lobe bearing on the localization of the sense of smell
Gloor P, Vera CL, Sperti L (1964) Electrophysiological studies of hip- and on the interpretation of a particular variety of epilepsy.
pocampal neurons. 3. Responses of hippocampal neurons to Brain 12:346357.
repetitive perforant path volleys. Electroencephalogr Clin Johnston D, Wu M-S (1994) Foundations of cellular neurophysiology.
Neurophysiol 17:353370. Cambridge, MA: Bradford Books, MIT Press.
Goddard GV, McIntyre DC, Leech CK (1969) A permanent change in Johnston D, Magee JC, Colbert CM, Cristie BR (1996) Active proper-
brain function resulting from daily electrical stimulation. Exp ties of neuronal dendrites. Annu Rev Neurosci 19:165186.
Neurol 25:295330. Jung R (1949) Hirnelektrische Untersuchungen ber den Elekt-
Goldwitz D, Cotman CW (1978) Induction of extensive mbrial rokrampf: Die Erregungsablufe in corticalen Hirnregionen bei
branching in the adult rat brain. Nature 275:6467. Katze und Hund. Arch Psychiatr Nervenkr 183:206244.
Golgi C (1886) Sulla na anatomia degli organi centrali del sistema Jung R, Kornmller AE (1938) Eine methodik der Ableitung
nervosa. Milan: Hoepli. lokalisierter Potentialschwankungen aus subcorticalen
Golgi C, Bentivogli M, Swanson L (2001) On the ne structure of the Hirngebieten. Arch Psychiatr Nervenkr 109:130.
pes hippocampi major. Brain Res Bull 54:461483. Translated Kaada BR, Pribram KH, Epstein JA (1949) Respiratory and vascular
from Golgi C (1886) responses in monkeys from temporal pole, insula, orbital sur-
Grafstein B (1971) Transneuronal transfer of radioactivity in the cen- face and cingulate. Fed Proc 8:8384.
tral nervous system. Science 172:177179. Kaada BR, Jansen J Jr, Andersen P (1953) Stimulation of the hip-
Grastyn E, Lissk K, Madarsz I, Donhoffer H (1959) Hippocampal pocampus and medial cortical areas in unanesthetized cats.
electrical activity during the development of conditioned Neurology 3:844857.
reexes. Electroencephalogr Clin Neurophysiol 11:409430. Kaada BR, Rasmussen EW, Kveim O (1961) Effects of hippocampal
Gray EG (1959) Axo-somatic and axo-dendritic synapses of the cere- lesions on maze learning and retention in rats. Exp Neurol
bral cortex: an electron microscope study. J Anat (Lond) 3:333355.
93:420433. Kandel ER, Spencer WA (1961a) Electrophysiology of hippocampal
Gray JA (1982) The neuropsychology of anxiety: an enquiry into the neurons. II. After-potentials and repetitive ring. J Neurophysiol
functions of the septo-hippocampal system. New York: Oxford 24:243259.
University Press. Kandel ER, Spencer A (1961b) Excitation and inhibition of
Gray JA (2001) Emotional modulation of cognitive control: single pyramidal cells during hippocampal seizure. Exp Neurol
approach-withdrawal states double-dissociate spatial from ver- 4:162179.
bal two-back task performance. J Exp Psychol Gen 130:436452. Kandel ER, Spencer WA, Brinley FJ (1961) Electrophysiology of hip-
Green JD, Adey WR (1956) Electrophysiological studies of hip- pocampal neurons. I. Sequential invasion and synaptic organi-
pocampal connections and excitability. Electroencephalogr Clin zation. J Neurophysiol 24:225242.
Neurophysiol 8:245263. Kerns JG, Cohen JD, MacDonald AW, Cho RY, Stenger VA, Carter CS
Green JD, Arduini AA (1954) Hippocampal electrical activity in (2004) Anterior cingulate conict monitoring and adjustments
arousal. J Neurophysiol 17:533557. in control. Science 303:10231026.
Gross CG (1993) Huxley versus Owen: the hippocampus minor and Kimble DP (1963) The effects of bilateral hippocampal lesions in
evolution. Trends Neurosci 16:493498. rats. J Comp Physiol Psychol 56:273283.
34 The Hippocampus Book

Kimble DP (1968) Hippocampus and internal inhibition. Psychol ral activity (Cohen MM, Snider RS, eds), pp. 178191. New
Bull 70:285295. York: Harper & Row.
Klver H, Bucy PC (1937) Psychic blindness and other symptoms fol- Lynch G, Matthews DA, Mosko S, Parks T, Cotman C (1972) Induced
lowing bilateral temporal lobectomy in rhesus monkeys. Am J acetylcholinesterase-rich layer in rat dentate gyrus following
Physiol 119:352353. entorhinal lesions. Brain Res 42:311318.
Korsakov SS (1889) tude mdico-psychologique sur une forme des McIlwain H, Snyder SH (1970) Stimulation of piriform and neo-cor-
maladies de mmoire. Rev Philos 28:501530. tical tissues in an in vitro ow-system: metabolic properties and
Korsakov SS (1890) Eine psychische Strung kombiniert mit multi- release of putative transmitters. J Neurochem 17:521530.
plen Neuritis. Allg Z Psychiatr 46:475485. McNaughton BL, OKeefe J, Barnes CA (1983) The stereotrode: a new
Kramis R, Vanderwolf CH, Bland BH (1975) Two types of hip- technique for simultaneous isolation of several single units in
pocampal rhythmical slow activity in both the rabbit and the the central nervous system from multiple unit records. J
rat: relations to behavior and effects of atropine, diethyl ether, Neurosci Methods 8:391397.
urethane, and pentobarbital. Exp Neurol 49:5885. Marchi V, Algeri EG (1886) Sulle degenerazioni discendenti consecu-
Kristensson K, Olsson Y (1971) Retrograde axonal transport of pro- tive a lesioni in diverse zone della corteccia cerebrale. Riv Sper
tein. Brain Res 29:4347. Freniatr Med Leg Alien Ment 14:149.
Kuroiwa T, Terakado M, Yamaguchi T, Endo S, Ueki M, Okeda R Marr D (1969) A theory of cerebellum. J Physiol (Lond) 202:437470.
(1996) The pyramidal layer of sector CA1 shows the lowest hip- Marr D (1970) A theory for cerebral neocortex. Proc R Soc B
pocampal succinate dehydrogenase activity in normal and 176:161234.
postischemic gerbils. Neurosci Lett 206:117120. Marr D (1971) Simple memory: a theory for archicortex. Philos Trans
Kveim O, Setekleiv J, Kaada BR (1964) Differential effects of hip- R Soc B 262:2381.
pocampal lesions on maze and passive avoidance learning in Matthews DA, Cotman C, Lynch G (1976) An electron microscopic
rats. Exp Neurol 9:5972. study of lesion-induced synaptogenesis in the dentate gyrus of
Larrabee MG, Bronk DW (1939) Prolonged facilitation of synaptic the adult rat. I. Magnitude and time course of degeneration.
excitation in sympathetic ganglia. J Neurophysiol 10:139154. Brain Res 115:121.
Lasek R (1975) Axonal transport and the use of intracellular markers Mishkin M (1954) Visual discrimination performance following par-
in neuroanatomical investigations. Fed Proc 34:16031611. tial ablations of the temporal lobe. II. Ventral surface vs. hip-
Lewis PR, Shute CCD (1967) The cholinergic limbic system: projec- pocampus. J Comp Physiol Psychol 47:187193.
tion to the hippocampal formation, medial cortex, nuclei of the Mishkin M, Delacour J (1975) An analysis of short-term visual mem-
ascending cholinergic reticular system and the subfornical ory in the monkey. J Exp Psychol Anim Behav Process 1:326334.
organ and supra-optic crest. Brain 90:521537. Mosko S, Lynch G, Cotman CW (1973) The distribution of septal
Li C-L, McIlwain H (1957) Maintenance of resting membrane poten- projections to the hippocampus of the rat. J Comp Neurol 152:
tials in slices of mammalian cerebral cortex and other tissues in 163174.
vitro. J Physiol (Lond) 139:178190. Nadel L, OKeefe J (1974) The hippocampus in pieces and patches: an
Li XG, Somogyi P, Ylinen A, Buzsaki G (1994) The hippocampal CA3 essay on modes of explanation in physiological psychology. In:
network: an in vivo intracellular labeling study. J Comp Neurol Essays on the nervous system: a festschrift for Prof JZ Young
339:181208. (Bellairs R, Gray EG, eds), pp 367390. Oxford: Clarendon
Ling G, Gerard RW (1949) The normal membrane potential of frog Press.
sartorius bers. J Cell Physiol 34:383396. Nadel L, OKeefe J, Black AH (1975) Slam on the brakes: a critique of
Lloyd DPC (1949) Post-tetanic potentiation of response in monosy- Altmann, Brunner and Bayers response inhibition model of
naptic pathways of the spinal cord. J Gen Physiol 33:147170. hippocampal function. Behav Biol 14:151162.
Lmo T (1966) Frequency potentiation of excitatory synaptic activ- Nadler JV, Vaca KW, White WF, Lynch GS, Cotman CW (1976)
ity in the dentate area of the hippocampal formation. Acta Aspartate and glutamate as possible transmitters of excitatory
Physiol Scand 68(Suppl 277):128. hippocampal afferents. Nature 260:538540.
Lmo T (1971) Patterns of activation in a monosynaptic cortical Nauta WJH (1950) ber die sogenannte terminale Degeneration in
pathway: the perforant path input to the dentate area of the hip- der Zentralnervensystem und ihre Darstellung durch
pocampal formation. Exp Brain Res 12:1845. Silberimprgnation. Schweiz Arch Neurol Psychiatr 66:353376.
Lorente de N R (1934) Studies on the structure of the cerebral cor- Nauta WJH, Gygax PA (1954) Silver impregnation of degenerating
tex. II. Continuation of the study of the ammonic system. J axons in the central nervous system: a modied technic. Stain
Psychol Neurol (Lpz) 46:113177. Technol 29:9193.
Lorente de N (1947) Action potential of the motoneurons of the Nicoll RA, Eccles JC, Oshima T, Rubia F (1975) Prolongation of hip-
hypoglossus nucleus. J Cell Comp Physiol 29:207288. pocampal inhibitory postsynaptic potentials by barbiturates.
Lowry OH, Roberts NR, Leiner KY, Wu M-L, Farr AL (1954a) The Nature 258:625627.
quantitative histochemistry of brain. I. Chemical methods. J Obata K, Ito M, Ochi R, Sato N (1967) Pharmacological properties of
Biol Chem 207:117. the postsynaptic inhibition of the Purkinje cell axons and the
Lowry OH, Roberts NR, Wu M-L, Hixon WS, Crawford EJ (1954b) action of -aminobutyric acid on Deiters neurons. Exp Brain
The quantitative histochemistry of brain. II. Enzyme measure- Res 4:4357.
ments. J Biol Chem 207:1927. OKeefe J, Dostrovsky J (1971) The hippocampus as a spatial map:
Lowry OH, Roberts NR, Leiner KY, Wu M-L, Farr AL, Albers RW preliminary evidence from unit activity in the freely-moving
(1954c) The quantitative histochemistry of brain. III. Ammons rat. Brain Res 34:171175.
horn. J Biol Chem 207:3949. OKeefe J, Recce ML (1993) Phase relationship between hippocampal
Lowry O (1964) In: Morphological and biochemical correlates of neu- place units and the EEG theta rhythm. Hippocampus 3:317330.
 Historical Perspective 35

Olton DS, Walker JA, Gage FH (1978) Hippocampal connections and Schreiner L, Kling A (1956) Rhinencephalon and behavior. Am J
spatial discrimination. Brain Res 139:295308. Physiol 184:486490.
Orbach J, Milner B, Rasmussen T (1960) Learning and retention in Schwartzkroin P (1975) Characteristics of CA1 neurons recorded
monkeys after amygdala-hippocampus resection. Arch Neurol intracellularly in the hippocampal in vitro slice preparation.
3:230235. Brain Res 85:423436.
Palay SL, Palade GE (1955) The ne structure of neurons. J Biophys Schwartzkroin PA (1997) Origins of the epileptic state. Epilepsia
Biochem Cytol 1:6988. 38:853858.
Papez JW (1937) A proposed mechanism of emotion. Arch Neurol Schwartzkroin P, Andersen P (1975) Glutamic acid sensitivity of den-
Psychiatry 38:725743. drites in hippocampal slices in vitro. Adv Neurol 12:4551.
Peneld W, Erickson TC (1941) Epilepsy and cerebral localization: a Schwartzkroin P, Wester K (1975) Long-lasting facilitation of a
study of the mechanism, treatment, and prevention of epileptic synaptic potential following tetanization in the in vitro hip-
seizures. Springeld, IL: Charles C Thomas. pocampal slice. Brain Res 89:107119.
Phillips CG (1956) Intracellular records from Betz cells in the cat. Q Scoville W, Milner B (1957) Loss of recent memory after bilateral
J Exp Physiol 41:5869. hippocampal lesions. J Neurol Neurosurg Psychiatry 20:1121.
Raisman G (1969) Neuronal plasticity in the septal nuclei of the adult Segal M, Bloom FE (1974a) The action of norepinephrine in the rat
rat. Brain Res 14:2548. hippocampus. I. Iontophoretic studies. Brain Res 72:7997.
Raisman G, Field PM (1973) A quantitative investigation of the Segal M, Bloom FE (1974b) The action of norepinephrine in the rat
development of collateral reinnervation after partial deaf- hippocampus. II. Activation of the input pathway. Brain Res
ferentation of the septal nuclei. Brain Res 50:241264. 72:99114.
Raisman G, Field P (1990) Synapse formation in the adult brain after Semon R (1908) The mneme. English translation: London: 1908,
lesions and after transplantation of embryonic tissue. J Exp Biol 1911, and 1921. First German edition: 1904.
153:277287. Shapiro ML, Simon DK, Olton DS, Gage FH III, Nilsson O, Bjorklund
Raisman G, Cowan WM, Powell TPS (1965) The extrinsic, com- A (1989) Intrahippocampal grafts of fetal basal forebrain tissue
missural and association bres of the hippocampus. Brain alter place elds in the hippocampus of rats with mbria-fornix
88:963998. lesions. Neuroscience 32:118.
Raisman G, Cowan WM, Powell TPS (1966) An experimental analysis Shepherd GM (1972) The neuron doctrine: a revision of functional
of the efferent projection of the hippocampus. Brain 89:83108. concepts. Yale J Biol Med 45:584599.
Rall W, Shepherd GM (1968) Theoretical reconstruction of eld Shepherd MG, Raastad M, Andersen P (2002) General and vari-
potentials and dendrodendritic synaptic interactions in olfactory able features of varicosity spacing along unmyelinated axons in
bulb. J Neurophysiol 31:884915. the hippocampus and cerebellum. Proc Natl Acad Sci USA 99:
Ramon y Cajal S (1893) Estructura del asta de Ammon. Anal Sociedad 63406345
Espaol Historia Natural, vol 22. Translated to German by von Sherrington CS (1897) The central nervous system. In: A textbook of
Klliker A (1893). In Zeitschr Wiss Zool 56. physiology (Foster M, ed), vol 3, 7th ed. London: Macmillan.
Ramon y Cajal S (1911) Histologie du systme nerveux de lhomme et Shipley MT, Adamek GD (1984) The connections of the mouse olfac-
des vertbrs. Paris: A. Maloine. Reprinted (1955) Madrid: tory bulb: a study using orthograde and retrograde transport of
Instituto Ramon y Cajal. wheat germ agglutinin conjugated to horseradish peroxidase.
Ranck JB Jr (1973) Studies on single neurons in dorsal hippocam- Brain Res Bull 12:669688.
pal formation and septum in unrestrained rats. Exp Neurol 41: Skrede KK, Westgaard RH (1971) The transverse hippocampal slice:
461455. a well-dened cortical structure maintained in vitro. Brain Res
Renshaw B, Forbes A, Morison BR (1940) Activity of isocortex 35:589593.
and hippocampus: electrical studies with micro-electrodes. Sloan N, Jasper H (1950) Studies of the regulatory functions of
J Neurophysiol 3:74105. the limbic cortex. Electroencephalogr Clin Neurophysiol
Ribot T (1881) Les maladies de la mmoire. Paris: Ballire. 2:317327.
Richards CD, McIlwain H (1967) Electrical responses in brain sam- Sommer W (1890) Erkrankung des Ammonshorns als aetiologisches
ples. Nature 215:704707. Moment der Epilepsie. Arch Psychiatr Nervenkrank 10:631675.
Richardson TL, Turner RW, Miller JJ (1987) Action-potential dis- Spencer WA, Thompson RF, Neilson DR Jr (1966) Response decre-
charge in hippocampal CA1 pyramidal neuron: current source- ment of the exion reex in the acute spinal cat and transient
density analysis. J Neurophysiol 58:981996. restoration by strong stimuli. J Neurophysiol 29:221239.
Ries FA, Langworthy OR (1937) A study of the surface structure of Spielmeyer W (1925) Zur Pathogenese rtlich elektiver Gehirnvern-
the brain of the whale (Balaenoptera physalus and Physeter derungen. Z Ges Neurol Psychiatr 99:756776.
catodon). J Comp Neurol 68:147. Spruston N, Jonas P, Sakmann B (1995) Dendritic glutamate receptor
Rose M (1935) Cytoarchitektonik und Myeloarchitektonik der channels in rat hippocampal CA3 and CA1 pyramidal neurons.
Grosshirnsrinde. In: Handbuch der Neurologie (Bumke O, J Physiol (Lond) 482:325352.
Foerster O, eds), vol 1, pp 588778. Berlin. Squire LR (1998) Memory systems. C R Acad Sci III 321:153156.
Sagar JH, Cohen NJ, Corkin S, Growdon JH (1985) Dissociations Squire LR (2004) Memory systems of the brain: a brief history and
among processes in remote memory. Ann NY Acad Sci current perspective. Neurobiol Learn Mem 82:171177.
444:533535. Stevens CF (1966) Neurophysiology: a primer. New York: Wlley.
Sala L (1891) Zur Anatomie des grossen Seepferdfusses. Zeitschr Wiss Steward O, Cotman C, Lynch GS (1974) Growth of a new ber pro-
Zool 52. jection in the brain of adult rats: reinnervation of the dentate
Schaffer K (1892) Beitrag zur Histologie der Ammonshorn-forma- gyrus by the contralateral entorhinal cortex following ipsilateral
tion. Arch Mikr Anat 39:611632. entorhinal lesions. Exp Brain Res 20:4566.
36 The Hippocampus Book

Storm-Mathisen J, Fonnum F (1972) Localization of transmitter can- Von Gudden JBA (1896) Klinische und anatomische Beitrge zur
didates in the hippocampal region. Prog Brain Res 36:4158. Kenntniss der multiplen Alcoholneuritis nebst Bemerkungen
Storm-Mathisen J, Leknes AK, Bore AT, Vaaland JL, Edminson ber die Regenerationsvorgnge in peripheren Nervensystem.
P, Haug FM, Ottersen OP (1983) First visualization of glutamate Arch Psychiatr Nervenkr 28:643741.
and GABA in neurones by immunocytochemistry. Nature 301: Weiskrantz L (1956) Behavioral changes associated with ablation of
517520. the amygdaloid complex in monkeys. J Comp Physiol Psychol 49:
Stuart G, Spruston N, Sakmann B, Husser M (1997) Action poten- 381391.
tial initiation and backpropagation in neurons of the mam- Wernicke C (1881) Lehrbuch der Gehirnkrankheiten, vol 2. Berlin:
malian CNS. Trends Neurosci 20:125131. Theodore Fischer.
Swanson LW, Cowan, WM (1977) An autoradiographic study of the Wilson MA, McNaughton BL (1993) Dynamics of the hippocampal
organization of the efferent connections of the hippocampal ensemble code for space. Science 261:10551058.
formation in the rat. J Comp Neurol 172:4984. Winograd T (1975) Frame representations and the declarative/
Tulving E (1972) Episodic and semantic memory. In: Organization procedural controversy. In: Representation and understanding
and memory (Tulving E, Donaldson W, eds), pp 382403. San (Bobrow DG, Collins A, eds), pp 185210. San Diego: Academic
Diego: Academic Press. Press.
Vanderwolf CH (1969) Hippocampal electrical activity and volun- Yamamoto C, McIlwain H (1966a) Potentials evoked in vitro in
tary movement in the rat. Electroencephalogr Clin Neurophysiol preparations from the mammalian brain. Nature 210:
26:407418. 10551056.
Vinogradova O, Semyonova TP, Konovalov VP (1970) Trace phe- Yamamoto C, McIlwain H (1966b) Electrical activities in thin sec-
nomena in single neurons of hippocampus and mammillary tions from the mammalian brain maintained in chemically
bodies. In: Biology of memory (Pribram KH, Broadbent DE, dened media in vitro. J Neurochem 13:13331343.
eds), pp 191222. San Diego: Academic Press. Zimmer J, Ghwiler BH (1984) Cellular and connective organization
Von Euler C, Green JD, Ricci G (1958) The role of hippocampal den- of slice cultures of the rat hippocampus and fascia dentata. J
drites in evoked responses and after-discharges. Acta Physiol Comp Neurol 228:432446.
Scand 42:87111.
3 David Amaral and Pierre Lavenex

Hippocampal Neuroanatomy

century of neuroanatomical study and literally tens of thou-

3.1 Overview sands of research articles on the hippocampus, there is still no
consensus concerning certain facets of its nomenclature.
Where is the hippocampal formation and what does it look Hippocampal researchers tend to follow one of several
like? What types of neurons are located in this group of struc- implicit views about what the hippocampus is and what it is
tures, and what types of connections do they form? What is not. The view adopted in this book is that the hippocampus is
the difference between the hippocampus and the hippocam- one of several related brain regions that together comprise a
pal formation? Does the hippocampus look similar in the rat, functional system called the hippocampal formation.
monkey, and human brains? Are other components of the The hippocampus proper has three subdivisions: CA3,
hippocampal formation similar across species? Are the princi- CA2, and CA1. (CA comes from cornu ammonis; the deriva-
ples of neuroanatomical connectivity similar or different from tion of these terms is discussed shortly.) The other regions of
those found in other brain regions? These are only some of the the hippocampal formation include the dentate gyrus, subicu-
neuroanatomical issues that are explored in this chapter. The lum, presubiculum, parasubiculum, and entorhinal cortex.
elegant neuronal architecture of the hippocampal formation Thus, although the title for this book is The Hippocampus
and the simplicity and orderliness of its major connections Book, a more correct, albeit less melodious, title would have
have been seductive features to neuroscientists for decades. been The Hippocampal Formation Book. The rationale for
Yet, there are pragmatic reasons why a working knowledge of grouping these cytoarchitectonically distinct brain regions
its neuroanatomy is important. under the rubric hippocampal formation is developed
First, it is likely that certain peculiarities of the neural throughout this chapter.
organization of the hippocampus, such as its highly associa-
tional intrinsic connections, provide important clues to its 3.1.2 Similarities and Differences
function(s). Second, the design of functional, electrophysio- Between the Hippocampal Formation
logical, pharmacological, and behavioral studies require and other Cortical Areas
increasingly sophisticated knowledge of its boundaries, cell
types, connections, and chemical neuroanatomy. This chapter In some ways, such as the occurrence of large, pyramid-
lays the structural foundations for discussions of the func- shaped projection neurons and smaller interneurons, the neu-
tional organization of the hippocampal formation explored ral organization of some portions of the hippocampal
later in the book. formation resembles other cortical regions. Yet in important
ways, such as the largely unidirectional passage of information
3.1.1 Hippocampus: Part of a Functional Brain through intrahippocampal circuits and the highly distributed
System Called the Hippocampal Formation three-dimensional organization of intrinsic associational con-
nections, the neuroanatomy of the hippocampal formation is
One of the most captivating features of the hippocampus is its unique. It has often been touted as a heuristically simple
neuroanatomy. The relatively simple organization of its prin- model of neocortical organization, but this perspective
cipal cell layers coupled with the highly organized laminar dis- demeans both the hippocampal formation and the neocortex.
tribution of many of its inputs has encouraged its use as a It is more protable to highlight and investigate the distinctive
model system for modern neurobiology. Despite more than a neuroanatomical features of the hippocampal formation as

37
38 The Hippocampus Book

potential clues to its particular function(s) and the mecha- back to region A. As rst described by Ramn y Cajal (1893),
nisms by which these functions are realized. this is clearly not the case for the connections that link the var-
Our understanding of hippocampal neuroanatomy leads ious parts of the hippocampal formation (Fig. 31). The
to the prediction that whatever processes this group of struc- entorhinal cortex can, for convenience, be considered the rst
tures carries out they are likely to be quite different from those step in the intrinsic hippocampal circuit. The logic behind this
performed in other cortical regions. The hippocampal forma- is developed later in the chapter, but the priority afforded to
tion, for example, is one of only a few brain regions that the entorhinal cortex is based on the fact that much of the
receive highly processed, multimodal sensory information neocortical input reaching the hippocampal formation does
from a variety of neocortical sources. Moreover, its own sys- so through the entorhinal cortex. Cells in the supercial layers
tem of widely distributed intrinsic neuronal networks is ide- of the entorhinal cortex give rise to axons that project, among
ally suited for further mixing or comparing this information. other destinations, to the dentate gyrus. The projections from
This ability to integrate information from all sensory modali- the entorhinal cortex to the dentate gyrus form part of the
ties may thus be a unique attribute of the hippocampus con- major hippocampal input pathway called the perforant path.
ferred by the highly convergent-divergent organization of its Although the entorhinal cortex provides the major input to
connections. the dentate gyrus, the dentate gyrus does not project back to
the entorhinal cortex. This pathway is therefore nonrecipro-
3.1.3 Hippocampal Formation: With A Unique cated, or unidirectional.
Set of Unidirectional, Excitatory Pathways Likewise, the principal cells of the dentate gyrus, the gran-
ule cells, give rise to axons called mossy bers that connect
A common organizational feature of connections between with pyramidal cells of the CA3 eld of the hippocampus. The
regions of the neocortex is that they are largely reciprocal CA3 cells, however, do not project back to the granule cells.
(Felleman and Van Essen, 1991). If cortical region A projects The pyramidal cells of CA3, in turn, are the source of the
to cortical region B, region B often sends a return projection major input to the CA1 hippocampal eld (the Schaffer col-

Figure 31. The hippocampal formation.

A. Neurons in layer II of the entorhinal
cortex project to the dentate gyrus and the
CA3 eld of the hippocampus proper via
the perforant pathway. Neurons in layer III
of the entorhinal cortex project to the CA1
eld of the hippocampus and the subicu-
lum via the perforant and alvear pathways
(see text for a detailed description). The
granule cells of the dentate gyrus project
to the CA3 eld of the hippocampus via
mossy ber projections. Pyramidal neu-
rons in the CA3 eld of the hippocampus
project to CA1 via Schaffer collaterals.
Pyramidal cells in CA1 project to the
subiculum. Both CA1 and the subiculum
project back to the deep layers of the
entorhinal cortex. B. Projections along the
transverse axis of the hippocampal forma-
tion; the dentate gyrus is located proxi-
mally and the entorhinal cortex distally.
 Hippocampal Neuroanatomy 39

lateral axons). Following the pattern of its predecessors, CA1 (Amaral et al., 1984). Another example is the more complex
does not project back to CA3. The CA1 eld of the hip- organization of the primate entorhinal cortex, which appears
pocampus then projects unidirectionally to the subiculum, to be associated with stronger interconnections with the asso-
providing its major excitatory input. Again, the subiculum ciational areas of the neocortex, which are more developed in
does not project back to CA1. primates. Myriad other subtle species differences (e.g., the
Once one reaches CA1 and the subiculum, the pattern of chemical neuroanatomy of the hippocampal formation) have
intrinsic connections begins to become somewhat more elab- been noted in the literature. Certain neuroanatomical differ-
orate. CA1, for example, projects not only to the subiculum ences have even been described in different strains of mice
but also to the entorhinal cortex. Furthermore, whereas the that seem too subtle to be worthy of note yet could be of enor-
subiculum does project to the presubiculum and the para- mous practical importance given the current interest in the
subiculum, its more prominent cortical projection is directed use of transgenic techniques. A major future challenge is to
to the entorhinal cortex. Through these connections both CA1 determine whether and how these structural and neurochem-
and the subiculum close the hippocampal processing loop ical alterations affect the functioning of the hippocampal for-
that begins in the supercial layers of the entorhinal cortex mation.
and ends in its deep layers. Although this cursory survey of the Thus, to address the question posed in the title of this sec-
intrinsic connections of the hippocampal formation leaves tionis the hippocampus of humans and animals same or
out many of the facts that make the system somewhat more differentthe answer is both.
complex, it does serve to emphasize that the hippocampal for-
mation is organized in a fashion that is distinctly different 3.1.5 Synopsis of the Chapter
from most other cortical areas.
Our rationale for focusing on the rodent (mainly the rat) hip-
3.1.4 Hippocampus of Humans pocampal formation in the rst half of this chapter is that
and Animals: Same or Different? much of the available neuroanatomical data have been derived
from studies carried out in this animal and because of the
Once one has gained a familiarity with the neuroanatomical prominent role the rat plays in current functional analyses of
appearance of the rat hippocampal formation, it is not diffi- the hippocampal formation. Unfortunately, far less work has
cult to identify each of the major subdivisions in the monkey been carried out in the mouse, the clear choice for molecular
or human hippocampal formation (Fig. 32). Although the biology studies. It is widely presumed that the neuroanatomy
volume of the hippocampus is about 10 times larger in mon- of the mouse hippocampal formation is similar to that of the
keys than in rats and 100 times larger in humans than in rats, ratthough some additional conrmation would certainly be
the basic hippocampal architecture is common to all three welcome. We next compare and contrast the picture of hip-
species. Yet there are some striking species differences. The pocampal anatomy obtained from studies of the rat with that
compact pyramidal cell layer in the CA1 region of the rat, for of the monkey. We also describe the extensive cortical connec-
example, becomes thicker and more heterogeneous in the tions of the entorhinal cortex in the macaque monkey and the
monkey and human. Whereas this layer is only about 5 cells emerging prominence of adjacent related brain regions, such
thick in the rat, it can be more than 30 cells thick in the as the perirhinal and parahippocampal cortices. We then
human. The other region that demonstrates striking species move on to compare information on the monkey hippocam-
differences is the entorhinal cortex. In the rat, the entorhinal pal formation with what is known about the human hip-
cortex is typically divided into two main, cytoarchitectonically pocampal formation. The chapter concludes with a summary
distinct subdivisions. In the monkey there are seven subdivi- of the principles of hippocampal intrinsic circuitry and a dis-
sions, and in the human brain the classical cytoarchitectoni- cussion of how they may govern the ow of information
cists dened as many as 27 subdivisions (although recent through the hippocampal formation.
descriptions recognize only 8 subdivisions, similar to what is
observed in monkeys). Thus, despite the fact that the hip-
pocampal formation is often portrayed as a phylogenetically
primitive brain region, it nonetheless demonstrates substan- 3.2 Historical Overview of Hippocampal
tial species differences. NomenclatureWhats in a Name?
Although the patterns of connectivity appear to be gener-
ally similar in the rodent and primate brains, there are again The hippocampal formation has been prodded and sliced and
striking species differences. One example is the organization stained by anatomists for nearly 400 years. Like many brain
of the commissural connections of the dentate gyrus. In the regions, it has fallen victim to the imposition of various and
rat, there is a massive commissural system that provides nearly often confusing terminologies to describe its gross anatomical
one-sixth of the excitatory input to the dentate gyrus and histological structure. The term hippocampus (derived
(Raisman, 1965; Gottlieb and Cowan, 1973). In the macaque from the Greek word for sea horse) was rst coined during the
monkey and presumably in humans, however, commissural sixteenth century by the anatomist Arantius (1587), who
connections in the dentate gyrus are almost entirely absent considered the three-dimensional form of the human hip-
40 The Hippocampus Book

Figure 32. Nissl-stained sections and line drawings illustrating the general organization and
similarities of the subdivisions of the hippocampal formation in the rat, monkey, and human.
Note the differences in the relative position of the mbria in the rat (located lateroventrally)
and in the monkey and human (located mediodorsally). Bar in A  1 mm and applies to all
panels.

pocampus, lying in the oor of the inferior horn of the lateral both the names sea horse and silk worm to this region. Others
ventricle, to be reminiscent of this sea creature (Fig. 33). likened the arched structure of the hippocampus to a rams
Arantius also likened the shape of the hippocampus to a silk horn, and De Garengeot (1742) named the hippocampus
worm, but his term bombycini or bombyx never caught on. cornu ammonis or Ammons horn after the mythological
Because Arantius did not take the trouble to stake claim to the Egyptian god Amun Kneph, whose symbol was a ram.
hippocampus by illustrating its location and structure, other According to the Nomina Anatomica, the term hippocampus
anatomists felt free to devise their own terms. The main, or is currently acknowledged to be the standard term for the
intraventricular, portion of the human hippocampus was rst bulge occupying the oor of the lateral ventricle in the human
illustrated by Duvernoy in 1729, and he continued to assign brain. The term Ammons horn is now only rarely used,
 Hippocampal Neuroanatomy 41

Figure 33. Dorsolateral view of the human

hippocampus after removing the overlying
structures. The lateral ventricle has been
opened, exposing the shape of the hippocam-
pus. Bar  1 cm.

although a curious irony of current terminology is that, they could recruit or effect appropriate emotional responses
although hippocampus has become the standard term, subdi- (Papez, 1937). Despite the admirable job done by Papez in
visions of this structure are referred to using abbreviations of marshaling largely circumstantial evidence in support of his
cornu ammonis (CA3, CA2, CA1). Some of the terms men- view, there has been little modern substantiation of Papezs
tioned above, as well as many others, have found their way theory. The role of orchestrator of emotional expression is
into descriptions of the histological structure of the hip- now more closely linked with another prominent medial tem-
pocampal formation; hence it is helpful, at the start, to discuss poral lobe structure, the amygdaloid complex.
neuroanatomical terms currently associated with the hip- The hippocampus is also associated with the term limbic
pocampus. system. The origin of this term stems from the description by
A few terms are now of only historical interest. The hip- the neurologist Broca (1878) of le grande lobe limbique,
pocampus was for some time considered to be a prominent which comprised a series of contiguous cortical and subcorti-
part of the rhinencephalon, which, loosely speaking, referred cal structures located on the medial surface of the brain and
to a group of forebrain structures thought to be related espe- surrounding the ventricle at the border (limbus) of the corti-
cially to the sense of smell. In a classic article published in cal mantle. The limbic lobe was described initially as includ-
1947, however, Alf Brodal concluded that the hippocampus ing such circumventricular structures as the subcallosal,
could not be functioning solely as an olfactory structure cingulate, and parahippocampal gyri, as well as the underly-
because anosmic mammals such as dolphins had a substantial ing hippocampal formation. The number of structures
hippocampal formation (Brodal, 1947). It seemed unreason- included under the rubric limbic system has escalated dra-
able to Brodal that the hippocampus would be so robust in a matically since the 1950s, however, and has included such
nonsmelling creature if it functioned solely as a component of diverse structures as the amygdala, septal nuclei, and midbrain
the olfactory system (see Chapter 2). periaqueductal gray matter. The proliferation of brain regions
During the 1930s, the neurologist James Papez considered encompassed by the term limbic system has fueled a long-
the hippocampus to be a central component of a system for lasting debate about the utility of this term either as a neu-
emotional expression, and in many textbooks the hippocam- roanatomical or functional entity. Suffice it to say that in this
pus is still claimed to be the central component of the so- book the hippocampal formation is viewed as an independent
called Papez circuit. In Papezs view, the hippocampus was a functional system rather than part of the larger, ill-dened
conduit by which perceptions of emotionally salient situations group of structures collectively referred to as the limbic sys-
could be collected and channeled to the hypothalamus where tem. That said, the hippocampal formation obviously does
42 The Hippocampus Book

not function in isolation. Even as a functionally dened, inde- the main output projections and that these projections were
pendent system, it relies on interconnections with many other directed subcortically. With the discovery of robust projec-
brain systems to do its job effectively. tions from CA1 to the subiculum and entorhinal cortex, and
the major projections from the entorhinal cortex to the neo-
3.2.1 Denition of Hippocampal Areas: cortex, the trisynaptic circuit is now considered to be only a
Denition of Terms portion of the functional circuitry of the hippocampal forma-
tion. As it is now clear that the subiculum is the main source
When a histological section through the hippocampus is pre- of subcortical projections and the entorhinal cortex is the
pared and stained for cell bodies by a Nissl method, it is main source of projections to the neocortex, the concept of
immediately obvious that a number of cytoarchitectonically the trisynaptic circuit, although of great signicance in the
distinct structures are encompassed in the region grossly history of hippocampal research, is currently less inuential
dened as the hippocampus. A number of terminologies on theories of hippocampal function.
have been applied to some of these regions, and several syn- The reader who ventures from the relative safety of this
onymous terms are still commonly employed. Beyond the book into the primary hippocampal literature should be
problem of different terms being applied to the same region, aware that our usage of the term hippocampal formation is
the borders between several regions of the hippocampal for- widely, though not universally, accepted. Some authors
mation have not yet been rmly established. The terms include only the allocortical (a term applied to cortical regions
adopted in this book are based, in part, on converging evi- having fewer than six layers) regions as parts of the hip-
dence from cytoarchitectonic, histochemical, connectional, pocampal formation. Three-layered cortical regions typically
and functional data and in part on personal preference. have a single neuronal cell layer with ber-rich plexiform
The term hippocampus has historically been plagued by layers above and below the cell layer. In articles employing this
the fact that it is used as a term both for a gross anatomical usage, the hippocampal formation comprises the dentate
region of the brain (the bulge or protuberance in the oor of gyrus, hippocampus, and subiculum. The remaining elds
the human lateral ventricle) and for one of the cytoarchitec- presubiculum, parasubiculum, and entorhinal cortexare
tonically distinct entities that make up the region. As a result, then typically grouped together on the basis of their multil-
the meaning of the word hippocampus is context-dependent aminate structure under the term retrohippocampal (retro 
and may be ambiguous. We reserve the term hippocampus for behind) or parahippocampal (para  alongside or near) cor-
the region of the hippocampal formation that comprises the tex. In yet other variants, the terms hippocampus or hip-
CA elds (CA3, CA2, CA1) identied by the neuroanatomist pocampal complex are sometimes applied to the combination
Rafael Lorente de N (Lorente de N, 1934). The term hip- of the dentate gyrus and hippocampus proper. Clarication of
pocampal formation, in contrast, is applied to a group of the nomenclature is typically the rst sign of maturity for a
cytoarchitectonically distinct adjoining regions including the scientic enterprise, and we scrupulously adhere to the
dentate gyrus, hippocampus, subiculum, presubiculum, para- nomenclature in which the hippocampal formation com-
subiculum, and entorhinal cortex (Fig. 31). The adjective prises the six structures listed above.
hippocampal, by necessity, remains somewhat vague and Having outlined the major regions of the hippocampal for-
context-dependent. In general, it is used to refer to the larger mation, we now delve more deeply into the subdivisions of
area (as in hippocampal lesions) rather than to the cytoar- each of these regions. Before doing so, however, it is important
chitectonic region. If this seems confusing, the termi- to note that the terminology we apply to these regions is a
nological denitions of the next few paragraphs may provide hybrid derived from the analyses of classical and modern hip-
clarication. pocampal neuroanatomists. The two main contributors to the
The main justication for including the six regions named surviving hippocampal terminologies are Santiago Ramon y
above under the rubric hippocampal formation is that they Cajal (Ramn y Cajal, 1893) and his student Raphael Lorente
are linked, one to the next, by unique and largely unidirec- de N (Lorente de N, 1933, 1934). Some of their subdivi-
tional (functional) neuronal pathways. The early literature on sions, based solely on the analysis of Golgi-stained material,
the hippocampal formation emphasized the rst three links have not stood the test of time or the introduction of modern
in the hippocampal circuitry that were highlighted by apply- neuroanatomical methods. Many revisions of their nomencla-
ing the term trisynaptic circuit to the ensemble of pathways ture have been made as information has become available
(Andersen et al., 1971). concerning the connections and chemical architecture of the
hippocampal formation.
EC DG (synapse 1), DG CA3 (synapse 2),
CA3 CA1 (synapse 3)
3.2.2 Subdivisions of Hippocampal Areas
The denition of this powerful excitatory circuit was pro-
duced by a collaboration between early neuroanatomical and The dentate gyrus is a trilaminate cortical region with a char-
electrophysiological studies. It should be borne in mind, how- acteristic V or U shape. The dentate gyrus has a relatively sim-
ever, that the term trisynaptic circuit was coined during an era ilar structure at all levels of the hippocampal formation and is
when it was assumed that the hippocampus proper generated not typically divided into subregions. When discussing fea-
 Hippocampal Neuroanatomy 43

tures of the dentate gyrus, however, it is often useful to refer to are somewhat larger and less compact than those in the pre-
a particular portion of the V- or U-shaped structure. The por- subiculum.
tion of the granule cell layer that is located between the CA3 The entorhinal cortex is the only hippocampal region that
eld and the CA1 eld (separated by the hippocampal ssure) unambiguously demonstrates a multilaminate appearance.
is called the suprapyramidal blade; and the portion opposite Based on differences in the organization of these layers and
this is the infrapyramidal blade. The region bridging the two more recently on differences in connectional attributes, the
blades (at the apex of the V or U) is the crest. entorhinal cortex has been divided into two or more subre-
The hippocampus, especially in rodents, can easily be gions depending on the species. We shall return to a descrip-
divided into two major regions: a large-celled region that tion of the subdivisions of the entorhinal cortex later in the
abuts the dentate gyrus and a smaller-celled region that fol- chapter.
lows from it. Ramon y Cajal called these two regions regio Before moving on to descriptions of the three-dimensional
inferior and regio superior, respectively. The terminology of organization of the hippocampal formation in the rat, mon-
Lorente de N has achieved more common usage and is key, and human brains, we must say a few more words con-
employed here. He divided the hippocampus into three elds: cerning nomenclature. We often want to refer to a specic
CA3, CA2, and CA1. His CA3 and CA2 elds are equivalent to portion of one of the hippocampal regions. Given the com-
the large-celled regio inferior of Ramon y Cajal, and his CA1 plex shape of the hippocampal formation, no reference system
eld is equivalent to the regio superior. In addition to the is wholly adequate, and any description inevitably involves
greater size of the pyramidal cells in CA3 and CA2 compared arbitrary decisions about where to start or nish, which direc-
to CA1, the inputs and outputs of these areas are also differ- tion is up or down, and so on. We have adopted a reference
ent. The pyramidal cells of CA3, for example, receive the system in which the dentate gyrus is considered to be the
mossy ber input from the dentate gyrus, whereas the CA1 proximal pole of the hippocampal formation and the entorhi-
pyramidal cells do not. nal cortex is the distal pole (Fig. 31). A portion of any hip-
The CA2 eld has been the subject of substantial contro- pocampal eld can therefore be dened in relation to this
versy. As originally dened by Lorente de N, it is a narrow proximo-distal axis. For example, the proximal portion of
zone of cells interposed between CA3 and CA1. CA2 has large CA3 is located closer to the dentate gyrus, and the distal por-
pyramidal cell bodies similar to those in CA3 but, like CA1, it tion is located closer to CA2.
is not innervated by the mossy bers from the dentate gyrus. We also often need to specify subregions within the thick-
Although the existence of CA2 has often been questioned, the ness of a particular hippocampal region. As in most other cor-
bulk of available evidence indicates that there is indeed a tical areas, this radial dimension is usually described along a
narrow CA2 eld that can be distinguished from the other supercial-to-deep axis. In six-layer structures, layer I is close
hippocampal elds using a variety of criteria, including neu- to the pial surface, and layer VI is located close to the subcor-
rochemical markers. Lorente de N also dened a CA4 eld. tical white matter. Consistent with this convention, regions
As originally claried by Theodor Blackstad (1956) and then closer to the pia or hippocampal ssure are considered super-
by David Amaral (1978), the region that Lorente de N called cial and those in the opposite directioncloser to the alveus
CA4 is actually the deep, or polymorphic, layer of the dentate or ventricle where applicable) are considered deep. The
gyrus. molecular layer of the dentate gyrus, for example, is super-
The subiculum, presubiculum, and parasubiculum are cial to the granule cell layer. This supercialdeep nomencla-
sometimes grouped under the term subicular complex. ture has the merit of being applicable to all portions of the
Because each of these regions has distinct neuroanatomical hippocampal formation and to all of its cytoarchitectonic
features, they are better thought of as independent cortical elds. It has the drawback, however, of being somewhat coun-
areas. The border between CA1 and the subiculum occurs pre- terintuitive to neuroscientists, especially electrophysiologists,
cisely at the point where the Schaffer collateral projection whose electrodes approach the hippocampus from the alveus
from the CA3 eld ends. In the rodent, this occurs approxi- in the septal (or dorsal) portion of the hippocampus. As an
mately where the condensed pyramidal cell layer of CA1 electrophysiologist advances an electrode from the dorsal sur-
begins to broaden into the thicker layer of the subiculum. face of the brain toward the hippocampus, it would rst hit
The presubiculum lies adjacent to the subiculum and is the alveus and then the pyramidal cell layer. Although the
typically thought to have more than the three layers that char- alveus might seem like the supercial portion of the hip-
acterize the dentate gyrus, hippocampus, and subiculum. pocampus because it is closer to the surface of the brain, it
However, the exact delimitation of the deep layers of the pre- is actually its deep portion. As the electrode continues its
subiculum and the differentiation of cells belonging to the advance toward the hippocampal ssure, it enters the super-
presubiculum from those that belong to the deep layers of the cial portion of the hippocampus. Crossing the hippocampal
entorhinal cortex has never been clearly established. The most ssure, it then enters the supercial portion of the molecular
distinctive feature of the presubiculum is the densely packed layer of the dentate gyrus, and a further advance would bring
external cellular layer, which is populated by relatively small, it to the deep portion of the molecular layer, the suprapyra-
tightly packed pyramidal cells. The parasubiculum is charac- midal granule cell layer, the hilus, and nally the infrapyrami-
terized by a wedge-shaped layer II with cells that resemble but dal granule cell and molecular layers.
44 The Hippocampus Book

3.2.3 Major Fiber Bundles of many of the elds of the rodent hippocampal formation are
the Hippocampal Formation grossly C-shaped and vertically oriented, they tend to be much
more linear and horizontally oriented in primates.
Three major ber systems are associated with the hippocam-
pal formation (Fig. 34, see color insert). The rst is the angu- 3.3.1 Rat Hippocampal Formation
lar bundle, which carries bers between the entorhinal cortex
and the other elds of the hippocampal formation. The The rat hippocampal formation is an elongated, banana-
second is the mbria-fornix pathway through which the hip- shaped structure with its long axis extending in a C-shaped
pocampal formation is interconnected with the basal fore- manner from the midline of the brain near the septal nuclei
brain, hypothalamic, and brain stem regions. The third (rostrodorsally) over and behind the thalamus into the incipi-
comprises the dorsal and ventral commissures, through which ent temporal lobe (caudoventrally). The long axis of the hip-
the hippocampal formation of one hemisphere is connected pocampal formation is referred to as the septotemporal axis
with the hippocampal formation of the contralateral hemi- and the orthogonal axis as the transverse axis (Fig. 38). What
sphere. We deal with these in greater detail in Section 3.3.2. is not obvious from a surface view of the hippocampal forma-
tion is that different regions make up the structure at different
septotemporal levels. At extreme septal levels, for example,
only the dentate gyrus and the CA3CA1 elds of the hip-
3.3 Three-dimensional Organization pocampus are present. About a third of the way along the sep-
and Major Fiber Systems of the totemporal axis the subiculum first appears, and the
Hippocampal Formation presubiculum and parasubiculum are seen at progressively
more temporal levels. The entorhinal cortex is located even
The hippocampal formation is positioned quite differently in farther caudally and ventrally. The dorsolateral limit of the
the rodent and primate brains (Figs. 35 through 37, see entorhinal cortex occurs approximately at the rhinal sulcus,
color insert). This is due in part to the more developed cere- which forms a prominent, rostrocaudally/horizontally ori-
bral cortex in primates, which tends to force the dentate ented indentation on the ventrolateral surface of the rat brain.
gyrus and hippocampus into the temporal lobe. Whereas This sulcus nominally separates the entorhinal cortex ventrally

Figure 34. Major ber systems of the

rat hippocampal formation: angular bun-
dle; mbria-fornix; dorsal and central
commissures.
 Hippocampal Neuroanatomy 45

Figure 35. Magnetic resonance imaging

of the rat brain shows the position of the
hippocampus (dentate gyrus  hip-
pocampus proper  subiculum) in red
and the entorhinal cortex in green. A.
Oblique view. B. Dorsal view. C. Frontal
view. D. Lateral view. (Source: Courtesy of
Dr. G. Allan Johnson, Center for In Vivo
Microscopy, Duke University.)

Figure 36. Magnetic resonance imaging

of the monkey brain shows the position
of the hippocampus (dentate gyrus 
hippocampus proper  subiculum) in
red and the entorhinal cortex in green. A.
Oblique view. B. Dorsal view. C. Frontal
view. D. Lateral view.
46 The Hippocampus Book

Figure 37. Magnetic resonance imaging

of the human brain shows the position of
the hippocampus (dentate gyrus  hip-
pocampus proper  subiculum) in red
and the entorhinal cortex in green. A.
Oblique view. B. Dorsal view. C. Frontal
view. D. Lateral view.

Figure 38. Line drawing of the rat brain

shows the septotemporal and transverse
axes of the hippocampal formation.
 Hippocampal Neuroanatomy 47

from the perirhinal and postrhinal cortices dorsally. At rostral mbria-fornix to emphasize the continuity of bers in these
levels, however, the perirhinal cortex extends somewhat ven- bundles.
tral to the rhinal sulcus, and at caudal levels the entorhinal The ventricular, or deep, surface of the hippocampus is
cortex extends just slightly dorsal to the rhinal sulcus. covered by a thin sheet of myelinated bers called the alveus.
These bers form a white sheet overlying the hippocampus
3.3.2 Major Fiber Systems of the Rat that can be clearly seen when the overlying neocortex is
Hippocampal Formation aspirated. The alveus is a complex ber system with both
extrinsic afferent and efferent bers, and bers forming part
Angular Bundle of the intrahippocampal network travel within it (i.e., the
entorhinalCA1 alvear pathway, the CA1subiculum projec-
A major ber pathway associated with the hippocampal for- tion, and the CA1entorhinal projection). Some alvear bers
mation is the angular bundle, a ber bundle interposed originate from the pyramidal cells of the hippocampus and
between the entorhinal cortex and the presubiculum and subiculum and are en route to subcortical termination sites
parasubiculum. The angular bundle is the main route taken by (Meibach and Siegel, 1975). At temporal levels of the hip-
bers originating from the ventrally situated entorhinal cortex pocampal formation, the subcortically directed output bers
as they travel to all septotemporal levels of the other hip- extend obliquely in the alveus, from medial to lateral, over the
pocampal elds, particularly the dentate gyrus, hippocampus, surface of the hippocampus and collect in a bundle called the
and subiculum. In addition, the angular bundle contains com- mbria (from the Latin word for fringe), which becomes pro-
missural bers of entorhinal and presubicular origin and gressively thicker as it progresses from the temporal to the
bers to and from a variety of cortical and subcortical struc- septal level (i.e., as axons from more septally located pyrami-
tures that are interconnected with the entorhinal cortex. dal cells are added to the bundle). The rat mbria has a at-
The perforant path comprises the efferent entorhinal pro- tened appearance, contains approximately 900,000 axons, and
jections that traverse, or perforate, the subiculum on their way is situated along the lateral and rostral aspects of the hip-
to the dentate gyrus and the hippocampus. This is the main pocampus. The bers of the mbria are not randomly distrib-
route by which neocortical inputs reach the dentate gyrus and uted but are organized in a topographic fashion (Wyss et al.,
the hippocampus. We present a more detailed description of 1980). Axons located medially in the mbria (i.e., those clos-
the perforant path in the discussion of the inputs to the den- est to the hippocampus) tend to arise from more septal levels,
tate gyrus and the connectivity of the entorhinal cortex. whereas those located laterally arise from more temporal lev-
Entorhinal bers also reach the hippocampus via the els. Fibers from the subiculum are situated deeper to those
alveus, the temporoammonic alvear pathway rst described from the hippocampus.
by Cajal. At temporal levels, most of the entorhinal bers The fornix is the continuation of this bundle of hippocam-
reach the CA1 eld of the hippocampus after perforating the pal output bers to the subcortical target structures; it forms
subiculum (via the classic perforant pathway). At more septal a attened bundle located just below the corpus callosum very
levels, the number of entorhinal bers that take the alvear close to the midline (fornix is the Latin word for arch, which
pathway increases (Deller et al., 1996). In fact, in the septal is the shape of this tract over the diencephalon). Both mbria
portion of the hippocampal formation, most of the entorhinal and fornix carry bers from the hippocampus and subiculum;
bers to CA1 reach this subeld via the alveus. These bers the fornix, however, carries bers primarily from the septal
make sharp right-angle turns in the alveus, perforate the third of these structures. As the bers of the mbria leave
pyramidal cell layer, and nally terminate in the stratum the hippocampus and descend into the forebrain, they are
lacunosum-moleculare. The alveus is therefore also a major referred to as the columns of the fornix. The fornix splits
route by which entorhinal bers reach their targets in CA1. around the anterior commissure to form a rostrally directed
precommissural fornix, which innervates the septal nuclei and
Fimbria-Fornix Pathway nucleus accumbens, and a caudally directed postcommissural
fornix, which extends toward the diencephalon. As the post-
The mbria-fornix ber system provides the major conduit commissural fornix begins its course into the diencephalon
for subcortical afferent and efferent connections (Daitz and (ultimately reaching the mammillary nuclei of the posterior
Powell, 1954; Powell et al., 1957). It is perhaps easiest to hypothalamus), two smaller bundles split off. One, the medial
understand the mbria-fornix system by analogy with the corticohypothalamic tract, innervates a number of anterior
corticospinal ber system. The corticospinal bers are given hypothalamic areas. The other, called the subiculothalamic
different names at different points on their journey from the tract, carries bers to the anterior thalamic nuclei (Swanson
motor cortex to the spinal cord. Similarly, the subcortical and Cowan 1975; Canteras and Swanson, 1992).
afferent and efferent bers of the hippocampal formation are The mbria and fornix also carry bers that are traveling
given different names at different points in their trajectory to the hippocampal formation. Many of the subcortical inputs
from or toward the forebrain or brain stem. Because of this, to the hippocampal formation, including those from the sep-
and because the exact transition between mbria and fornix is tal nuclei (to the septal portion of the hippocampal forma-
difficult to dene, some hippocampal researchers use the term tion), the locus coeruleus, and the raphe nuclei enter via the
48 The Hippocampus Book

mbria-fornix pathway. Some subcortical structures have pro- is located caudally. Because the term septotemporal is not
jections that follow other pathways into the hippocampal for- appropriate for the monkey hippocampal formation (no por-
mation. Fibers from the anterior thalamus, for example, travel tion of it approaches the septal area), it is more common to
through the thalamic radiations and supracallosal stria to refer to the long axis in the primate as the rostrocaudal axis.
innervate the presubiculum. Still other subcortical projec- The orthogonal axis, however, is still referred to as the trans-
tions, particularly those from the amygdala, travel to the hip- verse axis.
pocampal formation via the external capsule. Two additional points should be made about the position
of the monkey hippocampal formation. First, at the rostral
Dorsal and Ventral Hippocampal Commissures limit of the lateral ventricle, some elds of the monkey hip-
pocampal formation ex medially and then caudally. This is
A third major ber system associated with the hippocampal the monkey homologue of the pes hippocampi that is so
formation is the commissural system (Blackstad, 1956; prominent in the human brain. At the rostral levels where this
Raisman et al., 1965; Laatsch and Cowan, 1967; Laurberg, exure occurs, there are two representations of the hippocam-
1979). In the rat, there are both dorsal and ventral commis- pal formation in standard coronal views of the brain. It is
sures. Some 350,000 bers cross the midline in the ventral difficult in this exed region of the monkey hippocampal for-
hippocampal commissure, which is located just caudal to the mation, when viewed in standard coronal, Nissl-stained sec-
septal area and dorsocaudal to the anterior commissure. Many tions, to specify the identity and borders of the subdivisions
of these bers are true commissural bers and are directed to of the hippocampal formation. The most medial and caudal
both homotopic and heterotopic elds in the contralateral portion of the hippocampal formation (i.e., the part that is
hippocampal formation. A much smaller number of bers are bent backward) is the actual rostral pole of the monkey hip-
directed into the contralateral descending column of the pocampal formation, even though it is physically located
fornix and ultimately innervate the same structures on the somewhat caudal to the rostral extreme of the hippocampal
contralateral side of the brain that receive the ipsilateral formation. To make reference to different portions of the
pre- and postcommissural fornix. The dorsal hippocampal monkey hippocampal formation, we call the medial portion of
commissure crosses the midline just rostral to the splenium the hippocampal formation the uncal region (because it forms
(posterior part) of the corpus callosum and carries bers much of the medially situated bulge that in the human would
mainly originating from or projecting to the presubiculum be called the uncus). The exed, rostrally located portion that
and entorhinal cortex. The dorsal hippocampal commissure is runs mediolaterally is called the genu; and the laterally situ-
the route by which the presubiculum contributes a major pro- ated, main portion of the hippocampus is the body.
jection to the contralateral entorhinal cortex. The second point to note is that the monkey entorhinal cor-
tex is physically associated with only the rostral portion of the
3.3.3 Monkey Hippocampal Formation other hippocampal elds. The entorhinal cortex extends cau-
dally just to the level of the lateral geniculate nucleus, whereas
The rst point we address in this section is how the appear- the dentate gyrus, hippocampus, and subiculum extend well
ance and position of the nonhuman primate hippocampal caudal to this level. It is equally important to point out that the
formation differs from that of the rat. Much of the work on rostral half of the entorhinal cortex extends beyond the rostral
the monkey hippocampal formation has been carried out in limit of the other hippocampal elds, where it is located ven-
Old World monkeys, primarily the macaque monkey. The tromedial to the amygdaloid complex. Throughout virtually
description we provide here is for the hippocampal formation all of its rostrocaudal extent, the lateral border of the entorhi-
of typical research monkeys, such as Macaca fascicularis and nal cortex is at the rhinal sulcus, as in the rat.
Macaca mulatta (cynomolgus and rhesus macaque monkeys, The ber bundles of the monkey hippocampal formation
respectively). The hippocampal formation in the macaque are fundamentally similar to those in the rat, although there
monkey is not nearly as C-shaped in its long axis as it is in the are a number of minor differences. First, the mbria is located
rat. It lies almost horizontally in the temporal lobe; and, as in dorsomedially in the monkey rather than ventrolaterally as in
the human, it makes up the major portion of the oor of the the rat (Fig. 32). Second, the mbria leaves the substance
temporal horn of the fourth ventricle. The major determinant of the hippocampal formation at a point near the splenium of
of the change in position of the primate hippocampal forma- the corpus callosum. Thus, the compact bundle of bers that
tion is the massive development of the associational cortices is called the body of the fornix travels rostrally as a pendulous,
of the frontal and temporal lobes. As a result of the caudal and attened cable hanging under the corpus callosum. Upon
ventral transposition of the temporal lobes that takes place reaching the level of the anterior commissure, the bundle
developmentally to accommodate the larger cortical surface, descends as the columns of the fornix and follows the same
the primate hippocampal formation comes to lie almost trajectories outlined for the rat. Third, the ventral and dorsal
entirely within the medial temporal lobe. Because of the ven- hippocampal commissures are relatively less prominent in the
trorostral rotation of the monkey hippocampal formation, the monkey than in the rat, reecting the more restricted com-
homologue of the temporal pole of the rat hippocampal for- missural connections observed in the monkey brain (Amaral
mation is located rostrally in the monkey brain, and the et al., 1984; Demeter et al., 1985). The more prominent of the
equivalent of the septal pole of the rat hippocampal formation two commissures in the monkey is the dorsal hippocampal
 Hippocampal Neuroanatomy 49

commissure. It carries bers from the presubiculum and encephalography, there is almost no commissural interaction
entorhinal cortex to the contralateral side; it also carries bers between the hippocampal formations located on each side of
originating in the parahippocampal cortex. the human brain (Wilson et al., 1987).
The subdivision of the hippocampal formation into vari- The ventral surface of the human temporal lobe is demar-
ous regions is fundamentally the same in rat and monkey cated into mediolateral strips by two prominent rostrocau-
brains. There are, however, substantial species differences in dally oriented sulci (Figs. 39 and 310). The more lateral of
certain subregions, especially CA1 and the entorhinal cortex, the two is the occipitotemporal sulcus, which is often broken
which are addressed in greater detail later in the chapter. by small, transverse gyri. The more medial of the sulci, and the
one that is more closely associated with the hippocampal for-
mation, is the collateral sulcus. The collateral sulcus is often
3.3.4 Human Hippocampal Formation continuous with the rostrally situated rhinal sulcus. Unlike the
situation in the rat and the monkey, however, the rhinal sulcus
The three-dimensional position of the human hippocampal is relatively insignicant in the human brain and is associated
formation is similar to that in the macaque monkey brain only with the most rostral portion of the entorhinal cortex. It
(Figs. 36 and 37, respectively). However, owing to the larger is thus not a useful border for the lateral boundary of the
development of the temporal association cortex, in particular entorhinal cortex.
that of the entorhinal and perirhinal cortices, the structure of The collateral sulcus forms most of the lateral border of
the ventromedial surface of the brain, including the gyral pat- what has classically been termed the parahippocampal gyrus.
terns, is substantially different in the human and monkey The parahippocampal gyrus is a complex region that contains
brains. After a brief summary of the gross anatomical attrib- a number of distinct cytoarchitectonic elds and has been
utes of the main or intraventricular portions of the human dened in different ways by different authors. In recent years,
hippocampal formation, we review some of the differences of the parahippocampal gyrus has often been broken up into an
the associated cortical regions. anterior part, which comprises mainly the entorhinal cortex
The classic gross anatomical image of the human hip- and associated perirhinal cortex, and a posterior part, which
pocampal formation is of a prominent bulge in the oor of includes the areas TF and TH of Von Economo (1929).
the temporal horn of the lateral ventricle (Fig. 33). As in the Unlike the situation in the monkey, where the rhinal sulcus
monkey, this portion of the hippocampal formation is widest forms a reasonably reliable lateral border for the entorhinal
at its rostral extent where the structure bends toward the cortex, the collateral sulcus does not provide a discrete lateral
medial surface of the brain. In this area, two to ve subtle gyri, boundary for the human entorhinal cortex. The entorhinal
or digitationes hippocampi, form the pes hippocampi (Gertz cortex actually ends approximately midway along the medial
et al., 1972). The substance of the pes hippocampi is formed bank of the collateral sulcus. The two elds of the perirhinal
by several of the hippocampal elds, and the constituents dif- cortex, areas 35 and 36 of Brodmann (1909), form the remain-
fer at different rostrocaudal levels (see below). Continuing der of the medial bank, fundus, and a portion of the lateral
caudally from the pes hippocampi, the main body of the hip- bank of the collateral sulcus. The perirhinal cortex is massively
pocampus gets progressively thinner as it bends dorsally enlarged in the human brain and may account, in part, for the
toward the splenium of the corpus callosum. prominence of the collateral sulcus. The border zone between
The mbria is situated on the medial surface of the human the entorhinal cortex and the perirhinal cortex was called the
hippocampus, as in the monkey (Fig. 32). At rostral levels, trans-entorhinal zone by Braak (1972, 1980). In this region,
the mbria is thin and at but becomes progressively thicker neuroanatomical markers that label layer II of the entorhinal
caudally as bers are continually added to it. As the mbria cortex demonstrate an oblique band of labeled cells, which
leaves the caudal extent of the hippocampus, it fuses with the makes it appear that layer II is diving underneath the layers of
ventral surface of the corpus callosum and travels rostrally in the perirhinal cortex. Because the perirhinal cortex terminates
the lateral ventricle. The portion of these mbrial bers at a variable point along the lateral bank of the collateral sul-
located between the caudal limit of the hippocampal forma- cus, it is extremely difficult to dene the borders of the
tion and the fusion with the corpus callosum is called the crus entorhinal and perirhinal cortices using imaging modalities
of the fornix, whereas the major portion of the rostrally such as magnetic resonance imaging. The only way it is cur-
directed ber bundle is, as in the monkey, called the body rently possible to dene the border of these elds accurately is
of the fornix. At the end of its rostral trajectory, the body of through histological analysis.
the fornix descends as the columns of the fornix. At about the Interestingly, much of the areal extent of the human
point where the mbria fuses with the posterior portion of entorhinal cortex (at least the rostral portions) can be visually
the corpus callosum, bers extend across the midline to identied on the surface of the brain by the conspicuous
form the hippocampal commissure. A variety of gross bumps, named verrucae (latin for warts), that mark its pial
anatomical terms have been applied to this commissure, but surface. These bumps mark the islands of cells that constitute
the term psalterium (alluding to a harp-like stringed instru- layer II of the entorhinal cortex. The dorsomedial aspect of
ment) is most common. As noted previously, the primate hip- the entorhinal cortex is marked by a conspicuous mound, or
pocampal commissural connections are much more limited secondary gyrus, referred to as the gyrus ambiens (Fig. 310).
than in the rodent; and as suggested by stereotaxic depth The dorsomedial limit of the entorhinal cortex with the amyg-
50 The Hippocampus Book

Figure 39. Ventromedial view of the human tem-

poral lobe. Bar  1 cm.

daloid complex is marked by the shallow sulcus semiannu- (or hook) in gross anatomical descriptions (Fig. 39). The
laris, which is located dorsal to the gyrus ambiens. most rostral of the uncal bulges is often separately labeled
There is a series of prominent bulges on the medial surface the gyrus uncinatus. Histological sections through the gyrus
of the hemisphere just caudal to the dorsomedial portion of uncinatus indicate that it is made up of the amygdalo-
the entorhinal cortex; this is generally labeled the uncus hippocampal region (of the amygdaloid complex) and the

Figure 310. Ventral surface of the human brain

after the brain stem and cerebellum have been
removed. Bar  1 cm.
 Hippocampal Neuroanatomy 51

transition zone between the amygdala and hippocampus. The

middle bulge, called the band of Giacomini, is generally com-
posed of the most medial part of the dentate gyrus. The most
caudal of the uncal bulges, called the intralimbic gyrus, is
composed mainly of a portion of the CA3 eld of the hip-
pocampus.
As the parahippocampal gyrus meets the retrosplenial
region caudally, a group of small bumps known as the gyri
Andreae Retzii are seen on the medial surface of the brain
(Fig. 39). This irregular region of human cortex marks
the caudal limit of the hippocampal formation and is com-
posed principally of CA1 and the subiculum. Two obliquely
oriented small gyri located deep to the gyri Andreae Retzii are
called the fasciola cinerea (which corresponds to the caudal-
most part of the dentate gyrus) and the gyrus fasciolaris
(which corresponds to the caudal pole of the CA3 eld). A
thin remnant of the hippocampal formation surrounds the
splenium of the corpus callosum and ascends dorsally over the
corpus callosum to form the induseum griseum (or supra-
commissural hippocampal formation), which comprises rem-
nants of the subiculum and other elds of the hippocampal
formation.

3.4 Neuroanatomy of the Rat

Hippocampal Formation

We have now concluded our overview of the gross or three-

dimensional organization of the rat, monkey, and human hip-
pocampal formations. In the sections that follow, we review
the neuroanatomical organization of each of the elds of the
hippocampal formation of the rat, beginning with the dentate
gyrus and ending with the entorhinal cortex. In each section,
we describe the cytoarchitectonic organization followed by a
summary of the resident neurons and nally a survey of the
pattern of intrinsic and extrinsic connections. The amount of
information available for each of the cytoarchitectonic elds
of the hippocampal formation varies substantially. The most
Figure 311. Horizontal section through the rat hippocampal for-
detailed anatomical information is available for the dentate
mation. A. Nissl-stained section. B. Line drawing shows the various
gyrus, perhaps because it is the simplest of the hippocampal regions, layers, and ber pathways. C. Timms sulde silver-stained
elds. For this reason and because the dentate gyrus is the rst section. Note the three bands of the molecular layer of the dentate
structure to receive direct entorhinal input and does not proj- gyrus in C. The outer band corresponds to the terminal zone of the
ect outside the hippocampal formation, we begin our descrip- lateral perforant pathway; the middle unstained region corresponds
tion here. We then progress through the hippocampus and to the terminal zone of the medial peforant pathway; the inner band
other hippocampal elds. corresponds to the zone of termination of the associational and
We have prepared a series of illustrations to demonstrate commissural pathways of the dentate gyrus. ab, angular bundle; al,
the cytoarchitectonic organization of the rat hippocampal alveus; CA1, CA1 eld of the hippocampus; CA2, CA2 eld of the
formation (Figs. 311 through 314). The rst gure (Fig. hippocampus; CA3, CA3 eld of the hippocampus; DG, dentate
gyrus; EC, entorhinal cortex; , mbria; gcl, granule cell layer of the
311) is a photomicrograph of a section cut in the horizontal
dentate gyrus; hf, hippocampal ssure; ml, molecular layer of the
plane. This plane has the advantage of illustrating all compo-
dentate gyrus; Para, parasubiculum; pcl, pyramidal cell layer of the
nents of the rat hippocampal formation. A Nissl-stained sec- hippocampus; pl, polymorphic layer of the dentate gyrus; Pre, pre-
tion is shown at the top, and a Timms sulde silver section is subiculum; sl, stratum lucidum of CA3; sr, stratum radiatum of the
shown at the bottom. The latter has been used extensively for hippocampus; sl-m, stratum lacunosum moleculare of the hip-
highlighting various ber and terminal systems in the hip- pocampus. Roman numerals: cortical layers. Bar in A  500 m
pocampal formation. After Figure 311, we provide images of and applies to all panels.
52 The Hippocampus Book

Figure 312. Coronal sections at three rostrocaudal levels through the rat brain show the rela-
tive position of the hippocampal formation. The three panels on the left are Nissl-stained sec-
tions; the three panels on the right are adjacent sections stained with Timms sulde silver
stain; the line drawings (middle column) highlight the regions of the rat hippocampal forma-
tion seen in the stained sections. For abbreviations see Figure 311. Bar  1 mm.
 Hippocampal Neuroanatomy 53

Figure 313. Horizontal sections at three dorsoventral levels through the rat brain show the
relative position of the hippocampal formation. The three panels on the left are Nissl-stained
sections; the three panels on the right are adjacent sections stained with Timms sulde silver
stain; the line drawings (middle column) highlight the regions of the rat hippocampal forma-
tion seen in the stained sections. For abbreviations see Figure 311. Bar  1 mm.
54 The Hippocampus Book

Figure 314. Sagittal sections at three mediolateral levels through the rat brain show the
relative position of the hippocampal formation. The three panels on the left are Nissl-stained
sections; the three panels on the right are adjacent sections stained with Timms sulde silver
stain; the line drawings (middle column) highlight the regions of the rat hippocampal forma-
tion seen in the stained sections. For abbreviations see Figure 311. Bar  1 mm.
 Hippocampal Neuroanatomy 55

similar stains of coronal (Figure 312), horizontal (Fig. 313), the branches directed toward the supercial portion of the
and sagittal (Fig. 314) sections through different levels of the molecular layer; most of the distal tips of the dendritic tree
hippocampal formation. Finally, we must point out that to end just at the hippocampal ssure or at the ventricular sur-
keep this chapter readable we have used citations sparingly. face. The dendritic trees of granule cells located in the
Detailed references for the information provided in this chap- suprapyramidal blade tend, on average, to be larger than those
ter can be found in Witter and Amaral (2004). of cells located in the infrapyramidal blade (3500 m vs. 2800
m). Desmond and colleagues (Desmond and Levy, 1985)
3.4.1 Dentate Gyrus provided estimates for the number of dendritic spines on the
granule cell dendrites. They found that cells in the suprapyra-
Cytoarchitectonic Organization midal blade have 1.6 spines/m, whereas cells in the
infrapyramidal blade have 1.3 spines/m. With these num-
The dentate gyrus is comprised of three layers. Supercially, bers and the mean dendritic lengths given above, an estimate
closest to the hippocampal ssure is a relatively cell-free layer of the number of spines on the average suprapyramidal gran-
called the molecular layer. In the rat, this layer has an average ule cell would be 5600 and for an infrapyramidal cell 3640.
thickness of approximately 250 m. The principal cell layer The total number of granule cells in one dentate gyrus
(granule cell layer) lies deep to the molecular layer and is of the rat is about 1.2106 (West et al., 1991; Rapp and
made up of a densely packed layer that is four to eight granule Gallagher, 1996). Although cell proliferation and neurogenesis
cells thick. The granule cell and molecular layers (which in the dentate gyrus persist into adulthood and appear to be
together are sometimes referred to as the fascia dentata) form under environmental control, modern stereological studies
a V- or U-shaped structure (depending on the septotemporal have shown that the total number of granule cells does not
position) that encloses a cellular region, the polymorphic cell vary in adult animals (Rapp and Gallagher, 1996). Only infant
layer, which constitutes the third layer of the dentate gyrus and juvenile mice exposed to running wheels or socially com-
(Fig. 311). plex, enriched environments demonstrate a larger dentate
gyrus and a greater number of granule cells that persist into
Neuron Types adulthood than animals raised in standard laboratory cages
(Kempermann et al., 1997). Similar manipulations performed
Dentate Granule Cell in adult animals can affect cell proliferation and/or survival of
The principal cell type of the dentate gyrus is the granule cell newly generated neurons, but they have no signicant impact
(Figs. 315 and 316) The dentate granule cell has an ellipti- on the volume of the dentate gyrus or the total number of
cal cell body with a width of approximately 10 m and a granule cells (Kempermann et al., 1998).
height of 18 m (Claiborne et al., 1990). Each cell is closely Before leaving the topic of granule cell number, we should
apposed to other granule cells, and in most cases there is no note that the packing density and thickness of the granule cell
glial sheath intervening between the cells. The granule cell has layers varies somewhat along the septotemporal axis of the
a characteristic cone-shaped tree of spiny dendrites with all dentate gyrus (Gaarskjaer, 1978a). The packing density of

Figure 315. Dendritic arborization of

the principal cells in the rat dentate gyrus
(granule cells) and hippocampus proper
(pyramidal cells). See the text for details.
56 The Hippocampus Book

Figure 316. Dendritic organization

of dentate granule cells. Computer-
generated reconstructions of horserad-
ish peroxidase (HRP)-lled granule
cells from the suprapyramidal blade.
Bar  100 m. (Source: Adapted from
Claiborne et al., 1990.)

granule cells is higher septally than temporally. Because the cal dendrite directed into the molecular layer (where it divides
packing density of CA3 pyramidal cells follows an inverse gra- into several aspiny branches) and several principal basal den-
dient, the net result is that at septal levels of the hippocampal drites that ramify and extend into the polymorphic cell layer.
formation the ratio of granule cells to CA3 pyramidal cells is Most of these cells contain biochemical markers for the
something on the order of 12:1, whereas at the temporal pole inhibitory transmitter -aminobutyric acid (GABA) and are
the ratio drops to 2:3. Because the CA3 pyramidal cells are the thus presumably inhibitory (Ribak et al., 1978; Ribak and
major recipients of granule cell innervation and the number Seress, 1983). The number of basket cells is not constant
of mossy ber synapses is roughly the same along the sep- throughout the transverse or septotemporal extents of the
totemporal axis, contact probability is much lower septally dentate gyrus (Seress and Pokorny, 1981). At septal levels, the
than temporally. ratio of basket cells to granule cells is 1:100 in the suprapyra-
The granule cell is the only principal cell of the dentate midal blade and 1:180 in the infrapyramidal blade. At tempo-
gyrus; that is, it is the only cell type that gives rise to axons that ral levels, the number is 1:150 for the suprapyramidal blade
leave the dentate gyrus to innervate another hippocampal and 1:300 for the infrapyramidal blade. These data raise a
eld (i.e., CA3). There is one type of neuron, called the mossy theme that is repeated throughout the chapter: that despite
cell, whose axon leaves the dentate gyrus of one side of the the apparent cytoarchitectonic homogeneity of the hip-
brain only to innervate the dentate gyrus of the other side. pocampal elds, there are several differences (especially
There are numerous other types of neuron, most of which are regarding neurochemical innervation) at different septotem-
inhibitory interneurons. We describe them in turn. poral levels of the hippocampal formation.

Pyramidal Basket Cell Other Interneurons

The most intensively studied interneuron is the pyramidal There has been an explosion in the number of interneurons
basket cell (Fig. 317). These cells are generally located along identied in the rat hippocampal formation. A detailed
the deep surface of the granule cell layer. They have pyramid- overview of the characteristics of the various hippocampal
shaped cell bodies 25 to 35 m in diameter and are wedged interneurons has been published by Freund and Buzsaki
slightly into the granule cell layer. The basket portion of the (1996). Hippocampal interneurons, which form a heteroge-
name refers to the fact that the axon of these cells forms peri- neous population, are designed to carry out different func-
cellular plexuses that surround and form synapses with the tions. Many of the cell types can be distinguished on the basis
cell bodies of granule cells. Ramon y Cajal rst described the of the distribution of their axonal plexus. Some have axons
pyramidal basket cells as having a single, principal aspiny api- that terminate on cell bodies, whereas others have axons that
 Hippocampal Neuroanatomy 57

Figure 317. Morphological classication of the interneurons of ing an indication of which domain is innervated by which inter-
the rat dentate gyrus. Filled circles indicate the location of the cell neuron groups. The laminar distribution of various inputs, often
bodies, and thick lines indicate the predominant orientation and showing correspondence with the interneuron type or axon distri-
laminar distribution of the dendritic tree. The dentate granule bution, is also indicated. (Source: Adapted from Freund and
cells (principal neurons) are illustrated in the background, provid- Buzsaki, 1996.)

terminate exclusively on the initial segments of other axons. active for GABA. They form symmetrical synaptic contacts
Interneurons have also been distinguished on the basis of with the cell bodies, proximal dendrites, and occasionally
their inputs. Some are preferentially innervated (e.g., by the axon initial segments of granule cells and therefore function
serotonergic fibers originating from the raphe nuclei). as inhibitory interneurons. These cells are not neurochemi-
Interneurons can also be differentiated from principal cells on cally homogeneous, however, as subsets appear to colocalize
the basis of their electrophysiological characteristics. At least distinct categories of other neuroactive substances.
some interneurons have high rates of spontaneous activity
and re in relation to the theta rhythm. For this reason, Neurons of the Molecular Layer
interneurons are often called theta cells (see Chapter 11). The molecular layer is occupied primarily by dendrites of the
A major challenge is to determine if different classes of granule, basket, and polymorphic cells as well as axons and
interneurons demonstrate distinct electrophysiological terminal axonal arbors from the entorhinal cortex and other
response proles. sources. At least two neuron types are also present in the
Within the same subgranular region occupied by the cell molecular layer. The rst is located deep in the molecular
bodies of the pyramidal basket cells are several other cell types layer, has a multipolar or triangular cell body, and gives rise to
with distinctly different somal shapes, as well as different den- an axon that produces a substantial terminal plexus largely
dritic and axonal congurations (Amaral, 1978). Some of limited to the outer two-thirds of the molecular layer. This
these cells are multipolar with several aspiny dendrites enter- neuron, which has aspiny dendrites that remain mainly within
ing the molecular and polymorphic layers, whereas others the molecular layer, has been called the MOPP cell (molecular
tend to be more fusiform-shaped with a similar dendritic dis- layer perforant path-associated cell). This terminology was
tribution. As Ribak and colleagues pointed out, many of these proposed by Han et al. (1993) to bring some order to naming
cells share ne structural characteristics such as infolded interneurons in the hippocampal formation. The lettering sys-
nuclei, extensive perikaryal cytoplasm with large Nissl bodies, tem refers to the location of the cell body and to the region
and intranuclear rods. Moreover, it appears that all of these where the axon is distributed.
cells give rise to axons that contribute to the basket plexus in Frotscher and colleagues described a second type of neu-
the granule cell layer. Many of these neurons are immunore- ron in the molecular layer that resembles the so-called chan-
58 The Hippocampus Book

delier, or axo-axonic, cell originally found in the neocortex most of the daughter dendritic branches remain within the
(Soriano and Frotscher, 1989). These cells are generally polymorphic layer, an occasional dendrite pierces the granule
located immediately adjacent to or even within the supercial cell layer and enters the molecular layer. The mossy cell den-
portion of the granule cell layer. The axo-axonic cell is named drites virtually never enter the adjacent CA3 eld.
for the fact that its axon descends from the molecular layer The most distinctive feature of the mossy cell is that all of
into the granule cell layer, collateralizes profusely, and then its proximal dendrites are covered by large, complex spines
terminates, with symmetrical synaptic contacts, exclusively on evocatively called thorny excrescences. These spines are the
the axon initial segments of granule cells. Thus, their shape distinctive sites of termination of the mossy ber axons (i.e.,
resembles that of a chandelier. Each axo-axonic cell may axons of the dentate granule cells). Although thorny excres-
innervate the axon initial segments of as many as 1000 gran- cences are also observed on the proximal dendrites of pyram-
ule cells. Because these cells are immunoreactive for markers idal cells in CA3, they are never as dense as the ones on the
of GABAergic neurons and make symmetrical synapses, it is mossy cells. The distal dendrites of the mossy cell have typical
likely that they provide a second means of inhibitory control pedunculate spines that appear to be less densely distributed
of granule cell output. The inputs to the axo-axonic cells are than those on the distal dendrites of the pyramidal cells in the
currently unknown although their dendrites remain mainly in hippocampus. The mossy cells are immunoreactive for gluta-
the molecular layer, where they are likely to receive perforant mate and give rise to axons that project to the inner third of
path input from the entorhinal cortex. the molecular layer of the ipsilateral and contralateral dentate
Other neurons with cell bodies located in the molecular gyrus, making asymmetrical terminations on the dendrites of
layer are members of the IS (interneuron-specic) class of granule cells. The mossy cells thus appear to be the major
interneuron, which are specialized for termination on other source of the excitatory associational/commissural projection
interneurons. These IS neurons are demonstrated using to the dentate gyrus. For this reason, the mossy cell does not
immunohistochemistry for vasoactive intestinal peptide t the classic description of an interneuron.
(VIP), and their axons overlap with the dendrites of the O-LM There are also a number of fusiform cells in the polymor-
and HIPP cells (these cell types are dened shortly). phic layer. The main difference between the fusiform cell types
is whether they have spines. One type, the long-spined multi-
Neurons of the Polymorphic Cell Layer polar cell rst described by Amaral (1978), has recently been
The polymorphic layer harbors a variety of neuron types, but called the HIPP cell (hilar perforant path-associated cell). This
little is known about many of them (Amaral, 1978). The most cell type has two or three principal dendrites that originate
common type, and certainly the most impressive, is the mossy from the poles of the cell, run mainly parallel to the granule
cell (Fig. 318). This cell type is probably what Ramon y Cajal cell layer, and can extend for nearly the entire transverse
referred to as the stellate or triangular cells located in his length of one blade of the granule cell layer. The conspicuous
subzone of fusiform cells; and it is undoubtedly what Lorente feature of this cell is the distribution of copious, long, often
de N referred to as modied pyramids. The cell bodies of branched spines over its cell body and dendrites. Intracellular
the mossy cells are large (2535 m) and are often triangular staining techniques demonstrate that these cells have axons
or multipolar in shape. Three or more thick dendrites origi- that ascend into the outer two-thirds of the molecular layer
nate from the cell body and extend for long distances in the (i.e., the perforant path zone) and terminate with symmetri-
polymorphic layer. Each principal dendrite bifurcates once or cal and presumably inhibitory synapses on the dendrites of
twice and generally gives rise to a few side branches. Although granule cells. An amazing feature of these neurons is that their

Figure 318. Mossy cell of the poly-

morphic layer of the rat dentate gyrus.
A. Camera lucida drawing of the soma
and dendritic arbor. Note that the den-
drites extend widely in the polymor-
phic layer but do not penetrate the
granule cell layer. B. Camera lucida
drawing of the soma and axonal plexus.
The axon ramies throughout the poly-
morphic layer and enters the dentate
granule cell layer at considerable dis-
tances from the soma. A single collat-
eral projects through the CA3 cell body
layer and toward the mbria. Bar 
100 m. (Source: Adapted from
Buckmaster et al., 1992.)
 Hippocampal Neuroanatomy 59

axonal plexus can extend for as much as 3.5 mm along the the perforant path bers originating from the medial entorhi-
septotemporal axis of the dentate gyrus (the entire length of nal area terminate in the middle third of the molecular layer.
the dentate gyrus in the rat is only about 10 mm) and may These terminal zones are readily distinguished by the classic
generate as many as 100,000 synaptic terminals. Because Timms staining method for visualizing heavy metals; this
inhibitory interneurons typically have aspiny dendrites and method demonstrates dense staining in the outer third of the
relatively local axonal plexuses, this long-spined multipolar molecular layer, a near absence of staining in the middle third,
HIPP cell is a highly atypical interneuron. At least some of and dark staining in the inner third that is associated with
these HIPP cells appear to correspond to the somatostatin/ the commissural/associational connection (Fig. 311). Projec-
GABA cells that give rise to the somatostatin innervation of tions from both areas of the entorhinal cortex innervate the
the outer portion of the molecular layer. entire transverse extent of the molecular layer. The thin axon
There are also multipolar or triangular cells in the poly- branches (0.1 m) in the molecular layer of the dentate gyrus
morphic layer with thin, aspiny dendrites that extend within show periodic varicosities with a thickness of 0.5 to 1.0 m.
both the hilus and the molecular layer. The axons of these Most of entorhinal cortex layer II spiny stellate cells project up
HICAP cells (hilar commissural-associational pathway- to 2 mm in the septotemporal direction, forming a sheet-like
related cells) extend through the granule cell layer and branch axon arbor in the molecular layer (Fig. 319) (Tamamaki
profusely in the inner third of the molecular layer. A variety of and Nojyo, 1993). There is surprisingly little quantitative
other neuron types exist in the polymorphic layer of the den- information about the termination of the perforant path pro-
tate gyrus whose axonal plexus has not yet been well jections.
described. Because the perforant path bers terminate exclusively in
Before leaving the typing of interneurons in the dentate the molecular layer, certain neurons of the dentate gyrus are
gyrus, it is perhaps worth noting that the eld has come a long not innervated by this input. Cells with dendrites conned to
way from believing that interneurons merely damp down the polymorphic layer do not receive entorhinal input. The
neuronal activity. Rather, most current hippocampologists mossy cells, for example, are thus likely to receive little or no
highlight the heterogeneity of interneurons and see them as direct perforant path input.
integral components of normal information processing in the It has often been assumed that the perforant path bers
hippocampal formation. from the entorhinal cortex are the only hippocampal input
reaching the dentate gyrus, but it is now clear that at least a
Extrinsic Connections minor projection also arises in the presubiculum and para-
subiculum (Kohler, 1985). These bers enter the molecular
Entorhinal Cortex Projection to the Dentate Gyrus layer of the dentate gyrus and ramify in a zone that is inter-
The dentate gyrus receives its major input from the entorhinal spersed between the lateral and medial perforant path projec-
cortex via the so-called perforant pathway (Ramn y Cajal, tions. The presubicular axons tend to be thicker than those
1893). The projection to the dentate gyrus arises mainly from from the entorhinal cortex and give rise to collaterals that take
cells located in layer II of the entorhinal cortex, although a
minor component of the projection also comes from layers V
and VI (Steward and Scoville, 1976). In the molecular layer of
the dentate gyrus, the entorhinal terminals are strictly con-
ned to the supercial (outer) two-thirds, where they form
asymmetrical synapses that account for nearly 85% of the
total axospinous terminations (Nafstad, 1967; Hjorth-
Simonsen and Jeune, 1972). These contacts occur primarily
on the dendritic spines of granule cells, although a small num-
ber of perforant path bers also form asymmetrical synapses
on the shafts of GABA-positive interneurons. The organiza-
tion of the perforant path projection in the mouse is similar
to that in the rat, although there are some detectable species
differences, such as a paucity of commissural connections in
the mouse (van Groen et al., 2002, 2003).
The perforant pathway can be divided into two parts based
on the region of origin, pattern of termination, and appear-
Figure 319. Perforant path projections. Distribution of labeled
ance in histochemical and immunohistochemical prepara- axon branches of a layer II spiny stellate neuron in the molecular
tions. In the rat, the two divisions have been called the lateral layer of the dentate gyrus and the stratum lacunosum-moleculare
and medial perforant paths because they originate from the of the CA2-CA3 elds of the hippocampus observed in a parasagit-
lateral and medial entorhinal areas, respectively. Perforant tal section. Bar  500 m. (Source: Tamamaki and Nojyo, 1993.
path bers originating in the lateral entorhinal area terminate With permission of Wiley-Liss, a subsidiary of John Wiley &
in the most supercial third of the molecular layer, whereas Sons.
60 The Hippocampus Book

a radial course in the molecular layer, in contrast to the pre- pocampal formation are cholinergic. Although it had long
dominantly transverse orientation of the entorhinal perforant been assumed that the septal projection to the hippocampal
pathway bers. Virtually nothing is currently known about formation was entirely cholinergic, it is now clear that many
which cells these bers innervate or what type of transmitter of the septal cells that project to the dentate gyrus are actually
they use. Because the presubiculum receives the only direct GABAergic. The most interesting facet of this heterogeneous
input from the anterior thalamic nucleus, these bers provide septal projection is that the cholinergic and GABAergic com-
a potential link by which thalamic information could reach ponents target different cell types. Fibers of the septal
the dentate gyrus. GABAergic projections terminate preferentially on other
GABAergic nonpyramidal cells, such as the basket pyramidal
Basal Forebrain Inputs: Projections from the Septal Nuclei cells of the dentate gyrus, and form symmetrical, presumably
The dentate gyrus receives relatively few inputs from subcor- inhibitory, contacts. The heaviest GABAergic septal termina-
tical structures. Certainly the most robust and longest studied tion is on interneurons located in the polymorphic layer. The
is the projection from the septal nuclei (Mosko et al., 1973; cholinergic septal projection to the dentate gyrus, in contrast,
Swanson, 1978a; Baisden et al., 1984; Amaral and Kurz, 1985; terminates mainly on granule cells, making asymmetrical,
Wainer et al., 1985; Nyakas et al., 1987). The septal projection presumably excitatory, contacts on dendritic spines, chiey in
arises from cells of the medial septal nucleus and the nucleus the inner third of the molecular layer; only 5% to 10% of the
of the diagonal band of Broca, and it travels to the hippocam- cholinergic synapses are on interneurons.
pal formation via four routes: the mbria, dorsal fornix, The septal projection to the dentate gyrus and to the
supracallosal stria, and a ventral route through and around remainder of the hippocampal formation is topographically
the amygdaloid complex. Septal bers heavily innervate cells organized. Cells located medially in the medial septal nucleus
of the polymorphic layer, particularly in a narrow region just tend to project preferentially to septal or dorsal levels of the
subjacent to the granule cell layer. The large mossy cells are dentate gyrus, whereas cells located laterally in the medial sep-
innervated by cholinergic bers (Lubke et al., 1997). Septal tal nucleus tend to project to temporal levels. Because the
bers are lightly distributed throughout the molecular layer medially situated neurons in the medial septal nucleus tend to
(Fig. 320a). be GABAergic rather than cholinergic, septal levels of the den-
A large portion of the bers of the septal projection to the tate gyrus receive most of their cholinergic input from the
dentate gyrus are cholinergic. Altogether, 30% to 50% of the nucleus of the diagonal band. In contrast, temporal levels of
cells in the medial septal nucleus and 50% to 75% of the cells the dentate gyrus receive their cholinergic innervation prima-
in the nucleus of the diagonal band that project to the hip- rily from the medial septal nucleus. The ventral septal pathway

Figure 320. Septal and hypothalamic

inputs to the rat hippocampal formation.
A. Distribution of PHA-L-labeled bers
in the hippocampal formation resulting
from an injection focused in the medial
portion of the medial septal nucleus.
B. Distribution of PHA-L-labeled bers
resulting from an injection in the supra-
mammillary nucleus. Note the dense
plexus of bers located just supercial to
the granule cell layer and in the CA2
region of the hippocampus (between the
small arrows). (Sources: A. From
Gaykema et al., 1990. B. Adapted from
Haglund et al., 1984.)
 Hippocampal Neuroanatomy 61

appears to arise mainly from the nucleus of the diagonal band that project to the dentate gyrus. Although taken together
of Broca. these cells constitute a sizable input to the hippocampal for-
mation, their diffuseness and lack of any distinguishing bio-
Supramammillary and Other Hypothalamic Inputs chemical marker has made it difficult to study their patterns
The major hypothalamic projection to the dentate gyrus arises of termination in the hippocampal formation.
from a population of large cells, the supramammillary area,
which caps and partially surrounds the medial mammillary Brain Stem Inputs
nuclei (Wyss et al., 1979; Dent et al., 1983; Vertes, 1993; The dentate gyrus receives a particularly prominent nora-
Magloczky et al., 1994). The supramammillary projection ter- drenergic input from the pontine nucleus locus coeruleus
minates heavily in a narrow zone of the molecular layer (Pickel et al., 1974; Swanson and Hartman, 1975; Loughlin et
located just supercial to the granule cell layer and only lightly al., 1986). The noradrenergic bers terminate mainly in the
in the polymorphic layer or the rest of the molecular layer polymorphic layer of the dentate gyrus and extend into
(Fig. 320b). Most of the supramammillary bers terminate the stratum lucidum of CA3, as if preferentially terminating
on the proximal dendrites of granule cells. There has been in the zones occupied by mossy bers (Fig. 3-21).
some controversy about the neurotransmitter of the supra- The dentate gyrus receives a minor, diffusely distributed
mammillary projection. There is substantial evidence that this dopaminergic projection that arises mainly from cells located
projection is excitatory and is likely using glutamate as a pri- in the ventral tegmental area. The dopaminergic bers termi-
mary neurotransmitter (Kiss et al., 2000). Most, but not all, of nate mainly in the polymorphic layer.
the glutamatergic suprammamillary neurons that project to The serotonergic projection that originates from median
the dentate gyrus also colocalize calretinin; some of these cells and dorsal divisions of the raphe nuclei also terminates most
also colocalize substance P (Borhegyi and Leranth, 1997). heavily in the polymorphic layer in an immediately subgran-
In addition to the supramammillary cells, there are cells ular portion of the layer (Conrad et al., 1974; Moore and
scattered in several hypothalamic nuclei (many of which are Halaris, 1975; Khler and Steinbusch, 1982; Vertes et al.,
in a perifornical position or in the lateral hypothalamic area) 1999). A number of GABAergic interneurons appear to be

Figure 321. Line drawing of horizontal sections through the rat hippocampal formation
shows the distribution of A. noradrenergic, B. serotonergic, and C. dopaminergic bers.
(Source: Adapted from Swanson et al., 1987.
62 The Hippocampus Book

preferentially innervated by the serotonergic bers. The tar- Given the small number of basket cells relative to granule
gets are often the pyramidal basket cells. Fusiform neurons cells, the question arises as to how widespread the inuence of
in the region, particularly those that stain for the calcium- a single basket cell is. Analysis of Golgi-stained axonal
binding protein calbindin, are also heavily innervated. As with plexuses from single basket cells indicates that they extend for
the cholinergic projection from the septum, many of the cells distances of more than 900 m in the transverse axis and
in the raphe nuclei that project to the hippocampal formation about 1.5 mm in the septotemporal axis. This widely distrib-
appear to be nonserotonergic, but their transmitter is not uted axonal plexus would allow a single basket cell to inu-
known. ence as many as 10,000 granule cells (about 1%). The other
inhibitory input to granule cells originates with the chandelier
(axo-axonic) cells located in the molecular layer. These cells
Intrinsic Connections form symmetrical contacts exclusively with the axons initial
segment of granule cells.
Basket Cell and Axo-axonic Cell Innervation
of the Granule Cell Layer Granule Cell Projection to the Polymorphic Layer
As already noted, a variety of basket cells are located just The granule cells give rise to distinctive unmyelinated axons
below the granule cell layer and appear to contribute to a that Ramon y Cajal called mossy bers. The mossy bers have
extremely dense terminal plexus that is conned to the gran- unusually large boutons that form en passant synapses with
ule cell layer (Struble et al., 1978; Sik et al., 1997). The termi- the CA3 pyramidal cells; we describe this projection in more
nals in this basket plexus are GABAergic and form detail shortly. What is not generally appreciated is that the
symmetrical, inhibitory contacts located primarily on the cell mossy ber axons form a distinctive set of collaterals that also
bodies and proximal dendritic shafts of apical dendrites of the innervate cells in the polymorphic layer of the dentate gyrus
granule cells. GABAergic neurons in the polymorphic layer (Fig. 322). Each principal mossy ber (which is on the order
are themselves innervated by other GABAergic terminals, of 0.20.5 m in diameter) gives rise to about seven thinner
some of which arise from extrinsic sources, such as the collaterals in the polymorphic layer before entering the CA3
GABAergic septal input. This polysynaptic cascade of eld of the hippocampus. As much as 2300 m of the collat-
inhibitory interconnections indicates that the hippocampal eral axonal plexus is generated by a single mossy ber in the
circuitry provides intricate inhibitory and disinhibitory con- polymorphic layer (Claiborne et al., 1986). In the polymor-
trol of cell excitability, an issue discussed in Chapters 5 and 6. phic layer, the mossy ber collaterals branch extensively, and

Figure 322. Granule cell projections to the

polymorphic cell layer. Axon arbors of two
adjacent granule cells (gray and black) and
their mGluR1a immunoreactive targets
reconstructed from three neighboring 60 m
thick sections. Of the 175 small terminals and
lopodiae (arrowheads) 52 innervated
mGluR1a immunoreactive targets, whereas
large mossy terminals contacted none. gcl,
granule cell layer; pl, polymorphic layer. Bar
 50 m. (Source: Adapted from Acsady et
al., 1998.)
 Hippocampal Neuroanatomy 63

the daughter branches bear two types of synaptic varicosity. Most of the synaptic terminals of the associational/com-
Numerous small (approximately 2 m) spherical synaptic missural pathway form asymmetrical, presumably excitatory
varicosities are distributed unevenly along these collaterals. synaptic terminals on the spines of proximal dendrites of
There are about 160 to 200 of these varicosities distributed granule cells. Many, perhaps all, of the axons contributing to
throughout the axonal collateral plexus of a single granule these synaptic terminals originate from the mossy cells of the
cell, and they form contacts on dendrites located in the poly- polymorphic layer, and individual mossy cells contribute a
morphic layer. At the end of each of the collateral branches projection to both the ipsilateral associational and commis-
there are also larger (35 m diameter), irregularly shaped sural projections. The fact that the mossy cells are immunore-
varicosities that resemble, although are smaller than, the active for glutamate adds credence to the notion that the
mossy ber boutons found in CA3. The mossy ber terminals associational/commissural projection is excitatory.
in the polymorphic layer establish contacts with the proximal There are a number of interesting features of this feed-
dendrites of the mossy cells (Ribak et al., 1985), the basal den- back projection from the mossy cells to the granule cells.
drites of the pyramidal basket cells, and other, unidentied First, the projection from mossy cells located at any particular
cells. Acsady et al. (1998) made the surprising discovery that level of the dentate gyrus is distributed widely along the lon-
most of the granule cell collaterals in the polymorphic cell gitudinal axis, both septally and temporally from the point of
layer terminate on GABAergic interneurons. Because there are origin. Axons from any particular septotemporal point in the
160 to 200 such varicosities (compared with about 20 of the dentate gyrus may innervate as much as 75% of the long axis
larger thorny expansions), mossy ber axons synapse on a of the dentate gyrus (Amaral and Witter, 1989). Second, the
larger number of interneurons than do mossy cells or CA3 projection to the molecular layer at the septotemporal level of
pyramidal cells. Mossy ber collaterals may enter the granule origin is extremely weak but becomes increasingly stronger at
cell layer, but they virtually never enter the molecular layer. levels that are progressively more distant from the cells of ori-
The collaterals that enter the granule cell layer appear to ter- gin. Remembering that mossy cells are the recipients of mas-
minate preferentially on the apical dendritic shafts of pyram- sive innervation from the granule cells at their same level (via
idal basket cells. This lack of mossy ber innervation of the the mossy ber collaterals into the polymorphic layer), it
molecular layer is of some importance because in the presence appears that the mossy cells pass on the collective output of
of kindling and some pathological conditions, such as granule cells from one septotemporal level to granule cells
epilepsy, mossy bers can be induced to sprout into the located at distant levels of the dentate gyrus.
molecular layer. The full impact of this longitudinal organization of the
associational projection cannot be fully appreciated without
Mossy Cell Projection to the Associational/Commissural Zone one further piece of information: The associational bers con-
The inner third of the molecular layer receives a projection tact not only the spines of the granule cell dendrites but also
that originates exclusively from neurons in the polymorphic the dendritic shafts of the GABAergic basket cells, which in
layer (Laurberg and Sorensen, 1981; Buckmaster et al., 1992, turn innervate the granule cells. Thus, the associational pro-
1996; Frotscher, 1992). Because in the rat this projection orig- jection may function both as a feedforward excitatory path-
inates in both the ipsilateral and contralateral sides of the hip- way to distant granule cells and as a disynaptic feedforward
pocampus, it has been called the associational/commissural inhibitory pathway, with the pyramidal basket cell as the inter-
projection. This projection was initially thought to arise from mediary.
both the cells of the polymorphic layer and the CA3 pyrami-
dal cells located within the connes of the dentate gyrus. If GABA/Somatostatin Projection from the Polymorphic
this were true, it would mean that at least one of the intrinsic Layer to the Outer Molecular Layer
hippocampal connections would be bidirectional. It is now Although the neuroanatomy of the hippocampal formation
clear, however, that the associational/commissural projection has been analyzed for many years, new components of its
arises exclusively from cells in the polymorphic layer, and CA3 intrinsic circuitry continue to be discovered. This is often the
cells do not project to the molecular layer. Although this state- result of applying new neuroanatomical techniques. One
ment is generally true, it appears that at least some CA3 neu- example is the discovery of a second projection from the poly-
rons in the most temporal extreme of the hippocampal morphic layer to the molecular layer. Antibodies directed
formation might send collaterals into the molecular layer of against the peptide somatostatin have revealed that neurons
the dentate gyrus. Aside from the existence of this projection, scattered throughout the polymorphic layer are immunoreac-
almost nothing is known concerning its magnitude. Nor is it tive for this peptide and account for approximately 16% of the
known why this projection is observed only in the temporal GABAergic cells in the dentate gyrus (Morrison et al., 1982;
portion of the hippocampal formation. Thus, the principle of Bakst et al., 1986; Freund and Buzsaki, 1996; Sik et al., 1997;
unidirectionality, although generally true, may have excep- Boyett and Buckmaster, 2001). As noted earlier, the somato-
tions to the rule. For all practical purposes, the organization of statin immunoreactive cells may correspond, in part, to the
the commissural projection is similar to that of the ipsilateral HIPP cells. The somatostatin-positive cells all colocalize with
associational connection. GABA and are the source of the somatostatin immunoreactive
64 The Hippocampus Book

bers and terminals in the outer two-thirds of the molecular intra- and infrapyramidal bundles are largely eliminated, and
layer. This system of bers, which forms contacts on the distal virtually all mossy bers travel in the stratum lucidum.
dendrites of the granule cells, provides a third means of releas- Although our focus is primarily on human, monkey, and rat
ing inhibitory control of granule cell activity (in addition to hippocampus in this chapter, it is worth noting in passing that
the basket cell plexus and the axo-axonic terminals provided variations in the relative size of the infrapyramidal mossy
by the chandelier cells). Because electron microscopic studies bers have been noted across different strains of mice, includ-
have demonstrated that somatostatin cells are contacted by ing mouse strains used widely in transgenic experiments; this
mossy ber terminals, the projection to the outer molecular variation sometimes correlates with different behavioral pro-
layer thus constitutes a local feedback inhibitory circuit. les (Lipp et al., 1987; Hausheer-Zarmakupi et al., 1996).
Interestingly, unlike the mossy cell associational projec- Granule cells at all transverse positions in the granule
tion, which terminates more heavily at distant levels of the cell layer generate mossy bers that extend for the full prox-
dentate gyrus, the GABA/somatostatin projection terminates imo-distal distance of CA3 (Fig. 323). Cells located in the
most heavily at the level of the cells of origin; moreover, ter- infrapyramidal blade of the granule cell layer have axons that
mination rapidly decreases within approximately 1.5 mm sep- tend to enter CA3 in the infrapyramidal bundle but ultimately
tally and temporally to the cells of origin. Thus, the mossy cell cross the pyramidal cell layer to enter the deep portion of the
projection and the somatostatin/GABA cell projection have stratum lucidum. The axons of granule cells located in the
terminal elds that are spatially complementary in both radial crest of the dentate gyrus tend to enter CA3 in the intrapyra-
and horizontal axes; the distribution suggests that the two cell midal bundle and also ultimately ascend into the stratum
types mediate distal excitation and local inhibition, respec- lucidum. Cells located in the suprapyramidal blade of the
tively. dentate gyrus give rise to axons that enter CA3 in the stratum
lucidum and continue within the most supercial portion of
Dentate Gyrus Efferent Projection: Mossy Fibers the stratum lucidum (Claiborne et al., 1986).
The mossy bers give rise to unique, complex en passant
The dentate gyrus does not project to any brain region other presynaptic terminals called mossy ber expansions (Amaral
than the CA3 eld of the hippocampus. The axons that and Dent, 1981). Part of the uniqueness of these terminals is
project to CA3, the mossy bers, arise exclusively from the their size; they can be as large as 8 m in diameter but more
granule cells and terminate in a relatively narrow zone mainly typically range from 3 to 5 m in greatest dimension. Their
located just above the CA3 pyramidal cell layer (Blackstad et large size has attracted the attention of physiologists interested
al., 1970; Gaarskjaer, 1978b; Swanson et al., 1978; Claiborne et in using them for patch-clamp studies of transmitter release
al., 1986). In the proximal portion of CA3, mossy bers are (Henze et al., 2000). The mossy ber expansions form highly
also located below and within the pyramidal cell layer. The irregular, complex, interdigitated attachments with the intri-
layer of mossy ber termination located just above the pyram- cately branched spines called thorny excrescences that are
idal cell layer is called the stratum lucidum because the lack of located on the proximal dendrites of the CA3 pyramidal cells.
myelin on the mossy bers gives the layer a relatively clear The thorny excresences are so distinctive they clearly mark the
appearance in fresh tissue (as one might visualize in a hip- location of mossy ber synaptic termination. In the proximal
pocampal slice experiment). There is no indication that den- portion of CA3, for example, thorny excrescences are located
tate neurons other than the granule cells project to CA3; in on both the basal and apical proximal dendrites of pyramidal
particular, cells in the polymorphic layer do not project to the cells, which are therefore in contact with both the infra- and
hippocampus, at least in the rodent. The dentate projection to suprapyramidal mossy ber bundles. In the mid and distal
CA3 stops precisely at the border of CA3 with CA2, and the portions of CA3, however, thorny excresences are almost
lack of granule cell input is one of the main features that dis- entirely restricted to the apical dendritic processes that
tinguishes CA3 from CA2 pyramidal cells. traverse the stratum lucidum (Fig. 324). The CA2 pyramidal
All dentate granule cells project to CA3, and the axon tra- cells, which do not receive any mossy ber input, are devoid of
jectory is partially correlated with the position of the parent thorny excrescences (Fig. 325).
cell body. Before describing features of the mossy ber projec- Another distinctive feature of the mossy ber expansion is
tion, it is worth making a few points concerning the organiza- the number of active synaptic zones they demonstrate. A sin-
tion of the terminal regions in CA3. In the proximal portion gle mossy ber expansion can make as many as 37 synaptic
of CA3 (close to the dentate gyrus), mossy bers are distrib- contacts with a single CA3 pyramidal cell dendrite. Three-
uted below, within, and above the pyramidal cell layer. The dimensional analysis of serial sections through these synapses
bers located below the layer (i.e., those that are in the area indicates that although a mossy ber expansion may be in
occupied primarily by basal dendrites) are generally called the synaptic contact with more than one complex spine originat-
infrapyramidal bundle. The bers located in the pyramidal ing from the same parent dendrite it does not typically contact
cell layer are called the intrapyramidal bundle, and those spines on two different dendrites. Thus, one mossy ber
located above the pyramidal cell layer (i.e., in the area occu- expansion does not typically contact two pyramidal cells.
pied mainly by proximal apical dendrites) are called the The large mossy ber expansions occur approximately
suprapyramidal bundle; the suprapyramidal bundle occupies every 135 m along the parent axon (Fig. 323), and each
the stratum lucidum. At mid and distal portions of CA3, the mossy ber axon forms about 15 of these complex boutons.
 Hippocampal Neuroanatomy 65

Figure 323. Topography of the mossy bers in the CA3 region. A. fourth granule cell located posteriorly. The original coronal images
Camera lucida drawing of mossy bers of three adjacent granule are rotated to emphasize the spatial characteristics of the bers.
cells (truncated). Note the numerous lopodial extensions of the C. Wire diagram of the three mossy bers shown in A, depicting
large mossy terminals in A (arrowheads) and thin stalks of large the distribution of mossy terminals. Note that a shorter interbouton
mossy terminals (arrows). Boxed area in the inset in A shows the distance prevails in the proximal portion of CA3. sl, stratum
position of the bers in CA3. B. Neurolucida reconstruction of the lucidum; pcl, pyramidal cell layer. Bars: A, 50 m; B, 400 m; C,
same three axons shown in A and an additional mossy ber of a 200 m. (Source: Adapted from Acsady et al., 1998.)
66 The Hippocampus Book

Early Golgi anatomists indicated that the mossy ber axons

were mainly oriented perpendicular to the long axis of the
hippocampus. Theodor Blackstad and colleagues conrmed
the largely transverse trajectory of the mossy bers using
degeneration track tracing methods, and Andersen and col-
leagues demonstrated this electrophysiologically. Modern
studies, such as those carried out by Claiborne and colleagues,
also indicate that mossy bers do not collateralize once they
enter the hippocampus; they stay largely within the same sep-
totemporal level as their cells of origin. There is one peculiar-
ity of the mossy ber projection, however, that has eluded
explanation. Lorente de N (1934) and McLardy (McLardy
and Kilmer, 1970) indicated that the mossy bers actually do
change their course and take a longitudinal direction but not
until they reach the distal portion of CA3. The extent of this
distal longitudinal projection was further claried by Swanson
and colleagues using the autoradiographic method of neural
tract tracing (Swanson et al., 1978). They showed that granule
cells located at septal levels give rise to mossy bers that travel
throughout most of the transverse extent of CA3 at the same
septotemporal level. Just at the CA3/CA2 border, however,
they abruptly change course and travel toward the temporal
pole for nearly 2 mm. The extent of this longitudinal compo-
nent appears to depend on the septotemporal location of the
cells of origin. Granule cells in the mid to temporal portions
of the dentate gyrus have mossy bers that exhibit only a
slight temporal inclination at their distal extremity; mossy
bers that originate at the extreme temporal pole of the den-
tate gyrus barely extend to the CA3/CA2 border and have lit-
tle or no longitudinal component. Thus, in the septal half of
the hippocampus, there appears to be some overlap of mossy
Figure 324. Camera lucida drawing of a CA3 pyramidal neuron bers originating from different septotemporal levels, but it
located in the midportion of the eld. As this neuron lies outside occurs only at a restricted distal portion of the CA3 eld. On
the zone of the infrapyramidal mossy ber bundle, most of the
the face of it, this indicates that some CA3 pyramidal cells
thorny excrescences are located on the proximal apical dendrites.
located extremely close to the border with CA2 may be con-
The axon of this neuron is indicated by an arrowhead. pcl, pyrami-
dal cell layer; sl, stratum lucidum; so, stratum oriens; sr, stratum
tacted by mossy ber axons from granule cells spread out over
radiatum. Bar  100 m. (Source: Adapted from Ishizuka et al., a much broader septotemporal extent of the dentate gyrus.
1995.) They might therefore form a special class of integrator CA3
cells. Elegant neuroanatomical verication of this has come
from the work of Acsady and colleagues (1998), who stained
Thus, each granule cell communicates with only 15 CA3 dentate granule cells using in vivo intracellular techniques.
pyramidal cells. It is important to point out, however, that Not only do mossy bers travel 2 to 3 mm in the longitudial
these 15 pyramidal cells are distributed throughout the full direction, they demonstrate typical mossy ber expansions
proximo-distal length (transverse axis) of CA3. Because there along this portion of their trajectory; these terminals do, in
are approximately 2.5  105 CA3 pyramidal cells and approx- fact, make contact with thorny excrescences.
imately 1.2  106 granule cells, each pyramidal cell is expected Before leaving this description of the mossy bers, it
to receive input from about 72 granule cells. That is, an indi- should be noted that in addition to the mossy ber expansions
vidual dentate granule cell has the potential to activate about there are infrequent thin collaterals that emanate from the
15 CA3 pyramidal cells, whereas an individual pyramidal cell parent axon or from the mossy ber expansions. These collat-
receives input from about 72 granule cells. This pattern, not erals appear to terminate preferentially on cells other than the
seen elsewhere in the brain, has attracted considerable interest pyramidal cells, such as the GABAergic pyramidal basket cells
among computational neuroscientists. The fact that each of in CA3.
the mossy ber expansions has many release sites might There is substantial evidence indicating that the granule
ensure highly efficient depolarization of the innervated cells use glutamate as their primary transmitter, and the asym-
pyramidal cells. This is, however, currently controversial (see metrical contacts made between the mossy ber expansions
Chapters 5 and 6). and the thorny excrescences tend to conrm this notion.
 Hippocampal Neuroanatomy 67

Figure 325. Camera lucida drawing

of a CA2 pyramidal neuron. Note that
although the size and general characteris-
tics of this neuron are similar to those in
CA3, there are no thorny excrescences on
the proximal apical dendrites. Two basal
dendrites cut at the surface of the slice
are indicated by asterisks. The axon of
this neuron is indicated by an arrowhead.
CA1, CA1 eld of the hippocampus;
CA3, CA3 eld of the hippocampus; hf,
hippocampal ssure; pcl, pyramidal cell
layer; sl, stratum lucidum; sl-m, stratum
lacunosum-moleculare; so, stratum
oriens; sr, stratum radiatum. Bar  100
m. (Source: Adapted from Ishizuka et
al., 1995.)

Nonetheless, the mossy bers are also immunoreactive for narrow, relatively cell-free layer located deep to the pyramidal
several other neuroactive substances. At least some of the cell layer is called the stratum oriens. This layer contains the
mossy bers demonstrate immunoreactivity for the opioid basal dendrites of the pyramidal cells and several classes of
peptide dynorphin and are also immunoreactive for GABA interneurons. Stratum oriens can be dened as the infrapyra-
(Walker et al., 2002). Interestingly, although the granule cells midal region in which some of the CA3 to CA3 associational
do not normally demonstrate mRNA for the synthetic connections and the CA3 to CA1 Schaffer collateral connec-
enzymes GAD65 or GAD67, prolonged stimulation of the tions are located. Deep to the stratum oriens is the thin, ber-
perforant path can induce GAD messenger expression in rat containing alveus. In the CA3 eld, but not in CA2 or CA1, a
granule cells. The possible role of GABA or opioid peptides in narrow acellular zone, the stratum lucidum, is located just
the synaptic economy of the mossy bers is discussed in rela- above the pyramidal cell layer and is occupied by the mossy
tion to synaptic plasticity at these synapses in Chapter 10. bers. There is a slight thickening of the stratum lucidum at
its distal end, where, at least at septal levels of the hippocam-
3.4.2 Hippocampus pus, the mossy bers bend temporally and travel longitudi-
nally. This zone, called the end bulb, is more prominent in
Cytoarchitectonic Organization species such as the guinea pig than in the rat; it marks the
CA3/CA2 border. The stratum radiatum is located supercial
An overview of the major elds of the hippocampus was given to the stratum lucidum in CA3 and immediately above the
earlier in the chapter. We now go into more detail about its pyramidal cell layer in CA2 and CA1. The stratum radiatum
laminar organization, which is generally similar for all the can be dened as the suprapyramidal region in which the
elds of the hippocampus. The principal cellular layer is called CA3 to CA3 associational connections and the CA3 to
the pyramidal cell layer. The pyramidal cell layer is tightly CA1 Schaffer collateral connections are located. The most
packed in CA1 and more loosely packed in CA2 and CA3. The supercial layer of the hippocampus is called the stratum
68 The Hippocampus Book

lacunosum-moleculare. It is in this layer that bers from the

entorhinal cortex terminate. Afferents from other regions,
such as the nucleus reuniens of the midline thalamus, also
terminate in the stratum lacunosum-moleculare. Both the
stratum radiatum and the stratum lacunosum-moleculare
contain a variety of interneurons.
In the following sections, we sometimes deal with the
organization of the CA3 and CA2 elds of the hippocampus
rst and then move on to the CA1 eld. For other topics, such
as the interneurons of the hippocampus, there is not enough
known to warrant discussing CA3, CA2, and CA1 separately.
For these topics, we discuss data for the entire hippocampus.
Moving on from the dentate gyrus, we begin our discussion
with the CA3 eld.

Pyramidal Cells of CA3 and CA2

The principal neuronal cell type of the hippocampus is the

pyramidal cell, which makes up most of the neurons in the
pyramidal cell layer. Pyramidal cells have a basal dendritic tree
that extends into the stratum oriens and an apical dendritic
tree that extends to the hippocampal ssure.
Ishizuka and colleagues (Ishizuka et al., 1995) demon-
strated that the dendritic length and organization of CA3
pyramidal cells are quite variable (Figs. 324 and 327). The
smallest cells (with a soma size of about 300 m2 or 20 m in
diameter) are located in the limbs of the dentate gyrus and
have a total dendritic length of 8 to 10 mm. The largest cells
(with a soma size of about 700 m2 or 30 m in diameter),
which are located distally in the eld, have total dendritic
Figure 326. Camera lucida drawing of a CA1 pyramidal neuron
lengths of 16 to 18 mm. The distribution of the dendritic trees from the midportion of the eld. Note that side branches originate
of CA3 pyramidal cells also varies depending on where the cell from the primary dendrites throughout the full extent of the stra-
body is located. Cells located in the limbs of the dentate gyrus, tum radiatum. Note also the curved and irregular trajectories of
for example, have few or none of their dendrites extending dendritic branches in the stratum lacunosum-moleculare. The axon
into the stratum lacunosum-moleculare, and thus these cells of this neuron is indicated by an arrowhead. pcl, pyramidal cell
receive little or no direct input from the entorhinal cortex. The layer; sl-m, stratum lacunosum-moleculare; so, stratum oriens; sr,
cells, however, receive a larger number of mossy ber termi- stratum radiatum. Bar  100 m. (Source: Adapted from Ishizuka
nals on their apical and basal dendritic trees and are thus et al., 1995.)
under greater inuence of the granule cells than distally
located CA3 cells, which receive only apical mossy ber input. dendrite, and others have two. Cells with two apical dendrites
tend to have slightly greater total dendritic length in the apical
Pyramidal Cells of CA1 direction. Neurons with a single apical dendrite, however, tend
to have slightly larger basal dendritic trees; thus, overall, the
In contrast to the substantial heterogeneity of dendritic dendritic tree in all CA1 neurons have about the same total
organization characteristic of CA3 pyramidal cells, investiga- length. This anatomical homogeneity, however, cannot reect
tors such as Norio Ishizuka and Dennis Turner and colleagues functional homogeneity because, as we shall see shortly, there
(Pyapali et al., 1998) demonstrated that the CA1 pyramidal are differences in the entorhinal cortex inputs received at dif-
cells show remarkable homogeneity of their dendritic trees ferent locations along the transverse axis of CA1.
(Figs. 326 and 327). As well as being more homogeneous,
they are also, on average, smaller than CA3 cells. The total Interneurons
dendritic length averages approximately 13.5 mm, and the
average size of CA1 cell somata is about 193 m2 or 15 m in Pyramidal neurons are by far the most numerous neurons in
diameter. Regardless of where a pyramidal cell is located in the hippocampus. As in the dentate gyrus, however, there is a
CA1, it has about the same total dendritic length and the same fairly heterogeneous group of interneurons that are scattered
dendritic conguration. Some pyramidal cells have one apical through all layers (Fig. 328) (Freund and Buzsaki, 1996).
 Hippocampal Neuroanatomy 69

Figure 327. Summary of the organization

of hippocampal pyramidal cells produced as
a composite of computer-generated line
drawings of neurons from CA3, CA2, and
CA1. Dashed line in CA3 marks the region
occupied by the infrapyramidal and
suprapyramidal mossy bers. Source:
(Adapted from Ishizuka et al., 1995.)

Figure 328. Morphological classication of the interneurons in ron groups. The laminar distribution of the inputs, which often
the hippocampus proper. Filled circles indicate the location of the show correspondence with the interneuron type or axon distribu-
cell bodies, and thick lines indicate the predominant orientation tion, is also indicated. pcl, pyramidal cell layer; sl-m, stratum
and laminar distribution of the dendritic tree. The pyramidal cells lacunosum-moleculare; so, stratum oriens; sr, stratum radiatum.
(principal neurons) are illustrated in the background, providing an (Source: Adapted from Freund and Buzsaki, 1996.)
indication of which domain is innervated by the various interneu-
70 The Hippocampus Book

Most types of interneuron are found in all the hippocampal located in the zones occupied by recurrent pyramidal cell col-
subelds. The pyramidal basket cell resides in or close to laterals. In CA3 this includes all strata except the stratum
the pyramidal cell layer, and its dendrites extend into the lacunosum-moleculare, but in CA1 it includes only the stra-
stratum oriens, stratum radiatum, and stratum lacunosum- tum oriens. The axons of the O-LM cell leave the stratum
moleculare. The dendrites are beaded and aspiny, and they oriens (or in whichever layer the cell body is located) and rise
receive both asymmetrical and symmetrical synapses. Most of directly to the stratum lacunosum-moleculare, ramifying
the excitatory inputs are known to arise from hippocampal there to form a dense plexus. These axons form symmetrical
pyramidal cells. In fact, the dendritic tree of a pyramidal bas- synapses with the distal apical dendrites of pyramidal neu-
ket cell receives at least 2000 excitatory inputs. Because each rons. Because most of the excitatory input to the O-LM cells
pyramidal cell contributes only a single synapse to a particu- appears to arise from recurrent collaterals of the pyramidal
lar basket cell, the degree of pyramidal cell convergence on an cells, this class of interneuron inhibits activity in the distal
individual basket cell is enormous. Neurons with basket cell- dendrites of pyramidal cells in a disynaptic, feedback manner.
like axons have a variety of morphologies. There are fusiform- Another class of hippocampal interneuron, the bistratied
shaped basket cells in the stratum oriens and stellate-shaped cell, also has its cell body located close to the pyramidal cell
basket cells in the stratum radiatum. In all cases, the axons of layer. The dendritic trees of these neurons are multipolar but
these cell types innervate the soma and proximal dendrites of do not reach the stratum lacunosum-moleculare. The axon of
the pyramidal cells. The transverse extent of the basket cell the bistratied cell sends collaterals into the stratum oriens
axonal plexus (in a 400-m in vitro slice preparation) is and the deep portion of the stratum radiatum, where a dense
between 900 and 1300 m. Within this plexus, there are as terminal plexus is produced. These neurons generate an enor-
many as 10,000 synaptic varicosities; and because a basket cell mous axonal plexus, on the order of 80 mm in total length and
makes only 2 to 10 synapses on each pyramidal cell, a typical generating up to 16,000 synaptic varicosities. The bistratied
basket cell innervates as many as 1000-plus pyramidal cells. A cells have axons that terminate on both the dendritic shafts
single basket cell could thus have substantial inhibitory inu- and the dendritic spines of pyramidal cells. Although the
ence over a large population of pyramidal cells. inputs to these cells have not been thoroughly investigated,
A second type of hippocampal interneuron is the chande- their dendrites reside in the zone of associational connections
lier, or axo-axonic, cell. The axo-axonic cells found in the hip- in CA3 and the Schaffer collateral bers in CA1. It is likely,
pocampus are similar to the ones described in the dentate therefore, that they are driven in both a feedforward and feed-
gyrus. Their cell bodies, like those of the basket cells, are back manner.
located in or adjacent to the pyramidal cell layer, and their There are other interneurons located in the stratum radia-
dendrites span all the hippocampal strata. The axons of the tum, and their stellate or multipolar dendritic plexus is con-
chandelier cells have a transverse spread of approximately 1 ned to the layer. The axons of these cells tend to ramify
mm. They travel just supercial to the pyramidal cell layer and locally in the stratum radiatum and terminate primarily on
periodically give rise to collaterals that enter the pyramidal the dendrites of pyramidal cells.
cell layer and terminate on the proximal axons of the pyrami- A fairly sizable population of interneurons is located in the
dal neurons. Each axo-axonic cell terminates on approxi- stratum lacunosum-moleculare or at the border between the
mately 1200 pyramidal cell axon initial segments, and each stratum lacunosum-moleculare and the stratum radiatum
initial segment is innervated by 4 to 10 axo-axonic cells. (Lacaille and Schwartzkroin, 1988). These LM neurons (stra-
The early Golgi studies of Ramon y Cajal and Lorente de tum lacunosum-moleculare interneurons) have dendrites that
N made it abundantly clear that there are a variety of non- are oriented horizontally (i.e., within the layer) but occasion-
pyramidal cell types in the stratum oriens, stratum radiatum, ally have branches that extend into the pyramidal cell layer.
and stratum lacunosum-moleculare of the hippocampus. The axon also takes a predominantly horizontal orientation
Ribak and colleagues (1978) were the rst to discover that and ramies mainly in the stratum lacunosum-moleculare or
most of these neurons are immunoreactive for GABAergic the supercial portion of the stratum radiatum. Although the
markers, and most are thought to be interneurons. Freund connectivity of this class of neurons is not well worked out,
and Buzsaki (1996) described the location, dendritic organi- their axons do form symmetrical synapses on the distal den-
zation, and axonal distribution of these cells based on a classi- drites of pyramidal cells.
cation system that is dependent on the region of innervation. An additional type of interneuron in the hippocampus is
One class of cells (Lacaille et al., 1987) has been called the the IS neuron (interneuron-selective). The IS neurons cell
O-LM cell (oriens lacunosum-moleclare(associated cell) and bodies are located in all layers and can be identied by stain-
has as its dening feature a dense axonal arbor that is conned ing with antibodies to the calcium-binding protein calretinin.
to the stratum lacunosum-moleclare (also known as cells The IS cells have a number of notable features, including the
terminating in conjunction with entorhinal afferents). The propensity for their dendrites to form bundles with dendrites
location of the cell body of this class of interneuron varies of other IS neurons. The major unifying feature, however, is
depending on which hippocampal eld it inhabits. The prin- that their axons terminate exclusively on other interneurons.
ciple seems to be that the cell body and dendritic tree are Little is yet known concerning their input/output characteris-
 Hippocampal Neuroanatomy 71

tics. Their dendrites, however, are disposed so they could be originate from cells in layer II, and collaterals of the same layer
innervated by all major afferent and intrinsic connections of II cells reach both the dentate gyrus and CA3/CA2, implying
the hippocampus, and their axons could potentially innervate that similar information reaches these structures.
all of the interneurons described above.
All of the interneurons we have described are immunopos- Entorhinal Cortex Projection to CA1
itive for markers of GABA. As in other cortical regions, the Although the entorhinal cortex also projects to the CA1 eld
interneurons of the hippocampus colocalize a number of of the hippocampus, the organization of this projection is
other neuroactive substances. Thus, many of these neurons fundamentally different from the projection to CA3/CA2.
can be visualized with antibodies to peptides such as somato- First, the cells of origin are in layer III rather than in layer II.
statin, VIP, cholecystokinin, neuropeptide Y (NPY), and cal- Second, the pattern of terminal distribution is organized
cium-binding proteins such as parvalbumin, calbindin, and in a topographical fashion rather than in a laminar fashion
calretinin. It remains something of a mystery what these (Fig. 329). As in CA3/CA2, the entorhinal bers terminate
neuroactive substances are doing in GABAergic neurons, throughout the full width of the stratum lacunosum-molecu-
but at the very least they provide useful markers for establish- lare of CA1. However, bers originating in the lateral entorhi-
ing subcategories of the large population of GABAergic nal area terminate in the distal portion of CA1 (close to the
interneurons. subiculum), whereas bers originating in the medial entorhi-
nal area terminate in the proximal portion of CA1 (close to
Extrinsic Connections CA2). Thus, depending on where a CA1 pyramidal cell is
located in the transverse axis of the hippocampus, it receives
One of the distinguishing features of the connectivity of the inputs from a different portion of the entorhinal cortex. We
hippocampus is that most of its synaptic input arises from noted earlier that CA1 cells are remarkably homogeneous
within its own boundaries. CA3 and CA2 are heavily inner- in their dendritic length; the inputs they receive, however,
vated by collaterals of their own axons (i.e., associational con- differ as a function of their location along the proximo-
nections) and from axons of the contralateral CA3 and CA2 distal axis.
(i.e., commissural connections). CA1, in turn, receives its
heaviest input from CA3. A relatively lighter projection arises Hippocampal Projections to the Entorhinal Cortex
from the entorhinal cortex and terminates on the most distal Early studies using autoradiographic tract-tracing techniques
dendrites of the pyramidal cells as well as on interneurons suggested that all elds of the hippocampus send a return pro-
with dendrites in the stratum lacunosum-moleculare. There jection to the entorhinal cortex, but it is now clear that only
are relatively few other extrinsic inputs to the hippocampus, cells located in CA1 give rise to this projection (Naber et al.,
and the ones that do exist generally account for a relatively 2001). The projecting cells appear to send their axons to
small number of synapses. roughly the same region of the entorhinal cortex from which
they receive their input (Fig. 329). Thus, proximal CA1 cells
Entorhinal Cortex Projection to CA3/CA2 project to the medial entorhinal cortex, whereas distal CA1
Although the entorhinal innervation of CA3 is mentioned in cells project to the lateral entorhinal area. We return to a more
most studies of the perforant path projection, the organization detailed description of these projections in Section 3.4.5.
of this component of the projection is generally not dealt with
in much detail. Indeed, there has been little in-depth research Hippocampal Connections with the Neocortex
on the entorhinal projections to the hippocampus despite the and Amygdaloid Complex
fact that its prominence suggests that it is extremely important. Although this book focuses on the hippocampal formation,
The perforant path takes its name from the observation that it we repeatedly emphasize that no brain structure can be seen
perforates the subiculum and hippocampal ssure en route to in isolation. The hippocampus sends projections to and
the dentate gyrus, but it is now quite clear that collaterals of receives projections from numerous other brain regions, and
bers that project to the dentate gyrus also project to the hip- these interconnections are vital to understanding its function.
pocampus. In fact, the origin and laminar terminal distribu- Having said that, and notwithstanding the extremely impor-
tion of the perforant path projection to CA3 are similar to tant functional relationship between hippocampus and neo-
those to the dentate gyrus (Witter, 1993). Entorhinal terminals cortex, it turns out that only selected parts of the hippocampal
are distributed throughout the width of the stratum lacuno- formation have discrete, monosynaptic connections with the
sum-moleculare. As with the projection to the molecular layer neocortex. The CA3 and CA2 elds of the hippocampus, for
of the dentate gyrus, projections from the lateral entorhinal example, have no known connections with the neocortex.
area terminate supercially in the stratum lacunosum-molec- Sensitive tracing methods have recently shown that CA3, in
ulare, and those from the medial entorhinal area terminate in particular its temporal parts, receives input from the amyg-
the deep half of the layer. The laminar origin, types, and num- daloid complex, which was previously thought to send projec-
bers of synaptic contacts of this projection are similar to those tions only to CA1 and the subiculum (Pikkarainen et al., 1999;
to the dentate gyrus. The entorhinal projections to CA3/CA2 Pitkanen et al., 2000). These inputs originate mainly from the
72 The Hippocampus Book

Figure 329. Reciprocal entorhinalhippocampal connections. A. injections into the lateral entorhinal cortex and the distal portion of
Distribution of labeled bers in the hippocampus and entorhinal the CA1 eld of the hippocampus. The terminal bers originating
cortex after combined anterograde tracer injections into the medial in the entorhinal cortex overlap with the injection site in CA1, and
portion of the entorhinal cortex and the proximal portion of the the terminal bers originating in CA1 overlap with the injection site
CA1 eld of the hippocampus. The terminal bers originating in in the entorhinal cortex. The subdivisions of the entorhinal cortex
the entorhinal cortex overlap with the injection site in CA1, and the follow the nomenclature of Insausti et al. (1997). Areas CE and ME
terminal bers originating in CA1 overlap with the injection site in constitute the medial entorhinal cortex, whereas DLE, DIE, VIE, and
the entorhinal cortex. B. Distribution of labeled bers in the hip- AE constitute the lateral entorhinal cortex. (Source: Adapted from
pocampus and entorhinal cortex after combined anterograde tracer Naber et al., 2001.)

caudomedial portion of the parvicellular division of the basal stratum lacunosum-moleculare and overlaps the projection
nucleus and terminate heavily in the stratum oriens and stra- arising from the lateral entorhinal area. The same CA1 cells
tum radiatum. The best-documented cortical connection, give rise to a return projection to the perirhinal cortex. CA1
other than with the entorhinal cortex, which we have dened cells located in the septal portion of the hippocampus have
as part of the hippocampal formation, is with the perirhinal also been reported to project to the retrosplenial cortex, and
(areas 35 and 36) and postrhinal cortices. Cells in the perirhi- those located at mid-septotemporal levels provide a fairly sub-
nal cortex give rise to a relatively selective projection to the stantial projection to the medial frontal lobe.
most distal CA1 pyramidal cells (i.e., those located at the bor- The temporal two-thirds of the distal portion of CA1 is
der with the subiculum). This projection terminates in the reciprocally connected with the amygdaloid complex.
 Hippocampal Neuroanatomy 73

Projections from the basal nucleus of the amygdala terminate The suparamammillary region projects weakly, if at all, to
in the stratum oriens and stratum radiatum of the CA1/ CA3 and CA1.
subiculum border region. In addition, the accessory basal and
cortical nuclei projects to the stratum lacunosum-moleculare. Thalamic Connections: Nucleus Reuniens
CA1 cells in the same region give rise to a return projection to and Other Midline Nuclei
the basal nucleus of the amygdala. The thalamic inputs to the hippocampal formation have
received relatively little attention. It has been known for some
Basal Forebrain Connections time that the anterior thalamic complex is intimately inter-
The septum provides the major subcortical input to CA3. As connected with the presubiculum. However, Herkenham and
with the dentate gyrus, the septal projection originates mainly others demonstrated fairly prominent projections from mid-
in the medial septal nucleus and the nucleus of the diagonal line (nonspecic) regions of the thalamus to several elds of
band of Broca. The projection terminates most heavily in the the hippocampal formation (Herkenham, 1978; Wouterlood et
stratum oriens and to a lesser extent in the stratum radiatum al., 1990; Dolleman-Van der Weel and Witter, 2000). The
(Fig. 320). The CA1 eld receives a substantially lighter sep- nucleus reuniens, located on the midline, gives rise to a promi-
tal projection than CA3, but the bers are also most densely nent projection to the stratum lacunosum-moleculare of CA1,
distributed in the stratum oriens. where it overlaps with bers from the entorhinal cortex. The
Until the mid-1970s, it was commonly assumed that the nucleus reuniens projection travels to CA1 via the internal
hippocampal elds gave rise to both the precommissural and capsule and cingulum bundle rather than through the fornix
postcommissural projections to the basal forebrain and dien- and mbria. This projection innervates all septotemporal lev-
cephalon. Indeed, Ramon y Cajals classic diagram of the els of CA1, with a preference for the mid-septotemporal levels.
hippocampus shows bers from CA1 coursing toward the The nucleus reuniens bers terminate with asymmetrical
mbria. Swanson and Cowan demonstrated, however, that synapses on spines and thin dendritic shafts in the stratum
most of these projections originate from the subiculum lacunosum-moleculare on both principal neurons and
(Swanson and Cowan, 1975). It is now quite clear that the only GABAergic interneurons.
sizable subcortical projection from CA3 is to the lateral septal
nucleus. The CA3 projection to the lateral septal nucleus trav- Brain Stem Inputs
els via the mbria and precommissural fornix. The CA3 The hippocampus, like the dentate gyrus, receives noradren-
projection to the septal complex is bilateral; and some CA3 ergic and serotonergic inputs from brain stem nuclei (Fig.
bers cross in the ventral hippocampal commissure to inner- 321). Noradrenergic bers and terminals arising from the
vate the homologous region of the contralateral lateral septal locus coeruleus are most densely distributed in the stratum
nucleus. This pathway is topographically organized such that lucidum and the most supercial portion of stratum lacuno-
septal portions of CA3 project dorsally in the lateral septal sum-moleculare. A much thinner plexus of axons is distrib-
nucleus, and progressively more temporal portions of CA3 uted throughout the other layers of CA3. Serotonergic bers
project more ventrally; proximal CA3 cells tend to project are distributed more diffusely and sparsely in CA3 than in the
medially in the lateral septal nucleus, and distally situated CA3 noradrenergic bers. Despite the rather low number of bers,
cells terminate more laterally. Interestingly, virtually all CA3 the serotonergic innervation of the hippocampus demon-
cells give rise to projections to both CA1 and the lateral septal strates several interesting features. First, there are two calibers
nucleus. of axon, thick and thin, arising from the raphe nuclei that
Essentially all CA1 pyramidal cells also project to the lateral innervate the hippocampus. Most of the serotonergic vari-
septal nucleus. However, the CA1 projection is strictly ipsilat- cosities, which are located on the thin bers, do not appear to
eral, and some of the bers travel to the septal nuclei via the have standard synaptic junctions and may release transmitter
dorsal fornix rather than through the mbria. into the extracellular space. The varicosities on the thicker
bers, in contrast, form standard asymmetrical synapses that
Hypothalamic Connections preferentially terminate on GABAergic inhibitory neurons,
There has been little work dealing specically with the extrin- specically on the classes of interneuron that project to the
sic inputs and outputs of CA2. In general, CA2 appears to dendrites of hippocampal neurons. Thus, even the relatively
have the same connections as CA3. However, the CA2 eld few serotonergic bers that innervate the hippocampus
receives particularly prominent innervation from the poste- may have a profound action by enhancing the GABAergic
rior hypothalamus, in particular from the supramammillary inhibitory activity of the hippocampal interneurons. There
area (Fig. 320B) and the tuberomammillary nucleus. These are few, if any, dopaminergic bers in CA3. In general, CA1
projections terminate mainly in and around the pyramidal receives much lighter monoaminergic innervation than CA3.
cell layer and mainly on principal cells (Magloczky et al., The functional implications of monoaminergic inputs to
1994). There is no evidence that CA2 returns the projection to the hippocampus, particularly in relation to long-term poten-
the supramammillary region. In fact, none of the hippocam- tiation (LTP), is considered in more detail in Chapters 5, 6
pal elds appears to project into the postcommissural fornix. and 7.
74 The Hippocampus Book

Intrinsic Connections: CA3 Associational Connections as extending only through the stratum radiatum, it should be
and Schaffer Collaterals emphasized that both the stratum radiatum and stratum
oriens of CA1 are heavily innervated by CA3 axons. Thus, the
As mentioned earlier, the major source of input to the hip- Schaffer collaterals are as highly associated with the apical
pocampus is the hippocampus itself. The CA3 to CA3 associ- dendrites of CA1 cells in the stratum radiatum as they are
ational connections and the CA3 to CA1 Schaffer collateral with the basal dendrites in the stratum oriens. Moreover,
connections are unique in many respects. Perhaps the major CA3 cells located at any particular septotemporal level dis-
distinguishing feature of these projections, however, is their tribute some of their collaterals to much of the full septotem-
extensive spatial distribution. Through these connections a poral extent of CA1. This projection is not chaotic, however,
particular pyramidal cell in CA3 can, in theory, interact with and its topographical organization develops a network in
other hippocampal neurons distributed throughout much of which certain CA3 cells are more likely to contact certain CA1
the ipsilateral and contralateral hippocampus. The massive cells.
potential for association in the hippocampus is undoubtedly The major organizational features of this projection are as
linked to its function. The hippocampal connections, follows: CA3 cells located close to the dentate gyrus (proximal
although widely distributed, are nonetheless systematically CA3), although projecting both septally and temporally, proj-
organized. These projections have been studied by Ishizuka ect more heavily to levels of CA1 located septal to their loca-
and colleagues (Ishizuka et al., 1990) and Buzsaki and col- tion. CA3 cells located closer to CA1, in contrast, project more
leagues (Li et al., 1994). heavily to the levels of CA1 located temporally (Fig. 330). At
All CA3 and CA2 pyramidal cells give rise to highly diver- or close to the septotemporal level of the cells of origin, those
gent projections to all portions of the hippocampus. CA3 cells located proximally in CA3 give rise to collaterals that
pyramidal cells give rise to highly collateralized axons that dis- tend to terminate supercially in the stratum radiatum.
tribute bers both within the ipsilateral hippocampus (to Conversely, cells located more distally in CA3 give rise to pro-
CA3, CA2, and CA1), to the same elds in the contralateral jections that terminate deeper in the stratum radiatum and
hippocampus (the commissural projections), and subcorti- stratum oriens. At or close to the septotemporal level of ori-
cally to the lateral septal nucleus. Some CA3 (especially those gin, CA3 pyramidal cells located near the dentate gyrus tend
located proximally) and CA2 cells contribute a small number to project somewhat more heavily to distal portions of CA1
of collaterals that innervate the polymorphic layer of the den- (near the subiculum), whereas CA3 projections arising from
tate gyrus. Although claims of other hippocampal connec- cells located distally in CA3 terminate more heavily in por-
tions are found in the literature, it is now quite clear that CA3 tions of CA1 located closer to CA2. The truly thick Schaffer
does not project to the subiculum, presubiculum, parasubicu- collaterals (those that Schaffer originally described) originate
lum, or entorhinal cortex. only from the proximal CA3 cells. These cells give rise to a
The CA3 projections to CA3 and CA2 are typically called thick axon that ascends from the stratum oriens into the most
the associational connections, and the CA3 projections to the supercial portion of the stratum radiatum and travels to
CA1 eld are typically called the Schaffer collaterals. This ter- the distal part of CA1, where it contributes many collaterals.
minology, however, may be misleading, as one should remem-
ber that these two projections are true collaterals and thus Figure 330. Organization of the projections from the CA3 eld to
potentially carry the same information. Both the CA3 to CA3 the CA1 eld of the hippocampusthe Schaffer collaterals. The
associational projections and the CA3 to CA1 Schaffer collat- location of the cells of origin is indicated by small triangles in the
erals demonstrate a systematic gradient-like projection pat- middle coronal section. Terminals from these cells are indicated by
tern. Although it is somewhat out of sequence to discuss the different shades of gray similar to those in the triangles.
CA3 to CA1 projections rst, we do so because they have been
worked out in somewhat better detail and provide a clear
model for understanding CA3 projections. Moreover, the
organization of the CA3 to CA1 projection shares many orga-
nizational similarities with the CA3 to CA3 projection.

CA3 to CA1 Connections: Schaffer Collaterals

All portions of CA3 and CA2 project to CA1. The pattern of
terminal distribution, however, depends on the location of the
CA3/CA2 cells of origin. The older notion that a typical CA3
pyramidal cell sends a single axon to CA1, traveling linearly
through the eld with equal contact probability at all regions
in CA1 is clearly incorrect. In fact, each CA3 pyramidal cell
gives rise to highly collateralized axons that follow both trans-
verse and oblique orientations through CA1 (Ishizuka et al.,
1990). Although Schaffer collaterals are typically illustrated
 Hippocampal Neuroanatomy 75

The axons of distal CA3 cells tend to be much thinner and to To summarize, the CA3 to CA1 projection is the major
project directly to CA1, either within the stratum oriens or input to CA1 pyramidal cells. The projection terminates on
through the deep portion of the stratum radiatum. the basal dendrites in the stratum oriens and the apical den-
The position of the terminal eld in CA1 varies in a sys- drites in the stratum radiatum. Individual CA3 axons distrib-
tematic fashion relative to the distance from the cells of origin. ute extensively and may innervate neurons throughout as
Regardless of the septotemporal or transverse origin of a pro- much as two-thirds of the entire septotemporal extent of the
jection, the highest density of terminal and ber labeling in hippocampus. The probability that a particular CA1 cell is
CA1 shifts to deeper parts of the stratum radiatum and stra- contacted by a particular CA3 cell depends, in part, on the
tum oriens at levels septal to the cells of origin and shifts away transverse positions of the two cell bodies and their septotem-
from the stratum oriens and into supercial parts of the poral distance. At one particular septotemporal level, a distal
stratum radiatum at levels temporal to the cells of origin. CA3 cell is more likely to interact with a proximal CA1 cell,
Moreover, the highest density of ber and terminal labeling in whereas a proximal CA3 cells is more likely to interact with a
CA1 shifts proximally (toward CA3) at levels septal to the ori- distal CA1 cell. The proximal CA3 cells are the only ones with
gin and distally (toward the subiculum) at levels temporal to classic thick Schaffer collaterals, and the thickness of these ini-
the origin. tial axons is likely to reect the longer distance the axon must
The entire axonal plexus of several CA3 pyramidal cells travel to innervate distal CA1 cells.
have been labeled by intracellular staining techniques devised
by Buzsaki and colleagues (Li et al., 1994). These studies CA3 to CA3 Associational Connections
provide convincing evidence that the axons of individual The associational projections from CA3 to CA3 are also
CA3 cells can distribute to as much as two-thirds of the organized in a highly systematic fashion. One somewhat idio-
septotemporal extent of the ipsilateral and contralateral CA1 syncratic facet of this projection is that cells located proxi-
elds. The plexus from a single CA3 neuron comprises as mally in CA3 communicate only with other proximally
much as 150 to 300 mm of total axonal length, on which located CA3 cells. Associational projections arising from mid
30,000 to 60,000 synaptic varicosities are formed. Although and distal portions of CA3, however, project throughout
single neurons have not been evaluated at all septotemporal much of the transverse extent of CA3 and also project much
levels, it appears that the extent of the CA3 connections is more extensively along the septotemporal axis. The density of
more restricted at temporal levels. Here neurons may give rise CA3 associational projections also shifts along the septotem-
to axonal plexuses that innervate only the temporal third of poral axis. The radial gradient of termination (supercial to
the CA1 eld. deep in the stratum radiatum and stratum oriens) is similar to
A number of pieces of fundamental information concern- that described for the CA3 to CA1 projection. The transverse
ing the CA3 projection to CA1 are still unknown. For exam- gradient, however, is the reverse; CA3 projections shift proxi-
ple, it is not clear how many synapses a single CA3 cell makes mally in CA3 at levels located temporal to the cells of origin
on a typical CA1 cell. To answer this question using neu- and shift distally in CA3 at more septal levels.
roanatomical procedures would be a monumental task. First,
a CA3 and CA1 neuron would have to be impaled with dye- Commissural Connections of the Hippocampus
bearing pipettes, and the two cells would have to be tested for In the rat, the CA3 pyramidal cells give rise to commis-
connectivity by electrophysiological methods. The two cells sural projections to the CA3, CA2, and CA1 elds of the con-
would then have to be lled and processed histologically for tralateral hippocampal formation (Blackstad, 1956; Fricke
visualization. That is the easy part. The neurons would then and Cowan, 1978). In fact, the same CA3 cells give rise to
have to be sectioned serially for electron microscopy and all both the ipsilateral and commissural projections. Although
putative synapses evaluated. Only in this way would one be the commissural projections follow roughly the same topo-
able to determine how many synapses are made by all of the graphical organization as the ipsilateral projections and
collaterals of a particular CA3 neuron on an identied CA1 generally terminate in homologous regions on both sides,
pyramidal cell. A number of laboratories have estimated the there are minor differences in the distribution of terminals.
extent of connectivity between CA3 and CA1 cells, and the If a projection is heavier to the stratum oriens on the ipsilat-
numbers are always quite low. Work from Harris and col- eral side, for example, it may be heavier in the stratum radia-
leagues (Sorra and Harris, 1993), which has been replicated by tum on the contralateral side. The detailed topography of
Trommaid et al. (1996), indicates that there are perhaps as few the commissural connections has not been as thoroughly
as two to four contacts between a single axon in the stratum investigated as the ipsilateral connections. As with the com-
radiatum and a particular dendritic tree and certainly no missural projections from the dentate gyrus, CA3 bers to the
more than 10 synapses between typical CA3 and CA1 neu- contralateral hippocampus form asymmetrical synapses
rons. The surprising state of affairs, however, is that we simply on the spines of pyramidal cells in CA3 and CA1 but also
do not know with certainty how many contacts a single CA3 terminate on the smooth dendrites of interneurons. As noted
neuron makes with a single CA1 cell. These data are critical earlier, hippocampal commissural connections are much less
for interpreting some of the quantal analysis studies described abundant in the monkey and are likely to be absent in
in Chapter 10. humans.
76 The Hippocampus Book

CA1 Lacks Associational Connections

Unlike CA3, pyramidal cells in CA1 give rise to a much more
limited associational projection. As the CA1 axons travel in
the alveus or the stratum oriens toward the subiculum (see
below), occasional collaterals are generated that enter the stra-
tum oriens and the pyramidal cell layer, most likely contacting
interneurons such as basket cells in the stratum oriens and in
turn inhibiting CA1 pyramidal cells. It is also conceivable that
these collaterals might contact the basal dendrites of other
CA1 cells. What is clear, however, is that the massive associa-
tional network, which is so apparent in CA3, is largely missing
in CA1. Similarly, although a weak commissural projection to
the contralateral CA1 appears to be present, there is no exten-
sive commissural projection originating in CA1, as is the case
in CA3. Figure 331. Organization of the projections from the CA1 eld of
the hippocampus to the subiculum. A. Coronal section shows the
cells of origin (triangles) and their respective terminal elds, indi-
CA1 Projection to the Subiculum
cated by similar shades of gray. B. Two-dimensional unfolded map
of CA1 and the subiculum. The septal portion of these elds is at
The CA1 eld gives rise to two intrahippocampal projections.
the top of the gure, and the temporal pole is at the bottom; the
The rst is a topographically organized projection to the adja-
border between CA2 and CA1 is on the left.
cent subiculum. The second is to the deep layers of the
entorhinal cortex. The latter projection is discussed in Section
3.4.5. CA1 (dened as the region that receives CA3 projections) also
Axons of CA1 pyramidal cells descend into the stratum ends at this border and is replaced with the molecular layer of
oriens or the alveus and bend sharply toward the subiculum the subiculum. This layer can be subdivided into a deep por-
(Amaral et al., 1991). The bers reenter the pyramidal cell tion that is continuous with the stratum radiatum of CA1 and
layer of the subiculum and ramify profusely in the pyramidal a supercial portion that is continuous with the stratum
cell layer and the deep portion of the molecular layer. Unlike lacunosum-moleculare of CA1 and the molecular layer of the
the CA3 to CA1 projection, which distributes throughout CA1 presubiculum. The deep portion receives bers from CA1,
in a gradient fashion, the CA1 projection ends in a topo- whereas the supercial portion receives bers from the
graphical fashion in the subiculum. Proximally located CA1 entorhinal cortex. The stratum oriens of CA1 is no longer
cells project to the distal portion of the subiculum, whereas present in the subiculum.
distally located CA1 cells project just across the border into
the proximal portion of the subiculum; the midportion of Neuron Types
CA1 projects to the midportion of the subiculum (Fig. 331).
Nobuaki Tamamaki (Tamamaki et al., 1987) injected horse- Research on the cytology and connectivity of the subiculum
radish peroxidase into single CA1 pyramidal cells and demon- has lagged behind the research on other areas of the hip-
strated that individual axonal plexuses distribute to about pocampal formation; and until recently (Funahashi and
one-third of the transverse extent of the subicular pyramidal Stewart, 1997a,b; Harris et al., 2001) much of the information
cell layer and for approximately one-third of the septotempo- on subicular cell types came from the classic Golgi studies of
ral length. It appears, therefore, that the CA1 to subiculum Ramon y Cajal and Lorente de N in young mice. The princi-
projection segments these structures roughly into thirds. The pal cell layer of the subiculum is populated by large pyramidal
CA1 to subiculum projection, like the CA3 to CA1 projection, neurons (Fig. 332). This layer starts just underneath the dis-
is organized in a divergent fashion, as even a small portion of tal end of CA1 and continues in a position deep to layer II of
CA1 projects to about a third of the septotemporal extent of the presubiculum. These cells are relatively uniform in shape
the subiculum. and size, and they extend their apical dendrites into the
molecular layer and their basal dendrites into deeper portions
3.4.3 Subiculum of the pyramidal cell layer. A subdivision into at least two cell
types has been proposed based on their ring characteristics:
Cytoarchitectonic Organization regular spiking cells and intrinsically bursting cells (Greene
and Totterdell, 1997). Although these two cell types do not
The cytoarchitectonic characteristics of the rat subiculum exhibit distinct morphological characteristics (but see below),
have been studied to a limited extent (OMara et al., 2001). they show a differential distribution in the pyramidal cell
The CA1/subiculum border is marked by abrupt widening of layer. Bursting cells are more numerous deep in the pyramidal
the pyramidal cell layer and an increased staining intensity cell layer, whereas regular spiking cells are more common
with the Timm stain (Fig. 311). The stratum radiatum of supercially. The two populations can also be distinguished by
 Hippocampal Neuroanatomy 77

Figure 332. Neurolucida reconstructions of supercial and deep subicular cells. Somata and
dendrites are shown in black. Axons are shown in gray. Local collaterals tend to be longer in the
cell layer for supercially located cells; deep cells tend to have axon collaterals that ascend close
to the primary dendrite. CA1 is at the left, and the presubiculum is at the right. Bar  500 m.
(Source: Adapted from Harris et al., 2001.)

preferential staining for either NADPH-diaphorase/nitric density of this local intrinsic connection, as estimated by the
oxide synthetase (regular spiking neurons) or somatostatin number of varicosities on locally distributed axon collaterals,
(bursting cells). Both cell types are projection neurons, but is much higher than in CA1. In addition, the two types of
they might differ with respect to their connectivity because subicular pyramidal neurons differ with respect to their local
there is evidence suggesting that only bursting cells project to connectivity. Intracellular labeling of electrophysiologically
the entorhinal cortex. Intermingled among the pyramidal cells identied bursting cells generally show an axonal distribution
are many smaller neurons, presumably representing the that remains in the region circumscribed by their apical den-
interneurons of the subiculum. Little is known, however, drites (i.e., a columnar organization), whereas the regular
about whether the interneurons seen in the subiculum are spiking cells generally give rise to an axon that shows more
similar to those observed in the hippocampus. Subpopula- widespread distribution along the transverse axis (Fig. 333).
tions of these neurons appear to have characteristics similar to It is not known whether differences exist with respect to pos-
those described for CA1; among these subpopulations are sible septotemporal spread. Although much work remains to
GABAergic cells that stain for the calcium-binding protein be done, available data indicate that the organization of the
parvalbumin. intrinsic connectivity of the subiculum is different from that
of CA3 and CA1. There is both crude columnar and laminar
Intrinsic Connections organization, such that the bursting cells form a set of
columns and the regular spiking neurons integrate columnar
The subiculum gives rise to a longitudinal associational pro- activity along the transverse axis.
jection that extends from the level of the cells of origin to
much of the subiculum lying temporally (or ventrally). Extrinsic Connections: Subiculum,
Interestingly, this projection seems to be largely unidirec- a Major Output Structure
tional, as few if any associational projections course septally or
dorsally from their point of origin (Harris et al., 2001). The The subiculum is a major source of efferent projections from
associational bers terminate diffusely in all layers of the the hippocampal formation. Following the discovery by
subiculum. Recent studies have consistently found that the rat Swanson and Cowan that the subiculum, rather than the
subiculum neither gives rise to nor receives commissural con- hippocampus, is the origin of the major subcortical connec-
nections. tions to the diencephalon and brain stem (via the postcom-
Subicular pyramidal cells provide a strong local input in missural fornix), evidence has mounted that the subiculum is
the pyramidal cell layer and just supercial to it, targeting one of the two primary output structures of the hippocampal
proximal portions of the apical dendrites. Interestingly, the formation (Swanson and Cowan, 1975; Swanson et al., 1981;
78 The Hippocampus Book

collective hunch was that bers simply passed through the

subiculum and probably did not terminate in it. Earlier
anterograde tracer studies indicated that perforant path bers
are directed toward the molecular layer of the subiculum, but
proof that these bers formed a terminal plexus among the
subicular pyramidal cells was lacking. Witter and colleagues,
however, have provided evidence, at both light and electron
microscopic levels, that the subiculum receives a strong pro-
jection from the entorhinal cortex. The topography of the
projection is similar to that described for CA1. Fibers are
directed toward restricted transverse portions of the subicu-
lum and terminate in the outer two-thirds of the molecular
layer. The lateral component of the perforant pathway prefer-
entially projects to the proximal part of the subiculum (i.e.,
the part of the subiculum that borders CA1), and the medial
Figure 333. Model of intrinsic organization of the subiculum. component distributes to more distal portions of the subicu-
Deep cells have ascending axon collaterals that remain in close prox- lum (i.e., closer to the presubiculum). The projection origi-
imity to their apical dendrites, giving rise to a columnar pattern for nates mainly from layer III, although some of the axons of
the deep cells. Supercial cells have axon collaterals that can run layer II cells that cross the subiculum on their way to the den-
long distances in the cell layer perhaps serving to integrate activity
tate gyrus and CA3 may also give off collaterals that terminate
between columns. (Source: Adapted from Harris et al., 2001.)
in the subiculum. Entorhinal bers target dendritic spines of
presumed principal neurons with asymmetrical synapses
(80%), whereas 5% to 10% of the asymmetrical synapses ter-
Donovan and Wyss, 1983; Groenewegen et al., 1987; Witter
minate on dendritic shafts, most likely belonging to interneu-
and Groenewegen, 1990; Witter et al., 1990; Canteras and
rons. Subicular interneurons may also receive a minor
Swanson, 1992; Naber and Witter, 1998; Ishizuka, 2001;
inhibitory perforant path input in view of the symmetrical
Kloosterman et al., 2003).
synapses onto dendritic shafts.
The subiculum reciprocates the entorhinal input (Fig.
Subiculum Projects to the Presubiculum and Parasubiculum 334). Projections from the subiculum reach all parts of the
The subiculum projects to the presubiculum, but this projec- entorhinal cortex and terminate in the deep layers; termina-
tion is not of the magnitude of other intrinsic hippocampal tion is particularly dense in layer V. A minor component of the
formation projections. It is probably better to think of the subicular projection also extends supercial to the lamina dis-
subiculum projection to the presubiculum as one of a series of secans, predominantly to layer III. Subicular bers generally
pathways that distributes information processed in the den- form asymmetrical synapses with spines and dendrites located
tate gyrus, hippocampus, and subiculum to a series of cortical in these layers. A small number of these bers form symmet-
and subcortical structures. The subiculum can be viewed, rical synapses, suggesting small inhibitory input from the
therefore, as the last step in the large loop of processing subiculum to layer V. Therefore, the overall organization of
through the hippocampal formation. the subiculo-entorhinal projection mimics that of the CA1-
Subicular bers terminate mainly in layer I of the pre- entorhinal projection. The presence of asymmetrical synapses
subiculum. The projection to the dorsal part of the pre- at the termination of this pathway is consistent with its
subiculum, however, terminates deep to the prominent layer reported excitatory inuences on the entorhinal cortex.
II, with weaker projections to layers I and II. The projections
from the subiculum to the parasubiculum mainly terminate in Subicular Connections with the Neocortex
layer I and the supercial portion of layer II. These projections and Amygdaloid Complex
are topographically organized. Septal (or dorsal) portions of The subiculum gives rise to prominent projections to the
the subiculum project to dorsal and caudal aspects of both the medial and ventral orbitofrontal cortices and to the prelimbic
presubiculum and the parasubiculum, and temporal (or ven- and infralimbic cortices (Verwer et al., 1997). In the orbito-
tral) portions of the subiculum project to ventral and rostral frontal and prelimbic cortices subicular bers mainly inner-
aspects of the pre- and parasubiculum. vate the deep layers, whereas in the infralimbic cortex the
projection extends into the supercial layers. Subicular pro-
Subiculum Reciprocally Connected with the Entorhinal Cortex jections also reach medial portions of the anterior olfactory
Because the perforant path bers traverse the subiculum on nucleus.
their way to the dentate gyrus and hippocampus, the question The subiculum provides a meager projection to the ante-
of whether some of these bers might terminate in the subicu- rior cingulate cortex, whereas the projection to the retrosple-
lum has long been a matter of controversy. Until recently, the nial cortex is substantial (Wyss and Van Groen, 1992). The
 Hippocampal Neuroanatomy 79

Figure 334. Topographical organization of the subicular projec- tion of the subiculum projects to medial portions of the lateral and
tions to the parahippocampal region (entorhinal, perirhinal, and medial entorhinal cortex. B. Relation between the proximo-distal
postrhinal cortices). A. Relation between the septotemporal origin origin in the subiculum with a rostrocaudal termination in the
in the subiculum with a lateral-to-medial termination in the parahippocampal region. The proximal subiculum projects to the
parahippocampal region. The septal portion of the subiculum proj- perirhinal and lateral entorhinal cortex; the distal subiculum proj-
ects to the perirhinal and postrhinal cortices and the lateral portion ects to the postrhinal and medial entorhinal cortex. (Source:
of both the lateral and medial entorhinal cortex. The temporal por- Adapted from Kloosterman et al., 2003.)

subiculo-retrosplenial projection terminates predominantly subiculum, but this projection terminates only in the proxi-
in layers II and III. The projections to the retrosplenial cortex mal third of the eld (i.e., at the border region with CA1).
originate predominantly from the septal two-thirds of the The proximal portion of the subiculum also receives an
subiculum. The perirhinal cortex receives a strong input from input from the parvicellular portion of the basal nucleus of
the subiculum, which terminates in both supercial and deep the amygdaloid complex, the posterior cortical nucleus, and
layers. the adjacent amygdalohippocampal area. These amygdaloid
In the rat, there is a paucity of detailed information regard- inputs terminate mainly at the CA1/subiculum border region,
ing direct cortical inputs to the subiculum. Many of the corti- where they preferentially innervate the molecular layer of the
cal regions that project fairly heavily to the entorhinal cortex subiculum and stratum lacunosum-moleculare of CA1
(see below) do not project to the subiculum. No inputs, for (Pitkanen et al., 2000).
example, have been reported from the pre- and infralimbic The temporal one-third of the subiculum gives rise to
cortices. Similarly, no portion of the retrosplenial cortex proj- return projections to the amygdaloid complex. The major
ects to the subiculum. Reports of projections from the cingu- component of this projection terminates in the accessory
late cortex have been somewhat contradictory. In a combined basal nucleus, with more moderate projections reaching sev-
electrophysiological and neuroanatomical study, White et al. eral other nuclei but not the lateral nucleus. The ventral
(1990) reported a projection from the anterior cingulate cor- subiculum also projects heavily to the bed nucleus of the stria
tex to the subiculum. However, this projection has not been terminalis and moderately to the ventral part of the claustrum
observed consistently. The perirhinal cortex projects to the or endopiriform nucleus.
80 The Hippocampus Book

Basal Forebrain Connections: Septal Nucleus Interestingly, the projections to the subiculum and to CA1
and Nucleus Accumbens originate from different but intermingled populations of
The most prominent subcortical subicular projections are neurons in the nucleus reuniens. Although earlier studies
those to the septal complex, the adjacent nucleus accumbens, had indicated that the subiculum was interconnected with
and the mammillary nuclei. The projection to the septal area the anterior nuclear complex, it is now clear that the ante-
terminates predominantly in the lateral septal nuclei. Closely rior nucleus projects almost exclusively to the pre- and para-
associated with the septal projection is the equally robust subiculum.
projection to the nucleus accumbens and adjacent portions Some of the thalamic regions that project to the subiculum
of the olfactory tubercle. Subicular bers terminate through- receive a return projection from the subiculum. Subicular
out the nucleus accumbens, with the projection to its caudo- bers terminate bilaterally in the nucleus reuniens, the
medial part being most dense. As with other striatal nucleus interanteromedialis, the paraventricular nucleus, and
structures, the subicular projection to the nucleus accumbens the nucleus gelatinosus. Although subicular projections to
is unidirectional. Whereas the subicular projections to the parts of the anterior thalamic complex have been described in
lateral septal nucleus are almost entirely conned to the ipsi- the literature, more recent retrograde tracing studies have
lateral side, those to the nucleus accumbens show a weak con- shown that these projections arise from the presubiculum.
tralateral component. The subiculum receives a relatively
weak cholinergic projection from the septal complex; bers Brain Stem Inputs
originating from the medial septal nucleus and the nucleus of Monoaminergic ascending pathways from the noradrenergic
the diagonal band terminate in the pyramidal cell and molec- locus coeruleus, the dopaminergic ventral tegmental area, and
ular layers. the serotonergic median and dorsal raphe nuclei reach the
subiculum, but they do not show preferential innervation of
Hypothalamic Connections: Mammillary Nuclei this region. Few details are available concerning the regional
The subiculum provides the major input to the mammillary localization of these pathways in the subiculum.
nuclei. The projection is heavy and is distributed bilaterally in
nearly equal density. The subiculomammillary bers originate Topography of Subicular Efferent Projections
mainly from the septal two-thirds of the subiculum. Although As with the projections from CA3 to CA1 and from CA1 to the
the temporal one-third of the subiculum also contributes to subiculum, the subicular efferent projections are topographi-
the mammillary projection, the major hypothalamic target of cally organized. In large part, the subicular projections pre-
this portion of the subiculum is the ventromedial nucleus of serve the transverse topography established by the CA1 to
the hypothalamus. The subicular projections to the mammil- subiculum projection. It is clear that different projections
lary nuclei reach all portions of the medial nucleus but are originate from at least the proximal and distal halves of the
topographically organized; the lateral mammillary nucleus is subiculum. The subiculum also demonstrates marked sep-
only sparsely innervated by the subiculum. Subicular bers totemporal topography, such that the projections that arise
also project to the lateral hypothalamic region located adja- from the septal or dorsal two-thirds of the subiculum are dif-
cent to the lateral mammillary nucleus. ferent from those that arise from the temporal or ventral third
Whereas the subiculum does not receive a return projec- (Witter and Amaral, 2004).
tion from the medial or lateral mammillary nuclei, the supra- Turning rst to the septotemporal topography, it appears
mammillary region projects heavily to the subiculum, that the projections to the entorhinal cortex, the lateral septal
particularly to its temporal levels. This portion of the subicu- complex, the nucleus accumbens, and the medial mammillary
lum also receives an input from the premammillary nucleus. nucleus originate from the entire septotemporal extent of the
It is not clear whether there are any local connections between subiculum (Witter and Groenewegen, 1990; Ishizuka, 2001).
the medial mammillary nucleus and the supramammillary Different septotemporal levels of the subiculum, however,
area or the premammillary nucleus that might complete the project to different portions of these elds. In the entorhinal
subiculohypothalamic loop. cortex, for example, the septal-to-temporal origin in the
subiculum is related to a lateral-to-medial termination in the
Thalamic Connections: Nucleus Reuniens entorhinal cortex. Septal levels of the subiculum project pref-
and Other Midline Nuclei erentially to lateral and caudal parts of the entorhinal cortex
The thalamic inputs to the subiculum are similar to those to (i.e., the parts that lie adjacent to the rhinal sulcus).
CA1. Thalamic inputs originate mainly in the nucleus Progressively more temporal levels of the subiculum project to
reuniens, the paraventricular nucleus, and the parataenial more medially located parts of the entorhinal cortex.
nucleus. The septal and temporal extremes of the subiculum Although addressed in more detail below, it is important to
appear to be devoid of input from the nucleus reuniens. The point out that this topography is completely in register (i.e.,
midline thalamic projections terminate mainly in the molec- they are point-to-point reciprocal) with the projections from
ular layer of the subiculum, whereas in CA1 they are coexten- the entorhinal cortex to the subiculum. Thus, cells in the
sive with the projections from the entorhinal cortex. subiculum that receive input from a subregion in the entorhi-
 Hippocampal Neuroanatomy 81

nal cortex give rise to a return projection to the same region ization. Thus, the proximal portions of the subiculum project
in the entorhinal cortex. In some respect this breaks the rule to rostral medial mammillary nuclei, and distal portions of
that all structures in the hippocampal formation have unidi- the subiculum project more caudally. A similar situation exists
rectional connectivity, but other aspects of the neuroanatomy for the subicular projections to the entorhinal cortex. The
of these areas completely justify their inclusion in the func- proximal half of the subiculum projects to the lateral entorhi-
tionally dened hippocampal formation. nal area, and the distal half of the subiculum projects to the
In the nucleus accumbens, the septotemporal axis of origin medial entorhinal area.
in the subiculum determines a caudomedial to rostrolateral Because the proximal third of the subiculum gives rise to
axis of termination. Dorsomedial portions of the lateral septal projections to at least several cortical and subcortical regions,
complex receive inputs from septal levels of the subiculum, the question arises as to whether it is the same or different
and ventral portions of the lateral septal complex are inner- populations of subicular cells that innervate each structure.
vated by bers originating in more temporal parts of the The answer initially appeared to be that projections to the sep-
subiculum. tal complex, entorhinal cortex and mammillary complex arise,
Similar septotemporal topography has also been described at least in part, as collaterals from single subicular neurons;
for the subicular projections to the presubiculum and the but it now appears that largely independent populations of
medial mammillary nuclei. The latter projection arises mainly intermixed neurons in the subiculum project to each of its ter-
from the septal two-thirds of the subiculum, whereas the ven- minal regions (Naber and Witter, 1998). Given our earlier
tral one-third gives rise to projections to other hypothalamic assertion that the subiculum is the last staging post of hip-
regions such as the ventromedial nucleus. pocampal processing, this state of affairs seems to create the
This dichotomy between the septal two-thirds and the possibility that the outputs destined for different target struc-
temporal one-third of the subiculum is reected in the organ- tures can be carrying distinctly different information.
ization of other projections. Projections to the amygdala and
the bed nucleus of the stria terminalis, for example, originate 3.4.4 Presubiculum and Parasubiculum
exclusively from the temporal one-third of the subiculum,
whereas projections to the retrosplenial and perirhinal cor- Cytoarchitectonic Organization and Neuron Types
tices originate predominantly from the septal two-thirds. The
subicular projections to the midline thalamus demonstrate The presubiculum, Brodmanns area 27, is relatively easily
even greater septotemporal topography. The most septal part differentiated from the subiculum in standard Nissl-stained
of the subiculum projects preferentially to the interanterome- material. It has a distinct, densely packed external cell layer
dial nucleus; mid-septotemporal levels of the subiculum pref- that consists mainly of darkly stained, small pyramidal cells.
erentially project to the nucleus reuniens; and the temporal The most supercial cells are the most densely packed (layer
third of the subiculum projects most heavily to the paraven- II), whereas the deeper cells have a somewhat looser arrange-
tricular nucleus. ment (layer III). The differentiation between layers II and III
Whereas the septotemporal topography appears to be is more clear-cut at dorsal levels of the presubiculum.
organized in a gradient, or gradual, fashion, the transverse The dorsal presubiculum (sometimes called the post-
organization of subicular efferents is remarkably discrete. subiculum) has clearly distinguishable supercial and deep
Along the transverse axis of the subiculum, two essentially cell layers. In the ventral portion of the presubiculum, how-
nonoverlapping populations of cells can be differentiated that ever, the deep layers are difficult to distinguish from the deep
give rise to projections to specic sets of brain structures. This layers of the entorhinal cortex or from the principal cell layer
transverse organization of the outputs of the subiculum is of the subiculum. Deep to the lamina dissecans there are one
consistently observed along its entire septotemporal axis, or two layers of large, darkly stained pyramidal cells; and deep
although it is clearer septally than temporally. to these cells is a rather heterogeneous collection of pyramidal
Neurons in the proximal half of the subiculum (closest to and polymorphic cells. The latter cells have not been studied
CA1) project to the infralimbic and prelimbic cortices, the anatomically in great detail (Fig. 335), although electrophys-
perirhinal cortex, the nucleus accumbens, the lateral septum, iological discoveries about their receptive eld characteristics
the amygdaloid complex, and the core of the ventromedial are important to one prominent theory of hippocampal func-
nucleus of the hypothalamus. Cells in the distal half of the tion (see Chapter 8).
subiculum project mainly to the retrosplenial cortex and the The parasubiculum (Brodmanns area 49) lies adjacent to
presubiculum. Cells projecting to the midline thalamic nuclei the presubiculum. Layers II and III of the parasubiculum con-
are mainly located in the midportion of the subiculum. sist of rather densely packed, lightly stained, large pyramidal
The subicular projections to the entorhinal cortex and to cells. This and other characteristics, such as the distinctive
the medial mammillary nucleus do not follow a strictly trans- staining for heavy metals observable with the Timms stain
verse organization, as cells in all proximo-distal portions of method, are the major features that differentiate the para-
the subiculum project to these areas. However, the topography subiculum from the presubiculum (Fig. 311). There is no
of these projections indicates a more subtle transverse organ- clear differentiation between layers II and III; and as with the
82 The Hippocampus Book

Commissural projections from the dorsal portion of the pre-

subiculum are relatively sparse. The parasubiculum also gives
rise to a commissural projection that terminates most densely
in layers I and III.

Extrinsic Connections

Presubiculum and Parasubiculum Reciprocally

Connected with Anterior Thalamic Nuclei
The presubiculum and parasubiculum receive a number of
subcortical inputs, but the one that is unique to these portions
of the hippocampal formation is their interconnection
with the anterior thalamic nuclear complex, primarily the
anteroventral and anterodorsal nuclei and the closely related
laterodorsal nucleus (Kaitz and Robertson, 1981; Robertson
and Kaitz, 1981). Although earlier studies had suggested that
the presubiculum also receives inputs from the anteromedial
Figure 335. Cell types of the presubiculum and parasubiculum. nucleus, the principal projections of this nucleus appear to be
Camera lucida drawings of neurobiotin-lled cells in a horizontal directed to the anterior parts of the cingulate cortex rather
section through the hippocampal formation. A. Pyramidal cell in than to the presubiculum or parasubiculum.
layer II of the parasubiculum. B. Stellate cell in layer II of the pre- There are topographical differences in the thalamic con-
subiculum. C. Pyramidal cell in layer III of the presubiculum. nections of the ventral and dorsal portions of the presubicu-
D. Stellate cell in layer V of the presubiculum. E. Pyramidal cell
lum. The ventral portion receives most of its input from the
in layer V of the presubiculum. F. Pyramidal cell in layer V of the
laterodorsal and anteroventral nuclei, whereas the dorsal
parasubiculum. G. Stellate cell in layer V of the parasubiculum.
Bar  100 m. (Source: Adapted from Funahashi and Stewart,
part receives projections mainly from the laterodorsal and
1997a.) anterodorsal nuclei. The thalamic projections mainly termi-
nate in layers I, III, and IV. The nucleus reuniens also projects
to layer I of the presubiculum, but this is a much lighter pro-
presubiculum, the deep layers are continuous with those of jection.
the entorhinal cortex. Like most other thalamocortical connections, the pre-
subiculum sends a return projection back to the thalamus.
Intrinsic Connections The presubiculum projects massively and bilaterally to the
anterior nuclear complex. The presubicular projections arise
There are well developed associational connections in the pre- mainly from cells located deep to the lamina dissecans (layer
subiculum that interconnect all dorsoventral levels in a highly VI). The fact that these deep cells are related to the anterior
directional manner. Layer II cells in ventral portions of the nuclei of the thalamus lends some credibility to the idea that
presubiculum project to more dorsal levels of the presubicu- they are, in fact, deep layers of the pre- and parasubiculum
lum. Projections in the opposite direction (dorsal to ventral), rather than of the entorhinal cortex.
however, arise preferentially from cells in the deep layers.
The parasubiculum also gives rise to associational connec- Presubiculum Projection to Layer III of the Entorhinal Cortex
tions that distribute both dorsally and ventrally from the cells The most prominent intrahippocampal projection from the
of origin. The dorsally directed projections extend for short presubiculum is to the entorhinal cortex (Shipley, 1975;
distances and are quite weak. The ventrally directed projec- Caballero-Bleda and Witter, 1993). This projection has a
tions are substantially denser and extend for long distances number of interesting features. First, the presubicular projec-
along the long axis of the parasubiculum. The parasubiculum tion is directed only to the medial entorhinal area. Second, the
gives rise to a particularly dense projection to the dorsally projection terminates almost exclusively in layer III and to a
located area 29e. This small wedge-shaped region was initially much lesser extent in the deep part of layer I. Third, the
described by Blackstad (1956) and later characterized in more crossed homotopic projection from the presubiculum to the
detail by Haug (1976). Although the neuroanatomy of area contralateral entorhinal cortex is every bit as dense as the ipsi-
29e remains sketchy, it might be part of the parasubiculum. lateral projection. The projection to the entorhinal cortex is
topographically organized. The location of the presubicular
Commissural Connections terminal eld in the entorhinal cortex is determined by both
the proximo-distal and dorsoventral location of the presubic-
The presubiculum gives rise to strong commissural projec- ular cells of origin.
tions to layers I and III of the homotopic part of the con- The parasubiculum selectively innervates layer II of the
tralateral presubiculum (Van Groen and Wyss, 1990). entorhinal cortex. In contrast to the presubiculum, the para-
 Hippocampal Neuroanatomy 83

subiculum projects to both the medial and lateral entorhinal a radial manner, which is quite distinct from the predomi-
areas, although the projection to the lateral entorhinal area is nantly transverse orientation of the entorhinal perforant
less robust. The parasubiculum projects to the contralateral pathway bers.
entorhinal cortex, but these projections are much weaker than A fact that has not been generally appreciated is that the
the ipsilateral ones. The topographical organization of the parasubiculum gives rise to a fairly substantial projection to
parasubicular projection to the entorhinal cortex is compara- the molecular layer of the dentate gyrus (Kohler, 1985). Like
ble to that of the presubiculo-entorhinal projection. the lighter projection from the presubiculum, this projection
occupies the supercial two-thirds of the molecular layer
Presubiculum and Parasubiculum Connected (with a preference for the midportion of the molecular layer),
with Some Neocortical Regions and the bers have a predominantly radial orientation.
The presubiculum receives relatively few extrahippocampal Because the parasubiculum receives a projection from the
cortical inputs. The most prominent one originates in the ret- anterior thalamic nuclei, its projection to the molecular layer
rosplenial cortex (Van Groen and Wyss, 1990; Wyss and Van provides a route by which thalamic input might inuence the
Groen, 1992). Cells located in layer V of the retrosplenial cor- very early stages of hippocampal information processing. The
tex give rise to projections that terminate in layers I and III/V parasubiculum projects weakly to the stratum lacunosum-
of the presubiculum. A second cortical input originates from moleculare of the hippocampus and to the molecular layer of
layer V of the visual area 18b. This projection mainly distrib- the subiculum. The parasubiculum also projects bilaterally to
utes to the dorsal half of the presubiculum and terminates in layers I and III of the presubiculum.
layers I and III. Minor cortical inputs originate in the prelim-
bic cortex and in a dorsal portion of the medial prefrontal Basal Forebrain Connections
cortex. The presubiculum and parasubiculum receive heavy choliner-
Extrahippocampal projections from layer V of the pre- gic input. The medial septal nucleus and the vertical limb of
subiculum reach the granular retrosplenial cortex, where they the diagonal band of Broca mainly innervate layer II of the
terminate preferentially in layers I and II. These projections presubiculum.
are topographically organized, such that the ventral pre-
subiculum projects mainly to the ventral part of the granular Hypothalamic Connections: Mammillary Nuclei
retrosplenial cortex, and the dorsal part of the presubiculum The deep layers of the presubiculum project bilaterally to the
projects more dorsally. These projections also exhibit a rostro- medial and lateral mammillary nuclei. The projections to the
caudal organization, such that rostral portions of the pre- medial mammillary nuclei are topographically organized in a
subiculum project to rostral parts of the retrosplenial cortex, manner similar to those that originate in the subiculum
and caudal portions of the presubiculum project to caudal (Thompson and Robertson, 1987; Allen and Hopkins, 1989;
parts of the retrosplenial cortex. It has been suggested that the Van Groen and Wyss, 1990).
dorsal presubiculum projects to the deep layers of the most The presubiculum receives input from the area surround-
caudal portion of the perirhinal cortex, although an alterna- ing the mammillary nuclei. Fibers from the supramammillary
tive interpretation is that this projection is directed to the nucleus terminate preferentially in the deeper cell layers of the
most caudodorsal portion of the medial entorhinal cortex, presubiculum, although those characterized as being positive
which is difficult to differentiate from the caudal perirhinal for -melanocyte-stimulating hormone (an opiate peptide
cortex. The work of Burwell et al. (1998a,b) supports the expressed in neurons whose somata are located in the lateral
latter idea. hypothalamic area) terminate in the molecular layer.
With the exception of the relatively light projections from
the retrosplenial cortex and the occipital visual cortex, there Brain Stem Inputs
are no other known extrahippocampal cortical inputs to the The presubiculum receives input from various nuclei in the
parasubiculum. The laminar distribution of these inputs is brain stem. A particularly dense innervation arises from the
similar to that described for the presubiculum. dorsal and ventral raphe nuclei; at least a component of this
projection is serotonergic and innervates layer I. The nora-
Other Intrahippocampal Connections drenergic locus coeruleus innervates the plexiform layer.
The presubiculum projects to layers I and II of the para-
subiculum bilaterally. Anterograde tracing studies with the 3.4.5 Entorhinal Cortex
lectin tracer PHA-L have demonstrated that the presubiculum
and perhaps to a greater extent the parasubiculum contribute The entorhinal cortex plays an extraordinarily important role
projections, albeit modest ones, to many of the other regions in the ow of information through the hippocampal forma-
of the hippocampal formation. For example, there is a modest tion. It is not only the main entry point for much of the sen-
bilateral projection from the presubiculum to the subiculum. sory information processed by the hippocampal formation, it
There is also a weak projection to all elds of the hippocam- provides the main conduit for processed information to be
pus and to the molecular layer of the dentate gyrus. The pre- relayed back to the neocortex. As portrayed in this chapter, the
subicular bers to the dentate molecular layer are arranged in entorhinal cortex is the beginning and the end point of an
84 The Hippocampus Book

extensive loop of information processing that takes place in layer that can be subdivided into bands, and layer Va, which
the hippocampal formation. Although neuroanatomical forms a band of large, darkly stained pyramidal neurons and
investigation of the entorhinal cortex has historically lagged is most conspicuous in the central parts of the entorhinal cor-
behind the work conducted in other elds of the hippocampal tex. At other levels, the packing density of cells is not high, and
formation, many new ndings have been forthcoming in the smaller cells of the deeper part of this layer (Vb) inter-
recent years. This progress has been spurred on, in part, by the mingle with it. Finally there is layer VI, containing a highly
appreciation, initially by van Hoesen et al. (1991), that the heterogeneous population of cell sizes and shapes. This cell
entorhinal cortex is a site of early, devastating pathology in density decreases toward the border with the white matter.
degenerative diseases such as Alzheimers disease. It has also The cells of layer VI appear to blend gradually into the subja-
been fostered by the reemergence of the notion rst put forth cent subcortical white matter and the overlying layer V.
by Ramon y Cajal that (to paraphrase) whatever the rest of
the hippocampal formation is doing depends on what the Regional Organization
entorhinal cortex has done. We begin our description of There have been several attempts to subdivide the rat entorhi-
the entorhinal cortex with a discussion of some lingering con- nal cortex, and unfortunately there have been almost an equal
troversies concerning its laminar and regional organization. number of differing opinions concerning the number and ter-
minology of the subelds. The subject has been discussed and
Cytoarchitectonic Organization reviewed by Menno Witter, Ricardo Insausti, and their col-
leagues (Witter, 1993; Insausti et al., 1997; Witter et al., 2000;
Laminar Organization Burwell and Witter, 2002). Nevertheless, it is now generally
There are currently two schemes of cortical lamination accepted that the entorhinal cortex can be subdivided into two
applied to the entorhinal cortex. As one might expect, this general areas: the lateral entorhinal area (LEA) and the medial
causes substantial confusion, especially to the neophyte entorhinal area (MEA) (Fig. 336). Layer II is more clearly
hippocampologist. One nomenclature, which divided the demarcated in the LEA than in the MEA, and the cells are
entorhinal cortex into seven layers, was rst suggested by extremely densely packed and tend to be clustered in islands.
Ramn y Cajal and later modied to more closely resemble The cells in layer II of the MEA are somewhat larger and do
the standard six-layer scheme applied to the isocortex. not show a distinct clustering into islands; the border between
According to this scheme, there are four cellular layers (II, III, layers II and III is not as sharp as in the LEA. In both entorhi-
V, VI) and two acellular or plexiform layers (I, IV). The acel- nal areas, however, the overall differences in cell size between
lular layer IV is also called lamina dissecans. Ramon y Cajals layers II and III facilitate the delineation of the two layers. The
scheme with slight modication has been employed by other cell layers, particularly layers IV to VI, can be better dif-
Amaral and colleagues in several primate studies. The other ferentiated from each other in the MEA than in the LEA, and
commonly used scheme was proposed by Lorente de N, who cells in the MEA generally show a more radial or columnar
also differentiated six layers. Five of Lorente de Ns layers arrangement. The lamina dissecans of the MEA is sharply
were cellular (II, III, IV, V, VI) with a cell-free lamina dissecans delineated but is less clear in the LEA.
(layer IIIb) between layers III and IV. This scheme was used in It should also be stressed that the terms lateral and medial
most of the older studies of the entorhinal cortex. It is still entorhinal areas do not relate in a simple manner to the car-
used in rodent studies and in at least some studies of the dinal transverse plane of the rat brain (Fig. 337). Both the
human entorhinal cortex, particularly those of Van Hoesen LEA and the MEA have a more or less triangular shape. The
and colleagues. LEA occupies the rostrolateral part of the entorhinal cortex;
Primarily to emphasize the lack of an internal granular cell its base is oriented rostrally and its tip caudolaterally, next to
layer in the entorhinal cortex, we have decided to adopt the rhinal ssure. The MEA occupies the remaining triangular
Ramon y Cajals nomenclature and have labeled the cell-poor area, which has its base caudally and its tip rostromedially
layer IV lamina dissecans. Starting from the pial surface, the such that the tip lies medial to the LEA. A different nomencla-
layers include layer I, the most supercial plexiform or molec- ture has recently been proposed by Insausti and colleagues to
ular layer, which is cell-poor but rich in transversely oriented accommodate the oblique orientation of the rat entorhinal
bers; layer II, containing mainly medium-sized to large stel- cortex and to address the need for subdemarcation of the LEA
late cells and a population of small pyramidal cells that tend and MEA (Insausti et al., 1997).
to be grouped in clusters (cell islands) particularly in the lat-
eral entorhinal area; layer III, containing cells of various sizes Neuron Types
and shapes but predominantly pyramidal cells; layer IV (or
lamina dissecans), a cell-free layer located between layers III Our current knowledge of the cytology of the entorhinal cor-
and V that is most apparent in those portions of the entorhi- tex is based largely on the classic Golgi studies of Ramon y
nal cortex that lie close to the rhinal ssure, particularly at the Cajal and Lorente de N in young mice. A few intracellular
caudal levels of the entorhinal cortex. In the remainder of the labeling studies have also been conducted in rats, and they
entorhinal cortex, groups of cells invade this layer so it has an have contributed important new information concerning
incomplete or patchy appearance. Next are layer V, a cellular entorhinal cell types (Hamam et al., 2000, 2002). It is proba-
 Hippocampal Neuroanatomy 85

Figure 336. Cytoarchitectonic characteristics of the rat entorhinal crographs of LEA (D) and MEA (E) taken from portions of the
cortex. AC. Photomicrographs of Nissl-stained coronal sections entorhinal cortex enclosed by boxes in B and C, respectively.
through three selected rostrocaudal levels of the entorhinal cortex, The layers of the entorhinal cortex are indicated with roman
arranged from rostral (A) to caudal (C), showing the lateral (LEA) numerals. PcoA, posterior cortical nucleus of the amygdala. Bars:
and medial (MEA) entorhinal area subdivisions. Arrowheads indi- AC, 1 mm; D and E, 250 m (Source: Dolorfo and Amaral,
cate LEA and MEA boundaries. D, E. High magnication photomi- 1998b.)
86 The Hippocampus Book

bly safe to predict that a number of new entorhinal cell types

will be discovered as these techniques are applied more inten-
sively to analysis of the entorhinal cortex. Other recent studies
have focused on determining which particular neuron types
project to the dentate gyrus, hippocampus, and subiculum.
These studies have employed retrograde tracing techniques
either alone or in conjunction with intracellular lling. The
following description provides a short overview of the various
cell types of each of the entorhinal layers and a brief comment
on the major characteristics of their dendritic and axonal
organization.
Layer I is populated by a small number of widely dispersed
neurons that have been further differentiated using various
criteria. There are stellate and horizontal GABAergic neurons
that have been shown electrophysiologically to terminate on
the dendrites of layer II cells that project to the dentate gyrus.
Layer II is populated by stellate and pyramidal cells (Fig.
338), both of which project to the dentate gyrus and CA3.
These cells give rise to extensive associational projections to
layer II of other regions of the entorhinal cortex, and they
contribute a few collaterals to deeper layers of the entorhinal
cortex. Although there are some exceptions, most layer II cells
have a dendritic tree that is conned to supercial layers I and
II. A class of axo-axonic cells, similar to the cortical chandelier
Figure 337. Relative positions of the lateral (LEA) and medial
cell, is also located in layer II; although some of these cells are
(MEA) entorhinal cortex and the perirhinal and postrhinal cortices. present in layer III, they have not been observed in the deep
A. Lateral view of the rat brain. B. Unfolded map of the entorhinal, layers.
perirhinal, and postrhinal cortices. Bar  1 mm. (Source: Adapted In layer III, the most numerous neurons are the pyramidal
from Burwell and Amaral, 1998b.) cells (Fig. 339), which give rise to the perforant and alvear

Figure 338. Morphological characteristics of entorhinal cortex III. B. Camera lucida drawing of a typical layer II stellate cell. Note
layer II neurons. A. Camera lucida drawing of a typical layer II the multiple thick primary dendrites and the widely diverging
pyramidal cell. Note the thick apical dendrite branching above the upper and lower dendritic trees, with supercially directed den-
border with layer I, the thin basal dendrites arising radially from the drites reaching the topmost portion of layer I. The axon (truncated)
soma, and the limited extent of the upper dendritic tree. The axon emerges from the base of the soma. Bar  100 m. (Source:
(arrow) emerges from the base of the soma and branches in layer Adapted from Klink and Alonso, 1997.)
 Hippocampal Neuroanatomy 87

Layer IV, the lamina dissecans, although generally referred

to as a cell-free layer, does contain scattered cell bodies
described as fusiform or pyramidal cells that have apical den-
drites reaching up to layer I and an axon reaching the deep
white matter, a characteristic of projection neurons. Some of
the cells in the lamina dissecans stain positively with antibod-
ies against GABA, NPY, calretinin, calbindin, or somatostatin.
Layer V contains three main classes of neuronpyramidal
cells, small spherical cells, and fusiform neuronsaccording
to the Golgi-based description of Lorente de N. Layer Va is
characterized by its large, darkly staining pyramidal cells.
These cells have large apical dendrites that ascend toward the
supercial portion of layer II and into layer I. The axons of
these cells run into the deep white matter and the angular
bundle, and additional collaterals innervate the supercial
layers of the entorhinal cortex. According to Lorente de N,
the collaterals of such cells form a column-like plexus situated
close to the location of the parent cell body and span the
entire thickness of the entorhinal cortex. These collaterals
mainly distribute to the deep cell layers (V/VI), although
Figure 339. Morphological characteristics of entorhinal cortex occasional collaterals reach layer II. Layer Va also contains var-
layer III neurons. Camera lucida of a typical layer III pyramidal cell. ious small neurons that have been characterized as horizontal,
The axon is indicated by an arrow. Bar  100 m. (Source: Adapted stellate, and multipolar cells. Golgi studies indicate that the
from Gloveli et al., 1997.) axonal plexus of the latter group either remains in layer Va or
extends to the supercial layers I to III. Angel Alonso and col-
path projections to CA1 and the subiculum. Their axons give leagues have used intracellular staining techniques to study
rise to collaterals that distribute mainly to layers I to III. The the morphology of layer V neurons, particularly in the deeper
apical dendrites of layer III pyramidal cells ascend, give off portion of layer V, or layer Vb (Fig. 340). They also found
branches to layer II, and ultimately form a terminal tuft in that there are three main types of neuron. Large and small
layer I. Layer III also contains multipolar, stellate, fusiform, pyramidal cells, neurons with dendrites oriented horizontally
horizontal, and bipolar cells, all of which appear to contribute and mainly conned to layer V, and an atypical form of mul-
to the perforant pathway. This large, heterogeneous group tipolar neuron with long wavy dendrites. A surprising aspect
of cells undoubtedly coincides with what Lorente de N of the latter neurons is that some of their dendrites meander
described as atypical layer III neurons. from the entorhinal cortex, through the angular bundle, and

Figure 340. Morphological characteristics of entorhinal cortex layer V neurons. Computer-

aided reconstructions of pyramidal, horizontal, and polymorphic cells. Dashed lines represent
the borders of the indicated layers. Bar  100 m. (Source: Adapted from Hamam et al., 2002.)
88 The Hippocampus Book

into the subiculum. Some of these cells may thus contribute to tial way to the portions of the entorhinal cortex that project to
the projections to the dentate gyrus and the hippocampus. other levels of the dentate gyrus. Thus, the associational con-
The overall picture emerging from these recent studies is that nections seem to be organized to integrate all of the informa-
all three types of layer V neuron should be considered projec- tion that comes into a particular portion of the entorhinal
tion neurons in that they send an axon to the white matter. cortex and are relayed to a particular septotemporal level of
They also may function as local circuit neurons, connecting the dentate gyrus.
the deep layers to the supercial layers. Not much is known about the detailed microcircuitry of
Layer VI contains a wide variety of neuron types. Based on the entorhinal associational connections. Presumably, the
the predominant distribution of their axonal plexus, these layer II cells that project to other portions of layer II terminate
cells can be grouped into three categories: cells that mainly on the same stellate and pyramidal cells that give rise to the
inuence other cells in layer VI or Vb; cells that by means perforant path projection. It is not known, however, if associ-
of their highly collateralized axons can inuence a vertical ational connections also terminate on interneurons in layer II.
column of cells in layers I to III; and cells whose axons Of even greater functional signicance is the issue of whether
are directed toward the deep white matter and are there- the deep layer cells project to the cells of layers II and III that
fore likely to be projection neurons. Some of these cells con- give rise to the perforant pathway. This is a critical missing
tribute to the projections to the dentate gyrus and the hip- piece of information. Because the deep layer neurons receive
pocampus. feedback projections from CA1 and the subiculum, an associ-
On the basis of their restricted axonal distribution, several ational connection from deep cells to supercial cells would
of the smaller cell types in the entorhinal cortex have been provide the link for completing the loop through the hip-
classied as interneurons. Most of these interneurons are pocampal formation. Another issue is whether these associa-
likely GABAergic. GABAergic neurons are found in all layers tional connections are excitatory or inhibitory. If a substantial
of the entorhinal cortex, although they are most abundant portion of the deep to supercial pathway originates from
in the supercial layers. GABAergic interneurons can be sub- GABAergic neurons, or if the pathway is excitatory but termi-
categorized on the basis of their colocalization with various nates preferentially on inhibitory cells located in layer II, the
neuroactive substances (e.g., peptides) or their expression of output of the hippocampal formation from CA1 and the
one or more of the various calcium-binding proteins. The subiculum could inhibit layer II or III neurons that project to
supercial layers contain a large number of cells that are the dentate gyrus, hippocampus, and subiculum.
immunoreactive for parvalbumin, whereas the number of
calbindin-D28-positive neurons is much lower. Although Commissural Connections
most of the GABAergic neurons are undoubtedly interneu-
rons, at least some of the GABAergic neurons in layers II and Relatively strong commissural connections, arising from all
III project to the dentate gyrus. portions of the entorhinal cortex, terminate predominantly in
layers I and II of the homotopic area of the entorhinal cortex
Intrinsic/Associational Connections (Goldowitz et al., 1975; Hjorth-Simonsen and Zimmer, 1975;
Deller, 1998). The entorhinal cortex also gives rise to a com-
The entorhinal cortex contains a substantial system of associ- missural projection to other components of the contralateral
ational connections. Intraentorhinal bers are organized in hippocampal formation. The largest component of this pro-
three rostrocaudally oriented bands, and connections that link jection is directed toward the dentate gyrus, but elds CA3
different transverse (or mediolateral) regions of the entorhi- and CA1 of the hippocampus and the subiculum also receive
nal cortex are rather restricted (Dolorfo and Amaral, 1998b). a contralateral input. The crossed entorhinal projection is
Associational connections originate in both supercial and heaviest to septal portions of the hippocampal subelds and
deep layers. Projections originating from layers II and III tend rapidly diminishes in strength at more temporal levels. This
to terminate mainly in the supercial layers, whereas projec- crossed projection apparently arises exclusively from layer
tions originating from the deep layers terminate in both the III cells.
deep and supercial layers.
The global organization of the associational connections in Organization of the Perforant
the entorhinal cortex can be best understood in relation to the and Alvear Pathway Projections
topography of the entorhinal projections to other elds of the
hippocampal formation (Fig. 341). As we describe shortly, As a reminder of our earlier description of the perforant path
different parts of the entorhinal cortex project to different input to the dentate gyrus, hippocampus, and subiculum,
septotemporal levels of the dentate gyrus. The portions of the both the lateral and medial entorhinal areas project to all three
lateral and medial entorhinal areas that project to the septal areas. The lateral and medial components of the perforant
half of the dentate gyrus, for example, are located laterally and path terminate along the supercial-to-deep gradient in the
caudally in the entorhinal cortex, close to the rhinal sulcus. molecular layer of the dentate gyrus and the stratum lacuno-
Cells located in this region give rise to associational connec- sum-moleculare of CA3 and along the transverse axis of CA1
tions to other cells in the same region but not in any substan- and the subiculum (Fig. 342). Layer II cells give rise to the
 Hippocampal Neuroanatomy 89

Figure 341. Summary of the intrinsic connections of the entorhi- projection zones has substantial associational connections (large
nal cortex and the association of the entorhinal cortex with the hip- arrows) that remain largely in the zone of origin. Projections
pocampal formation and neocortex. A. Organization of entorhinal between each band are less prominent (small arrows). C. Entorhinal
projections to the dentate gyrus. A band of layer II cells located in intrinsic projections and entorhinal afferent and efferent projec-
the lateral and caudomedial portion of the entorhinal cortex (light tions. Entorhinal associational connections arise from both the deep
gray) projects to the septal half of the dentate gyrus; a band in the and supercial layers of the entorhinal cortex (arrows in the boxes)
mid-mediolateral entorhinal cortex (medium gray) projects to the and terminate mainly in the supercial layers. A less prominent
third quarter of the dentate gyrus; and a band located in the most associational connection in the deep layers is apparent primarily in
rostromedial entorhinal cortex (dark gray) projects to the temporal lateral portions of the entorhinal cortex. (Source: Adapted from
pole of the dentate gyrus. B. Organization of entorhinal intrinsic Dolorfo and Amaral, 1998b.)
projections: projections within and between bands. Each of these
90 The Hippocampus Book

Figure 342. Laminar and topographical organization of the the subiculum, layer III entorhinal cortex projections are organized
entorhinal projection to the dentate gyrus, the hippocampus, and topographically; the LEA projects to the distal CA1 and proximal
the subiculum. The surface of the entorhinal cortex is represented subiculum (i.e., at the CA1/subiculum border), and the MEA proj-
on the left (the rhinal sulcus is to the left). Layer II entorhinal cor- ects to the proximal CA1 and distal subiculum. Laterally situated
tex projections to the dentate gyrus and the CA3 and CA2 elds of portions of the entorhinal cortex project to septal levels of the hip-
the hippocampus terminate in a laminar fashion; the LEA projects pocampal formation, whereas progressively more medial portions
supercially in the molecular layer and the stratum lacunosum- of the entorhinal cortex project to more temporal levels of the hip-
moleculare, and the MEA projects deeper. In contrast in CA1 and pocampal formation.

projection to the dentate gyrus and CA3, and layer III cells and medial entorhinal areas) projects to the septal half of
project to CA1 and the subiculum. Although current usage the dentate gyrus. Most of the cells in this domain project to
applies the term perforant path to all entorhino-hippocampal all portions of the septal half of the dentate gyrus; thus, the
projections, entorhinal bers also reach CA1 via the alveus projection is both divergent and convergent. The next domain
(i.e., the alvear pathway originally described by Ramon y is more medially situated and projects to the third quarter of
Cajal). In the temporal portion of the hippocampus, most of the dentate gyrus. The last domain is medially and rostrally
the entorhinal bers reach CA1 after perforating the subicu- situated and projects to the temporal quarter of the dentate
lum (classic perforant pathway). At more septal levels, how- gyrus.
ever, the number of entorhinal bers that take the alvear There are a number of functional implications of the
pathway increases; and in the septal portion of the hippo- organization of these projections. First and foremost, because
campus, most of the entorhinal bers reach CA1 via the alvear the associational connections of the entorhinal cortex seem to
pathway. These bers make a sharp turn in the alveus, perfo- respect this tripartite organization, it is reasonable to think of
rate the pyramidal cell layer, and terminate in the stratum three functional, parallel systems encompassed within the
lacunosum-moleculare. Both pathways demonstrate the same entorhino-hippocampal system. Second, when conducting
septotemporal organization of their projections. Thus, certain stimulation or lesion experiments designed to evaluate the
portions of the entorhinal cortex project to certain septotem- contributions of the medial and lateral perforant paths to hip-
poral levels of the other hippocampal elds. Following this pocampal function, one must be careful to bear in mind that
brief overview, we now delve more deeply into the topograph- the cells giving rise to the medial perforant path are located
ical organization of the perforant path projection. caudal (not medial) to the cells that give rise to the lateral per-
Dolorfo and Amaral (1998a) have shown that cells located forant path.
laterally and caudally in the entorhinal cortex project to septal One nal comment must be made on the medial and lat-
levels of the hippocampal elds, whereas cells located progres- eral perforant path projections. Although the cellular charac-
sively more medially and rostrally (in the medial entorhinal teristics of layers II and III of the medial and lateral entorhinal
cortex) project to more temporal levels of the hippocampal areas are similar, the lateral and medial perforant path pro-
subelds (Fig. 341). It now appears that there may be three jections demonstrate a number of differential features. Both
largely nonoverlapping domains of the entorhinal cortex that components of the perforant path use glutamate as their pri-
project to three distinct levels of the dentate gyrus. The later- mary transmitter, but they exhibit differential distributions of
ally situated domain (encompassing cells of both the lateral glutamate receptors. The medial perforant path bers, for
 Hippocampal Neuroanatomy 91

example, are immunoreactive for the metabotropic glutamate the raw material the hippocampal formation uses to accom-
receptor mGLUR 2/3, whereas the lateral bers are not. In plish its purported function(s). This is an area in which there
contrast, the lateral perforant path demonstrates dynorphin are substantial species differences. Keeping to the format that
immunoreactivity, whereas the medial perforant path does we have followed so far, we give an overview of the inputs to
not. One would have thought that some of these differences the rat entorhinal cortex. However, doing so does a disservice
would show up more clearly in the cells of origin, but thus far to the signicance of the connections between the entorhinal
there is no distinctive marker at the level of the entorhinal cor- cortex and the neocortex, as these connections are so much
tex for cells that give rise to the lateral and medial perforant more prominent in the monkey brain. Thus, when we provide
path projections. our comparison of the organization of the hippocampal for-
mation in the rat, monkey, and human in a later portion of the
Feedback Projections from the Hippocampus chapter, we return to a detailed overview of the cortical inputs
and Subiculum of the monkey entorhinal cortex.
Burwell has carried out a thorough quantitative analysis of
We already mentioned that the dentate gyrus and the CA3 the organization of cortical inputs to the rat entorhinal cortex
eld of the hippocampus do not project back to the entorhi- and compared the results with inputs to the perirhinal and
nal cortex. Thus, the recipients of the layer II projection do postrhinal cortices (Fig. 343) (Burwell and Amaral, 1998a).
not have any direct inuence over the activities of the entorhi- She found many similarities in the complement of cortical
nal cortex. It is only after the layer II and layer III projection inputs to the lateral and medial entorhinal areas. Each receive
systems are combined in CA1 and the subiculum that return about one-third of its total input from the piriform cortex
projections to the entorhinal cortex are generated. The return (LEA 34%, MEA 31%). These areas also receive roughly equal
projections mainly terminate in the deep layers (V and VI), proportions from temporal (LEA 26%, MEA 21%) and frontal
although some bers ascend into layer I (Witter et al., 1988; (LEA 11%, MEA 10%) regions. Some differences are observed
Naber et al., 2001; Kloosterman et al., 2003). It is not known in the proportions of insular, cingular, parietal, and occipital
which entorhinal neurons are the recipients of these return inputs. The lateral entorhinal area receives more input from
projections. What is clear, however, is that the projections insular cortex (LEA 21%, MEA 6%). In contrast, the medial
from CA1 and the subiculum to the entorhinal cortex are also entorhinal area receives more input from cingulate (LEA 3%,
topographically organized (Figs. 329 and 334). Septal por- MEA 11%), parietal (LEA 3%, MEA 9%), and occipital (LEA
tions of CA1 and the subiculum project chiey to lateral parts 2%, MEA 12%) regions. Not all portions of the rat entorhinal
of the entorhinal cortex, and more temporal parts of CA1 and cortex receive substantial cortical input. In fact, it is only the
the subiculum project to more medial parts of the entorhinal lateral and caudal parts of the entorhinal cortex (those pro-
cortex. Moreover, the transverse location of the cells of origin jecting to septal levels of the dentate gyrus) that are heavily
in CA1 and the subiculum also determines whether these pro- innervated by the neocortex. This neuroanatomical organiza-
jections terminate in the medial or lateral entorhinal cortex. tion has strong functional implications and likely underlies
The projections from the proximal part of CA1 and the distal the behavioral dissociations observed with dorsal (septal) ver-
part of the subiculum distribute exclusively to the medial sus ventral (temporal) hippocampal lesions.
entorhinal cortex, whereas cells located in the distal part of The neocortical inputs to the entorhinal cortex of the rat
CA1 and the proximal part of the subiculum project mainly to comprise two groups: those that terminate in the supercial
the lateral entorhinal cortex. layers (IIII) and those that terminate in the deep layers
The important point about these return projections is (IVVI). The rst category delivers information to the super-
that they are exactly in register (i.e., they are point-to-point cially located entorhinal neurons, which are the source of the
reciprocal) with the entorhinal inputs to these areas. Thus, projections to the dentate gyrus, hippocampus, and subicu-
at the global level, all of the circuitry is available for rever- lum. The second group has greater inuence on the deeply
beratory circuits to be established through the loop, starting located cells of the entorhinal cortex, which receive processed
and ending at the entorhinal cortex. This remarkable topogra- information from the other hippocampal elds and give rise
phy conrms the critical role of the entorhinal cortex with to feedback projections to certain cortical regions. In general,
respect to the input to and output from the hippocampal for- the cortical afferents that reach the deep layers terminate
mation. rather diffusely, whereas those that terminate supercially
have a more restricted mediolateral and/or rostrocaudal dis-
Extrinsic Connections tribution. Although speculative, the rst group may constitute
a set of information-bearing inputs, whereas the second group
Interconnections of the Entorhinal Cortex may have more of a modulatory inuence on the output of
with Neocortical Regions the hippocampal formation.
If the entorhinal cortex is viewed as the rst step of processing A substantial input to the supercial layers of the entorhi-
in the hippocampal formation, it is reasonable to wonder nal cortex originates from olfactory structures such as the
what types of information it receives. In other words, what is olfactory bulb, the anterior olfactory nucleus, and the piri-
92 The Hippocampus Book

Figure 343. Pattern and strength of cortical and intrinsic connec- AId/v/p, dorsal, ventral, and posterior agranular insular cortices;
tivity of the rat parahippocampal region (entorhinal, perirhinal, and AUD, primary auditory cortex; MOs, secondary motor area; Pir, pir-
parahippocampal cortices). The thickness of the solid lines repre- iform cortex; PTLp, posterior parietal cortex; RSPd and RSPv, retro-
sents the relative strength of the connections based on the densities splenial cortex, dorsal and ventral; SSp and SSs, primary and
of retrogradely labeled neurons. Open lines represent reported con- supplementary somatosensory areas; Tev, ventral temporal area;
nections for which no comparable quantitative data are available. VISC, visceral granular insular cortex; VIS1 and VISm, lateral and
The weakest projections ( 250 labeled cells/mm3) are not shown medial visual association cortex; VISp, primary visual cortex.
here. ACAd and ACAv, dorsal and ventral anterior cingulate cortex; (Source: Adapted from Burwell and Amaral, 1998b.)

form cortex. These olfactory projections terminate through- sively in our discussion of the monkey entorhinal cortex, but
out most of the rostrocaudal extent of the entorhinal cortex, it is important to give an overview of this area because it pro-
mainly in layer I and the supercial portion of layer II. Only vides major input to the rat entorhinal cortex. The perirhinal
the most caudal portion of the rat medial entorhinal area does cortex in the rat is made up of two areas, 35 and 36, which
not receive any olfactory input. In addition to terminating on appear to receive slightly different complements of neocorti-
the principal cells of these layers, olfactory bers also termi- cal inputs (Fig. 343) (Burwell et al., 1995; Burwell, 2001).
nate on layer I GABAergic neurons, which presumably inter- Area 36 of the perirhinal cortex receives more, higher-level
act with the principal cells in layers II and III. cortical input than does area 35. Most of this input comes
A second prominent cortical input to the supercial layers from the ventral temporal associational area (Tev), which is
of the entorhinal cortex arises from the laterally adjacent located dorsally adjacent to area 36. Other major inputs are
perirhinal and postrhinal cortices. The perirhinal and postrhi- from the postrhinal cortex and the entorhinal cortex. The
nal cortices are polysensory convergence areas that receive predominant inputs to area 35 arise from the piriform,
inputs from a variety of unimodal and polymodal sensory entorhinal, and insular cortices. Area 35 receives more than
cortices. The perirhinal cortex terminates mainly in the lateral one-fourth of its input from the piriform cortex and only
entorhinal area, and the postrhinal cortex terminates heavily, slightly less from the lateral entorhinal area. About one-fth of
though not exclusively, in the medial entorhinal area. In both the total input arises from insular cortices. Interestingly, the
cases, the projections terminate preferentially in layers I and perirhinal projections to the entorhinal cortex arise preferen-
III of the entorhinal cortex. tially from area 35, and the intrinsic projections of the perirhi-
We address the issue of the perirhinal cortex more exten- nal cortex seem to be organized to funnel information into
 Hippocampal Neuroanatomy 93

area 35, which in turn gives rise to the main perirhinal projec- additional sources of information. The substantial input from
tions to the entorhinal cortex. the amygdaloid complex, which originates mainly from the
Cortical afferents to the deep layers of the entorhinal cor- lateral and basal nuclei, is presumably conveying information
tex arise from a variety of cortical areas. These areas include about the emotional state of the organism (Pikkarainen et al.,
projections from the agranular insular cortex, the medial pre- 1999a). The densest termination of the amygdaloid bers is in
frontal region (particularly the infralimbic, prelimbic, and the ventrolateral part of the entorhinal cortex (Fig. 344).
anterior cingulate cortices and the retrosplenial cortex Efferents from the lateral amygdaloid nucleus distribute most
(Insausti et al., 1997). intensely to the deep portion of layer III but end also between
Efferents of the entorhinal cortex return projections to the cell islands of layer II and in layer I. The bers from the
many of the cortical areas that provide input to the entorhinal basal nucleus terminate diffusely in layers III to V, whereas
cortex. There are projections to olfactory areas, originating those from the cortical nuclei and the periamygdaloid cortex
predominantly from layers II, III, and Va. An important issue preferentially project to layers I and II. The entorhinal cortex
is whether the entorhinal cortex of the rat, like that of the sends feedback projections to the amygdala that terminate
monkey (see below), gives rise to prominent, widespread pro- mainly in the basal nucleus. These projections originate from
jections to multimodal association cortices. Initial studies in cells in layer V, although a few cells in more supercial layers
the rat indicated that the entorhinal cortex projects mainly to may also contribute to the projection. Dense inputs also orig-
adjacent, limited portions of the temporal cortex. Swanson inate from the ventral part of the claustrum or endopiriform
and Khler (1986), however, suggested that the rat entorhinal nucleus, but the topographical and laminar organization of
cortex gives rise to projections that reach a much larger these inputs is not well understood.
domain of the cortical surface. Whether one believes that the The entorhinal cortex projects bilaterally to the striatum,
entorhinal cortex has widespread neocortical connections particularly to the nucleus accumbens and adjacent parts of
seems to hinge on the demarcation of the entorhinal cortex the olfactory tubercle. These projections originate mainly
from the adjacent perirhinal cortex. Insausti and colleagues from layer V and are topographically organized; medial parts
conrmed the report by Sarter and Markowitsch (1985) that of the entorhinal cortex project to the caudomedial portion of
the only cells that give rise to the extensive connections the nucleus accumbens, and more lateral portions of the
reported by Swanson and Khler are located in the most dor- entorhinal cortex project to more lateral parts of the nucleus.
solateral region of the entorhinal cortex (i.e., on the border of
the entorhinal and perirhinal cortices). Whether these cells Basal Forebrain and Hypothalamic Connections
indeed belong to the deep layers of the entorhinal cortex or The entorhinal cortex receives its cholinergic innervation
form a population of perirhinal cells is not yet clearly estab- mainly from the septum. This projection is topographically
lished. Based on the distribution of immunoreactivity for par- organized such that cells in the horizontal limb of the nucleus
valbumin or for certain glutamate receptor subunits, the of the diagonal band preferentially project to the most lateral
border between the entorhinal and perirhinal cortex appears part of the entorhinal cortex, whereas the medial septal
to be oblique and perhaps intermixed in this critical region. nucleus and the vertical limb of the nucleus of the diagonal
Thus, it remains unresolved at this point whether it is the band project to more medial parts of the entorhinal cortex.
entorhinal cortex or the perirhinal cortex that projects widely Septal projections terminate densely in the cell-sparse lamina
to the neocortex. One can safely conclude, however, that the dissecans and less densely in layer II.
major portion of the entorhinal cortex does not contribute Like the hippocampus and the subiculum but unlike the
projections to the unimodal areas of the neocortex, and that pre- and parasubiculum, the entorhinal cortex projects back
the bulk of neocortically directed projections are to higher- to the septal region. The projection originates from cells in
order associational and polysensory cortices and not to sen- layer Va, although some layer II cells also contribute to these
sory or motor regions. In this respect, the situation in the rat projections. The bers from the entorhinal cortex preferen-
closely resembles the connectivity observed in the monkey. tially terminate in the lateral septal complex.
Among the cortical areas that do receive entorhinal input are The entorhinal cortex receives diffuse inputs from various
the infralimbic, prelimbic, orbitofrontal, agranular insular, structures in the hypothalamus. They include inputs from the
perirhinal, and postrhinal cortices; these projections originate supramammillary nucleus that terminate rather diffusely with
mainly from cells in layer Va. Only weak projections have been some preference for layers III to VI, from the tuberomammil-
reported to the retrosplenial cortex. lary nucleus distributing diffusely throughout the entorhinal
cortex, and from the lateral hypothalamic area reaching the
Other Telencephalic Connections: deep layers of the entorhinal cortex.
Amygdala, Claustrum, Striatum
In addition to the cortical inputs just described, the entorhi- Thalamic Connections: Nucleus Reuniens
nal cortex receives a number of subcortical inputs. Whereas and Other Midline Nuclei
some of them, such as the monoaminergic and cholinergic The major thalamic inputs to the entorhinal cortex originate
inputs, may be viewed as largely modulatory, others, such as in the nucleus reuniens and the nucleus centralis medialis
the input from the amygdaloid complex, might also provide (Van der Werf et al., 2002). The rhomboid, paraventricular,
94 The Hippocampus Book

Figure 344. Summary of the reciprocal connections between the of the medial entorhinal cortex do not receive amygdala inputs.
amygdala and the hippocampal formation and the perirhinal and Areas DLE, DIE, VIE, and AE of the lateral entorhinal cortex receive
parahippocampal cortices. The subdivisions of the entorhinal cortex inputs from the various amygdala nuclei. (Source: Pitknen et al.,
follow the nomenclature of Insausti et al. (1997). Areas CE and ME 2000.)

and parataenial nuclei contribute minor projections. The plies the entorhinal cortex with a diffusely organized nora-
nucleus reuniens bers densely innervate the deep portion of drenergic input that exhibits slightly more dense termination
layers I and III and give rise to a few collaterals that extend in layer I.
into layer II. Separate populations of nucleus reuniens cells
project to the entorhinal cortex, CA1, and the subiculum.
There is no evidence that the entorhinal cortex projects back
to the thalamus. 3.5 Chemical Neuroanatomy

Brain Stem Inputs 3.5.1 Transmitters and Receptors

The entorhinal cortex receives dopaminergic input from
cells located in the ventral tegmental area. The projection Detailed, comprehensive treatment of the chemical neu-
preferentially terminates in a restricted rostrolateral part of roanatomy of the hippocampal formation would demand
the lateral entorhinal area, where bers are arranged in dense, a lengthy chapter of its own. We have already described the
columnar patches in layers I to III. The serotonergic innerva- distribution of systems dened by their neurotransmitter con-
tion arises from the central and dorsal raphe nuclei and tentnoradrenaline (norepinephrine), serotonin, acetyl-
terminates diffusely in all layers but with a preference for choline, GABAand a variety of reviews are available on the
the supercial layers. The noradrenergic locus coeruleus sup- chemical neuroanatomy of the hippocampus (Swanson et al.,
 Hippocampal Neuroanatomy 95

1987; Kobayashi and Amaral, 1998). We thus restrict ourselves eralcorticoid receptors. Receptors for both steroids were
here to providing an overview of the diversity of neurochem- highly expressed in cells of the dentate gyrus, CA3, and CA2.
ical substances in the hippocampal formation. Lower levels of expression were observed in CA1. This is the
A variety of peptides and other chemical markers have opposite of the distribution observed in the rat and the tree
been shown to subdivide the population of GABAergic shrew.
interneurons. Among the peptides that colocalize with partic-
ular populations of GABAergic interneurons are VIP, somato-

statin, NPY, corticotropin-releasing factor (CRF), substance P,
3.6 Comparative Neuroanatomy of the Rat,
cholecystokinin, galanin, and the opioid peptides dynorphin
Monkey, and Human Hippocampal Formation
and enkephalin. It is worth commenting further on the dis-
tribution of the opioid peptides because, in addition to being
When one views Nissl-stained sections of the hippocampus
localized to certain populations of interneurons, they are
from the rat, monkey, and human, it is immediately apparent
observed in intrinsic excitatory pathways of the hippocampal
that one is looking at the same brain region (Figs. 32 and
formation. The bers of the lateral perforant path, for exam-
345). The densely packed granule cell layer is obvious in all
ple, are immunoreactive for Leu-enkephalin, whereas the
three species, as is the progressively more complex lamination
mossy bers of the dentate gyrus are positive for dynorphin.
when one progresses from the dentate gyrus to the entorhinal
Another class of substances that appear to mark certain
cortex. On closer inspection, however, a number of differences
subsets of GABAergic neurons selectively is the family of
make it clear that the hippocampal formation of the monkey
calcium-binding proteins, including parvalbumin, calbindin,
or human is not simply a scaled-up version of the rat hip-
and calretinin. Whereas parvalbumin immunoreactivity
pocampal formation. Some of the hippocampal elds, such as
appears to be exclusively conned to a subset of GABAergic
CA1 and the entorhinal cortex, are disproportionately larger
interneurons, the other calcium-binding proteins can be
in the primate. The entorhinal cortex has many more subdivi-
found in both interneurons and principal neurons. Although
sions in the monkey and human than in the rat; and the lam-
the precise function of the various calcium-binding proteins
inar organization is much more distinct in the primate brain.
has not been well established, their existence has provided
In the following sections, we review some of the similarities
a useful anatomical tool. Although standard immunohisto-
and differences in the organization and connections of the
chemistry of GABAergic neurons with antibodies to glutamic
acid decarboxylase (GAD) or to GABA does not label den-
drites very well, parvalbumin-immunoreactive neurons are
fully labeled in a Golgi-like fashion. Thus, even though par-
valbumin does not label all GABAergic neurons, those that are
labeled can be subjected to precise analyses of their inputs.

3.5.2 Steroids

Neurons in the hippocampal formation have been shown to

concentrate glucocorticoids and mineralcorticoids (McEwen
and Wallach, 1973). Studies using uptake of 3H-corticosterone
show that cells in CA2 and CA1 exhibit the greatest uptake.
There is slightly less uptake in CA3 cells and cells of the poly-
morphic layer of the dentate gyrus; dentate granule cells also
demonstrate some 3H-corticosterone uptake. The amount of
uptake in other elds of the hippocampal formation has not
yet been described. What is perhaps most surprising is the fact
that, in the rat, the hippocampus demonstrates the highest
level of uptake of any brain region. The distribution of gluco-
corticoid receptor mRNA has also been evaluated in the tree
shrew. Here the highest density of mRNA was observed in the
pyramidal cells of the subiculum and in the granule cells of
the dentate gyrus. In the hippocampus, higher densities of
mRNA were observed in CA1 than in CA3. Mineralcorticoid
receptor mRNA is also expressed in the tree shrew hippocam-
pus. It is expressed most strongly in CA1 and less strongly in
CA3; this contrasts with the pattern in the rat, where mineral- Figure 345. Nissl-stained coronal sections through the hippocam-
corticoid expression is higher in CA3. pal formation of the rat (A), monkey (B), and human (C) presented
In situ hybridization has also been used in postmortem at the same magnication. Bar in C  2 mm and applies to all
humans to study the distribution of glucocorticoid and min- panels.
96 The Hippocampus Book

hippocampal formation in these three species. Of course, Cytoarchitectonic Organization

although we can address certain issues concerning cell num-
ber and distribution in the human brain, we are unable to say There are several cytoarchitectonic differences between the rat
anything about patterns of connectivity in the human hip- and monkey hippocampal formation. The polymorphic layer
pocampal formation. of the dentate gyrus is relatively smaller in the monkey, and
much of the territory located enclosed within the blades of the
3.6.1 Neuron Numbers granule cell layers is occupied by the CA3 eld. This hilar
portion of the CA3 region is so expansive that some
The number of neurons present in the various subdivisions of researchers have confused it for an enlarged polymorphic
the hippocampal formation have been counted by several layer. Cells in this region bear the gold standard of inclusion
investigators, but rarely has the same author analyzed the rat, in CA3, however, as they give rise to Schaffer collaterals to
monkey, and human hippocampal formations. Moreover, CA1.
there is a lot of variability in the estimates published in the lit- The other obvious difference between the rat and monkey
erature, and few investigators have used modern stereological hippocampal formation is the much thicker pyramidal cell
techniques to carry out these assessments. We have compiled layer in the monkey CA1 (Fig. 349). In the rat, this layer is
a summary of current estimates of neuron number obtained tightly packed and is typically about 5 cells thick. In the mon-
with modern stereological techniques. The data presented key, the cell layer is much more diffusely organized and is 10
here are derived from published and unpublished work from to 15 cells thick. As a result, not only is the boundary between
three laboratories: Mark West at Aarhus University, Peter the pyramidal cell layer and the stratum radiatum less clear, so
Rapp at New York University, and our own at the University of is the boundary between CA1 and the subiculum. In the mon-
California Davis (Table 3-1). key, at least some of the Schaffer collaterals from CA3 termi-
In our initial comparison of the hippocampal formation in nate in the pyramidal cell layer, presumably on the apical
the rat, monkey, and human, the volume of the hippocampal dendrites of cells located deep in the layer or on the basal den-
complex (dentate gyrus  hippocampus) was said to be about drites of neurons located supercially in the layer.
10 times larger in monkeys than in rats and 100 times larger The entorhinal cortex exhibits major cytoarchitectonic dif-
in humans than in rats (rat 32 mm3, monkey 340 mm3; ferences between the rat and the monkey. The laminar organ-
human 3300 mm3). If, however, we compare the total number ization of the monkey entorhinal cortex is much clearer than
of neurons in the various hippocampal areas, we observe that that in the rat. Throughout much of the entorhinal cortex, for
certain regions are comparatively more developed than others example, there is a clear distinction between layers V and VI in
in the monkey and human brain. In the dentate gyrus, there the monkey, whereas these layers tend to blur together in the
are approximately 10 times more granule cells in the monkey rat. The monkey entorhinal cortex is also much more differ-
than in the rat, a ratio that parallels the overall volume differ- entiated than in the rat. As noted previously, Amaral and col-
ences. However, there are only 15 times more dentate granule leagues typically divided the rat entorhinal cortex into only
cells in humans compared to rats, whereas the volume of the lateral and medial areas, whereas they divided the monkey
dentate gyrus plus the hippocampus is about 100 times larger entorhinal cortex into seven distinct subdivisions.
in humans than in rats. In CA1, however, there are only three
times more pyramidal cells in the monkey than in the rat, Neuron Types
whereas there are 35 times more cells in humans than in rats.
We must add, however, that some of these estimates are still Until recently, there has been little direct comparison of the
provisional and need to be replicated before drawing solid morphology of similar neurons in the rat and monkey hip-
inferences on the relative development of the various hip- pocampus. A few studies, using either the Golgi technique or
pocampal regions in rats, monkeys, and humans. intracellular staining techniques, have compared easily identi-
ed cell types.
3.6.2 Comparison of Rat and Monkey The granule cells of the monkey dentate gyrus have been
Hippocampal Formation examined with both the Golgi technique and intracellular ll-
ing techniques in the in vitro slice preparation (Duffy and
In monkeys, the hippocampal formation lies entirely within Rakic, 1983; Seress and Mrzljak, 1987, 1992). Depending on
the temporal lobe (Figs. 346 through 348) and lacks the which aspects of these studies one wishes to emphasize, it
pronounced C shape along its septotemporal axis that is so could be concluded that granule cells are basically similar in
characteristic in the rat. The region equivalent to the septal rats and monkeys or that there have been substantial modi-
pole of the rat hippocampus is located caudally in the mon- cations of at least some granule cells. In general, dentate gran-
key, and the equivalent of the temporal pole is located ros- ule cells have the same unipolar apical dendritic tree in the
trally (Figs. 35 and 36). Much of the entorhinal cortex is monkey as in the rat. The total dendritic length of each gran-
located rostral to the remainder of the hippocampal forma- ule cell is also relatively similar in the rat and the monkey.
tion. In fact, the rostral half of the entorhinal cortex lies ven- Seress and colleagues were the rst to point out that at least
tral to the amygdaloid complex rather than the hippocampus. some monkey granule cells have basal dendrites (Fig. 350),
 Hippocampal Neuroanatomy 97

Figure 346. Coronal sections at three rostrocaudal levels through the monkey brain show the
relative position of the hippocampal formation. The three panels on the left are Nissl-stained
sections; the three panels on the right are adjacent sections stained with Timms sulde silver
stain. The line drawings (middle column) highlight the regions of the monkey hippocampal
formation seen in the stained sections. Bar  2 mm.

and a similar observation has been made by Scheibel and col- the dentate gyrus (Buckmaster et al., 1992; Buckmaster and
leagues. This feature has never been reported for normal rat Amaral, 2001). In the rat, these cells give rise to the associa-
granule cells. Thus, it appears that there are some species dif- tional-commissural connections to the inner portion of the
ferences in dentate granule cell morphology, but the func- molecular layer, and its dendrites are generally conned to the
tional signicance of these differences is yet unclear. polymorphic area (i.e., the dendrites extend neither into the
Another example of differences in an ostensibly similar molecular layer nor into the adjacent CA3 eld). The mossy
cell type in the rat and the monkey has emerged from intra- cell in the monkey is quite different. First, there appear to be
cellular staining of mossy cells of the polymorphic region of at least two forms. One is very much like the rat mossy cell,
98 The Hippocampus Book

Figure 347. Horizontal sections at three dorsoventral levels through the monkey brain show
the relative position of the hippocampal formation. The three panels on the left are Nissl-
stained sections; the three panels on the right are adjacent sections stained with Timms sulde
silver stain. The line drawings (middle column) highlight the regions of the monkey hippocam-
pal formation seen in the stained sections. Bar  2 mm.

with dendrites conned to the polymorphic cell layer and tion in the molecular layer of the dentate gyrus, whereas stan-
axons directed to the molecular layer. There is a second type, dard mossy cells are not (because the perforant path does not
however, that extends much of its dendritic arbor into the enter the polymorphic layer). Moreover, in rats the granule
molecular layer (Fig. 351). Moreover, many of these cells give cells are the only input to CA3, whereas in the monkey the
rise to projections into the adjacent CA3 region. The implica- mossy cells appear to contribute an additional projection.
tion of this altered morphology in the monkey is that these Although these structural alterations must be conrmed by
mossy cells are capable of receiving perforant path innerva- functional studies, they suggest that there are fundamental
 Hippocampal Neuroanatomy 99

Figure 348. Sagittal sections at three mediolateral levels through the monkey brain show the
relative position of the hippocampal formation. The three panels on the left are Nissl-stained
sections; the three panels on the right are adjacent sections stained with Timms sulde silver
stain. The line drawings (middle column) highlight the regions of the monkey hippocampal
formation seen in the stained sections. Bar  2 mm.
100 The Hippocampus Book

Figure 349. High magnication photomicrographs of Nissl- dal cell layer is approximately 15 cells thick and shows some sub-
stained coronal sections through the CA1 eld of the hippocampus lamination, with the top half of the layer having a slightly higher
in the rat (A), monkey (B), and human (C). Note that the CA1 density of neurons. The pyramidal cell layer in the human CA1 is
pyramidal cell layer in the rat (which is the darkly stained layer at even thicker and has a more laminated structure. Bar in A  200
the bottom of panel A) is about 5 cells thick. The monkey pyrami- m and applies to all panels.

differences regarding the circuit characteristics of the dentate little work has been carried out in the monkey, however, to
gyrus between the rat and the monkey. know if the topographical organization of intrinsic circuits is
the same in the two species. So far, minor differences have been
Connections observed. In the monkey, for example, perforant path bers
arising from the rostral entorhinal area (the equivalent of the
Basic Organization of the Intrinsic Hippocampal Circuitry rat lateral entorhinal area) terminate, as in the rat, mainly in
To the extent that it has been examined, the basic principles the outer third of the molecular layer of the dentate gyrus.
of organization of the intrinsic circuitry of the monkey hip- Some terminations, however, also continue in a decreasing
pocampal formation resemble those observed in the rat. Too gradient fashion into the middle third of the molecular layer.

Table 31.
Number of principal neurons in the subdivisions of the rat, monkey,
and human hippocampal formation (in millions)

Cells Rat Monkey Human

Gcl DG 1.20a.b 12c 18d

Hilus 0.05b 0.23e 1.72d
CA3 0.25a,b 1.27e 2.83d
CA1 0.39a,b 1.30e 14d
Subiculum 0.29b 0.75e 5.95d
Pre/parasubiculum 0.70f 2.08e
EC II 0.11f 0.26e 0.66g
EC III 0.25f 3.66g
EC V  VI 0.33f 3.78g
EC total 0.69f 2.16e 8.1g

Sources: aRapp and Gallagher (1996); bWest et al. (1991); cLavenex et al. unpublished data; dWest et al. (1994); eRapp et al.
unpublished data; fMulders et al. (1997); gWest and Slomianka (1998).
 Hippocampal Neuroanatomy 101

Figure 350. Camera lucida drawings of

primate granule cells. A. Monkey granule
cell in the upper part of the granule cell
layer (gcl). The dendrites are cone-shaped
in the molecular layer (ml); the axon (a)
originates from the base of the soma and
extends into the polymorphic layer (pl).
B. Monkey granule cell in the deep part
of the granule cell layer with a basal den-
drite. Both apical and basal dendrites are
fully covered with spines. The apical den-
drites extend through the molecular
layer, and the basal dendrites extend into
the polymorphic layer. The axon origi-
nates from the base of the soma and
extends into the polymorphic layer. C.
Three types of human granule cell in the
human dentate gyrus. The neuron on the
right has dendrites in the molecular layer
only, whereas the other two neurons have
both apical dendrites extending into the
molecular layer and basal dendrites
extending into the polymorphic layer.
Bar in C  20 m and applies to all
panels. (Source: Adapted from Seress
and Mrzljak, 1987.)

Projections from the caudal entorhinal cortex (the region from the polymorphic layer of the dentate gyrus to the con-
equivalent to the rat medial entorhinal area) terminate in a tralateral molecular layer of the dentate gyrus and from the
similar fashion: heaviest in the middle third and gradually CA3 eld of the hippocampus to the contralateral CA3 and
decreasing in the outer third of the molecular layer. Thus, CA1 elds. In the monkey, these connections are virtually
the border between the lateral and medial entorhinal termina- absent. Only the most rostral part of the dentate gyrus and
tions is much less distinct in the monkey than in the rat. hippocampus (corresponding to the most temporal portion
in the rat) demonstrates any commissural connections, and
Lack of Commissural Connections in they are limited to the homotopic regions on the contralateral
the Monkey Hippocampal Formation side. Interestingly, whereas the commissural connections of
One of the most striking connectional differences between the the dentate gyrus and hippocampus are largely absent, the
rat and the monkey relates to the organization of the com- connection originating in the presubiculum and terminating
missural connections (Amaral et al., 1984; Demeter et al., in layer III of the contralateral entorhinal cortex appears to be
1985). In the rat, there are extensive commissural projections as robust in the monkey as in the rat.
102 The Hippocampus Book

Figure 351. Mossy cell of the polymorphic layer of the monkey dentate gyrus. A. Camera
lucida drawing of the soma and dendritic arbor. Note that the dendrites extend widely in the
polymorphic layer and extend into the molecular layer, in contrast to what is observed in the
rat. B. Camera lucida drawing of the soma and axonal plexus. Bar  100 m. (Source: Adapted
from Buckmaster and Amaral, 2001.)

Increased Cortical Interconnectivity of communication with the neocortex (Insausti et al., 1987). For
the Monkey Hippocampal Formation essentially all cortical regions, the return projections from the
The big difference between the rat and monkey brain is the entorhinal cortex originate in layers V and VI and terminate
much greater amount of neocortex in the primate. Much of both deeply and supercially in a manner typical of other cor-
this neocortex is dedicated to visual processing, but there is tical feedback projections. The laminar organization of the
also a substantial increase in the amount of association cortex, inputs to the entorhinal cortex varies according to the cortical
particularly in the frontal and temporal lobes. A substantial area of origin. Projections from some regions, such as the
portion of this association cortex is polysensory, and most of perirhinal and parahippocampal cortices, terminate preferen-
the polysensory cortical regions are interconnected with the tially in layers I to III, which give rise to the perforant path
hippocampal formation via connections with the entorhinal projections. Projections from other cortical areas, such as
cortex. The more robust neocortical connectivity has given rise the orbitofrontal and insular cortices, project to the deep lay-
to the suggestion that processing in the primate hippocampal ers of the entorhinal cortex, which generate the major output
formation is more highly dependent on and integrated with pathways to the neocortex and other regions. These latter con-
processing of sensory information in the neocortex. nections, therefore, might preferentially be involved in modu-
In the monkey temporal lobe, there is massive expansion of lating the output of the hippocampal formation. However,
the perirhinal and parahippocampal regions that border the they may also contribute to the perforant pathway via some of
hippocampal formation. Rostrally, this region includes the the few layer V and VI neurons that contribute to this pathway
perirhinal cortex, which comprises areas 35 and 36 of or through intrinsic deep to supercial connections in the
Brodmann. The perirhinal cortex starts at a level through the entorhinal cortex.
rostral hippocampal formation and extends rostrally to form Whereas the perirhinal and parahippocampal cortices pro-
the medial portion of the temporal polar cortex. The caudal vide the major input to the hippocampal formation, other
continuation of the perirhinal cortex is the parahippocampal substantial inputs originate in the retrosplenial cortex (area 29
cortex, which includes areas TF and TH of von Bonin and in the caudal cingulate gyrus), the polysensory region of the
Bailey, both of which are bordered laterally by the unimodal superior temporal gyrus (along the dorsal bank of the supe-
visual areas TE or TEO. The perirhinal and parahippocampal rior temporal gyrus), and the orbitofrontal cortex (mainly
cortices are important players in the economy of the neocor- caudal area 13). The general rule is that the entorhinal cortex
tical innervation of the hippocampal formation, as they pro- receives input only from cortical regions with demonstrated
vide nearly two-thirds of the neocortical inputs to the polysensory convergence. Thus, only olfactory input (which
entorhinal cortex. originates from all levels of the olfactory system including the
The major inputs to the monkey entorhinal cortex are olfactory bulb in the primate) has direct access to the hip-
summarized in Figures 352 and 353. In both rats and mon- pocampal formation. The olfactory connection, however, pro-
keys, the entorhinal cortex provides the major interface for vides another example of a difference between the rat and
 Hippocampal Neuroanatomy 103

Figure 352. Summary of the major entorhinal cortical afferents in the macaque monkey.
(Source: Adapted from Insausti et al., 1987.)

monkey hippocampal formations. Whereas in the rat almost ing. From the unimodal association cortices, information
the entire entorhinal cortex receives direct input from the converges on a few polysensory cortical regions, which in turn
olfactory bulb, in the monkey this is restricted to about 10% project in a convergent fashion on the entorhinal cortex
of the surface area of the entorhinal cortex. (Lavenex and Amaral, 2000). The entorhinal cortex relays this
Given this pattern of neocortical inputs to the monkey information to the other hippocampal elds via the perforant
entorhinal cortex, the hippocampal formation can be viewed path projection. Once the information processed by the hip-
as the nal stage in a cascade of neocortical sensory process- pocampal formation is returned to the deep layers of the
104 The Hippocampus Book

Figure 353. Summary of the organization and strength of cortical inputs to the monkey
entorhinal, perirhinal (areas 35 and 36), and parahippocampal (areas TF and TH) cortices.
(Source: Adapted from Suzuki and Amaral, 1994.)

entorhinal cortex, they return this highly processed informa- human brain do need to be developed. Perhaps the com-
tion to many of the polysensory cortical regions that provide bination of transcranial electromagnetic stimulation during
feedback projections to the unimodal sensory regions. functional magnetic resonance imaging (fMRI) can provide
There are a number of implications of this cascade of pro- some information on pathways in the human brain, and one
jections into the entorhinal cortex and the organization of the can only hope that the wizardry of molecular biology will
entorhinal projections to the other hippocampal elds. One is someday provide new pathway-selective markers that allow
that the information the hippocampal formation is using has comparative studies of the rat, monkey, and human hip-
been highly preprocessed and integrated. Except for the olfac- pocampal formation. That being said, we next provide a short
tory input, little elemental or unimodal sensory information overview of the neuroanatomy of the human hippocampal
is directed to the hippocampal formation. A second, related formation.
implication is that it is unlikely that there is appreciable segre-
gation of sensory information processing in the hippocampal Cytoarchitectonic Organization
formation (i.e., visual information is not processed separately
from auditory information). In fact, the pattern of neu- The general appearance of the human hippocampal forma-
roanatomical connectivity predicts that the hippocampal for- tion is reminiscent of the monkey hippocampal formation
mation carries out whatever functions it accomplishes with (Figs. 32, 345, 354). Yet, differences are also apparent. For
high level, multimodal representations of sensory experiences. example, the CA1 eld of the hippocampus is even thicker in
humans than in monkeys (Fig. 349); in some regions, it is
3.6.3 Comparison of Monkey and as much as 30 cells thick. In addition to being thicker, the
Human Hippocampal Formation CA1 pyramidal cell layer takes on a distinctly multilaminate
appearance, with cells of different size and shape predominat-
At the risk of making points that may seem obvious, our ing at different depths of the layer. Unfortunately, there is lit-
understanding of the human hippocampus is necessarily tle information concerning the exact morphological or
primitive compared to that of the rat or the monkey. Because neurochemical characteristics of these neurons; there is not
standard experimental tract-tracing studies cannot be even a comprehensive Golgi analysis of the neurons of the
done, novel approaches for mapping connections in the human dentate gyrus and hippocampus.
 Hippocampal Neuroanatomy 105

Figure 354. Nissl-stained coronal sections through the human the hippocampus; ML, molecular layer of the dentate gyrus; o,
hippocampal formation, arranged from rostral (A) to caudal (F). A, stratum oriens of the hippocampus; ot, optic tract; p, pyramidal
amygdaloid complex; a, alveus; ab, angular bundle; ac, anterior cell layer of the hippocampus; PaS, parasubiculum; PHG, parahip-
commissure; AHA, amygdalo-hippocampal area; CA1, CA1 eld of pocampal gyrus (areas TF and TH); pl, polymorphic layer of the
the hippocampus; CA2, CA2 eld of the hippocampus; CA3, CA3 dentate gyrus; PRC, perirhinal cortex (areas 35 and 36); PrS,
eld of the hippocampus; cas, calcarine sulcus; cf, choroidal ssure; presubiculum; r, stratum radiatum of the hippocampus; RSP,
cos, collateral sulcus; DG, dentate gyrus; EC, entorhinal cortex; f, retrosplenial cortex; S, subiculum; ssa, sulcus semiannularis; V,
mbria; GL, granule cell layer of the dentate gyrus; hc, hippocampal temporal horn of the lateral ventricle. The numbers in B indicate
commissure; hf, hippocampal ssure; irs, intrarhinal sulcus; LGN, the layers of the entorhinal cortex. Bar in A  2 mm and applies
lateral geniculate nucleus: l-m: stratum lacunosum-moleculare of to all panels.
106 The Hippocampus Book

Another obvious difference is the cellular composition of 1997a,b). They found that both the mossy ber projection and
the region enclosed within the limbs of the granule cell layer. the Schaffer collateral system (from CA3 to CA1) can be
There are obviously far more cells in this region, and most labeled over short distances by DiI and that these projections
appear to be cells of the CA3 eld. The polymorphic layer of resemble those observed in the monkey. As important as these
the dentate gyrus (the hilus) forms a narrow band that lies just limited observations are, they highlight the fact that we do not
subjacent to the granule cell layer. Little is known about the currently know whether the neuroanatomical connections
morphology of cells in the human dentate gyrus. As indicated that have been described for the rat and the monkey are appli-
in the comparison between the rat and the monkey, however, cable to the human brain.
there are clear differences in the morphology of certain cell
types. As many as 30% of granule cells in the human dentate Magnetic Resonance Imaging
gyrus have basal dendrites (Fig. 350); this number may be of the Human Hippocampal Formation
even larger than in the monkey.
Even less is known about the human subiculum, pre- With the emergence of MRI studies of the human brain, there
subiculum, and parasubiculum. Although the distribution of have been numerous structural (Jack et al., 2003) and func-
various neurochemicals or receptors has been incidentally tional (Maguire et al., 2003) studies of the human hippocam-
illustrated during the course of large survey studies, almost pal formation. Studies have used both manual tracing
nothing has been written on the cellular morphology of these techniques as well as algorithms for automated segmentation.
regions. The human entorhinal cortex has attracted consider- Of course, without the benet of cytoarchitectonic guidance,
ably more attention because of its vulnerability in Alzheimers it has generally proven difficult if not impossible to differenti-
disease. Although the same general regions identied in the ate the component parts of the hippocampal formation. It has
monkey can be identied in the human, there is evidence even been complicated to dene accurately the boundaries of
for additional elds in the human entorhinal cortex. Classic regions, such as the entorhinal cortex (Insausti et al., 1998).
cytoarchitectonicists, for example, have dened as many as 27 Figure 355 presents a coronal section stained by the Nissl
divisions. method. To get a sense of the dimensions of this region and
what can be distinguished during a standard functional imag-
Connections ing study, we have placed a grid of 4-mm square boxes over
the medial temporal lobe. With resolution at this level (which
There is little information concerning the organization of might be typical for a standard functional imaging study), one
connections in the human hippocampal formation owing to might expect to have one voxel focused over the subiculum or
the fact that most neuroanatomical tracing techniques require perhaps two voxels over the entorhinal cortex. Other voxels,
injection of tracers into the living brain. Some investigators however, would clearly overlap adjacent elds, such as the
have relied instead on techniques such as local application of dentate gyrus and CA3. Positron emission tomography stud-
the lipophilic dye DiI to map short pathways in human post- ies have even poorer spatial resolution. Thus, given current
mortem material. Clifford Saper and colleagues have used functional imaging technologies, it is difficult to dene spe-
this technique to study the connections of the granule cells cic activations for dened subelds of the hippocampal
and the projections of the enclosed portion of CA3 (Lim et al., formation.

Figure 355. Nissl-stained coronal sections through

the human hippocampal formation at the caudal
aspect of the uncus. The grid overlay consists of
squares that are 4 mm on edge. For abbreviations, see
Figure 311. (Source: Adapted from Amaral, 1999.)
 Hippocampal Neuroanatomy 107

Critical Comparison of Neuroanatomy of Rat, Monkey, and formation is both massively divergent and convergent. Cells
Human Hippocampal Formation Must Await Comparable located in a focused point in the entorhinal cortex, for exam-
Quantitative Data for Each Species ple, can inuence neurons in as much as 50% of the entire
This chapter has dealt primarily with the rat hippocampal for- septotemporal axis of the dentate gyrus. Cells in the polymor-
mation because, as we pointed out at the beginning, much of phic layer of the dentate gyrus, in turn, can interconnect dif-
the neuroanatomical information available has been gained ferent levels of the dentate gyrus, providing further spread of
from studies of the rat. Furthermore, much of the electro- information to as much as 75% of the septotemporal axis.
physiological and behavioral literature discussed throughout Similarly, divergent projections have been described for the
the rest of the book is based on studies in the rat. With the CA3 to CA1 connections and for the CA1 to subiculum
dramatic increase in the use of noninvasive imaging of the connections. Thus, a perspective most consistent with the
human hippocampal formation, particularly in studies of known neuroanatomy is that the hippocampal formation
memory, the need has grown for more detailed information contains a series of three-dimensional networks of connec-
about the neuroanatomical organization of the human hip- tions. Depending on the synaptic interactions of these con-
pocampal formation. This is also important as increasingly nections (excitatory versus inhibitory), as well as the setting of
sophisticated models of hippocampus-related pathology are the myriad physiological parameters described in Chapter 5
introduced based primarily on rodent studies. A variety of (e.g., density and efficiency of synapses, transmitter release,
models for temporal lobe epilepsy, for example, have been synaptic plasticity), they might focus information processing
advanced based on cell degeneration and ber sprouting in to a specic level, recruit neurons at distant levels into the
the rat hippocampus. However, if the fundamental circuit dia- ongoing processing, or have no effect at all. Presumably, the
gram of the rat hippocampal formation is different from that role of this distributed network will become better under-
of humans, models based on the rat may be misleading. stood as increasingly massive parallel electrophysiological
Unfortunately, the extent of the similarities and differences recording procedures are employed.
in the hippocampal formation of the rat, monkey, and human
cannot yet be accurately gauged. Based on the few examples of 3.7.2 Functional Implications of the
established species differences described previously, it would Septotemporal Topography of Connections
not be surprising if substantial differences exist in the cellular
morphology, connectivity, and chemical neuroanatomy of the Given the septotemporal topography observed throughout
hippocampal formation across species. The relevance of these the hippocampal formation, there may still be some segre-
differences to the normal function and the pathology of the gation of functional divisions in this system. Although
hippocampal function will be an interesting area of compara- the massively divergent/convergent nature of hippocampal
tive neuroanatomy in the future. connections is inconsistent with a high degree of spatially
restricted information processing, the possibility remains that
there are a few septotemporally positioned domains of rela-
tively restricted information processing, at least at early stages,
3.7 Principles of Hippocampal Connectivity through the hippocampal circuitry. It appears that there are at
and Implications for Information Processing least three separable domains of neurons in the entorhinal
cortex: a lateral region, a mid region and a medial region. The
Neuroanatomy has the unfortunate characteristic of being most laterally situated domain of neurons innervates the
detail-oriented, and it is easy to lose sight of the forest for the septal half of the dentate gyrus and other innervated hip-
trees when absorbing and recounting information on the cells pocampal elds; the mid region innervates the next quarter;
and connections of the hippocampal formation. In the and the medial domain innervates the temporal quarter of the
following sections, we take a somewhat more integrative dentate gyrus (Fig. 341). Because the lateral portion of the
approach and attempt to draw implications from what we rat entorhinal cortex receives the bulk of the input from other
have described so far. These implications are generally quite neocortical areas, it is reasonable to assume that septal levels
speculative at this point, but they concern the important issue of the hippocampal formation are more highly involved with
of the functional correlates of the known hippocampal neu- processing exteroceptive sensory information. Because the
roanatomy. medial portions of the rat entorhinal cortex are preferentially
innervated by structures such as the amygdaloid complex, the
3.7.1 Highly Distributed Three-Dimensional temporal portion of the hippocampal formation may prefer-
Network of Intrinsic Connections entially deal with interoceptive or emotional information.
Because virtually all of the in vivo electrophysiological analy-
What is the hallmark of hippocampal connectivity? Beyond sis of the rat hippocampal formation has been carried out in
the dening aspect of unidirectional connectivity between the septal portion (and most of the in vitro work was carried
individual components, perhaps the facet that is most striking out to give slices from the rostal and mid-septotemporal
is that each of the intrinsic connections of the hippocampal levels), it is of interest to determine whether fundamentally
108 The Hippocampus Book

different types of response patterns are observed when tem- jections to these elds. The lateral entorhinal cortex projects
porally situated neurons are probed. Several lines of behav- to the border region of CA1 and the subiculum, whereas the
ioral data from lesioned animals are already consistent with medial entorhinal cortex projects either more proximally in
this idea (see Chapter 14). CA1 or more distally in the subiculum. Layer III neurons of
the medial entorhinal cortex are heavily inuenced by pre-
3.7.3 Functional Implications of the Transverse subicular input and thus by the thalamus, whereas layer III
Topography of Connections neurons in the lateral entorhinal area are heavily innervated
by the amygdaloid complex.
Despite the fact that intrinsic hippocampal connections have Taken together, this organization suggests that CA1 cells
an extensive septotemporal distribution, they are not uni- located close to the CA3 eld receive information from a sub-
formly distributed along the transverse axis. This organization set of CA3 cells located close to CA2 and a subset of layer III
raises the possibility of partially independent channels of cells located in the medial entorhinal cortex. CA1 cells located
information ow through and out of the hippocampal forma- closer to the subiculum, in contrast, receive preferential input
tion (Fig. 356). As we mentioned previously, it is not the case from CA3 cells located close to the dentate gyrus and from
that every CA1 cell has equal probability of input from every layer III cells in the lateral entorhinal cortex. A prediction
CA3 cell. The intrinsic circuitry of the hippocampal formation would be, therefore, that the response properties of CA1 cells
appears to be organized such that cells located at a particular at different transverse positions, at a particular septotemporal
transverse position in a eld are much more likely to be con- level, should be different.
nected with cells located at a particular transverse position of The notion of transverse topography of hippocampal con-
the innervated eld. This allows the possibility, therefore, that nections is made all the more compelling when it is appreci-
there is some channeling of information processing through ated that the subicular output is also organized in a columnar
the various hippocampal elds. As described fully above, cells fashion, with projections to different brain regions, or differ-
located proximally in CA3 tend to project to the most distal ent parts of the same brain region, originating from the prox-
CA1 cells. Projections from the distal portion of CA3, in con- imal, middle, and distal thirds of the subiculum. Neurons in
trast, terminate mainly in the proximal portion of CA1. The the proximal third of the subiculum project to the infralimbic
midportion of CA3 lls in the spaces between these two pro- and prelimbic cortices, the nucleus accumbens, and the lateral
jections. The CA1 projection to the subiculum demonstrates septal region. Projections from this portion of the ventral part
even more striking transverse topography. The proximal por- of the subiculum also project to the ventromedial nucleus of
tion of CA1 projects discretely to the distal third of the subicu- the hypothalamus and to the amygdala. The midtransverse
lum; the distal portion of CA1 projects just across the border portion of the subiculum projects mainly to the midline thal-
into the proximal third of the subiculum; and the middle por- amic nuclei, and neurons in the distal portion of the subicu-
tion of CA1 projects to the midregion of the subiculum. lum project to the retrosplenial portion of the cingulate cortex
The tripartite transverse organization in CA1 and the and the presubiculum. Although all portions of the subiculum
subiculum also appears to be respected by the entorhinal pro- project to the entorhinal cortex, the pattern of projections

Figure 356. Summary of the transverse

organization of the connections through
the hippocampal formation. This gure
highlights the possibility that informa-
tion is segregated, or channeled,
through the hippocampal formation and
ultimately reaches different recipients of
hippocampal output.
 Hippocampal Neuroanatomy 109

reciprocates the topography of the perforant path projections

to the subiculum. Thus, the proximal portion of the subicu-
lum projects to the lateral entorhinal area, and more distal
portions of the subiculum project to the medial entorhinal
area. Thus, from CA3 on through the entorhinal cortex, there
exists the potential for information that arrives at the earliest
point in this chain of connections to maintain some uniquess
from information that arrives at different portions of CA3.
The functional signicance of this neuroanatomical pathway
segregation is unknown at present.
To summarize, although the intrinsic connectivity of the
hippocampal formation is highly divergent, the network that
is formed is not without some topography. The septotempo-
ral and transverse components of this topography provide the
increased probability of communication between certain neu-
rons in each eld and thus the possibility of some form of
channeling or segregation of information processing in the
hippocampal formation.

3.7.4 Serial and Parallel Processing Figure 357. Summary of the serial and parallel pathways through
in the Hippocampal Formation the hippocampal formation. Although the intrinsic hippocampal
circuitry is largely unidirectional, it contains both serial and parallel
A unique feature of the intrinsic hippocampal circuitry is the projections. See text for details.
largely unidirectional organization of the projections that
interconnect the various hippocampal regions. A popular
notion is that these unidirectional projections also imply an
exclusively serial or sequential ow of informationrst
from the entorhinal cortex to the dentate gyrus, then to the 3.8 Conclusions
CA3 eld of the hippocampus, and so on. However, the data
summarized in the body of this chapter clearly indicate that Although this chapter has presented a substantial amount of
the intrinsic hippocampal circuitry has both serial and paral- neuroanatomical information, it has only begun to scratch
lel projections (Fig. 357). the surface of what is available in the primary literature. How-
The entorhinal cortex, in particular, contributes many ever, beause we have focused on general features and princi-
parallel projections to several elds of the hippocampal for- ples of organization, it should be more than adequate for
mation. The same layer II entorhinal cells probably give rise delving into the molecular and cellular aspects of hippocam-
to projections that terminate in both the dentate gyrus and pal organization (see Chapters 5 and 7) as well as the function
the CA3 eld of the hippocampus. Thus, whatever the of the hippocampal formation in the living organism. The
information conveyed by layer II entorhinal cells, it arrives neuroanatomy of the hippocampal formation is unique and
both monosynaptically and disynaptically (through mossy predicts that the behavioral functions it subserves are
ber intermediaries) at CA3. It is not known, of course, undoubtedly unique as well.
whether information from a single entorhinal cell reaches
a particular CA3 cell both monosynaptically and disynapti-
cally. This is an important question to resolve in future ACKNOWLEDGMENTS
studies.
The existence of prominent associational connections in The authors are indebted to the many colleagues who have worked in
the dentate gyrus, hippocampus, and entorhinal cortex also our laboratory over the years. Much of the knowledge they have
provides the substrate for polysynaptic activation in hip- uncovered has gone into the pages of this chapter. We are grateful
pocampal circuits. The functional implication of this more particularly to Dr. Menno Witter and Dr. Ricardo Insausti, who not
complex circuitry is that each hippocampal region is not only co-authored earlier comprehensive summaries on the neu-
roanatomy of the rat and human hippocampal formation, but also
entirely dependent on the preceding region for input, which
reviewed the current chapter to help ensure its accuracy. We also
raises the prospect that each region may be acting semi-inde-
thank Dr. Lazlo Seress for additional comments on the manuscript.
pendently from, as well as in concert with, other hippocampal The original research summarized in this chapter has been supported
elds. Hippocampal neuroanatomy is thus consistent with the by the National Institutes of Health and the Human Frontier Science
electrophysiological ndingsfor example, that CA1 place Program. Finally, we are grateful to our families who understand that
elds are apparently normal even after pharmacological inac- science requires some sacrice of time with them. Their support has
tivation of the dentate gyrus. made this chapter possible.
110 The Hippocampus Book

REFERENCES Brodal A (1947) The hippocampus and the sense of smell: a review.
Brain 70:179222.
Acsady L, Kamondi A, Sik A, Freund T, Buzsaki G (1998) GABAergic Brodmann K (1909) Vergleichende Lokalisationslehre der Grosshirn-
cells are the major postsynaptic targets of mossy bers in the rat rinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues.
hippocampus. J Neurosci 18:33863403. Leipzig.
Allen GV, Hopkins DA (1989) Mamillary body in the rat: topography Buckmaster PS, Amaral DG (2001) Intracellular recording and label-
and synaptology of projections from the subicular complex, ing of mossy cells and proximal CA3 pyramidal cells in macaque
prefrontal cortex, and midbrain tegmentum. J Comp Neurol monkeys. J Comp Neurol 430:264281.
286:311336. Buckmaster PS, Strowbridge BW, Kunkel DD, Schmiege DL,
Amaral DG (1978) A Golgi study of cell types in the hilar region of Schwartzkroin PA (1992) Mossy cell axonal projections to the
the hippocampus in the rat. J Comp Neurol 182:851914. dentate gyrus molecular layer in the rat hippocampal slice.
Amaral DG (1999) Introduction: what is where in the medial tempo- Hippocampus 2:349362.
ral lobe? Hippocampus 9:16. Buckmaster PS, Wenzel HJ, Kunkel DD, Schwartzkroin PA (1996)
Amaral DG, Dent JA (1981) Development of the mossy bers of Axon arbors and synaptic connections of hippocampal mossy
the dentate gyrus. I. A light and electron microscopic study of cells in the rat in vivo. J Comp Neurol 366:271292.
the mossy fibers and their expansions. J Comp Neurol Burwell RD (2001) Borders and cytoarchitecture of the perirhinal
195:5186. and postrhinal cortices in the rat. J Comp Neurol 437:1741.
Amaral DG, Kurz J (1985) An analysis of the origins of the choliner- Burwell RD, Amaral DG (1998a) Perirhinal and postrhinal cortices of
gic and noncholinergic septal projections to the hippocampal the rat: interconnectivity and connections with the entorhinal
formation of the rat. J Comp Neurol 240:3759. cortex. J Comp Neurol 391:293321.
Amaral DG, Witter MP (1989) The three dimensional organization Burwell RD, Amaral DG (1998b) Cortical afferents of the perirhinal,
of the hippocampal formation: a review of anatomical data. postrhinal, and entorhinal cortices of the rat. J Comp Neurol
Neuroscience 31:571591. 398:179205.
Amaral DG, Insausti R, Cowan WM (1984) The commissural con- Burwell RD, Witter MP (2002) Basic anatomy of the parahippocampal
nections of the monkey hippocampal formation. J Comp Neurol region in monkeys and rats. Oxford: Oxford University Press.
224:307336. Burwell RD, Witter MP, Amaral DG (1995) Perirhinal and postrhinal
Amaral DG, Dolorfo C, Alvarez-Royo P (1991) Organization of CA1 cortices of the rat: a review of the neuroanatomical literature
projections to the subiculum: a PHA-L analysis in the rat. Hip- and comparison with findings from the monkey brain.
pocampus 1:415436. Hippocampus 5:390408.
Andersen P, Bliss TVP, Skrede KK (1971) Lamellar organization of Caballero-Bleda M, Witter MP (1993) Regional and laminar organi-
hippocampal excitatory pathways. Exp Brain Res 13:222238. zation of projections from the presubiculum and parasubicu-
Arantius G (1587) De humano foetu . . . Ejusdem anatomicorum obser- lum to the entorhinal cortex: an anterograde tracing study in the
vationum liber. Venice. rat. J Comp Neurol 328:115129.
Baisden RH, Woodruff ML, Hoover DB (1984) Cholinergic and non- Canteras N, Swanson L (1992) Projections of the ventral subicu-
cholinergic septo-hippocampal projections: a double-label lum to the amygdala, septum, and hypothalamus: a PHAL
horseradish peroxidase-acetylcholinesterase study in the rabbit. anterograde tract-tracing study in the rat. J Comp Neurol
Brain Res 290:146151. 324:180194.
Bakst I, Avendao C, Morrison JH, Amaral DG (1986) An experi- Claiborne BJ, Amaral DG, Cowan WM (1986) A light and electron
mental analysis of the origins of the somatostatin immunoreac- microscopic analysis of the mossy bers of the rat dentate gyrus.
tive bers in the dentate gyrus of the rat. J Neurosci 6:14521462. J Comp Neurol 246:435458.
Blackstad TW (1956) Commissural connections of the hippocampal Claiborne BJ, Amaral DG, Cowan WM (1990) A quantitative three-
region in the rat, with special reference to their mode of termi- dimensional analysis of granule cell dendrites in the rat dentate
nation. J Comp Neurol 105:417537. gyrus. J Comp Neurol 302:206219.
Blackstad TW, Brink K, Hem J, Jeune B (1970) Distribution of hip- Conrad LCA, Leonard CM, Pfaff DW (1974) Connections of the
pocampal mossy bers in the rat: an experimental study with median and dorsal raphe nuclei in the rat: an autoradiographic
silver impregnation methods. J Comp Neurol 138:433450. and degeneration study. J Comp Neurol 156:179206.
Borhegyi Z, Leranth C (1997) Distinct substance P- and calretinin- Daitz HM, Powell TPS (1954) Studies of the connexions of the fornix
containing projections from the supramammillary area to the system. J Neurol Neurosurg Psychiatry 17:7582.
hippocampus in rats: a species difference between rats and De Garengeot RJC (1742) Splanchnologie ou lanatomie des visceres,
monkeys. Exp Brain Res 115:369374. 2nd ed, pp 250251. Paris: C. Osmond.
Boyett JM, Buckmaster PS (2001) Somatostatin-immunoreactive Deller T (1998) The anatomical organization of the rat fascia dentata:
interneurons contribute to lateral inhibitory circuits in the den- new aspects of laminar organization as revealed by anterograde
tate gyrus of control and epileptic rats. Hippocampus tracing with Phaseolus vulgaris-Leucoagglutinin (PHAL). Anat
11:418422. Embryol (Berl) 197:89103.
Braak H (1972) Zur Pigmentarchitektonic der Grosshirnrinde des Deller T, Adelmann G, Nitsch R, Frotscher M (1996) The alvear path-
Menschen. I. Regio entorhinalis. Z Zellforsch Mikrosk Anat way of the rat hippocampus. Cell Tissue Res 286:293303.
127:407438. Demeter S, Rosene DL, Van Hoesen GW (1985) Interhemispheric
Braak H (1980) Architectonics of the human telencephalic cortex. New pathways of the hippocampal formation, presubiculum and
York: Springer. entorhinal and posterior parahippocampal cortices in the rhe-
Broca P (1878) Anatomie compare des circonvolutions crbrales: le sus monkey: the structure and organization of the hippocampal
grand lobe limbique. Rev. Anthropol 1:385498. commissures. J Comp Neurol 233:3047.
 Hippocampal Neuroanatomy 111

Dent JA, Galvin NJ, Staneld BB, Cowan WM (1983) The mode of pocampus and dentate gyrus of the rat. J Comp Neurol
temination of the hypothalamic projection to the dentate gyrus: 149:393420.
an EM autoradiographic study. Brain Res 258:110. Greene JRT, Totterdell S (1997) Morphology and distribution of elec-
Desmond NL, Levy WB (1985) Granule cell dendritic spine density trophysiologically dened classes of pyramidal and nonpyrami-
in the rat hippocampus varies with spine shape and location. dal neurons in rat ventral subiculum in vitro. J Comp Neurol
Neurosci Lett 54:219224. 380:395408.
Dolleman-Van der Weel MJ, Witter MP (2000) Nucleus reuniens Groenewegen HJ, Vermeulen-Van Der Zee E, Te Kortschot A, Witter
thalami innervates gamma aminobutyric acid positive cells in M (1987) Organization of the projections from the subiculum
hippocampal eld CA1 of the rat. Neurosci Lett 278:145148. to the ventral striatum in the rat: a study using anterograde
Dolorfo CL, Amaral DG (1998a) Entorhinal cortex of the rat: topo- transport of Phaseolus vulgaris leucoagglutinin. Neuroscience
graphic organization of the cells of origin of the perforant path 23:103120.
projection to the dentate gyrus. J Comp Neurol 398:2548. Haglund L, Swanson LW, Khler C (1984) The projection of the
Dolorfo CL, Amaral DG (1998b) Entorhinal cortex of the rat: organ- supramammillary nucleus to the hippocampal formation: an
ization of intrinsic connerctions. J Comp Neurol 398:4982. immunohistochemical and anterograde transport study with
Donovan MK, Wyss JM (1983) Evidence for some collateralization the lectin PHA-L in the rat. J Comp Neurol 229:171185.
between cortical and diencephalic efferent axons of the rat Hamam BN, Amaral DG, Alonso AA (2002) Morphological and elec-
subicular cortex. Brain Res 259:181192. trophysiological characteristics of layer V neurons of the rat lat-
Duffy CJ, Rakic P (1983) Differentiation of granule cell dendrites in eral entorhinal cortex. J Comp Neurol 451:4561.
the dentate gyrus of the rhesus monkey: a quantitative Golgi Hamam BN, Kennedy TE, Alonso A, Amaral DG (2000)
study. J Comp Neurol 214:224237. Morphological and electrophysiological characteristics of layer
Felleman DJ, Van Essen DC (1991) Distributed hierarchical process- V neurons of the rat medial entorhinal cortex. J Comp Neurol
ing in the primate cerebral cortex. Cereb Cortex 1:147. 418:457472.
Freund TF, Buzsaki G (1996) Interneurons of the hippocampus. Han ZS, Buhl EH, Lorinczi Z, Somogyi P (1993) A high degree of
Hippocampus 6:347470. spatial selectivity in the axonal and dendritic domains of physi-
Fricke R, Cowan WM (1978) An autoradiographic study of the com- ologically identied local-circuit neurons in the dentate gyrus of
missural and ipsilateral hippocampo-dentate projections in the the rat hippocampus. Eur J Neurosci 5:395410.
adult rat. J Comp Neurol 181:253269. Harris E, Witter MP, Weinstein G, Stewart M (2001) Intrinsic con-
Frotscher M (1992) Application of the Golgi/electron microscopy nectivity of the rat subiculum. I. Dendritic morphology and
technique for cell identication in immunocytochemical, retro- patterns of axonal arborization by pyramidal neurons. J Comp
grade labeling, and developmental studies of hippocampal neu- Neurol 435:490505.
rons. Microsc Res Tech 23:306323. Haug FM (1976) Sulphide silver pattern and cytoarchitectonics of
Funahashi M, Stewart M (1997a) Presubicular and parasubicular cor- parahippocampal areas in the rat. Special reference to the subdi-
tical neurons of the rat: electrophysiological and morphologi- vision of area entorhinalis (area 28) and its demarcation from
cal properties. Hippocampus 7:117129. the pyriform cortex. Adv Anat Embryol Cell Biol 52:373.
Funahashi M, Stewart M (1997b) Presubicular and parasubicular Hausheer-Zarmakupi Z, Wolfer DP, Leisinger-Trigona MC, Lipp HP
cortical neurons of the rat: functional separation of deep and (1996) Selective breeding for extremes in open-eld activity of
supercial neurons in vitro. J Physiol (Lond) 501:387403. mice entails a differentiation of hippocampal mossy bers.
Gaarskjaer FB (1978a) Organization of the mossy ber system of the Behav Genet 26:167176.
rat studied in extended hippocampi. I. Terminal area related to Henze DA, Urban NN, Barrioneuvo G (2000) The multifarious hip-
number of granule and pyramidal cells. J Comp Neurol pocampal mossy eer pathway: a review. Neuroscience 98:
178:4972. 407427.
Gaarskjaer FB (1978b) Organization of the mossy ber system of the Herkenham M (1978) The connections of the nucleus reuniens thal-
rat studied in extended hippocampi. II. Experimental analysis of ami: evidence for a direct thalamo-hippocampal pathway in the
ber distribution with silver impregnation methods. J Comp rat. J Comp Neurol 177:589610.
Neurol 178:7388. Hjorth-Simonsen A, Jeune B (1972) Origin and termination of the
Gaykema RPA, Luiten PGM, Nyakas C, Traber J (1990) Cortical pro- hippocampal perforant path in the rat studied by silver impreg-
jection patterns of the medial septum-diagonal band complex. nation. J Comp Neurol 144:215232.
J Comp Neurol 293:103124. Hjorth-Simonsen A, Zimmer J (1975) Crossed pathways from the
Gertz SD, Lindenberg R, Piavis GW (1972) Structural variations entorhinal area to the fascia dentata. J Comp Neurol 161:
in the rostral human hippocampus. Johns Hopkins Med J 130: 5770.
367376. Insausti R, Amaral DG, Cowan WM (1987) The entorhinal cortex of
Gloveli T, Schmitz D, Empson RM, Dugladze T, Heinemann U (1997) the monkey. II. Cortical afferents. J Comp Neurol 264:356395.
Morphological and electrophysiological characterization of Insausti R, Herrero MT, Witter MP (1997) Entorhinal cortex of the
layer III cells of the medial entorhinal cortex of the rat. rat: cytoarchitectonic subdivisions and the origin and distribu-
Neuroscience 77:629648. tion of cortical efferents. Hippocampus 7:146183.
Goldowitz D, White WF, Steward O, Lynch G, Cotman C (1975) Insausti R, Juottonen K, Soininen H, Insausti AM, Partanen K, Vainio
Anatomical evidence for a projection from the entorhinal cortex P, Laakso MP, Pitkanen A (1998) MR volumetric analysis of the
to the contralateral dentate gyrus of the rat. Exp Neurol human entorhinal, perirhinal, and temporopolar cortices. AJNR
47:433441. Am J Neuroradiol 19:659671.
Gottlieb DI, Cowan WM (1973) Autoradiographic studies of the Ishizuka N (2001) Laminar organization of the pyramidal cell layer
commissural and ipsilateral association connections of the hip- of the subiculum in the rat. J Comp Neurol 435:89110.
112 The Hippocampus Book

Ishizuka N, Weber J, Amaral DG (1990) Organization of intrahip- hippocampal formation in humans. I. The mossy ber pathway.
pocampal projections originating from CA3 pyramidal cells in J Comp Neurol 385:325351.
the rat. J Comp Neurol 295:580623. Lim C, Mufson EJ, Kordower JH, Blume HW, Madsen JR, Saper
Ishizuka N, Cowan WM, Amaral DG (1995) A quantitative analysis CB (1997b) Connections of the hippocampal formation in
of the dendritic organization of pyramidal cells in the rat hip- humans. II. The endfolial ber pathway. J Comp Neurol 385:
pocampus. J Comp Neurol 362:1745. 352371.
Jack CR Jr, Slomkowski M, Gracon S, Hoover TM, Felmlee JP, Stewart Lipp HP, Schwegler H, Heimrich B, Cerbone A, Sadile AG (1987)
K, Xu Y, Shiung M, OBrien PC, Cha R, Knopman D, Petersen Strain-specic correlations between hippocampal structural
RC (2003) MRI as a biomarker of disease progression in a ther- traits and habituation in a spatial novelty situation. Behav Brain
apeutic trial of milameline for AD. Neurology 60:253260. Res 24:111123.
Kaitz SS, Robertson RT (1981) Thalamic connections with limbic cor- Lorente de N R (1933) Studies on the structure of the cerebral cor-
tex. II. Corticothalamic projections. J Comp Neurol 195:527545. tex. I. The area entorhinalis. J Psychol Neurol 45:381438.
Kempermann G, Kuhn G, Gage FH (1997) More hippocampal neu- Lorente de N R (1934) Studies on the structure of the cerebral
rons in adult mice living in an enriched environment. Nature cortex. II. Continuation of the study of the ammonic system.
386:493495. J Psychol Neurol 46:113177.
Kempermann G, Kuhn HG, Gage FH (1998) Experienced-induced Loughlin SE, Foote SL, Grzanna R (1986) Efferent projections of
neurogenesis in the senescent dentate gyrus. J Neurosci nucleus locus coeruleus: morphologic subpopulations have dif-
18:32063212. ferent efferent targets. Neuroscience 18:307319.
Kiss J, Csaki A, Bokor H, Shanabrough M, Leranth C (2000) The Lubke J, Deller T, Frotscher M (1997) Septal innervation of mossy
supramamillo-hippocampal and supramammillo-septal gluta- cells in the hilus of the rat dentate gyrus: an anterograde tracing
matergic/aspartatergic projections in the rat: a combined and intracellular labeling study. Exp Brain Res 114:423432.
[3H]D-aspartate autoradiographic and immunohistochemical Magloczky Z, Acsady L, Freund TF (1994) Principal cells are the post-
study. Neuroscience 97:657669. synaptic targets of supramammillary afferents in the hippocam-
Klink R, Alonso A (1997) Morphological characteristics of layer II pus of the rat. Hippocampus 4:322334.
projection neurons in the rat medial entorhinal cortex. Maguire EA, Spiers HJ, Good CD, Hartley T, Frackowiak RS, Burgess
Hippocampus 7:571583. N (2003) Navigation expertise and the human hippocampus: a
Kloosterman F, Witter Menno P, van Haeften T (2003) Topo- structural brain imaging analysis. Hippocampus 13:250259.
graphical and laminar organization of subicular projections McEwen BS, Wallach G (1973) Corticosterone binding to hippocam-
to the parahippocampal region of the rat. J Comp Neurol 455: pus: nuclear and cytosol binding in vitro. Brain Res 57:373386.
156171. McLardy T, Kilmer WL (1970) Hippocampal circuitry. Am Psychol
Kohler C (1985) Intrinsic projections of the retrohippocampal region 25:563566.
in the rat brain. I. The subicular complex. J Comp Neurol Meibach RC, Siegel A (1975) The origin of fornix bers which proj-
236:504522. ect to the mammillary bodies in the rat: a horseradish peroxi-
Khler C, Steinbusch H (1982) Identication of serotonin and non- dase study. Brain Res 88:508512.
serotonin-containing neurons of the mid-brain raphe project- Moore RY, Halaris AE (1975) Hippocampal innervation by serotonin
ing to the entorhinal area and the hippocampal formation: a neurons of the midbrain raphe in the rat. J Comp Neurol
combined immunohistochemical and uorescent retrograde 164:171184.
tracing study in the rat brain. Neuroscience 7:951975. Morrison JH, Benoit, R., Magistretti PJ, Ling N, Bloom FE (1982)
Laatsch RH, Cowan WM (1967) Electron microscopic studies of the Immunohistochemical distribution of prosomatostatin-related
dentate gyrus of the rat. II. Degeneration of commissural affer- peptides in hippocampus. Neurosci Lett 34:137142.
ents. J Comp Neurol 130:241262. Mosko S, Lynch G, Cotman CW (1973) The distribution of septal
Lacaille JC, Schwartzkroin PA (1988) Stratum lacunosum-molecu- projections to the hippocampus of the rat. J Comp Neurol
lare interneurons of hippocampal CA1 region. I. Intracellular 152:163174.
response characteristics, synaptic responses, and morphology. Naber PA, Witter MP (1998) Subicular efferents are organized mostly
J Neurosci 8:14001410. as parallel projections: a double-labeling, retrograde-tracing
Lacaille JC, Mueller AL, Kunkel DD, Schwartzkroin PA (1987) Local study in the rat. J Comp Neurol 393:284297.
circuit interactions between oriens/alveus interneurons and Naber PA, da Silva FHL, Witter MP (2001) Reciprocal connections
CA1 pyramidal cells in hippocampal slices: electrophysiology between the entorhinal cortex and hippocampal elds CA1 and
and morphology. J Neurosci 7:19791993. the subiculum are in register with the projections from CA1 to
Laurberg S (1979) Commissural and intrinsic connections of the rat the subiculum. Hippocampus 11:99104.
hippocampus. J Comp Neurol 184:685708. Nafstad PHJ (1967) An electron microscope study on the termina-
Laurberg S, Sorensen KE (1981) Associational and commissural col- tion of the perforant path bres in the hippocampus and the
laterals of neurons in the hippocampal formation (hilus fasciae fascia dentata. Z Zellforsch Und Mikrosk Anat 76:532542.
dentatae and subeld CA3). Brain Res 212:287300. Nyakas C, Luiten PGM, Spencer DG, Traber J (1987) Detailed pro-
Lavenex P, Amaral DG (2000) Hippocampal-neocortical interaction: jection patterns of septal and diagonal band efferents to the hip-
a hierarchy of associativity. Hippocampus 10:420430. pocampus in the rat with emphasis on innervation of CA1 and
Li X-G, Somogyi P, Ylinen A, Buzsaki G (1994) The hippocampal dentate gyrus. Brain Res Bull 18:533545.
CA3 network: an in vivo intracellular labeling study. J Comp OMara SM, Commins S, Anderson M, Gigg J (2001) The subiculum:
Neurol 339:181208. a review of form, physiology and function. Prog Neurobiol
Lim C, Blume HW, Madsen JR, Saper CB (1997a) Connections of the 64:129155.
 Hippocampal Neuroanatomy 113

Papez JW (1937) A proposed mechanism of emotion. Arch Neurol Soriano E, Frotscher M (1989) A GABAergic axo-axonic cell in the
Psychiatry 38:725743. fascia dentata controls the main excitatory hippocampal path-
Pickel VM, Segal M, Bloom FE (1974) A radioautographic study of way. Brain Res 503:170174.
the efferent pathways of the nucleus locus coeruleus. J Comp Sorra KE, Harris KM (1993) Occurrence and three-dimensional
Neurol 155:1542. structure of multiple synapses between individual radiatum
Pikkarainen M, Ronkko S, Savander V, Insausti R, Pitkanen A (1999) axons and their target pyramidal cells in hippocampal area CA1.
Projections from the lateral, basal, and accessory basal nuclei of J Neurosci 13:37363748.
the amygdala to the hippocampal formation in rat. J Comp Steward O, Scoville SA (1976) Cells of origin of entorhinal cortical
Neurol 403:229260. afferents to the hippocampus and fascia dentata of the rat. J
Pitkanen A, Pikkarainen M, Nurminen N, Ylinen A (2000) Reciprocal Comp Neurol 169:347370.
connections between the amygdala and the hippocampal for- Struble RG, Desmond NL, Levy WB (1978) Anatomical evidence
mation, perirhinal cortex and postrhinal cortex in the rat. Ann for interlamellar inhibition in the fascia dentata. Brain Res 152:-
NY Acad Sci 911:369391. 580585.
Powell TPS, Guillery RW, Cowan WM (1957) A quantitative study of Suzuki WA, Amaral DG (1994) Perirhinal and parahippocampal cor-
the fornix-mamillo-thalamic system. J Anat 91:419437. tices of the macaque monkey: cortical afferents. J Comp Neurol
Pyapali GK, Sik A, Penttonen M, Buzsaki G, Turner DA (1998) 350:497533.
Dendritic properties of hippocampal CA1 pyramidal neurons in Swanson LW (1978a) The anatomical organization of septo-hip-
the rat: intracellular staining in vivo and in vitro. J Comp Neurol pocampal projections. In: Functions of the septo-hippocampal
391:335352. system, pp 2548. Amsterdam: Elsevier North Holland.
Raisman G, Cowan WM, Powell TPS (1965) The extrinsic afferent, Swanson LW, Cowan, WM (1975) Hippocampo-hypothalamic con-
commissural and association bres of the hippocampus. Brain nections: origin in subicular cortex, not Ammons horn. Science
88:963997. 189:303304.
Ramn y Cajal S (1893) Estructura del asta de Ammon y fascia den- Swanson LW, Hartman BK (1975) The central adrenergic system:
tata. Ann Soc Esp Hist Nat 22. an immunouorescence study of the location of cell bodies
Rapp PR, Gallagher M (1996) Preserved neuron number in the hip- and their efferent connections in the rat utilizing dopamine-
pocampus of aged rats with spatial learning decits. Proc Natl -hydroxylase as a marker. J Comp Neurol 163:467506.
Acad Sci USA 93:99269930. Swanson LW, Khler C (1986) Anatomical evidence for direct projec-
Ribak CE, Seress L (1983) Five types of basket cell in the hippocam- tions from the entorhinal area to the entire cortical mantle in
pal dentate gyrus: a combined Golgi and electron microscopic the rat. J Neurosci 6:30103023.
study. J Neurocytol 12:577597. Swanson LW, Wyss JM, Cowan WM (1978) An autoradiographic
Ribak CE, Vaughn JE, Saito K (1978) Immunocytochemical localiza- study of the organization of intrahippocampal association path-
tion of glutamic acid decarboxylase in neuronal somata follow- ways in the rat. J Comp Neurol 181:681716.
ing colchicine inhibition of axonal transport. Brain Res Swanson LW, Sawchenko PE, Cowan WM (1981) Evidence for collat-
140:315332. eral projections by neurons in ammons horn, the dentate gyrus,
Ribak CE, Seress L, Amaral DG (1985) The development, ultrastruc- and the subiculum: a multiple retrograde labeling study in the
ture and synaptic connections of the mossy cells of the dentate rat. J Neurosci 1:548559.
gyrus. J Neurocytol 14:835857. Swanson LW, Khler C, Bjorklund A (1987) The limbic region. I. The
Robertson RT, Kaitz SS (1981) Thalamic connections with limbic septohippocampal system. In: Handbook of chemical neu-
cortex. I. Thalamocortical projections. J Comp Neurol roanatomy (Bjorklund A, Hkfelt T, Swanson LW, eds), pp
195:501525. 125227. Amsterdam: Elsevier.
Sarter M, Markowitsch HJ (1985) Convergence of intra- and inter- Tamamaki N, Nojyo Y (1993) Projection of the entorhinal layer II
hemispheric cortical afferents: lack of collateralization and evi- neurons in the rat as revealed by intracellular pressure-injection
dence for subrhinal cell group projecting heterotopically. J Comp of neurobiotin. Hippocampus 3:471480.
Neurol 236:283296. Tamamaki N, Abe K, Nojyo Y (1987) Columnar organization in the
Seress L (1992) Morphological variability and developmental subiculum formed by axon branches originating from single
aspects of monkey and human granule cells: differences CA1 pyramidal neurons in the rat hippocampus. Brain Res
between the rodent and primate dentate gyrus. Epilepsy Res 412:156160.
Suppl 7:328. Thompson SM, Robertson RT (1987) Organization of subcortical
Seress L, Mrzljak L (1987) Basal dendrites of granule cells are normal pathways for sensory projections to the limbic cortex. I.
features of the fetal and adult dentate gyrus of both monkey and Subcortical projections to the medial limbic cortex in the rat.
human hippocampal formations. Brain Res 405:169174. J Comp Neurol 265:175188.
Seress L, Pokorny J (1981) Structure of the granular layer of the rat Trommald M, Hulleberg G, Andersen P (1996) Long-term potentia-
dentate gyrus: a light microscopic and Golgi study. J Anat tion is associated with new excitatory spine synapses on rat den-
133:181195. tata granule cells. Learn Mem 3:218228.
Shipley MT (1975) The topographical and laminar organization of Van der Werf YD, Witter MP, Groenewegen HJ (2002) The intralam-
the presubiculums projection to the ipsi- and contralateral inar and midline nuclei of the thalamus: anatomical and func-
entorhinal cortex in the guinea pig. J Comp Neurol 160:127146. tional evidence for participation in processes of arousal and
Sik A, Penttonen M, Buzsaki G (1997) Interneurons in the hip- awareness. Brain Res Brain Res Rev 39:107140.
pocampal dentate gyrus: an in vivo intracellular study. Eur J Van Groen T, Wyss JM (1990) The connections of presubiculum and
Neurosci 9:573588. parasubiculum in the rat. Brain Res 518:227243.
114 The Hippocampus Book

Van Groen T, Kadish I, Wyss JM (2002) Species differences in the pro- Witter MP (1993) Organization of the entorhinal-hippocampal
jections from the entorhinal cortex to the hippocampus. Brain system: a review of current anatomical data. Hippocampus 3:
Res Bull 57:553-556. 3344.
Van Hoesen GW, Hyman BT, Damasio AR (1991) Entorhinal cortex Witter MP, Amaral DG (2004) The hippocampal region. In: The rat
pathology in Alzheimers disease. Hippocampus 1:18. brain (Paxinos G, ed, 3rd ed), pp 637670. San Diego, CA,
Van Groen T, Miettinen P, Kadish I (2003) The entorhinal cortex of London: Elsevier Academic Press.
the mouse: organization of the projection to the hippocampal Witter MP, Groenewegen HJ (1990) The subiculum: cytoarchitec-
formation. Hippocampus 13:133149. tonically a simple structure, but hodologically complex. In:
Vertes RP (1993) PHA-L analysis of projections from the supramam- Understanding the brain through the hippocampus (Storm-
millary nucleus in the rat. J Comp Neurol 326:595622. Mathisen J, Zimmer J, Ottersen OP, eds). Amsterdam:
Vertes RP, Fortin WJ, Crane AM (1999) Projections of the median Elsevier.
raphe nucleus in the rat. J Comp Neurol 407:555582. Witter MP, Griffioen AW, Jorritsma-Byham B, Krijnen JLM (1988)
Verwer RWH, Meijer RJ, Vanuum HFM, Witter MP (1997) Collateral Entorhinal projections to the hippocampal CA1 region in the
projections from the rat hippocampal formation to the lateral rat: an underestimated pathway. Neurosci Lett 85:193198.
and medial prefrontal cortex. Hippocampus 7:397402. Witter MP, Ostendorf RH, Groenewegen HJ (1990) Heterogeneity in
Von Economo C (1929) The cytoarchitectonics of the human cerebral the dorsal subiculum of the rat: distinct neuronal zones project
cortex. New York: Oxford University Press. to different cortical and subcortical targets. Eur J Neurosci
Wainer BH, Levey AI, Rye DB, Mesulam MM, Mufson EJ (1985) 2:718725.
Cholinergic and non-cholinergic septohippocampal pathways. Witter MP, Wouterlood FG, Naber PA, van Haeften T (2000)
Neurosci Lett 54:4552. Structural organization of the parahippocampal-hippocampal
Walker MC, Ruiz A, Kullmann DM (2002) Do mossy bers release network. Ann NY Acad Sci 911:125.
GABA? Epilepsia 43:196202. Wouterlood FG, Saldana E, Witter MP (1990) Projection from
West MJ, Slomianka L, Gundersen HJG (1991) Unbiased stereologi- the nucleus reuniens thalami to the hippocampal region: light
cal estimation of the total number of neurons in the subdivi- and electron microscopic tracing study in the rat with the
sions of the rat hippocampus using the optical fractionator. anterograde tracer Phaseolus vulgaris-leucoagglutinin. J Comp
Anat Rec 231:482497. Neurol 296:179203.
White TD, Tan AM, Finch DM (1990) Functional reciprocal connec- Wyss JM, Van Groen T (1992) Connections between the retrosplenial
tions of the rat entorhinal cortex and subicular complex with cortex and the hippocampal formation in the rat: a review.
the medial frontal cortex: an in vivo intracellular study. Brain Hippocampus 2:112.
Res 533:95106. Wyss JM, Swanson LW, Cowan WM (1979) Evidence for an input to
Wilson CO, Isokawa-Akesson M, Babb TL, Engel J Jr, Cahan L D, the molecular layer and the stratum granulosum of the dentate
Crandall PH (1987) Comparative view of local and interhemi- gyrus from the supramammillary region of the hypothalamus.
spheric limbic pathways in humans: an evoked potential analy- Anat Embryol (Berl) 156:165176.
sis. In: Fundamental mechanisms of human brain function (Engel Wyss JM, Swanson LW, Cowan WM (1980) The organization of the
J Jr, Ojemann GA, Lders HO, Williamson PD, eds), pp 2338. mbria, dorsal fornix and ventral hippocampal commissure in
New York: Raven Press. the rat. Anat Embryol (Berl) 158:303316.
4 Michael Frotscher and Lszl Seress

Morphological Development of the Hippocampus

neurons are formed, what molecules and mechanisms deter-

4.1 Overview mine the fate of newly generated cells? For example, postmi-
totic neurons migrate to their nal destination. How does this
A fundamental question in neurobiology is how the various happen? Upon arrival in their appropriate structure or layer,
brain structures and the enormous number of specic the postmigratory neuroblasts start to develop their processes.
interneuronal connections that serve the complex functions of Why does only one axon originate from the cell body, whereas
the mature CNS are formed during development. Work on the many dendrites emerge from it? Moreover, as these processes
hippocampal formation has shed light on the principles emerge, they do so in an organized manner. For example, in
underlying this process. With the advent of modern genetic the various hippocampal elds, principal neurons display a
approaches, including mutant mice lacking or overexpressing rather uniform structure. What are the molecules and mecha-
genes important for developmental processes, new insights nisms controlling this uniform differentiation? Why do the
into the development of the hippocampal formation are in the nonprincipal cells that invade the hippocampus from the
process of being obtained. ganglionic eminence (Pleasure et al., 2000) form such a het-
Not all aspects of hippocampal development are covered in erogeneous population of neurons? Finally, what are the
this chapter. Traditionally, developmental processes are stud- determinants of synapse formation? The hippocampus is a
ied in a descriptive manner by monitoring the formation of laminated structure in which all bers originating from a par-
mature structures from simpler, undifferentiated stages. ticular afferent source terminate at identical dendritic seg-
Major contributions have also come from studies of the phy- ments. How is this layer-specic termination of hippocampal
logenetic development of the hippocampal formation and afferents to be explained? To what extent is neuronal activity
have demonstrated that it is a form of phylogenetically old involved in the differentiation of dendrites, dendritic spines,
cortex, the archicortex, that develops in the medial wall of the and synaptic contacts?
telencephalic vesicle. Owing to the expansion of the cerebral An attempt is rst made to summarize our present knowl-
cortex in higher vertebrates, the hippocampal formation is edge on neuronal generation and migration in the hippocam-
translocated medially and ventrally into the inferior portion pus. We then address the development of major hippocampal
of the lateral ventricle. In rodents, the primary focus of this connections by describing determinants of pathnding, target
chapter, much of the hippocampus is located dorsally, the layer recognition, and synapse formation of hippocampal
curved structure of the hippocampus reecting this rotation afferents.
of the telencephalon (see Chapter 3). Phylogenetic develop-
ment is not dealt with in detail here, although we do present
some information on species other than rodents where appro-
priate. For a fuller treatment of phylogenetic differences, the 4.2 Neurogenesis and Cell Migration
reader is referred to the exhaustive coverage by Stephan
(Stephan, 1975). Most of our knowledge about cell formation in the rodent
What are the fundamental issues regarding the develop- hippocampal formation dates back to the seminal studies by
ment of the hippocampus? As in other organs, it is important Angevine (1965) and Altman and coworkers (Altman, 1966;
to understand the signals governing cell proliferation. Once Altman and Das, 1966; Altman and Bayer, 1975; Bayer, 1980).

115
116 The Hippocampus Book

These important papers provided the rst descriptions of the gin might explain their characteristic morphological differ-
origin, the time course of the generation, and the laminar dis- ences. On E 14.5 and E 15.5, the newly formed pyramidal cells
tribution of principal neurons, pyramidal neurons, and gran- already show eld-specic genetic markers characteristic for
ule cells. the CA1 or CA3 area of the mouse hippocampus (Tole et al.,
1997).
4.2.1 Pyramidal Neurons At birth, the pyramidal layer in the rat is thick and com-
posed of 6 to 10 rows of neuronal somata. As the hippocam-
Stem cells of both pyramidal neurons and granule cells origi- pus gets larger postnatally, the pyramidal layer becomes
nate from the ventricular germinal layers (or ventricular neu- thinner; and in adult rats it comprises two or three cell layers.
roepithelial layers) that are located below the ventricular wall This process of pyramidal cell rearrangement is unlikely to be
along the CA1 area. In contrast to the neocortical ventricular a passive one due, for example, to the volumetric increase of
germinal layers, there is no subventricular zone at the hip- the hippocampal formation. Postnatally generated glial cells
pocampal germinal area. Therefore the multiplying neurons might contribute to the reorganization. In fact, neonatal X-ray
directly migrate from the ventricular zone to their nal target irradiation, which has its greatest impact on glial cells, pre-
region (Altman and Bayer, 1990b). Pyramidal neurons of the vents formation of the mature pyramidal cell layer (Czurk et
rodent hippocampus are generated between embryonic days al., 1997). However, early-generated reelin-synthesizing Cajal-
(E) 10 and E 18 in the mouse (Angevine, 1965) and between Retzius cells also appear to play a role in the formation of the
E 16 and E 21 in the rat brain (Bayer, 1980). Pregnancy lasts pyramidal cell layer (see below).
19 days in mice and 21 days in rats. CA3 pyramidal cells are It is worth noting that the pyramidal cell layer of the hip-
generated with a peak of generation on E 17, whereas the peak pocampus also forms quite early in the human brain. The
of cell generation of CA1 pyramidal cells is on E 18 and E 19 pyramidal layer is formed during the rst half of pregnancy
(Bayer, 1980). (Arnold and Trojanowsky, 1996). In addition, the CA1 to C3
Except for the CA3 pyramidal cells, the route of migration subregions of the hippocampus can be differentiated as early
is short because the hippocampus closely follows the curve of as during the 16th embryonic week. Regional differentiation
the ventricle (Fig. 41). Future CA1 neurons form cell rows of the hippocampal formation thus substantially precedes that
oriented perpendicular to the ventricular area that end in of the neocortex (Kostovic et al., 1989).
the incipient CA1 pyramidal layer. From the ventricular ger-
minal layer, the early-generated CA3 pyramidal cells migrate 4.2.2 Granule Cells
3 to 4 days longer (than the CA1 cells) along the ventricular
wall in the only prenatally existing intermediate zone (Altman The formation of the granule cell layer of the dentate gyrus
and Bayer, 1990b). As a consequence, the later-generated differs in many respects from that of the pyramidal cell layer
CA1 pyramidal cells establish a distinct layer earlier than of the hippocampus. The rst granule cells of the mouse den-
the CA3 cells. It remains an open question whether the tate gyrus appear at approximately the same time as the rst
pyramidal neurons in CA1 and CA3 originate from the same pyramidal cells (i.e., on E 10) (Angevine, 1965). There is a
segment of the hippocampal neuroepithelium. A separate ori- suprapyramidal to infrapyramidal, or medial to lateral, gradi-
ent in the formation of the two blades of the granule cell layer;
that is, granule cells at the tip of the suprapyramidal blade are
Figure 41. Route of migration (arrows) of pyramidal cells from the rst to be generated. The generation of granule cells lasts
the ventricular germinative layer to the subiculum and to areas a much longer time than that of pyramidal neurons. There is
CA1 to CA3. DG, dentate gyrus; H, hilus. substantial evidence that it continues long into the postnatal
period and, at a reduced level, into adulthood. In the rat den-
tate gyrus, rst granule cells are generated 1 day later than the
rst pyramidal neurons (i.e., on E 17) (Bayer, 1980). In rats,
15% of the granule cells are generated before birth (Altman
and Bayer, 1975). Therefore, the time span of granule cell
generation is approximately three times longer than that of
pyramidal neurons and is likely to continue for the rest of
the animals life (for adult neurogenesis, see Chapter 9).
Interestingly, adult granule cell formation does not follow the
suprapyramidal to infrapyramidal sequence observed during
the embryonic and early postnatal period. Granule cells are
generated with no obvious pattern in both blades (Kemper-
mann et al., 1997a,b).
The prolonged postnatal formation of granule cells is not a
unique feature of rodents. In humans, granule cell formation
lasts more than 30 weeks, beginning at approximately the 13th
 Morphological Development of the Hippocampus 117

period of granule cell neurogenesis, and persisting hilar stem

cells may also form new neurons during adulthood (Altman
and Bayer, 1975, 1990a; Seress, 1978). In rodents, the total
number of hilar cells rapidly decreases between postnatal days
10 and 25, with a parallel increase in the number of granule
cells in the granule cell layer, suggesting that newly formed
cells from the hilus move to the granule cell layer (Seress,
1977). During the entire period of postnatal development,
labeled granule cells are always observed inside the granule cell
layer as early as 1 hour following the injection of bromod-
eoxyuridine (BrdU) or thymidine (Fig. 43, see color insert). It
is likely that stem cells from the hippocampal ventricular zone
move to the hilar region and form a secondary germinal zone
(Fig. 42). This possibility is supported by the fact that a sin-
gle neonatal X-ray irradiation prevents further cell formation
in both the ventricular zone and hilus without affecting the
Figure 42. Route of migration of dentate granule cells. There are already formed neuronal subpopulations (Czurko et al., 1997).
two possible ways for granule cells to reach the suprapyramidal To the extent that it has been studied, the developmental
or infrapyramidal blade of the dentate gyrus. 1, Granule cells characteristics of the rodent hilar region are also observed in
formed in the germinative layer move along the pyramidal layer
humans. In human fetuses, a large number of mitotic cells
of Ammons horn directly to the granule cell layer (arrows); 2, after
(indicated by the presence of Ki-67) are present in the hilar
migration from the ventricular germinative layer, stem cells form
a secondary germinative layer (black spot) in the hilus (H) and, region from the 13th week onward (Fig. 44A, see color
following further proliferation, move to the granule cell layer insert). These proliferating cells account for a large percentage
(arrows) of the dentate gyrus (DG). During the embryonic period of the total hilar cell population in 16- to 18-week-old fetuses
both routes are used, whereas the latter route prevails during the (Seress, unpublished observations). Proliferating cells can also
postnatal period. be found in the hilus during the late gestational period and
after birth (Seress et al., 2001) (Fig. 44B, see color insert).
Persisting stem cells have also been found in the hilus of the
embryonic week (Humphrey, 1967; Seress et al., 2001). There adult human hippocampal formation (Eriksson et al., 1998).
is also evidence for adult neurogenesis in the human dentate However, proliferative activity drops off considerably during
gyrus (Eriksson et al., 1998). the postnatal period (Seress et al., 2001).
The route of migration of the postmitotic granule cells is There is little morphological variability among granule
much longer than that of pyramidal neurons. Postmitotic cells cells even though they may be located at different places in the
must migrate from the ventricular germinal layer along the granule cell layer and may be born at different time points.
already formed hippocampus through the narrow intermedi- This morphological uniformity leads one to assume that all
ate zone between the alveus and future stratum oriens toward granule cells originate from the same stem cell population.
the hippocampus, where they form the cup-shaped dentate However, granule cells may originate from different germinal
gyrus that surrounds the tip of CA3 (Fig. 42). sources (Altman and Bayer, 1990a), and precursors of the
The two blades of dentate granule cells and the tip of the perinatally and postnatally forming granule cells can be dis-
CA3 area border the hilus of the dentate gyrus (see Chapter 3). tinguished (Altman and Bayer, 1990c). A newly discovered
During postnatal ontogenesis, the hilus (Fig. 43A, see color gene (Lef1) has been found to distinguish granule cell precur-
insert) contains large numbers of mitotic cells (Bayer, 1980). sors. This gene is expressed in the proliferating precursors of
Large hilar neurons, such as the mossy cells, are generated granule cells in the ventricular zone as well as in those of the
between E 15 and E 21 (Bayer, 1980). In addition, virtually all secondary germinal zone. Lef1-decient mice fail to form den-
hilar interneurons, representing approximately half of the tate granule cells (Galceran et al., 2000).
hilar cells (Seress and Ribak, 1983), are of subcortical origin Most of the granule cells display similar features in the
and invade the hilus from the lateral and medial ganglionic rodent and primate dentate gyrus. However, a subpopulation
eminence (Pleasure et al., 2000). Thus, the enduring cell pro- of granule cells in the primate dentate gyrus is characterized
liferation in the hilar region (Fig. 42) mainly represents local by the formation of basal dendrites (Seress and Mrzljak,
generation of granule cells (Altman and Bayer, 1990c). In fact, 1987). It is likely that local environmental factors, such as
the rate of postnatal cell formation in the hilus is similar to available space for dendritic expansion, may play an impor-
that in the periventricular germinal layer of the telencephalon tant role in the formation of basal dendrites on granule cells
(Seress, 1978). It is therefore reasonable to assume that the (Seress and Frotscher, 1990). Basal dendrites have also been
hilar region is a displaced secondary germinal layer of the regarded as a pathological phenomenon. For example,
dentate gyrus (Altman and Bayer, 1975; Seress, 1977). increased activity of the granule cells may evoke the formation
Proliferating cells persist in the hilar region during the entire of basal dendrites in epileptic rats (Spigelman et al., 1998).
118 The Hippocampus Book

Figure 43. A. Photomicrograph

showing 3H-thymidine-labeled cells
(black) in the hilus (H) and granule
cell layer (GCL) of the dentate gyrus
of a 5-day-old rat that received a single
injection of thymidine 1 hour before
sacrice. At this age, the infrapyrami-
dal blade is still very short and hardly
reaches the tip of the CA3 pyramidal
layer. The large number of labeled
cells above the granule cell layer sug-
gests that glial cells are also formed at
this age because the molecular layer
contains only a few interneurons,
which are born prenatally. Note the
lack of labeling in the CA3 area of
Ammons horn where pyramidal cells
are generated prenatally. n I set . Labeled
cells in the granule cell layer of this
animal at higher magnication.
Section is 10 m thick and stained
with toluidine blue. B. BrdU-labeled
cells (brown) in the suprapyramidal
blade of the dentate gyrus of a 3-
month-old rat that received daily
injections of BrdU from P4 to P6.
Bars: A, B, 50 m; inset, 10 m.

Epileptic seizures also appear to lead to an increased number uniform granule cells and pyramidal neurons, local circuit
of granule cells with basal dendrites in the human dentate neurons of the hippocampal formation form a heterogeneous
gyrus (von Campe et al., 1997). population (Freund and Buzski, 1996) (see Chapters 3 and
In summary, dentate granule cells in different mammalian 8). Even the classic basket cell can be subdivided into ve types
species form a largely homogeneous neuronal population. based on differences in location and dendritic arborization
Their dendritic morphology, however, may be modied to (Ribak and Seress, 1983). Neurochemical studies have con-
some extent depending on the local environmental or patho- rmed the existence of at least two basket cell types, which dis-
logical factors. play different receptors and co-transmitters (Hjos et al., 1998;
Katona et al., 1999). In addition to dendritic morphology, dif-
4.2.3 Local Circuit Neurons and Hilar Neurons ferences in axonal projections also distinguish subpopulations
of local circuit neurons in both the dentate gyrus (Halasy and
Local circuit neurons of the hippocampal formation differ Somogyi, 1993a,b) and the hippocampus proper (Soriano and
from the principal cells in both their morphological and Frotscher, 1989, 1993; Halasy et al., 1996; Spruston et al., 1997;
developmental features. In contrast to the morphologically Vida et al., 1998; Vida and Frotscher, 2000).
 Morphological Development of the Hippocampus 119

Figure 44. A. MIB-1 (Ki-67)-labeled

cells (brown) in the dentate gyrus
of a 16-week-old human fetus. A large
number of labeled cells are seen in
the hilus (H) underneath the granule
cell layer (GCL). Note the absence
of labeled cells in the CA3 area of
Ammons horn. In contrast, numerous
labeled cells, migrating from the ven-
tricular germinative layer along the
pyramidal layer toward the hilus, are
seen in the intermediate zone. B. MIB-
1 (Ki-67)-labeled cells are still frequent
in the hilus (H) of a 28-week-old fetus.
The granule cell layer and molecular
layer (ML) contain only a few labeled
cells. Sections are 10 m thick and are
stained with cresyl violet. Bars: A, 50
m; B, 20 m.

In contrast to the heterogeneity of the local circuit neuron cedes formation of the rst pyramidal cells by 1 day (Bayer,
phenotype, their neurogenetic prole seems much more 1980). Cell formation of the cells destined to reside in the den-
homogeneous; that is, there are no developmental events that dritic layers ceases relatively early. It is reasonable to suggest
would allow one to predict the large diversity among - that this exquisite timing may play some role in pyramidal and
aminobutyric acid (GABA)ergic cells. GABAergic interneu- granule cell morphogenesis, although there is no evidence for
rons are generated early in both the dentate gyrus and the this at this time.
hippocampus (Bayer, 1980). This conclusion was based on the Double-labeling studies have conrmed that GABAergic
observation that neurons in the molecular layer and hilus of neurons are generated earlier than the principal cells of the
the dentate gyrus as well as in the strata oriens, radiatum, and hippocampal formation (Amaral and Kurz, 1985; Lbbers
lacunosum-moleculare appear earlier than the neurons in the et al., 1985; Soriano et al., 1989a,b; Dupuy and Houser,
principal cell layers (Bayer, 1980). The peak cell formation in 1997, 1999). The site of origin of hippocampal local circuit
both the hilus and the molecular layer precedes the onset of neurons is in the ventricular/subventricular zone of the
granule cell generation. Similarly, in the CA1 to C3 areas, the medial and lateral ganglionic eminence. Available data do not
peak cell generation in the strata oriens and radiatum pre- allow one to correlate the site of generation with the diverse
120 The Hippocampus Book

morphological features of the local circuit neurons (Pleasure Although this scenario is certainly an oversimplication, it
et al., 2000). can be employed to explain the unusual lamination in the
Although GABAergic neurons are generated and posi- reeler hippocampus and dentate gyrus. In the hippocampus, a
tioned before their target cells, some of them change their double pyramidal cell layer is formed in reeler mice. As in the
position later on in development (Dupuy and Houser, 1999). neocortex, the marginal zone is located directly underneath
Moreover, the differentiation of axonal collaterals and the the pial surface, which in the hippocampus proper corre-
arborization of dendrites takes place relatively late in postna- sponds to the stratum lacunosum-moleculare. The stratum
tal life (Seress et al., 1989; Lang and Frotscher, 1990; Seress and lacunosum-moleculare contains Cajal-Retzius cells, and its
Ribak, 1990). This raises the possibility that the maturation of extracellular matrix is enriched in reelin. The absence of the
local circuit neurons is inuenced by interactions with their stop signal reelin in reeler mutants may allow some pyramidal
synaptic partners. cells to migrate deep into the stratum radiatum, forming there
a second pyramidal layer (Staneld and Cowan, 1979; Deller
4.2.4 Determinants of Neuronal et al., 1999b).
Migration in the Hippocampus The marginal zone of the dentate gyrus corresponds to the
outer portion of the molecular layer. Assuming that reelin acts
In the neocortex, cell layers are formed in an inside-out man- as a stop signal, one would expect that many granule cells
ner; that is, early-generated neurons form deep layers, whereas would invade the molecular layer in reeler mice because they
supercial layers are established by cells born late in ontoge- would not be stopped earlier. However, in the reeler dentate
netic development. As in the neocortex, the hippocampus fol- gyrus, the granule cells do not invade the molecular layer,
lows a similar inside-out migration. The dentate gyrus, in whereas many granule cells are found in the hilar region with
contrast, has an outside-in migration pattern; that is, early- their dendrites oriented in all directions (Staneld and
generated granule cells populate supercial portions of the Cowan, 1979; Drakew et al., 2002). What might be the expla-
granule cell layer and later-formed granule cells are located nation of the hilar location of the granule cells? Frster et al.
progressively deeper. (2002) found that the radial glial bers required for neuronal
What are the signals controlling neuronal migration that migration are dramatically altered in the dentate gyrus of
lead to such well dened cell layers of principal neurons? It reeler mice. Thus, it has been proposed that the granule cell
has been known for some time that these migrational migration defect seen in the reeler mutant is due to a malfor-
processes are under fairly strict genetic control. In 1951, a mation of the radial glial scaffold required for proper neu-
mouse mutant, the reeler mouse, was described (Falconer, ronal migration (Frster et al., 2002; Frotscher et al., 2003;
1951) in which neocortical lamination was reversed. Recently, Zhao et al., 2004).
the gene missing in reeler mice was discovered (dArcangelo et
al., 1995; Hirotsune et al., 1995). The gene product, reelin, was
found to be synthesized by a well known group of neurons in
the marginal zone, originally described by Cajal and Retzius 4.3 Development of Hippocampal
more than 100 years ago (Retzius, 1893, 1894; Ramn y Cajal, Connections
1911). These Cajal-Retzius cells are remarkable neurons. They
extend long dendrites horizontally in the marginal zone, par- When describing the development of neuronal connections, it
allel to the pial surface. Cajal-Retzius cells synthesize and is useful to distinguish at least three processes. First, there is
secrete reelin, so it forms a component of the extracellular the way an axon navigates to the target region, called axonal
matrix in the marginal zone (dArcangelo et al., 1997). pathnding. During this process, both soluble and membrane-
Available data are compatible with the proposal that reelin bound molecules and attractant and repellent signals are
acts as a stop signal for migrating neurons. Early-generated involved (see Tessier-Lavigne and Goodman, 1996, for review).
neurons are stopped as they approach the marginal zone, Next, the axon must recognize the appropriate target region
which has a high concentration of reelin. For later neurons to and target cell, called target recognition. As far as the hip-
reach this same stop signal, they must migrate past the early- pocampus is concerned, the segregated ber systems must rec-
generated neurons and approach the reelin-containing mar- ognize their appropriate layers. Finally, there is the process of
ginal zone. In this way the characteristic inside-out lamination synapse formation.
of the cortex is formed (Frotscher, 1997, 1998). Because the There is substantial and increasing evidence that a variety
marginal zone ultimately becomes layer I of the cortex, it of molecules are involved in these three processes. In princi-
remains an almost cell-free layer in normally developing mice. ple, membrane-bound molecules that attract or repel the
In reeler mice, in contrast, layer I (which does not have the growing tip of an axon, diffusible factors building up a gradi-
stop signal) is densely lled with neurons. Early-generated ent, and components of the extracellular matrix are likely to
neurons migrate until they reach the pial surface, and later- play a role. The involvement of various ligand/receptor fami-
generated neurons accumulate underneath. The normal lies in the formation of hippocampal connections has been
inside-out lamination is reversed (Fig. 45). summarized by Skutella and Nitsch (2001). Thus, the various
 Morphological Development of the Hippocampus 121

Figure 45. Hypothetical function of reelin in cortical lamination. tic inside-out pattern of cortical lamination is formed. Apical
A. In wild-type mice, the preplate splits into the marginal zone, dendrites, branching in the reelin-containing marginal zone, are
largely containing Cajal-Retzius cells (black cells), and the subplate extended by the increase in cortical thickness. D. In reeler mice
(white cells). Cajal-Retzius cells synthesize and secrete reelin (stip- lacking reelin, the separation of the subplate from the marginal
pled zone). B. The cortical plate develops between the marginal zone does not take place. E. In the absence of reelin, migrating cells
zone and the subplate. An early-generated cell (A) migrates from (A) are not stopped and reach the pial surface. F. Later-generated
the ventricular zone toward the marginal zone, where its migration neurons (B, C) accumulate underneath the early-generated neurons
is stopped by reelin. C. With increasing cortical thickness, the reelin- (A). As a result, the normal inside-out pattern of cortical lamination
containing marginal zone is moved further outward, allowing later- is reversed, and apical dendrites are not anchored in the marginal
generated neurons (B, C) to migrate farther than their predecessors zone. (Source: Frotscher, 1997.)
before their migration is stopped by reelin. This way the characteris-

semaphorins and their receptors neuropilin 1 and 2 and the gyrus. From a developmental point of view, one would expect
plexins netrin 1 and its receptor DCC (deleted in colon can- that the perforant path would utilize robust guidance signals
cer), the ephrin (Eph) family of tyrosine kinases and their lig- for axonal pathnding and target layer recognition. The
ands, the ephrins, and Slit and Robo were found to contribute entorhinal bers terminating in the outer two-thirds of the
to establishment of the hippocampal circuitry. dentate molecular layer and in stratum lacunosum-moleculare
In the following paragraphs, we deal with the develop- of the hippocampus proper form a sharp boundary toward the
ment of some of the main hippocampal afferent connections: adjacent commissural/associational bers in the inner molec-
those from the entorhinal cortex, the septum, and the con- ular layer of the dentate gyrus and in the stratum radiatum of
tralateral hippocampus. Whereas the bers from both the the hippocampus (see Chapter 3).
entorhinal cortex and the contralateral hippocampus are Recent studies have shed some light on the mechanisms
examples of a strictly laminated termination of hippocampal underlying the pathnding of entorhinal axons (del Rio et al.,
afferents, bers from the septum have a more diffuse dis- 1997; Chedotal et al., 1998; Frster et al., 1998; Frotscher,
tribution. 1998; Ceranik et al., 1999; Deller et al., 1999a,b; Savaskan et
al., 1999; Skutella et al., 1999; Steup et al., 1999). Investigators
4.3.1 Entorhinal Connections have focused on repulsive molecules expressed in the hip-
pocampus that would prevent the entorhinal afferents from
Neuroanatomists have been intrigued by the unique course of terminating in proximal layers close to the cell bodies of prin-
the entorhino-hippocampal projection, which perforates the cipal neurons. Similarly, molecules specically expressed in
subiculum and hippocampal ssure en route to the dentate the entorhinal cortex were studied for their capacity to repel
122 The Hippocampus Book

entorhinal axons, thereby directing them toward the hip-

pocampus. In particular, repulsive effects of semaphorins
(Sema3A) on entorhinal bers have been observed (Steup et
al., 1999), an effect that can be blocked by antibodies against
the semaphorin receptor neuropilin 1 (Chedotal et al., 1998).
As Sema3A is expressed by granule cells, it has been speculated
that this expression may be involved in the termination of
entorhinal bers on distal granule cell dendrites. Neuropilin 2,
the receptor for Sema3F, is strongly expressed in the hip-
pocampus, whereas the ligand is expressed in the developing
entorhinal cortex. As Sema3F is repulsive for growing axons
bearing its receptor, this particular pattern may be responsible
for the lack of ingrowth of hippocampal and dentate axons
into the entorhinal cortex (Chedotal et al., 1998; Steup et al.,
1999; Skutella and Nitsch, 2001). Similarly, Slit 1 and Slit 2,
being expressed in the entorhinal cortex, can repel axons of
hippocampal neurons expressing their receptors Robo 1 and
Robo 2 (Nguyen-Ba-Charvet et al., 1999). Also ephrins were
found to be involved in repulsive effects on entorhinal axons,
restricting their termination to distal dendritic portions of
hippocampal neurons (Stein et al., 1999). In addition to these
repulsive effects, neurotrophic factors with an attractive neu-
rotropic effect, components of the extracellular matrix, and a
guiding scaffold provided by pioneer neurons have been
found to navigate the axons of entorhinal neurons to their Figure 46. AC. Early-generated hippocampal Cajal-Retzius (CR)

hippocampal target structures. cells form a template for outgrowing axons of projection neurons in
the entorhinal cortex (EC). D. Later on in development when many
What could be the sequence of events underlying this com-
CR cells have disappeared, the axons of projection cells in the
plex process? Probably as a rst step, a guiding scaffold sup- entorhinal cortex establish denitive connections with distal den-
porting the directed growth of entorhinal axons toward the drites of dentate granule cells (GC).
hippocampus is formed. In fact, Ceranik et al. (1999) have
shown that early-generated Cajal-Retzius cells, located in the
marginal zones of the dentate gyrus (outer molecular layer) fact, Cajal-Retzius cells and their projections to the entorhinal
and hippocampus proper (stratum lacunosum-moleculare), cortex are also present in these mutants and may thus serve
give rise to a heavy projection to the entorhinal cortex that is their normal function as a template for entorhinal axons. In
established before the entorhino-hippocampal projection. normal mice, reelin affects the branching pattern of entorhi-
Thus, Cajal-Retzius cells in the outer molecular layer of the rat nal terminals. In the absence of reelin, entorhinal axons give
dentate gyrus are retrogradely labeled from the entorhinal rise to fewer collaterals and synapses (del Rio et al., 1997;
cortex as early as on E 17, whereas the rst entorhino- Borrell et al., 1999). The growth of entorhinal bers along
hippocampal bers were found to arrive in the dentate outer Cajal-Retzius cell axons is likely to be controlled by a variety
molecular layer only thereafter. Moreover, with combined of repulsive and attractive molecules (see above).
injections of two tracers, DiI and DiO, into the entorhinal cor- How do the entorhinal bers recognize their target layer?
tex and hippocampus, respectively, entorhinal axons were Current evidence strongly suggests that components of the
found to grow along axons of Cajal-Retzius cells. These nd- extracellular matrix play an important part in the segregation
ings suggest that the early-formed projection of hippocampal of hippocampal afferents. Frster et al. (1998), for example,
Cajal-Retzius cells to the entorhinal cortex provides a tem- demonstrated that uorescent beads coated with entorhinal
plate for outgrowing entorhinal axons Fig. 46A(C). A role of membranes precisely adhered to the correct entorhinal ter-
Cajal-Retzius cells in the pathnding of entorhinal axons is mination zone of the hippocampus. They then used this pat-
strongly supported by experiments in slice cultures, in which tern as an assay to study the role of extracellular matrix
Cajal-Retzius cells had been caused to degenerate by excito- components in this layer-specic adhesion. When slice cultures
toxic lesions (del Rio et al., 1997). In the absence of Cajal- of hippocampus were treated with hyaluronidase, the mem-
Retzius cells, entorhinal axons were unable to nd their way to brane-coated beads no longer adhered with layer specicity,
the hippocampus, whereas commissural bers invaded their indicating that hyaluronic acid and proteoglycans bound to it
termination elds as normal. may contribute to the layer-specic adhesion. Such adhesion
Because the entorhino-hippocampal projection develops seems also to be an important factor for target layer recogni-
almost normally in reeler mice, it does not appear that reelin tion of entorhinal bers. When hippocampal slices were
synthesized by Cajal-Retzius cells is essential to this process. In treated with hyaluronidase and then cocultured with entorhi-
 Morphological Development of the Hippocampus 123

nal cortex, entorhinal bers invaded the molecular layer but ber of long, immature spines, indicating that neuronal activ-
were no longer restricted to its outer two-thirds (Frster et al., ity may be required for the maturation of spines.
2001). It appears that hyaluronic acid and proteoglycans are Many questions concerning the layer-specic termination
essential for the formation of boundaries between afferent ter- of entorhinal bers remain open at present. As an example,
minal zones in the hippocampus. Interestingly enough, no the development of the ber lamination in the infrapyramidal
similar role of extracellular matrix components was found for blade of the dentate gyrus is different from that in the
the laminar specicity of commissural/associational bers suprapyramidal blade. Whereas the entorhinal bers arrive
(Zhao et al., 2003). In conclusion, secreted molecules such as before the commissural axons in the suprapyramidal blade,
the semaphorins, membrane-bound receptors, extracellular this sequence is reversed in the infrapyramidal blade
matrix components, and a template formed by Cajal-Retzius (Tamamaki, 1999). Do the bers to these different blades orig-
cell axons are likely to be involved in the directed growth and inate from the same cells of origin in the entorhinal cortex? If
layer-specic termination of entorhinal bers. so, what causes the delay in the ingrowth into the infrapyra-
What is the role of the postsynaptic target cell in the layer- midal blade?
specic termination of entorhinal afferents? In the rodent
brain, entorhinal axons innervate the hippocampus proper at 4.3.2 Commissural Connections
E 15 and the dentate gyrus at E 18/E 19 (Super and Soriano,
1994; Ceranik et al., 1999). The entorhinal bers arrive much Commissural bers, originating from CA3 pyramidal neurons
earlier than the commissural bers from the contralateral hip- and hilar mossy cells, project via the hippocampal commis-
pocampus (see below). At this early stage, pyramidal neurons sure to the contralateral hippocampus and dentate gyrus. In
and granule cells have not yet grown their distal dendritic tips the mouse hippocampus, the rst commissural axons arrive in
to the stratum lacunosum-moleculare and the outer molecu- the contralateral hippocampus at E 18 and in the dentate
lar layer of the dentate gyrus. Most of the granule cells have gyrus at P 2. This is considerably later than the arrival of the
not even been generated yet. Interestingly, the entorhinal entorhinal axons. The sequential generation of the entorhinal
bers recognize their appropriate target layers from the very neurons and commissurally projecting hippocampal cells and
beginning despite the absence of their appropriate target den- their sequential ingrowth into their target elds has led to a
drites at the stage of ber ingrowth. Evidently, the target cells hypothesis concerning their laminated termination in both
are unlikely to be involved in pathnding of entorhinal axons the hippocampus proper and the dentate gyrus: The later the
and target layer recognition. The early-generated Cajal- bers invade their target zones, the more proximally do they
Retzius cells, located in the termination zones of entorhinal terminate on their target cell dendrites (Bayer and Altman,
bers, may serve as primary targets. In fact, entorhinal bers 1987). Entorhinal axons, arriving early in development, con-
were found to establish synaptic contacts with cell bodies and tact distal dendrites of pyramidal cells and granule cells,
dendrites of Cajal-Retzius cells (del Rio et al., 1997; Frotscher whereas later-arriving commissural bers establish synapses
et al., 2001). A role of Cajal-Retzius cells as primary target in the stratum radiatum of the hippocampus proper and in
neurons of entorhinal bers is supported by X-irradiation the inner molecular layer of the dentate gyrus. One of the last
experiments. Irradiation of newborn rats, which eliminates projections to be formed, the ipsilateral mossy ber projec-
most of the postnatally generated granule cells but not the tion, accordingly terminates on the most proximal dendritic
early-born Cajal-Retzius cells, does not alter the layer-specic segments of CA3 pyramidal neurons.
termination of entorhinal bers (Laurberg and Hjorth- This hypothesis is not easy to test in vivo. However, it can
Simonsen, 1977; Frotscher et al., 2001). Conversely, ablation of be addressed in slice culture experiments in which the time of
Cajal-Retzius cells prevented layer-specic ingrowth (del Rio arrival of growing axons can be brought under precise exper-
et al., 1997). imental control. Frotscher and Heimrich (1993) cocultured a
Many Cajal-Retzius cells degenerate later during postnatal hippocampal target culture rst with an entorhinal culture
development (i.e., after the entorhino-hippocampal projec- and then with a second hippocampal culture, thus imitating
tion has been formed and pyramidal cells and granule cells the normal developmental sequence in this sequential culture
have grown their distal dendrites into the entorhinal termina- system. They then traced the entorhinal and commissural
tion elds). At that time, the contacts of entorhinal bers are connections to the hippocampal target culture and observed,
reorganized, and denitive synapses with distal dendritic por- as one would expect, that the entorhinal bers terminated on
tions of pyramidal cells and granule cells are made (Fig. the distal dendritic segments of their target neurons, as nor-
46D). Although pathnding and target recognition are not mal. Next, they reversed this sequence: First, the two hip-
affected by the absence of neuronal activity, there is an effect pocampal cultures were cocultivated; then, with a delay of a
on the maturation of synaptic structures. Drakew and col- couple of days, an entorhinal slice culture was added. Despite
leagues (Drakew et al., 1999; Frotscher et al., 2000b) observed the reversal of the sequence of ber ingrowth, the commis-
normal formation of the entorhino-hippocampal projection sural and entorhinal bers still terminated in their appropri-
in cocultures of entorhinal cortex and hippocampus incu- ate layers, indicating that the sequence of ber arrival in the
bated in the presence of tetrodotoxin (TTX), which blocks fast target region is not responsible for the laminated, proximo-
sodium channels. However, they observed an increased num- distal termination of these two afferents.
124 The Hippocampus Book

As the time of arrival is not the factor governing lamina- insert). This is in striking contrast to the entorhino-dentate
tion, the question arises as to potential cellular and molecular projection, which forms a compact termination zone with
factors determining the layer specicity of commissural bers. sharp borders in the outer molecular layer of the reeler mutant
Soriano and colleagues (Super et al., 1998) have suggested that (Deller et al., 1999a); moreover, this characteristic layer-
there are pioneer neurons, early-generated GABAergic cells, specic projection is nicely preserved in slice cultures from
that serve as primary targets for the commissural bers. Their wild-type animals and reeler mutants (Zhao et al., 2003) (Fig.
role is comparable to that of the Cajal-Retzius cells as primary 48, see color insert). Together, these ndings support the
targets of the entorhinal afferents. This idea became testable hypothesis that entorhinal bers, because of their early arrival
with the possibility of depleting the hippocampus of in the dentate gyrus at a time when their denitive target cells,
GABAergic neurons during development (Pleasure et al., the granule cells, have not yet been generated, do require pio-
2000). Contrary to Sorianos idea, commissural bers still neer target neurons. These neurons are likely to be the Cajal-
innervated their proper target areas in the absence of Retzius cells, which are, as normal, located in the outer
GABAergic neurons. molecular layer of reeler mice (see above). One is tempted to
As the commissural bers arrive relatively late in the hip- speculate that whereas the entorhinal bers require pioneer
pocampus and dentate gyrus, an alternative proposal was that neurons the commissural afferents establish contacts directly
pioneer neurons are not required. This is because the target with principal cell dendrites, which is reected in reeler mice
cells, the principal neurons, are already present at that time and by the irregular, loose distribution of both commissural bers
have already grown a dendritic arbor (Deller et al., 1999a). and granule cells (Fig. 49). In fact, when commissural bers
Evidence relevant to this idea has come from studies in reeler from a reeler culture are traced to a wild-type culture, they are
mice. In this mutant, the granule cells are loosely distributed found to form a normal, compact projection to the inner
throughout the hilar region (Staneld and Cowan, 1979; Deller molecular layer, precluding a cell-autonomous effect of the
et al., 1999a; Drakew et al., 2002) and the commissural bers reeler mutation on commissural neurons (Zhao et al., 2003).
do not form a compact layer but are, instead, distributed all Little is known about the molecules involved in pathnd-
over the hilar region (Deller et al., 1999a). This target-depend- ing and target recognition of commissural bers. As with
ent termination of commissural bers is nicely seen when other commissures, Netrin 1 and its receptor DCC are impor-
commissural axons from a wild-type culture are traced to both tant for formation of the hippocampal commissure. Thus, a
a second wild-type culture and a reeler culture: Whereas the hippocampal commissure does not form in Netrin 1-decient
commissural axons terminating in the wild-type culture form animals. Along this line, neurites of hippocampal neurons are
their normal compact projection to the inner molecular layer attracted by Netrin 1 in the stripe choice assay (Steup et al.,
of the dentate, the commissural bers that innervate the reeler 2000). The molecules governing the laminar specicity of the
culture terminate diffusely, corresponding to the scattered dis- commissural bers in their target zones remain to be deter-
tribution of their target granule cells (Fig. 47, see color mined.

Figure 47. Triplet cocultures of two wild-type

slices of hippocampus together with a reeler
hippocampal slice. The biocytin injection site
into the dentate gyrus of one of the wild-type
cultures (DG1) is marked with an asterisk.
Labeled commissural bers invade the two
other cultures, but only in the dentate gyrus of
the second wild-type culture (DG2) do they
show their characteristic compact termination
in the inner molecular layer (arrow). In the
reeler hippocampal culture, the commissural
bers arising from the same cells of origin have
lost their laminar specicity and terminate
throughout the hilar region of the dentate gyrus
(DG3). Arrowhead points to the mossy ber
projection in the wild-type culture that was also
labeled by the tracer injection into the hilus. p,
pyramidal layer; p1, p2 two pyramidal layers in
the reeler culture. Bar  100 m. (Source: Zhao
et al., 2003, 2003 by the Society for
Neuroscience.)
 Morphological Development of the Hippocampus 125

Figure 48. A. Entorhino-hippocampal projec-

tion in a complex culture of entorhinal cortex
(EC) and hippocampus. Biocytin injection sites
are labeled by asterisks. Note the layer-specic
termination of the entorhinal bers in the outer
molecular layer (OML) of the dentate gyrus. B.
In a slice culture of reeler hippocampus and
entorhinal cortex, entorhinal bers labeled by
biocytin injection into the entorhinal cortex
(asterisks) form a sharply segregated projection
to the outer portion of the molecular layer of the
dentate gyrus (DG) despite the malpositioned
granule cells, demonstrating that the trajectory
of entorhinal bers is not inuenced by the posi-
tion of their target neurons. p1, p2, two pyrami-
dal layers in the reeler hippocampus. Bar  100
m. (Source: Zhao et al., 2003; 2003 by the
Society for Neuroscience.)

4.3.3 Septal Connections pocampal bers can nd their way to the hippocampus. It has
in fact been shown that growth cones of septohippocampal
The septohippocampal projection consists of two parts: a bers grow along hippocampo-septal axons (Linke and
cholinergic one and a GABAergic one. The cells of origin for Frotscher, 1993). The molecular mechanisms underlying the
both parts are located in the medial septal nucleus/diagonal change of hippocampo-septal bers from the medial to the
band complex (MSDB) (see Chapter 3). The development of lateral septum remain to be elucidated.
the septohippocampal projection can be understood only by Unlike the entorhino-hippocampal and commissural pro-
taking into account the hippocampo-septal projection. The jections to the hippocampus, septohippocampal cholinergic
latter develops relatively early and surprisingly terminates in bers do not terminate in clearly demarcated layers. They are
the medial septum during development as early as on E 15. In found in all layers of the hippocampal formation but are cer-
adults, hippocampo-septal bers project to the lateral septum tainly more concentrated in the cell body layers. In contrast,
(Linke and Frotscher, 1993; Super and Soriano, 1994; Linke et the septohippocampal GABAergic projection cells display a
al., 1995). This early termination in the medial septal nucleus high target cell specicity terminating almost exclusively on
has led to the hypothesis that the early formed hippocampo- GABAergic interneurons in the hippocampus. Inhibitory neu-
septal projection serves as a template by which septohip- rons afferent to inhibitory neurons serve a disinhibitory role
126 The Hippocampus Book

Figure 49. Different signals control the laminar specicity of outgrowing granule cell dendrites. B. Findings in the reeler mutant,
entorhinal bers (black) and commissural bers (gray) to the den- where the granule cells are not aligned but scattered over the den-
tate gyrus. A. In wild-type mice, entorhinal bers, guided by tate gyrus, conrm that the entorhinal axons are not guided by their
Cajal-Retzius cell axons, reach the outer molecular layer (OML) target granule cells. As for the wild-type bers, entorhinal bers ter-
of the dentate gyrus before the granule cells in the granule cell minate with laminar specicity in the OML. In contrast, commis-
layer (GCL) have grown their distal dendritic tips to the OML. sural bers are loosely distributed over the dentate gyrus, thus
Molecules of the extracellular matrix control the laminar speci- following their scattered target cells. (Source: Adapted from
city of entorhinal axons. In contrast, later-arriving commissural Frotscher, 1998).
bers, terminating in the inner molecular layer (IML), meet the

with respect to the principal neurons (Freund and Antal, for axonal orientation (see above). A neurotropic effect of
1988). The septal cholinergic bers establish contacts with NGF was demonstrated some time ago (Gundersen and
both principal neurons and interneurons, and their contacts Barrett, 1979; see Tessier-Lavigne and Goodman, 1996, for
are on cell bodies, dendritic shafts, and spines. Two types of review), but it needs to be conrmed that the ingrowth of
contact are established: symmetrical and asymmetrical septohippocampal bers is controlled by neurotrophins.
(Frotscher and Leranth, 1985, 1986). It remains to be estab- One of the semaphorins, Sema3C, was found to repel septal
lished whether one of these types derives from the few bers (Steup et al., 2000); and neuropilin 2, its receptor, is
intrahippocampal cholinergic cells (Frotscher et al., 1986, present along the pathway of septal bers (Chedotal et al.,
2000a) that are found in the rat hippocampus but not in the 1998; Steup et al., 2000), suggesting involvement of this lig-
monkey hippocampus. andreceptor system in the formation of the septohippocam-
The rst septohippocampal bers arrive in the hippocam- pal projection.
pus at E 17 (i.e., 2 days after the hippocampo-septal projec-
tion is established) (Linke and Frotscher, 1993; Super and 4.3.4 General Principles Underlying
Soriano, 1994). Little is known about the molecules involved the Formation of Synaptic
in pathnding and target recognition of septohippocampal Connections in the Hippocampus
bers. Cholinergic bers are known to bear the p75 neu-
rotrophin receptor (p75NTR), suggesting that neurotrophins We are now beginning to understand at least some of the prin-
play a role. Nerve growth factor (NGF), one of the ligands for ciples underlying pathway formation in the hippocampus.
the p75 receptor, is synthesized by GABAergic hippocampal Different cellular and molecular factors, probably being effec-
neurons, which are distributed over all hippocampal layers, tive only during certain developmental time windows, are
compatible with the diffuse termination of septohippocampal involved in the formation of the different projections to the
bers. Moreover, NGF-synthesizing hippocampal nonprinci- hippocampus. When describing the formation of a pathway,
pal neurons project to the septum (Acsady et al., 2000), we found it useful to distinguish between axonal path nding ,
thus providing both a neurotrophic source and a template target recognition, and synapse formation. For pathnding and
 Morphological Development of the Hippocampus 127

target recognition, pioneer neurons provide a template by an 4.4.1 Neurogenesis

early-formed projection. Membrane-bound and soluble mol-
ecules and components of the extracellular matrix also seem The duration of embryonic development is 19 to 21 days in
to play a role at different times of development. A variety of mice and rats (life-span 1.53.0 years), 165 days in rhesus
ligandreceptor interactions has been elucidated that, by their monkeys (life span 2530 years), and 280 days in humans. As
repulsive or attractive effects, may guide the growth cone of a we have already noted, cell formation is a relatively late event
growing axon to the target region. Undoubtedly, new mole- in rodents: Pyramidal cells are generated during the second
cules or families of molecules of this kind will be discovered in half of the embryonic period, and granule cell formation
the future. Extracellular matrix molecules, proteoglycans such starts only a few days before birth. Furthermore, 85% of the
as neurocan, may be essential for the segregation of ber sys- dentate granule cells are formed postnatally. In contrast, neu-
tems in the hippocampus (Frster et al., 2001; Zhao et al., rogenesis in the primate hippocampal formation takes place
2003). relatively early. In rhesus monkeys, the rst neurons appear
Synapse formation is dealt with only briey here, and almost simultaneously in the various subregions of the hip-
the reader is referred to quantitative electron microscopic pocampal formation, from the entorhinal cortex to the den-
analyses and Golgi studies on the development of synapses tate gyrus, between E 36 and E 38 (Rakic and Nowakowski,
and dendritic spines (e.g., Crain et al., 1973; Minkwitz 1976; 1981). Except for the dentate gyrus, cell formation ceases dur-
Frotscher et al., 1977). This is a rapidly developing eld, ing the rst half of pregnancy, between E 62 and E 65. Granule
and more recent studies have provided evidence that the cell formation lasts until the end of the rst postnatal month,
mechanisms leading to new contacts or changes in the shape with only 15% of the granule cells forming postnatally (Rakic
of synapses or spines may not only be effective during and Nowakowski, 1981). Granule cell formation has also been
development but may also underlie plastic processes in the found in the dentate gyrus of adult monkeys (Kornack and
adult organism. A striking example is Engert and Bonhoeffers Rakic, 1999).
(1999) in vitro observation of new spines formed during In humans, the time of pyramidal cell neurogenesis is
the development of long-term potentiation (LTP). Similarly, not known and must be extrapolated from the time of the
Woolley and McEwen (1993) have shown that estrogen appearance of the cytoarchitectonic layers. Subelds of the
application to ovariectomized adult rats induces new spines hippocampal formation, such as the entorhinal cortex, subicu-
on CA1 pyramidal cell dendrites, an effect likely caused lum, and hippocampus, are discernible between E 70 and E 80,
by estrogen acting primarily on CA3 pyramidal cells and whereas the dentate granular layer appears around E 90
inducing the sprouting of their Schaffer collaterals to CA1 (Humphrey, 1967). Between E 160 and E 170, the cytoarchitec-
pyramidal neurons (Rune et al., 2002). Interestingly, recent tonic characteristics of all of the divisions of the hippocampal
studies have shown that estrogens synthesized in the formation are similar to what is observed in adults
hippocampus are important for hippocampal synaptic (Humphrey, 1967; Arnold and Trojanowski, 1996). With the
plasticity (Kretz et al., 2004). Future studies will help us estab- aid of the mitotic marker MIB-1 (Ki-67), we have directly
lish whether these plastic changes at spine synapses are shown that the ventricular zone and hilus contain a large num-
specic for hippocampal neurons or are a more general phe- ber of proliferating cells at E 100 (Fig. 44). The number of
nomenon occurring at many synapses in the central nervous labeled cells decreases rapidly from E 170 onward (24th week),
system. but MIB-1 (Ki-67)-positive cells are still observed in the hilus
and granule cell layer during the rst 6 months postnatally
(Seress et al., 2001).

4.4 Development of the Primate 4.4.2 Neuronal Differentiation

Hippocampal Formation
Differentiation of pyramidal neurons and granule cells con-
As in the previous chapter (see Chapter 3), one may raise the tinues for a relatively long period in rodents and primates.
question of how hippocampal development varies across phy- Spine density increases until the time of sexual maturation,
logenetically distant species such as rats, monkeys, and and total dendritic length is not reached before P 90 in the rat
humans. What is similar and what is different? Do the steps of CA1 area (Pokorny and Yamamoto, 1981). In the monkey, the
hippocampal development change proportionally with an spine density of CA3 and CA1 pyramidal cells does not reach
extended life-span, or are there some fundamental differences adult levels before sexual maturation (i.e., during the fourth
in the mechanisms and the timing of cell formation, cell year in rhesus monkeys) (Seress and Ribak, 1995b, unpub-
migration, dendritic development, and formation of neuronal lished observation). In humans, spine density and total den-
connectivity? We provide a short comparative analysis of hip- dritic length of CA1 pyramidal neurons increase until at least
pocampal development in primates. For details the reader the third postnatal year (Simic et al., 2001). In contrast to the
should consult a review on monkey and human hippocampal long-term differentiation of principal cells, local circuit neu-
development (Seress, 2001). rons in the primate hippocampal formation mature fast, as
128 The Hippocampus Book

indicated by the early fetal appearance of neurochemical Altman J (1966) Autoradiographic and histological studies of post-
markers of these neurons (Del Fiacco and Quartu, 1989; natal neurogenesis. II. A longitudinal investigation of kinetics,
Berger et al., 1993, 1999). migration and transformation of cells incorporating tritiated
Arrival of afferents at the rat hippocampal formation coin- thymidine in infant rats, with special reference to postnatal neu-
rogenesis of some brain regions. J Comp Neurol 128:431474.
cides with the development of their target cells. Thus, it takes
Altman J, Bayer SA (1975) Postnatal development of the hippocam-
place during the perinatal and early postnatal period, as pre-
pal dentate gyrus under normal and experimental conditions.
viously noted. In both the monkey and human, however, In: The hippocampus, vol 2 (Isaacson RL, Pribram KH, eds), pp.
afferent projections from the hypothalamus as well as from 95122. New York: Plenum Press.
the entorhinal cortex arrive early during the fetal period Altman J, Bayer SA (1990a) Mosaic organization of the hippocampal
(Kostovic et al., 1989; Hevner and Kinney, 1996; Berger et al., neuroepithelium and the multiple germinal sources of dentate
2001). In contrast to extrinsic afferents, intrinsic associational granule cells. J Comp Neurol 301:325342.
connections of the rodent hippocampus develop relatively Altman J, Bayer SA (1990b) Prolonged sojourn of developing pyram-
late, partly owing to the late formation of granule cells. idal cells in the intermediate zone of the hippocampus and their
Connections between dentate granule cells and CA3 pyrami- settling in the stratum pyramidale. J Comp Neurol 301:343364.
dal neurons as well as between dentate granule cells and hilar Altman J, Bayer SA (1990c) Migration and distribution of two popu-
lations of hippocampal granule cell precursors during the peri-
mossy cells develop postnatally (Ribak et al., 1985). This late
natal and postnatal periods. J Comp Neurol 301:365381.
development coincides with the prolonged postnatal develop-
Altman J, Das DG (1966) Autoradiographic and histological studies
ment of hippocampal functions (Nadel and Willner, 1989). In of postnatal neurogenesis. I. A longitudinal investigation of the
contrast, the development of associational connections is rel- kinetics, migration and transformation of cells incorporating
atively fast in rhesus monkeys, and the mainly prenatally tritiated thymidine in neonate rats, with special reference to
formed granule cells establish a well developed synaptic net- postnatal neurogenesis in some brain regions. J Comp Neurol
work by the fth to seventh postnatal month (Seress and 126:337390.
Ribak, 1995a,b). Surprisingly, this is not the case in humans, Amaral DG, Kurz J (1985) The time of origin of cells demonstrating
where the prenatally formed granule cells appear to form the glutamic acid-like immunoreactivity in the hippocampal for-
rst connections with hilar mossy cells and CA3 pyramidal mation of the rat. Neurosci Lett 59:3339.
neurons as late as during the 33rd gestational week, as indi- Angevine JB (1965) Time of neuron origin in the hippocampal
region: an autoradiographic study in the mouse. Exp Neurol
cated by the late appearance of excrescences on hilar mossy
Suppl 2:170.
cells and CA3 pyramidal neurons (Purpura, 1975). In fact,
Arnold SE, Trojanowski JQ (1996) Human fetal hippocampal devel-
there are almost no light microscopic signs of such connec- opment. I. Cytoarchitecture, myeloarchitecture and neuronal
tions at birth (Seress and Mrzljak, 1992; Seress, 2001). morphologic features. J Comp Neurol 367:274292.
Taken together, developmental events of the hippocampal Bayer SA (1980) Development of the hippocampal region in the rat.
formation are different in rodents and primates. However, I. Neurogenesis examined with 3H-thymidine autoradiography.
there are also common traits. In both rodents and humans the J Comp Neurol 190:87114.
late development of intrahippocampal associational connec- Bayer SA, Altman J (1987) Directions in neurogenetic gradients and
tions may inuence the maturation of the entire hippocampal patterns of anatomical connections in the telencephalon. Prog
network. The factors contributing to this late development are Neurobiol 29:57106.
likely to be different in rodents and primates. In rodents, the Berger B, Alvarez C, Goldman-Rakic PS (1993) Neurochemical devel-
opment of the hippocampal region in the fetal rhesus monkey.
late formation of granule cells may play a role. It remains to be
I. Early appearance of peptides, calcium binding proteins,
elucidated why the early-formed granule cells in humans do
DARPP-32 and the monoamine innervation in the entorhinal
not immediately form synapses with their target cells as is the cortex during the rst half of gestation (E47 to E90). Hippocam-
case in nonhuman primates. pus 3:279305.
Berger B, DeGrissac N, Alvarez C (1999) Precocious development of
parvalbumin-like immunoreactive interneurons in the hip-
ACKNOWLEDGMENTS pocampal formation and entorhinal cortex of the fetal
cynamolgus monkey. J Comp Neurol 403:309331.
This work was supported by Deutscher Akademischer Austausch- Berger B, Esclapez M, Alvarez C, Meyer G, Catala M (2001) Human
dienst (DAAD) with grants 323/2000 and 324/jo-H/2002, the Hun- and monkey fetal brain development of the supramammillary-
garian Ministry of Education (FKFP) with grant 0515/2000, hippocampal projections: a system involved in the regulation of
and the Deutsche Forschungsgemeinschaft (SFB 505 and Transregio theta activity. J Comp Neurol 429:515529.
SFB TR-3). Borrell V, Del Rio JA, Alcantara S, Derer M, Martinez A, DArcangelo
G, Nakajima K, Mikoshiba K, Derer P, Curran T, Soriano E
(1999) Reelin regulates the development and synaptogenesis of
REFERENCES the layer-specic entorhino-hippocampal connections. J Neu-
rosci 19:13451358.
Acsady L, Pascual M, Rocamora N, Soriano E, Freund TF (2000) Ceranik K, Deng J, Heimrich B, Lbke J, Zhao S, Frster E, Frotscher
Nerve growth factor but not neurotrophin-3 is synthesized by M (1999) Hippocampal Cajal-Retzius cells project to the
hippocampal GABAergic neurons that project to the medial entorhinal cortex: retrograde tracing and intracellular labelling
septum. Neuroscience 98:2331. studies. Eur J Neurosci 11:42784290.
 Morphological Development of the Hippocampus 129

Chedotal A, del Rio JA, Ruiz M He Z, Borrell V, de Castro F, Ezan F, Frster E, Tielsch A, Saum B, Weiss KH, Johanssen C, Graus-Porta
Goodman CS, Tessier-Lavigne M, Sotelo C, Soriano E (1998) D, Mller U, Frotscher M (2002) Reelin, disabled 1 and beta1-
Semaphorins III and IV repel hippocampal axons via two dis- integrins are required for the formation of the radial glial
tinct receptors. Development 125:43134323. scaffold in the hippocampus. Proc Natl Acad Sci USA 99:
Crain B, Cotman C, Taylor D, Lynch G (1973) A quantitative electron 1317813183.
microscopic study of synaptogenesis in the dentate gyrus of the Freund TF, Antal M (1988) GABA-containing neurons in the septum
rat. Brain Res 63:195204. control inhibitory interneurons in the hippocampus. Nature
Czurk A, Czeh B, Seress L, Nadel L, Bures J (1997) Severe spatial 336:170173.
navigation decit in the Morris watermaze after single high dose Freund TF, Buzski G (1996) Interneurons of the hippocampus.
of neonatal X-ray irradiation in the rat. Proc Natl Acad Sci USA Hippocampus 6:347470.
94:27662771. Frotscher M (1997) Dual role of Cajal-Retzius cells and reelin in cor-
DArcangelo G, Miao GG, Chen S-C, Soares HD, Morgan JI, tical development. Cell Tissue Res 290:315322.
Curran T (1995) A protein related to extracellular matrix Frotscher M (1998) Cajal-Retzius cells, reelin, and the formation of
proteins deleted in the mouse mutant reeler. Nature 374: layers. Curr pOin Neurobiol 8:570575.
719723. Frotscher M, Heimrich B (1993) Formation of layer-specic ber
DArcangelo G, Nakajima K, Miyata T, Ogawa M, Mikoshiba K, projections to the hippocampus in vitro. Proc Natl Acad Sci USA
Curran T (1997) Reelin is a secreted glycoprotein recognized by 90:1040010403.
the CR-50 monoclonal antibody. J Neurosci 17:2331. Frotscher M, Leranth C (1985) Cholinergic innervation of the rat
Del Fiacco M, Quartu M (1989) Ontogenetic variance of substance P hippocampus as revealed by choline acetyltransferase immuno-
in the human hippocampus. In: The hippocampus: newvistas cytochemistry: a combined light and electron microscopic
(Chan Palay V, Kohler C, eds), pp 131144. New York: Alan R. study. J Comp Neurol 239:237246.
Liss. Frotscher M, Leranth C (1986) The cholinergic innervation of the rat
Deller T, Drakew A, Frotscher M (1999a) Different primary target fascia dentata: identication of target structures on granule cells
cells are important for ber lamination in the fascia dentata: a by combining choline acetyltransferase immunocytochemistry
lesson from reeler mutant mice. Exp Neurol 156:239253. and Golgi impregnation. J Comp Neurol 243:181190.
Deller T, Drakew A, Heimrich B, Frster E, Tielsch A, Frotscher M Frotscher M, Hmori J, Wenzel J (1977) Transneuronal effects of
(1999b) The hippocampus of the reeler mutant mouse: ber entorhinal lesions in the early postnatal period on synapto-
segregation in area CA1 depends on the position of the postsy- genesis in the hippocampus of the rat. Exp Brain Res 30:
naptic target cells. Exp Neurol 156:254267. 549560.
Del Rio JA, Heimrich B, Borrell V, Frster E, Drakew A, Alcntara S, Frotscher M, Schlander M, Leranth C (1986) Cholinergic neurons in
Nakajima K, Miyata T, Ogawa M, Mikoshiba K, Derer P, the hippocampus: a combined light and electron microscopic
Frotscher M, Soriano E (1997) A role for Cajal-Retzius cells and immunocytochemical study in the rat. Cell Tissue Res 246:
reelin in the development of hippocampal connections. Nature 293301.
385:7074. Frotscher M, Vida I, Bender R (2000a) Evidence for the existence of
Drakew A, Frotscher M, Heimrich B (1999) Blockade of neuronal non-GABAergic, cholinergic interneurons in the rodent hip-
activity alters spine maturation of dentate granule cells but not pocampus. Neuroscience 96:2731.
their dendritic arborization. Neuroscience 94:767774. Frotscher M, Drakew A, Heimrich B (2000b) Role of afferent inner-
Drakew A, Deller T, Heimrich B, Gebhardt C, Del Turco D, Tielsch A, vation and neuronal activity in dendritic development and
Frster E, Herz J, Frotscher M (2002) Dentate granule cells in spine maturation of fascia dentata granule cells. Cereb Cortex
reeler mutants and VLDLR and ApoER2 knockout mice. Exp 10:946951.
Neurol 176:1224. Frotscher M, Seress L, Abraham H, Heimrich B (2001) Early gener-
Dupuy ST, Houser CR (1997) Developmental changes in GABAergic ated Cajal-Retzius cells have different functions in cortical
neurons of the rat dentate gyrus: an in situ hybridization and development. In: Brain stem cells (Miyan J, Thorndyke M,
birth dating study. J Comp Neurol 389:402418. Beesley PW, Bannister C, eds), pp. 4349. Oxford: BIOS
Dupuy ST, Houser CR (1999) Evidence for changing positions of Scientic Publishers.
GABA neurons in the developing rat dentate gyrus. Frotscher M, Haas CA, Frster E (2003) Reelin controls granule cell
Hippocampus 9:186199. migration in the dentate gyrus by acting on the radial glial scaf-
Engert F, Bonhoeffer T (1999) Dendritic spine changes associated fold. Cereb Cortex 13:634640.
with hippocampal long-term synaptic plasticity. Nature Galceran J, Miyashita-Lin EM, Devaney E, Rubenstein JLR,
399:6670. Grosschedl R (2000) Hippocampus development and genera-
Eriksson PS, Perlieva E, Bjrk-Eriksson T, Alborn A, Nordborg C, tion of dentate gyrus granule cells is regulated by LEF1.
Peterson DA, Gage FH (1998) Neurogenesis in the adult human Development 127:469482.
hippocampus. Nat Med 4:13131317. Gundersen RW, Barrett JN (1979) Neuronal chemotaxis: chick dorsal-
Falconer DS (1951) Two new mutants, trembler and reeler, with root axons turn toward high concentrations of nerve growth
neurological actions in the house mouse (Mus musculus L.). J factor. Science 206:10791080.
Genet P( aris) 50:192201. Hjos N, Papp E, Acsdy L, Levey AI, Freund TF (1998) Distinct
Frster E, Kaltschmidt C, Deng J, Cremer H, Deller T, Frotscher M interneuron types express m2 muscarinic receptor immunore-
(1998) Lamina-specic cell adhesion on living slices of hip- activity on their dendrites or axon terminals in the hippocam-
pocampus. Development 125:33993410. pus. Neuroscience 82:355376.
Frster E, Zhao S, Frotscher M (2001) Hyaluronan-associated adhe- Halasy K, Somogyi P (1993a) Distribution of GABAergic synapses
sive cues control fiber segregation in the hippocampus. and their targets in the dentate gyrus of rat: a quantitative
Development 128:30293039. immunoelectron microscopic analysis. J Hirnforsch 34:299308.
130 The Hippocampus Book

Halasy K, Somogyi P (1993b) Subdivisions in the multiple Nadel L, Willner J (1989) Some implications of postnatal maturation
GABAergic innervation of granule cells in the dentate gyrus of of the hippocampus. In: The hippocampus: newvistas (Chan-
the rat hippocampus. Eur J Neurosci 5:411429. Palay V, Kohler C, eds), pp. 1731. New York: Alan R Liss.
Halasy K, Buhl EH, Lorinczi Z, Tams G, Somogyi P (1996) Synaptic Nguyen-Ba-Charvet KT, Brose K, Marillat V, Kidd T, Goodman CS,
target selectivity and input of GABAergic basket and bistratied Tessier-Lavigne M, Sotelo C, Chedotal A (1999) Slit2-mediated
interneurons in the CA1 area of the rat hippocampus. chemorepulsion and collapse of developing forebrain axons.
Hippocampus 6:306329. Neuron 22:463473.
Hevner RF, Kinney HC (1996) Reciprocal entorhinal-hippocampal Pleasure SJ, Anderson S, Hevner R, Bagri A, Lowenstein DH,
connections established by human fetal midgestation. J Comp Rubenstein JLR (2000) Cell migration from the ganglionic emi-
Neurol 372:384394. nences is required for the development of hippocampal
Hirotsune S, Takahare T, Sasaki N, Hirose K, Yoshiki A, Ohashi T, GABAergic interneurons. Neuron 26:727740.
Kusakabe M, Murakami Y, Muramatsu M, Watanabe S, Nakao Pokorny J, Yamamoto Y (1981) Postnatal ontogenesis of hippocam-
K, Katsuki M, Hayashizaki Y (1995) The reeler gene encodes a pal CA1 area in rats. I. Development of dendritic arborization in
protein with an EGF-like motif expressed by pioneer neurons. pyramidal neurons. Brain Res Bull 7:113120.
Nat Genet 10:7783. Purpura DS (1975) Normal and aberrant neuronal development in
Humphrey T (1967) The development of the human hippocampal the cerebral cortex of human fetus and young infant. In: Brain
ssure. J Anat 101:655676. mechanisms in mental retardation: UCLA forum in medical sci-
Katona I, Sperlgh B, Sk A, Kofalvi A, Vizi ES, Freund TF (1999) ences pp. 141169. San Diego: Academic Press.
Presynaptically located CB1 cannabinoid receptors regulate Rakic P, Nowakowski RS (1981) The time of origin of neurons in the
GABA release from axon terminals of specic hippocampal hippocampal region of the rhesus monkey. J Comp Neurol
interneurons. J Neurosci 19:45444558. 196:99128.
Kempermann G, Kuhn HG, Gage FH (1997a) More hippocampal Ramn y Cajal S (1911) Histologie du systme nerveux de lhomme et
neurons in adult mice living in an enriched environment. des vertbrs , vol 2. Paris: Maloine.
Nature 386:493495. Retzius G (1893) Die Cajalschen Zellen der Grosshirnrinde beim
Kempermann G, Kuhn HG, Gage FH (1997b) Genetic inuence on Menschen und bei Sugetieren. Biol Unters 5:19.
neurogenesis in the dentate gyrus of adult mice. Proc Natl Acad Retzius G (1894) Weitere Beitrge zur Kenntnis der Cajalschen Zellen
Sci USA 94:1040910414. der Grosshirnrinde des Menschen. Biol Unters 6:2934.
Kornack DR, Rakic P (1999) Continuation of neurogenesis in the Ribak CE, Seress L (1983) Five types of basket cell in the hippocam-
hippocampus of the adult macaque monkey. Proc Natl Acad Sci pal dentate gyrus. J Neurocytol 12:577597.
USA 96:57685773. Ribak CE, Seress L, Amaral DG (1985) The development, ultrastruc-
Kostovic I, Seress L, Mrzljak L, Judas M (1989) Early onset of synapse ture and synaptic connections of the mossy cells of the dentate
formation in the human hippocampus: a correlation with Nissl- gyrus. J Neurocytol 14:717730.
Golgi architectonics in 15 and 16.5-week-old fetuses. Rune GM, Wehrenberg U, Prange-Kiel J, Zhou L, Adelmann G,
Neuroscience 30:105116. Frotscher M (2002) Estrogen up-regulates estrogen receptor
Kretz O, Fester L, Wehrenberg U, Zhou L, Brauckmann S, Zhao S, alpha and synaptophysin in slice cultures of rat hippocampus.
Prange-Kiel J, Naumann T, Jarry H, Frotscher M, Rune GM Neuroscience 113:167175.
(2004) Hippocampal synapses depend on hippocampal estro- Savaskan NE, Plaschke M, Ninnemann O, Spillmann AA, Schwab
gen synthesis. J Neurosci 24:59135921. ME, Nitsch R, Skutella T (1999) Myelin does not inuence the
Lang U, Frotscher M (1990) Postnatal development of nonpyramidal choice behaviour of entorhinal axons but strongly inhibits their
neurons in the rat hippocampus (areas CA1 and CA3): a com- outgrowth length in vitro. Eur J Neurosci 11:316326.
bined Golgi/electron microscope study. Anat Embryol 181: Seress L (1977) The postnatal development of rat dentate gyrus and
533545. the effect of early thyroid hormone treatment. Anat Embryol
Laurberg S, Hjorth-Simonsen A (1977) Growing central axons 151:335339.
deprived of normal target neurones by neonatal X-ray irradia- Seress L (1978) Divergent responses to thyroid hormone treatment of
tion still terminate in a precisely laminated fashion. Nature the different secondary germinal layers in the postnatal rat
269:158160. brain. J Hirnforsch 19:395403.
Linke R, Frotscher M (1993) Development of the rat septohip- Seress L (2001) Morphological changes in the human hippocampal
pocampal projection: tracing with DiI and electron microscopy formation from midgestation to early childhood. In: Handbook
of identied growth cones. J Comp Neurol 332:6988. of developmental cognitive neuroscience (Nelson CA, Luciana M,
Linke R, Pabst T, Frotscher M (1995) Development of the hippocam- eds), pp. 4558. Cambridge, MA: MIT Press.
poseptal projection in the rat. J Comp Neurol 351:602616. Seress L, Frotscher M (1990) Morphological variability is a charac-
Lbbers K, Wolff JR, Frotscher M (1985) Neurogenesis of GABAergic teristic feature of granule cells in the primate fascia dentata: a
neurons in the rat dentate gyrus: a combined autoradiographic combined Golgi-electron microscopic study. J Comp Neurol
and immunocytochemical study. Neurosci Lett 62:317322. 293:253267.
Minkwitz H-G (1976) Zur Entwicklung der Neuronenstruktur des Seress L, Mrzljak L (1987) Basal dendrites of granule cells are normal
Hippocampus whrend der pr- und postnatalen Ontogenese features of the fetal and adult dentate gyrus of both monkey and
der Albinoratte. III. Mitteilung: Morphometrische Erfassung human hippocampal formations. Brain Res 405:169174.
der ontogenetischen Vernderungen in Dendritenstruktur und Seress L, Mrzljak L (1992) Postnatal development of mossy cells in
Spinebesatz an Pyramidenneuronen (CA1) des Hippocampus. J the human dentate gyrus: a light microscopic Golgi study.
Hirnforsch 17:255275. Hippocampus 2:127142.
 Morphological Development of the Hippocampus 131

Seress L, Ribak CE (1983) GABAergic neurons in the dentate gyrus Staneld BB, Cowan WM (1979) The morphology of the hippocam-
appear to be local circuit and projection neurons. Exp Brain Res pus and dentate gyrus in normal and reeler mice. J Comp Neurol
50:173182. 185:393422.
Seress L, Ribak CE (1990) Postnatal development of the light and Stein E, Savaskan NE, Ninnemann O, Nitsch R, Zhou R, Skutella
electron microscopic features of basket cells in the hippocampal T (1999) A role for the Eph ligand ephrin-A3 in entorhino-
dentate gyrus of the rat. Anat Embryol 181:547566. hippocampal axon targeting. J Neurosci 15:88858893.
Seress L, Ribak CE (1995a) Postnatal development and synaptic con- Stephan, H (1975) Allocortex: Handbuch der mikroskopischen
nections of hilar mossy cells in the hippocampal dentate gyrus Anatomie des Menschen, vol IV/9. Berlin: Springer.
of rhesus monkeys. J Comp Neurol 355:93110. Steup A, Ninnemann O, Savaskan NE, Nitsch R, Pschel AW, Skutella
Seress L, Ribak CE (1995b) The postnatal development of CA3 T (1999) Semaphorin D acts as a repulsive factor for entorhinal
pyramidal neurons and their afferents in the Ammons horn of and hippocampal neurons. Eur J Neurosci 11:729734.
rhesus monkeys. Hippocampus 5:217231. Steup A, Lohrum M, Hamscho N, Savaskan NE, Ninnemann O,
Seress L, Frotscher M, Ribak CE (1989) Local circuit neurons in both Nitsch R, Fujisawa H, Pschel AW, Skutella T (2000) Sema3C
the dentate gyrus and Ammons horn establish synaptic connec- and netrin-1 differentially affect axon growth in the hippocam-
tions with principal neurons in ve day-old rats: a morpholog- pal formation. Mol Cell Neurosci 15:141155.
ical basis for inhibition in early development. Exp Brain Res Super H, Soriano E (1994) The organization of the embryonic and
78:19. early postnatal murine hippocampus. II. Development of
Seress L, brahm H, Tornczky T, Kosztolnyi GY (2001) Cell for- entorhinal, commissural and septal connections studied with
mation in the human hippocampal formation from mid- the lipophilic tracer DiI. J Comp Neurol 344:101120.
gestation to the late postnatal period. Neuroscience 105:3951. Super H, Martinez A, del Rio JA, Soriano E (1998) Involvement of
Simic G, Bjelos M, Rasin MR, Seress L, Kostovic I (2001) Postnatal distinct pioneer neurons in the formation of layer-specic con-
development of human CA1 pyramidal neurons. Soc Neurosci nections in the hippocampus. J Neurosci 18:46164626.
Abstr 27:Program 250.2. Tamamaki N (1999) Development of afferent ber lamination in the
Skutella T, Nitsch R (2001) New molecules for hippocampal develop- infrapyramidal blade of the rat dentate gyrus. J Comp Neurol
ment. Trends Neurosci 24:107113. 411:257266.
Skutella T, Savaskan NE, Ninnemann O, Nitsch R (1999) Target- and Tessier-Lavigne M, Goodman CS (1996) The molecular biology of
maturation-specic membrane-associated molecules determine axon guidance. Science 274:11231132.
the ingrowth of entorhinal bers into the hippocampus. Dev Tole S, Christian C, Grove EA (1997) Early specication and
Biol 211:277292. autonomous development of cortical elds in the mouse hip-
Soriano E, Frotscher M (1989) A GABAergic axo-axonic cell in the pocampus. Development 124:49594670.
fascia dentata controls the main excitatory hippocampal path- Vida I, Frotscher M (2000) A hippocampal interneuron associated
way. Brain Res 503:170174. with the mossy fiber system. Proc Natl Acad Sci USA
Soriano E, Frotscher M (1993) GABAergic innervation of the rat fas- 97:12751280.
cia dentate: a novel type of interneuron in the granule cell layer Vida I, Halasy K, Szinyei C, Somogyi P, Buhl EH (1998) Unitary IPSPs
with extensive axonal arborization in the molecular layer. J evoked by interneurons at the stratum radiatum/lacunosum-
Comp Neurol 334:385396. moleculare border in the CA1 area of the rat hippocampus in
Soriano E, Cobas A, Fairn A (1989a) Neurogenesis of glutamic acid vitro. J Physiol (Lond) 503:755773.
decarboxylase-immunoreactive cells in the hippocampus of the Von Campe G, Spencer DD, de Lanerolle NC (1997) Morphology of
mouse. I. Regio superior and regio inferior. J Comp Neurol dentate granule cells in the human epileptogenic hippocampus.
281:586602. Hippocampus 7:472488.
Soriano E, Cobas A, Fairen A (1989b) Neurogenesis of glutamic acid Woolley CS, McEwen BS (1993) Roles of estradiol and progesterone
decarboxylase-immunoreactive cells in the hippocampus of the in regulation of hippocampal spine density during the estrous
mouse. II. Area dentata. J Comp Neurol 281:603611. cycle in the rat. J Comp Neurol 336:293306.
Spigelman I, Yan XX, Obenaus A, Lee EY-S, Wasterlain CG, Ribak CE Zhao S, Frster E, Chai X, Frotscher M (2003) Different signals con-
(1998) Dentate granule cells form novel basal dendrites in a rat trol laminar specicity of commissural and entorhinal bers to
model of temporal lobe epilepsy. Neuroscience 86:109120. the dentate gyrus. J Neurosci 23:73517357.
Spruston N, Lbke J, Frotscher M (1997) Interneurons in the stratum Zhao S, Chai X, Frster E, Frotscher M (2004) Reelin is a positional
lucidum of the rat hippocampus: an anatomical and electro- signal for the lamination of dentate granule cells. Development
physiological characterization. J Comp Neurol 385:427440. 131:51175125.
This page intentionally left blank
5 Nelson Spruston and Chris McBain

Structural and Functional Properties

of Hippocampal Neurons

models that can reproduce or explain virtually any aspect of

5.1 Overview neuronal function. The purpose of such models, of course, is
to offer insight into the neurons function within a network
One of the most striking qualities of Cajals drawings is the and provide experimentally testable predictions.
great diversity of form exhibited by neurons in the central In this chapter, our aim is to summarize the wealth of
nervous system (CNS). Looking at these drawings, it is diffi- structural, histological, and physiological information that is
cult to escape the conclusion that neurons with different mor- available for hippocampal neurons. Most of the available data
phologies are likely to function differently. Indeed a plethora on this topic come from studies of the rat hippocampal for-
of anatomical, histological, and physiological studies support mation. Accordingly, we mention species primarily when
the view that diversity in form is mirrored by diversity of referring to data from species other than rat. We begin with
function. The hippocampus presents, in many ways, a micro- the CA1 pyramidal neuron because it is arguably the most
cosm of the cellular diversity exhibited on a larger scale studied class of neuron in the brain and probably better
throughout the CNS. An important step on the path to under- understood from both structural and functional points of
standing the function of the hippocampus is to reveal the view than any other type of neuron in the hippocampus. From
unique morphological and physiological properties of the there we proceed by comparing the function of the other
great variety of neurons it comprises. pyramidal neurons in the hippocampus: those in CA3 and the
Much can be learned by examining neuronal structure. The subiculum. We next cover two nonpyramidal excitatory neu-
dendritic tree denes which inputs can be received, and the rons in the hippocampus: granule cells of the dentate gyrus
axonal arborization indicates the targets of the output and mossy cells of the hilus. Finally, we conclude by describ-
(although dendro-dendritic and axo-axonal communication ing the properties of a variety of neurons in other regions
should not be forgotten). Electron microscopy provides comprising the hippocampal formation: the entorhinal cor-
details about the nature of each neurons inputs and output. It tex, presubiculum, and parasubiculum. In the nal section, we
allows us to infer which synapses are inhibitory (symmetrical) consider the many types of interneurons found in the hip-
and which are excitatory (asymmetrical) and where modula- pocampus.
tory machinery resides (e.g., endoplasmic reticulum for cal-
cium release, ribosomes for protein synthesis, vesicles for
neurotransmitter release or insertion of membrane proteins).
Histological studies (antibody or cDNA binding, histochemi- 5.2 CA1 Pyramidal Neurons
cal assays) provide further information, ranging from which
neurotransmitters are used to which voltage-gated channels The vastness of the literature on CA1 pyramidal neurons is
are present and where. Physiological studies, using electrodes largely attributable to a combination of structural considera-
and imaging both in vitro and in vivo, reveal the dynamic tions, cell viability, and historical accidents. Although long-
function of each neuron. term potentiation was rst described in the dentate gyrus,
A detailed picture of the function of each neuron can only most of the studies in the decades that followed its original
emerge by bringing together information from this variety of description have focused on the CA1 region. This is because of
techniques. Ultimately, complete understanding of the func- the relative ease of obtaining eld potential recordings and
tion of each type of neuron may be encapsulated by detailed intracellular recordings in this region. Also, the Schaffer col-

133
134 The Hippocampus Book

lateral axons from CA3 form a homogeneous pathway that is The number of these oblique branches varies from 9 to 30,
easily activated to study synaptic transmission and plasticity. with a mean of 17 (Bannister and Larkman, 1995a; Pyapali et
Studies of CA1 are more numerous than adjacent CA3 al., 1998) (Fig. 51C). Oblique dendrites branch no more than
because it is generally easier to keep cells in this region alive a few times, with a typical branch bifurcating just once at a
and healthy in slice preparations. Recently, CA1 pyramidal location close to its origin from the apical trunk. Despite their
neurons have been the focus of several studies of dendritic limited branching, however, oblique dendrites constitute most
integration because of the large primary apical dendrite, from of the dendritic length in the stratum radiatum (Bannister
which dendritic patch-clamp recordings can be obtained rou- and Larkman, 1995a; Megias et al., 2001). After the primary
tinely. These factors have contributed to tremendous advances apical trunk enters the stratum lacunosum-moleculare the
in understanding synaptic transmission, integration, and plas- apical dendrites continue to branch, forming a structure
ticity in the CNS. referred to as the apical tuft, which has an average of about 15
terminal branches (Bannister and Larkman, 1995a; Trommald
5.2.1 Dendritic Morphology et al., 1995).
Emerging from the base of the pyramidal soma are two to
Two elaborately branching dendritic trees emerge from the eight dendrites (a mean of ve). Most of these dendrites
pyramid-shaped soma of CA1 neurons (Fig. 51A). The basal branch several times (maximum 15 branch points), forming a
dendrites occupy the stratum oriens, and the apical dendrites basal dendritic tree with about 40 terminal segments
occupy the stratum radiatum (proximal apical) and stratum (Bannister and Larkman, 1995; Pyapali et al., 1998). Most
lacunosum-moleculare (distal apical). [Note: in this context branches in the basal dendrites occur rather close to the soma,
the terms proximal and distal refer to the relative distance so the terminal segments are quite long, constituting about
from the soma and should not be confused with the proximo- 80% of the total dendritic length (Bannister and Larkman,
distal axis of the hippocampal circuit (dentate-CA3-CA1- 1995a; Trommald et al., 1995). CA1 neurons differ consider-
subiculum-entorhinal cortex) introduced in Chapter 3.] Both ably in the position of the cell body, with some cells located in
the apical and basal dendritic trees occupy a roughly conical the stratum oriens as far as 100 m from the border between
(sometimes ovoid) volume (Pyapali et al., 1998). The size of the stratum pyramidale and the stratum radiatum. Cells with
the CA1 dendritic tree depends on species and age, but the best somata farther from this border tend to have more terminal
available data are from adult rats. The distance from the stra- dendrites in the stratum oriens (basal dendrites) and fewer
tum pyramidale to the hippocampal ssure is about 600 m, terminal dendrites in the stratum radiatum (apical oblique
and the distance from the stratum pyramidale to the alveus is dendrites) (Bannister and Larkman, 1995a).
about 300 m, yielding a distance of just under 1 mm from Many of the organelles found in the cell body extend into
the tips of the basal dendrites to the tips of the apical den- the proximal apical dendrites of CA1 neurons (Harris and
drites. The combined length of all CA1 dendritic branches is Stevens, 1989; Spacek and Harris, 1997). These include struc-
12.0 to 13.5 mm: basal dendrites contribute about 36% of the tures such as the smooth endoplasmic reticulum (SER) and
total length, apical dendrites in the stratum radiatum con- the Golgi apparatus. At greater distances from the soma, how-
tribute about 40%, and apical dendrites in the stratum lacuno- ever, a more limited set of organelles is found in dendrites.
sum-moleculare contribute the remaining 24% (Bannister Microtubules, neurolaments, and actin are prominent and
and Larkman, 1995a; Ishizuka et al., 1995; Trommald et al., serve transport and motility functions, and they are likely to
1995; Megias et al., 2001). CA1 dendites are studded with be important in synaptic plasticity. A network of SER is also
spines (Fig. 51B), the sites of most excitatory synaptic con- present in dendrites, where it is likely to serve important func-
tacts throughout the dendritic tree (Ramon y Cajal, 1904; tions in calcium buffering and release. The SER forms a con-
Lorente de No, 1934; Bannister and Larkman, 1995b; Megias et tinuous reticulum, which can extend into dendritic spines
al., 2001). (Spacek and Harris, 1997). Mitochondria are also numerous
Studies of dendritic morphology in adult rats suggest that in dendrites and are often associated with the SER, but glyco-
CA1 pyramidal neurons can be broadly classied into two gen granules are largely absent, except in sliced tissue, where
groups on the basis of dendritic morphology. In one group of they probably represent a response to stress (Fiala et al., 2003).
neurons, the primary apical dendrite extends all the way Dendritic mitochondria likely contribute to calcium handling
through the stratum radiatum before bifurcating in the stra- in addition to serving as the primary energy source. In apical
tum lacunosum-moleculare. The other group of CA1 neurons dendrites of CA1 pyramidal neurons, mitochondria ll about
has apical dendrites that bifurcate in the stratum radiatum 2% of the area of the main apical dendrite but 13% of smaller-
(Bannister and Larkman, 1995a). Most other features of these diameter branches (Nafstad and Blackstad, 1966). Sorting
two classes of CA1 neurons are similar, so most structural endosomes and multivesicular bodies form the endosomal
studies consider CA1 pyramidal neurons as a single morpho- pathway, which is responsible for sorting and recycling pro-
logical class (Fig. 51C). teins in dendrites as well as packaging them for transport to
Along the length of the primary apical dendrite, several the soma for degradation. This endosomal compartment
dendritic branches emerge obliquely in the stratum radiatum. forms a network that spans as many as 20 spines in CA1 den-
 Hippocampal Neurons: Structure and Function 135

A B
EC layer III
s.l.m.

Schaffer s.r.
collaterals
retrosplenial cortex
perirhinal cortex
al
s.p. pt lateral septal nucleus
se
diagonal band of Broca
taenia tecta
d
Schaffer
s.o. mi medial frontal cortex
collaterals anterior olfactory nucleus
olfactory bulb
axon o ral nucleus accumbens
subiculum mp basal nucleus of amygdala
te
EC layer V alveus anterior/dorsomedial hypothalamus
parasubiculum fimbria

h.f.

C CA3 s.l.m.
subiculum
s.r.
s.p.
s.o.

Figure 51. CA1 dendritic morphology, spines, and synaptic inputs from Bannister and Larkman, 1995a.) B. Three views of a three-
and outputs. A. Camera lucida drawing of a CA1 pyramidal neuron dimensional reconstruction of a stretch of dendrite from the s.r.,
from an adult rat, showing the cell body in the stratum pyramidale illustrating the density of dendritic spines and their diverse struc-
(s.p.), basal dendrites in the stratum oriens (s.o.), and apical den- ture on CA1 pyramidal neuron dendrites. Bar  1 m. (Source:
drites in the stratum radiatum (s.r.) and stratum lacunosum-molec- Adapted from Synapse Web, by Kristen Harris (http://synapses.mcg.
ulare (s.l.m.). The major excitatory inputs in each layer and the edu/anatomy/ca1pyrmd/radiatum/K18/K18.stm). C. Line drawings
major outputs are also indicated. For the mbrial projection, the of several CA1 pyramidal neurons illustrating their diversity of den-
septo-temporal positions noted indicate the source of CA1 cells dritic structure. Dashed line represents the hippocampal ssure
projecting to different target regions. For the alveus projection, the (h.f.). Bar  500 m. (Source: Adapted from Pyapali et al., 1998.)
subiculum is the major target. Bar  100 m. (Source: Adapted
136 The Hippocampus Book

drites (Cooney et al., 2002). Clathrin-coated pits and vesicles receive roughly twice as many inhibitory synapses as excita-
are also present in CA1 dendrites, often near the ends of tubu- tory synapses. The remaining 24% of inhibitory synapses
lar endosomal compartments. Ribosomes are present in CA1 contact the distal basal dendrites. These dendrites are there-
dendrites, where they are usually clustered in the form of fore similar to the apical obliques in terms of diameter,
polyribosomes (Steward et al., 1996; Steward and Worley, spine density, and numbers of excitatory versus inhibitory
2002). synapses.
The dendritic spines studding the surface of CA1 dendrites
5.2.2 Dendritic Spines and Synapses exhibit a broad range of size and morphological complexity
(Fig. 51B). Although quantitative analysis of spine shapes
CA1 pyramidal neurons are covered with about 30,000 den- does not indicate clear groups of spines (Trommald and
dritic spines. Electron microscopic, immunocytochemical, Hulleberg, 1997), several investigators have used a variety of
and physiological analyses have all converged on the conclu- names to describe spines of different shapes (Laatsch and
sion that most dendritic spines receive excitatory synaptic Cowan, 1966; Peters and Kaiserman-Abramof, 1970; Harris
inputs, indicating that spine density can be used as a reason- and Kater, 1994; Sorra and Harris, 2000). Thin spines are long,
able measure of excitatory synapse density (Gray, 1959; narrow protrusions terminating in a small, bulbous head.
Andersen et al., 1966; Megias et al., 2001). The distribution of Sessile spines are long, narrow protrusions that do not termi-
spines on different dendritic domains has been carefully nate in a head. Stubby spines are small protrusions lacking a
quantied (Andersen et al., 1980; Bannister and Larkman, clearly distinguishable neck and a head. Mushroom spines
1995b; Megias et al., 2001). The density of dendritic spines have a narrow neck and a large, bulbous head. Branched
and synapses on CA1 pyramidal neurons is highest in the stra- spines consist of a neck that branches and terminates in two
tum radiatum and stratum oriens but lower in the stratum bulbous heads, each of which receives synaptic input from dif-
lacunosum-moleculare. The soma and the rst 100 m of the ferent axons. These ve types of spines are not specically
main apical dendrite (1.82.5 m diameter) are almost com- localized to any particular region of the CA1 dendritic tree but
pletely devoid of spines. The next 150 m of the apical den- can be found in close apposition on virtually any dendritic
drite (1.62.2 m diameter) has a very low spine density, but branch.
the nal 150 m of the main apical dendrite (1.01.5 m Spine structure is not static but may change in response to
diameter) has a high spine density (about seven spines per lin- neurotransmitter receptor activation or environmental and
ear micrometer). Spine density is lower in the oblique apical hormonal signals (Hering and Sheng 2001; Bonhoeffer and
branches (0.50.6 m diameter; about three spines per Yuste, 2002; Nikonenko et al., 2002; Nimchinsky et al., 2002).
micrometer); but because these dendrites constitute a large Growth of new spines and changes in the structure of existing
fraction of the total dendritic length, about 47% of all CA1 spines are possible substrates of synaptic plasticity in the hip-
spines are located on these branches. Together with the spines pocampus (Geinisman, 2000; Popov et al., 2004). Although
on the main apical dendrite, about 54% of all excitatory the plasticity of spine structure in vivo is a matter of some
synapses contact spines in the apical dendrites in the stratum controversy (Grutzendler et al., 2002; Trachtenberg et al.,
radiatum. Spine density is considerably lower in the apical tuft 2002), imaging studies of hippocampal neurons in vitro have
(0.21.2 m diameter). Although these dendrites contribute revealed a rather dynamic picture of spine morphology
about 20% of the total dendritic length, they contain only (Hosokawa et al., 1995; Dailey and Smith, 1996; Engert and
about 6% of the dendritic spines. Asymmetrical synapses, Bonhoeffer, 1999; Maletic-Savatic et al., 1999; Matsuzaki et al.,
however, are also found on dendritic shafts in this region. If 2004). One particularly noticeable feature of time-lapse
these synapses are excitatory, as is usually presumed, about movies of hippocampal dendrites is the continuous extension
10% of all excitatory synapses are located in the apical tuft. and retraction of lopodia. Occasionally, these lopodia
The rst 30 to 50 m of the basal dendrites (0.50.9 m extend without retracting fully, leading to the hypothesis that
diameter) have a low spine density, whereas the distal basal they form the precursors for mature, stable spines, which
dendrites (0.30.5 m diameter) have a spine density compa- eventually establish functional synaptic connections with
rable to that of the apical oblique dendrites. Accordingly, presynaptic boutons (Dailey and Smith, 1996; Fiala et. al.
about 36% of all excitatory synapses on CA1 neurons contact 1998; Parnass et al., 2000). The motility of dendritic lopodia
spines in the basal dendrites. and spines, which notably lack microtubules and neurola-
The distribution of symmetrical, GABA-positive synapses ments, is facilitated by a network of lamentous actin (Sorra
has also been quantied in CA1 neurons (Megias et al., and Harris, 2000).
2001). About 24% of these synapses contact the soma and CA1 neurons have been the target of numerous studies of
spine-free proximal dendrites (apical and basal). Somewhat the mechanisms of calcium entry, buffering, and extrusion in
surprisingly, the sparsely spiny and densely spiny regions dendritic spines (Yuste and Denk, 1995; Yuste et al., 1999;
of the main apical dendrite (in the stratum radiatum) receive Majewska et al., 2000). Such studies have contributed to
similar numbers of inhibitory synapses, at about 3% of the important advances in our understanding not only of calcium
total each. About 26% of the inhibitory synapses contact handling in spines but the mechanisms of synaptic transmis-
oblique apical branches. About 20% of all inhibitory synapses sion at single synapses and a variety of forms of morphologi-
contact dendrites of the apical tuft. These dendrites therefore cal and functional plasticity (Emptage et al., 1999; Matsuzaki
 Hippocampal Neurons: Structure and Function 137

et al., 2001; Nimchinsky et al., 2002; Oertner et al., 2002; meability (Verdoorn et al., 1991; Burnashev et al., 1992;
Sabatini et al., 2002; Yasuda et al., 2003; Matsuzaki et al., 2004; Vissavajjhala et al., 1996; Wenthold et al., 1996). mGluRs are
Nimchinsky et al., 2004). located predominantly on the periphery of the PSD (Lujan et
Spines also contain numerous organelles, including al., 1996). These receptors are coupled directly to the SER via
smooth endoplasmic reticulum (SER). SER is found in about a molecule called Homer, which may be coupled to mecha-
half of the spines in CA1 but is present in most of the largest, nisms for releasing calcium from the SER (Xiao et al., 2000).
morphologically complex spines (Spacek and Harris, 1997; Individual excitatory synapses on CA1 neurons vary con-
Cooney et. al. 2002). SER is occasionally associated with a siderably in their expression of AMPA and NMDA receptors,
spine apparatus, which consists of stacks of SER associated even within particular dendritic domains. Indeed, silent
with other electron-dense material including polyribosomes. synapses, which are thought to contain NMDA receptors
CA1 spines also contain free polyribosomes and mRNA, lead- but few or no functional AMPA receptors, were rst discov-
ing to the hypothesis that proteins can be synthesized on ered and most extensively studied at SC synapses on CA1
demand in individual dendritic spines (Steward et al., 1996). neurons (Isaac et al., 1995; Liao et al., 1995; Isaac, 2003).
Endosomal organelles and coated vesicles are found in about Immunocytochemical analysis suggests that whereas the
one-third of CA1 spines (Cooney et al., 2002). Interestingly, number of NMDA receptors is relatively invariant, a tremen-
mitochondria are absent from most spines (Sorra and Harris, dous range exists in the number of AMPA receptors at indi-
2000). The presence of a large number of molecules and vidual synapses (Nusser et al., 1998; Racca et al., 2000).
organelles in spines, together with the separation that the Quantitative immunogold data suggest a correlation between
spine neck provides from the dendritic shaft and other spines, synapse size and AMPA receptor number (Nusser et al., 1998;
has led to the hypothesis that spines function as isolated Takumi et al., 1999; Ganeshina et al., 2004a,b) and recent
molecular compartments (Wickens, 1988; Koch and Zador, physiological data indicate that there are correlations between
1993; Harris and Kater, 1994; Sorra and Harris, 2000) and that spine morphology and glutamate receptor distribution, with
such compartmentalization is necessary for synapse-specic mushroom-shaped spines containing the most AMPA recep-
changes in synaptic strength (Wickens, 1988; Harris and tors and thinner spines and lopodia containing only NMDA
Kater, 1994). receptors (Matsuzaki et al., 2001).
A prominent feature of almost all CA1 spines (and spines
throughout the nervous system) is a postsynaptic density 5.2.3 Excitatory and Inhibitory Synaptic Inputs
(PSD), an electron-dense thickening of the postsynaptic
membrane. The PSD is located adjacent to the presynaptic Like most neurons in the CNS, CA1 neurons receive input
bouton(s) associated with the spine. Structurally, the size and from both excitatory and inhibitory presynaptic neurons. The
shape of the PSD denes two classes of axo-spinous synapses: principle excitatory inputs arrive from the entorhinal cortex
perforated synapses, which have large PSDs with complex (EC) and CA3 pyramidal neurons. Direct inputs from layer III
shapes, and nonperforated synapses, which have smaller, disk- pyramidal neurons in the EC project to CA1 neurons via the
like PSDs (Harris et al., 1992). Functionally, the PSD is a bio- perforant path (PP), so named because the bers leaving the
chemical specialization that allows numerous molecules (e.g., angular bundle perforate the subiculum (Cajal, 1911). The PP
receptors, kinases, cytoskeletal elements) to be associated in a input from the EC to CA1 (also referred to as the temporo-
structured array at the synapse (Sheng, 2001). Perforated ammonic path) selectively innervates the distal apical den-
PSDs on CA1 pyramidal neurons have more -amino- drites in the stratum lacunosum-moleculare (Blackstad,
3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) and N- 1958). Some of the PP bers forming synapses on CA1
methyl-D-aspartate (NMDA) receptors than their nonperfo- pyramidal neurons reach their targets in the stratum lacuno-
rated counterparts, suggesting that they may constitute a sum-moleculare via the temporo-alvear pathway (Deller et al.,
population of relatively powerful synapses (Ganeshina et al., 1996a). Additional inputs from the nucleus reuniens of the
2004a,b). thalamus and the basolateral nucleus of the amygdala also
A picture is now emerging concerning the organization of innervate CA1 neurons via synapses on the distal apical den-
the glutamate receptors in the PSD found in excitatory drites (Krettek and Price, 1977; Amaral and Witter, 1995;
synapses on CA1 spines (see Chapters 6 and 7). These synapses Dolleman-Van der Weel and Witter, 1996, 1997; Kemppainen
contain three types of glutamate receptors: NMDA, AMPA, et al., 2002). Inputs from CA3 pyramidal neurons on both
and metabotropic glutamate (mGluR). NMDA receptors, sides of the brain form the Schaffer collateral/commissural
which mediate a slow synaptic current blocked in a voltage- system (SC), which form synapses on the apical dendrites in
dependent manner by Mg2 (see Chapter 6), occupy a disk- the stratum radiatum and on the basal dendrites in the stra-
like space near the center of the PSD. AMPA receptors, which tum oriens (Schaffer, 1892; Blackstad, 1956; Storm-Mathisen
mediate a fast synaptic current, are distributed more evenly and Fonnum, 1972; Hjort-Simonsen, 1973).
throughout the PSD (Lujan et al., 1996; Racca et al., 2000). The A few notable differences between PP and SC synaptic
presence of the edited form of the GluR2 subunit, which is inputs have been identied. First, the PP synapses are located
expressed abundantly in CA1 neurons under normal condi- farther from the soma. Without a mechanism for compensat-
tions, renders most of these AMPA receptors impermeable to ing for this dendritic disadvantage, these synapses would be
Ca2, whereas NMDA receptors have a considerable Ca2 per- expected to have a weaker inuence on action-potential initi-
138 The Hippocampus Book

ation in the axon than the SC synapses (see Sections 5.2.8 and synapses (Otmakhova and Lisman, 1999). Similarly, the nora-
5.2.9). A second difference is that, relative to the stratum drenergic bers from the locus coeruleus project preferentially
radiatum, a greater proportion of synapses in the stratum to the stratum lacunosum-moleculare region of CA1
lacunosum-moleculare are formed on dendritic shafts, rather (Pasquier and Reinoso-Suarez, 1978).
than spines (Megias et al., 2001). Thus, any functions nor-
mally attributed to spines (e.g., biochemical compartmental- 5.2.4 Axon Morphology and Synaptic Targets
ization) must be absent at many of these synapses (but see
Goldberg et al., 2003a). The functional signicance of the A single axon emanates from the pyramidal soma of CA1
shaft synapses in the stratum lacunosum-moleculare is not pyramidal neurons and projects through the stratum oriens
known, but approximately equal numbers of these synapses and into the alveus. CA1 axons branch extensively, forming
are formed by the perforant path and thalamic nucleus collaterals with several targets, both within and beyond the
reuniens projections (Wouterlood et al., 1990). Another pos- hippocampus. In the CA1 subeld, CA1 axons form a limited
sible difference between PP and SC synapses is that there arbor restricted primarily to the stratum oriens; very few col-
appears to be more NMDA receptor activation at stratum laterals enter the stratum radiatum (Amaral et al., 1991).
lacunosum-moleculare synapses (Otmakhova and Lisman, Anatomical and physiological analyses indicate that CA1 neu-
1999), although immunogold EM studies do not reveal rons show a remarkably low connection probability of about
increased NMDA receptor density in the stratum lacunosum- 1% (Knowles and Schwartzkroin, 1981; Deuchars and
moleculare (Nicholson et al., 2006). Finally, PP synapses are Thomson, 1996). Thus, unlike CA3 pyramidal neurons, CA1
inhibited by dopamine, serotonin, and noradrenaline (norep- cells do not make many connections among themselves,
inephrine) to a much greater degree than SC synapses except in the developing hippocampus (Tamamaki et al.,
(Otmakhova and Lisman, 2000). 1987; Amaral et al., 1991; Aniksztejn et al., 2001). By contrast,
Numerous inhibitory neurons also target CA1 pyramidal CA1-interneuron connectivity is much higher, and the
neurons. Some of these interneurons target the soma and strength of excitatory postsynaptic potentials (EPSPs) on
axon, and others target the dendrites. This selective targeting interneurons is powerful (Gulys et al., 1993a; Ali et al., 1998;
suggests that the PP and SC synapses are under differential Csicsvari et al., 1998; Marshall et al., 2002). Because the CA1
control of different populations of interneurons. For example, axon does not enter the stratum radiatum, local synaptic con-
the oriens-alveus/lacunosum-moleculare (O-LM) interneu- nections onto other CA1 neurons occur on basal dendrites.
rons project a single axon to the stratum lacunosum-molecu- Interneurons outside the stratum oriens are not signicantly
lare region of CA1, where the axon collateralizes extensively contacted by CA1 axons.
and forms synapses on the same dendritic branches as the PP The most signicant intrahippocampal connection of CA1
synapses from entorhinal cortex. (For more details on neurons is to pyramidal neurons in the subiculum (Tamamaki
interneuron targeting, see Section 5.9 and Chapter 8.) Despite et al., 1987; Tamamaki and Nojyo, 1990; Amaral et al., 1991).
the differential targeting of the various interneurons in CA1, a This connection is likely to be especially important because
common feature is that they all use -aminobutyric acid the subiculum forms a powerful output of the hippocampus
(GABA) as the principal neurotransmitter. GABA activates (see Section 5.4). CA1 axons collateralize extensively in the
GABAA (ionotropic Cl
channels) and GABAB receptors (G subiculum but form a topographical projection. CA1 neurons
protein receptors coupled to K channel activation). GABAA closest to the subiculum contact their nearest neighbors in the
receptor function, however, is inhomogeneous, with more proximal subiculum, whereas CA1 neurons farther from the
rapid kinetics associated with proximal GABAA receptors and subiculum project to the most distal regions of the subiculum
slower kinetics associated with more distal GABAA receptors (see Chapter 3). Collaterals from individual CA1 axons form a
(Pearce, 1993). column extending the full height of the subiculum (500 m in
CA1 pyramidal neurons also receive neuromodulatory the rat) occupying about one-third the width of the subicu-
inputs from a number of subcortical nuclei. A large input lum (250300 m) and extending about 2 mm along the
arriving from the septum contains cholinergic afferents septo-temporal length of the hippocampus (Tamamaki et al.,
(Swanson et al., 1987). Projections from the locus coeruleus 1987; Tamamaki and Nojyo, 1990).
contain noradrenergic inputs; the raphe nuclei projections The extrahippocampal projections of CA1 neurons via the
contain serotonergic inputs; and the ventral tegmental area mbria depend on their position along the septo-temporal
sends dopaminergic afferents to the hippocampus (Storm- axis (Fig. 51A). Septal CA1 neurons project to retrosplenial
Mathisen, 1977). CA1 pyramidal neurons express numerous and perirhinal cortex as well as to the lateral septal nucleus and
receptor subtypes for each of these neuromodulators, but the diagonal band of Broca. Midsepto-temporal CA1 neurons
their distribution is not always uniform throughout various project to the taenia tecta and medial frontal cortex. Temporal
subcellular domains. For example, dopamine receptors are CA1 neurons project to the taenia tecta, medial frontal cortex,
localized preferentially in the stratum lacunosum-moleculare anterior olfactory nucleus, olfactory bulb, nucleus accumbens,
(Swanson et al., 1987; Goldsmith and Joyce, 1994), suggesting basal nucleus of the amygdala, and anterior and dorsomedial
that neuromodulation via this pathway is selectively posi- hypothalamus (Jay et al., 1989; van Groen and Wyss, 1990;
tioned to inuence PP synapses to a greater extent than SC Amaral and Witter, 1995). Importantly, however, not all CA1
 Hippocampal Neurons: Structure and Function 139

neurons project to each of these regions. For example, only a contributes to the fast AHP, whereas a slow, Ca2-activated K
subset of CA1 pyramidal neurons, along with a population of conductance contributes to the slow AHP (IAHP, mediated by
giant, nonpyramidal principal neurons, project to the olfac- SK channels). A K conductance reduced by muscarinic
tory bulb (van Groen and Wyss, 1990; Gulys et al., 1998). receptor activation (IM) and IC contribute to the medium AHP
No studies have explicitly quantied the total axonal length (Storm, 1990) (Fig. 52B). The ADP is mediated in part by a
or number of synapses formed by CA1 axon collaterals. More Ca2 tail current mediated by R-type Ca2 channels (Metz et
extensive data are available for the CA3 axon, which is dis- al., 2005) and may also have a contribution from persistent
cussed in Section 5.3.4. Light microscopic studies indicate, Na current (Yue et al., 2005).
however, that the CA1 axon, which has a diameter of less than Action potentials in CA1 pyramidal neurons typically have
1 m, has numerous en passant and terminal synaptic spe- a half-width of about 1 ms (Staff et al., 2000). In mature
cializations along its length (Tamamaki and Nojyo, 1990). The rats, repolarization is slowed by application of either
total number of synaptic connections is likely on the order of tetraethylammonium (TEA) or 4-aminopyridine (4-AP) at
several thousand. concentrations that block the delayed rectier K current, the
inactivating A-type and D-type K currents, and fast voltage-
5.2.5 Resting Potential and Action and Ca2-gated K currents (Storm, 1987; Golding et al.,
Potential Firing Properties 1999). Such results suggest that a variety of voltage-gated and
Ca2-sensitive K currents contribute to action-potential
CA1 pyramidal neurons have resting potentials, recorded in repolarization as well as the various phases of the AHP
slice preparations, in the range of
60 to
70 mV. Similar (Storm, 1990) (Fig. 52B).
values have been recorded using either sharp microelectrodes Longer depolarizing current injections via somatic record-
or patch pipettes (e.g., Storm, 1987; Spruston and Johnston, ing electrodes typically elicit continuous action potential r-
1992; Staff et al., 2000). At birth, resting potentials are 5 to 10 ing at a rate that is roughly proportional to the amount of
mV depolarized from this value but gradually hyperpolarize current injected. These spike trains usually exhibit spike-
to their adult value by 2 to 3 weeks of age (Spiegelman et al., frequency accommodation: a high frequency of action poten-
1992). In response to depolarizing current injections, action tials at the beginning of a step current injection followed by a
potentials typically have a threshold in the range of
40 to gradual reduction in spike frequency later in the current injec-

50 mV (Spruston and Johnston, 1992; Staff et al., 2000; but tion (Madison and Nicoll, 1984) (Fig. 52C1). Spike-fre-
see Fricker et al., 1999). Thus, CA1 neurons must be depolar- quency accommodation is caused by gradual activation of K
ized by about 20 mV before action potentials are triggered conductances such as IM and IAHP (Lancaster and Adams,
in vitro. The resulting action potentials have amplitudes of 1986; Lancaster and Nicoll, 1987). These K currents increase
about 100 mV. cumulatively during the action potential train because they do
The data discussed above are from recordings made in hip- not inactivate and also deactivate slowly, resulting in a larger
pocampal slices. Such measurements provide good estimates AHP and longer times to repolarize back to threshold for
of the intrinsic properties of CA1 neurons, but the values may the next spike. Trains of action potentials are followed by an
be slightly different in an active network. In vivo the ongoing after-hyperpolarization (AHP) lasting more than a second
synaptic activity causes action potentials at a rate of 1 to 10 (Fig. 52C2); and the amplitude of each of these AHPs
Hz, and the average membrane potential between spikes is increases with the number of action potentials in the preced-
sometimes as much as 10 mV depolarized from the in vitro ing train (Hotson and Prince, 1980; Madison and Nicoll,
measurements (Henze and Buzski, 2001). During locomo- 1984) owing to increased Ca2 entry and activation of
tion through a place eld, action potential ring transiently SK-type Ca2-activated K channels (Marrion and Tavalin,
increases to rates of around 8 Hz (theta frequency), with occa- 1998; Bowden et al., 2001).
sional bursts of high-frequency action potential ring ( 100 K currents that are activated below threshold for action
Hz) (Frank et. al. 2001) (see Chapter 11). potentials can also affect spike-ring patterns. The best exam-
Action potential ring patterns in vivo are determined in ple of this is the delay to ring of the rst action potential
part by the timing of synaptic inputs and in part by the intrin- observed with current injections just above threshold. K
sic ring properties of CA1 neurons. The intrinsic ring prop- conductances, such as A-type and D-type K currents, can be
erties of CA1 neurons have been studied extensively in the activated by subthreshold depolarizations, thus keeping the
slice preparation, in which it is relatively easy to obtain record- membrane potential from reaching the action potential
ings and to isolate the effects of intrinsic properties on spike threshold. As these conductances inactivate, however, their
ring due to the low rate of spontaneous background synap- hyperpolarizing inuence is removed, allowing the membrane
tic input. Under these conditions, brief depolarizing current to reach threshold. The delay to rst spike is partly deter-
injections typically elicit a single action potential followed by mined, therefore, by the activation and inactivation rates of
four after-potentials (Storm, 1987, 1990): (1) a fast after- subthreshold K currents (Storm, 1990).
hyperpolarization (AHP); (2) an after-depolarization (ADP); A common feature of action-potential ring patterns in
(3) a medium AHP; and (4) a slow AHP (Fig. 52A). A fast K vivo is bursting (Kandel and Spencer, 1961; Ranck, 1973; Fox
conductance gated by voltage and Ca2 (named IC or BK) and Ranck, 1975; Suzuki and Smith, 1985; Frank et al., 2001),
140 The Hippocampus Book

A B

C1 C2

Figure 52. CA1 spike frequency adaptation and slow after- delayed rectier K current (IK); M current (IM); fast Ca2-depend-
hyperpolarization (AHP). A. Four after-potentials following single ent K current (IC); transient Ca2-dependent K current (ICT);
spikes (S) in CA1 pyramidal neurons: (1) fast AHP (); (2) after- slow Ca2-dependent K current (IAHP); voltage-dependent Cl

depolarization (ADP) (); (3) medium AHP (); and (4) slow AHP current (ICl(V)); hyperpolarization-activated mixed cation current
(). The lower trace is on a slower time scale. Bar  5 mV, 50 ms (IQ  IH). (Source: Adapted from Storm, 1990.) C. Spike-frequency
(top) or 5 mV, 500 ms (bottom). Spikes are truncated. (Source: adaptation and the slow AHP in CA1 pyramidal neurons. C1 shows
Adapted from Storm, 1987.) B. Overview of a variety of voltage- a train of adapting action potentials in response to a step current
and Ca2-dependent currents in hippocampal pyramidal neurons: injection. Bar  40 mV and 2 nA vertical and 40 ms horizontal.
transient Na current (INaT); persistent Na current (INaP); T-type C2 shows the slow AHP corresponding to 1, 4, 5, and 7 action
Ca2 current (ICaT); N-type Ca2 current (ICaN); L-type Ca2 cur- potentials. Bar  5 mV, 1 second. (Source: Adapted from Madison
rent (ICaL); fast transient K current (IA); delay K current (ID); and Nicoll, 1984.

dened broadly as a brief period of high-frequency spiking increase the action potential ring rate. The intrinsic proper-
( 100 Hz) followed by a longer period of inactivity. The rel- ties of CA1 neurons may, however, provide an additional con-
ative contributions of synaptic drive and intrinsic membrane tribution to bursting in vivo.
properties to the burst ring of CA1 neurons in vivo are In vitro studies have identied two types of intrinsic burst-
unclear. Although CA1 cells exhibit some intrinsic bursting ing in CA1 neurons: Low-threshold bursting occurs in
under normal in vitro conditions (Wong and Prince, 1981; response to somatic current injections just above action
Masukawa et al., 1982), it is modest compared to other neu- potential threshold; they are generated when the ADP that fol-
rons, such as pyramidal neurons of subiculum, which burst lows an action potential is large enough to trigger additional
much more robustly (see Section 5.4.4). In the absence of a action potentials (Jensen et al., 1994; Metz et al., 2005; Yue et
strong intrinsic burst mechanism, bursting may occur as a al., 2005). High-threshold bursting occurs in response to
consequence of large synaptic inputs, which could transiently strong dendritic current injection or synaptic activation and is
 Hippocampal Neurons: Structure and Function 141

caused by the generation of Ca2 spikes in CA1 dendrites and inhibition of the AHP (Andrade and Nicoll, 1987; Ropert,
(Wong and Prince, 1978; Golding et al., 1999; Magee and 1988). Numerous other modulatory effects have been
Carruth, 1999) (see Section 5.2.11). reported in CA1 neurons, indicating that the resting and
A variety of factors inuence the spike-ring mode of CA1 active properties of these neurons in vivo are likely to depend
pyramidal neurons. Increasing extracellular K, decreasing on behavioral states.
extracellular Ca2, and decreasing osmolarity all enhance
bursting through different mechanisms (Jensen et al., 1994; 5.2.6 Resting Membrane Properties
Azouz et al., 1997; Su et al., 2002). Bursting is also likely to be
a target of modulation by neurotransmitters. For example, One key to understanding neurons at the cellular level is to be
cholinergic activation facilitates the induction of plateau able to predict action potential output in response to a given
potentials, which may be associated with some kinds of burst spatiotemporal pattern of synaptic inputs. In addition to the
ring in vivo (Fraser and MacVicar, 1996). Because so many properties of the synapses themselves, a neurons response to
factors inuence bursting, it is important to determine the a synaptic input depends on three things: its dendritic geom-
mechanism of bursting under conditions mimicking in vivo etry and the location of the synapse(s), its passive membrane
states as closely as possible. Intrinsic burst-ring mechanisms properties, and its active membrane properties. Passive mem-
have also been implicated as targets of epileptogenesis in the brane properties are those that do not depend on membrane
hippocampus (Wong et al., 1986), so understanding the patho- potential (i.e., they are linear). Active membrane properties,
physiology of bursting may lead to new treatments of epilepsy. by contrast, do depend on membrane potential (e.g., the
A number of aspects of pyramidal cell physiology are also voltage-gated Na and K channels that mediate the action
under modulatory control. Activation of muscarinic receptors potential, which are highly nonlinear). These properties are
causes depolarization and a reduction in spike-frequency very much interdependent, but physiologists attempt to treat
accommodation due to inhibition of IM, IAHP, and voltage- them separately, as we do in our discussion here. One justi-
insensitive K conductances (Benardo and Prince, 1982a,b,c; cation for doing this is that detailed neuronal simulations
Madison et al., 1987; Benson et al., 1988). Norepinephrine require a description of the passive membrane properties,
also reduces spike-frequency accommodation due to inhibi- which constitute a foundation on which active membrane
tion of IAHP (Madison and Nicoll, 1982; Pedarzani and Storm, properties are superimposed.
1996). Dopamine causes hyperpolarization and elevated CA1 pyramidal neurons, like most neurons in the brain,
action-potential threshold owing in part to activation of have long and extensively branching dendrites. A quantitative
Ca2-sensitive K conductances; it also raises the action- understanding of how membrane potential changes (such as
potential threshold in CA1 neurons (Benardo and Prince, EPSPs) spread through these structures requires a theoretical
1982d,e; Stanzione et al., 1984). Serotonin has biphasic effects treatment of dendrites combined with experimental measure-
on CA1 neurons, initially causing hyperpolarization owing to ments. Wilfrid Rall provided the theoretical framework with
activation of a K current, but later causing depolarization his seminal work on the cable theory (Box 51). Hippocam-

Box 51
Cable Theory

During the 1950s and 60s, Wilfrid Rall developed a theory for treating the ow of current in
passive dendrites. The seminal articles comprising this theory have been compiled and repub-
lished in book form (Segev et al., 1995). Because the theory was based on a theory similar to
that for trans-Atlantic telegraph cables, it is referred to as the cable theory. With this mathe-
matical treatment, neurons are considered to be long, leaky cables immersed in a conductive
medium (cerebrospinal uid). A synaptic current propagating along a passive dendrite is inu-
enced by three factors: membrane resistance, membrane capacitance, and axial (or intracellu-
lar) resistance. These properties depend in part on geometry (narrower dendrites have higher
membrane resistance, smaller capacitance, and higher axial resistance) and in part on the
composition of the membrane and cytoplasm. In his theory, Rall dened three geometry-
independent variables: membrane resistivity (Rm), specic membrane capacitance (Cm), and
intracellular resistivity (Ri). He also used two useful measures. The rst is called the space
constant (), which is equal to the length, in an innitely long cable, over which the mem-
brane potential decays to 1/e of its original value. The second is the electrotonic length
(L  l/), which is the ratio of the physical length to the space constant, a measure of the
total electrical length of a nite cable.
(Continued)
142 The Hippocampus Book

Box 51
Cable Theory (Continued)

Rall also showed that a class of branching dendritic trees with certain characteristics could be
collapsed to an equivalent cylinder with a characteristic L. He also described methods for esti-
mating L experimentally. Because most neurons deviate from the assumptions necessary to col-
lapse them to a single cable, however, the lasting impact of Ralls cable theory is that it provides
the foundation for modern computational analysis of neurons. Properties such as dendritic
structure and Rm (and possibly Ri, but less so Cm) differ across different types of neurons and
inuence their integrative properties, which can be predicted using computer models.

Box Fig. 51. A. Current owing along a dendrite may ow longitudinally (e.g., toward the
soma), or it may leak out across the dendritic membrane. B. Voltage attenuates with distance
along a dendritic cable in a way that depends on Rm, Ri, and dendritic geometry. Curve a
shows the decay of voltage in a semi-innite cylinder. Curves bd show attenuation in nite,
open-ended cylinders of decreasing length. Curves eg show attenuation in nite, closed-end
cylinders of decreasing length. All curves are for dendritic cylinders with Rm, Ri and diameter
yielding a space constant of 1. (Source: Adapted from Rall, 1959.) C. Dendritic trees with
particular geometries can be collapsed to an equivalent cylinder representation. Two criteria
must be met: At each branch point, the diameter of the parent dendrite, raised to the 3/2
power, must equal the sum of the daughter-dendrite diameters, each raised to the 3/2 power.
In addition, each dendritic branch must terminate at the same electrotonic length.
(Source: Adapted from Rall, 1964.)
 Hippocampal Neurons: Structure and Function 143

pal neurons were among the rst to have their cable proper- Another complication when determining Rm is the pres-
ties determined experimentally. Initially, microelectrode ence of voltage-gated channels that are open at the resting
recordings were used to estimate the electrotonic length (L) of membrane potential. Investigators typically seek to minimize
CA1 pyramidal neurons (Brown et al., 1981; Johnston, 1981). the contribution from activation or deactivation of these
These estimates proved to be of limited value, however, channels by making the evoked voltage change small (e.g., 15
because CA1 dendrites clearly violate the conditions required mV). In CA1 pyramidal neurons, however, this problem can-
to collapse them to a single cylinder (Mainen et al., 1996). For not be avoided entirely because of the presence of a powerful
example, basal, apical, and oblique dendritic branches termi- hyperpolarization-activated, nonspecic-cation current (Ih)
nate at different electrotonic distances (Turner, 1984; Pyapali that is active at rest (Halliwell and Adams, 1982; Spruston and
et al., 1998). Furthermore, there are indications that the mem- Johnston, 1992). This conductance introduces a noticeable
brane properties that contribute to L are not uniform over the sag in the voltage response. Hyperpolarizing voltage
surface of the dendritic tree (Golding et al., 2005). For these responses activate Ih, resulting in a gradual return of Vm
reasons, a more accurate picture of the passive behavior of the toward the resting potential during the current injection
CA1 dendritic tree is best achieved by computer simulation (hyperpolarizing sag). Depolarizing voltage responses deac-
methods, which facilitate the characterization based on the tivate Ih, similarly resulting in a gradual return of Vm toward
measured geometry and membrane properties of CA1 pyram- the resting potential (depolarizing sag). The presence of Ih-
idal neurons. mediated sag makes it difficult or impossible to determine m.
The most reliable measures of the parameters affecting To solve this problem, m is usually measured with Ih blocked
cable propertiesRm, Cm, Ricome from patch-clamp by bath application of 2 to 5 mM CsCl or 50 to 100 M
recordings made in hippocampal slices. Earlier estimates of ZD7288. Under these conditions, CA1 pyramidal neurons
these parameters using microelectrodes were affected by the from adult rats and guinea pigs have an RN of about 120 to 150
leak that is introduced by microelectrode penetration M and m of about 35 to 40 ms (Spruston and Johnston,
(Spruston and Johnston, 1992; Staley et al., 1992). Patch- 1992; Staff et al., 2000; Golding et al., 2005) (Fig. 53).
clamp recordings are not without their own problems, namely Assuming a Cm of 1 F/cm2, this corresponds to an Rm of
the possible complications introduced by dialysis of the cyto- about 40,000 cm2 (to understand the units of Rm, it may
plasm. There is good evidence, however, that this problem help to understand that the units of its inverse are conduc-
does not affect passive membrane properties substantially tance per unit area). It is important to realize, however, that
(Spruston and Johnston, 1992; Staff et al., 2000). More serious these values are strongly affected by blocking Ih. In the absence
problems are posed by the voltage-gated channels that are of its blockers, RN is 38% to 54% lower, and the membrane
active near the resting potential, as well as differences between potential changes decay accordingly faster (Spruston and
in vitro and in vivo conditions (see below). Johnston, 1992; Staff et al., 2000; Golding et al., 2005).
Passive membrane properties are typically estimated by In guinea pig CA1 neurons, hyperpolarizations or depolar-
measuring the response of a neuron to a step current injec- izations of about 5 mV from rest result in about a 20%
tion. Two features dene the response: the steady-state ampli- decrease and increase in RN, respectively (Hotson et al., 1979;
tude and the time course of the voltage change. Theoretically, Spruston and Johnston, 1992). Even after block of Ih, estimates
this time course can be described by the sum of several expo- of Rm and RN are sensitive to small changes in the resting
nentials, with the slowest component having a time constant membrane potential, but in the opposite direction of the volt-
equal to the membrane time constant (m), given by the prod- age dependence caused by Ih. These ndings indicate that the
uct RmCm. The steady-state voltage response (V) is deter- experimental estimates of these parameters are inuenced by
mined by the neurons input resistance (RN) in a way that multiple voltage-gated conductances; therefore, these param-
depends on Rm and geometry (large neurons with many den- eters may never be truly passive. Thus, it may be more appro-
drites and low Rm have the lowest RN). Input resistance can be priate to refer to the resting membrane properties of CA1
measured directly using Ohms law (RN  V/I), but deter- neurons, which should be regarded as approximations of the
mination of Rm (from m) requires a value of Cm. Because Cm theoretical passive membrane properties.
is largely dependent on the lipid composition of the mem- Despite the voltage dependence of these parameters, esti-
brane, its value has long been assumed to be nearly constant mates of their values near rest are essential because they pro-
at 1 F/cm2, an assumption that was validated by experimen- vide important parameters for computer models of neurons.
tal measurements of capacitance in cells with simple geome- One of the purposes of constructing such models, using
try (Gentet et al., 2000). Estimates of Cm from neurons with methods originally developed by Rall (Segev et al., 1995) and
more complex geometry are often close to this value; in currently incorporated into user-friendly programs such as
instances where substantial deviations are reported, however, NEURON (http://www.yale.edu) and GENESIS (http://www.
it is difficult to know if these reect real variability in Cm or genesis-sim.org/GENESIS/), is to estimate how synaptic
the extreme difficulty of accurately reconstructing the surface potentials attenuate between the dendrites and the soma. Such
area of neurons with branching dendrites studded with estimates, however, are highly dependent on the intracellular
spines. Most estimates of Rm have therefore been derived by resistivity, Ri. This parameter is best estimated using computer
measuring m and assuming a value for Cm of 1 F/cm2. models of carefully reconstructed neurons along with a meas-
144 The Hippocampus Book

A soma and distal dendrites is limited. One of the interesting

consequences of this is that it makes it extremely difficult to
study synaptic or voltage-gated conductances in dendrites
using electrodes placed at the soma. Space-clamp errors are
10 mV likely to be enormous for distal conductances and substantial
100 ms even for relatively proximal conductances (Spruston et al.,
1993, 1994; Major et al., 1994; Mainen et al., 1996). These
errors are greatest for transient conductance changes; and
even when measures of steady-state voltage control (such as
15 reversal potential measurements) indicate reasonable voltage
B
V(mV) control, errors can be very large for the transient case. The
10 presence of voltage-gated conductances in dendrites, though
presumably advantageous for the normal function of the neu-
5
I(pA) ron, makes it still more difficult to control dendritic mem-

200
100
brane potential using somatic electrodes. Thus, other methods
100 200 are needed to complement somatic recordings to determine

5
the functional properties of dendrites. Fortunately, several
Control such methods are now available (see Sections 5.2.11 through

10 CsCl 5.2.13).

15 5.2.8 Attenuation of Synaptic
Potentials in CA1 Dendrites
C
3 mV
100 ms The functional signicance of the dendritic tree, with these
nonuniform membrane properties and high Ri, is perhaps
best understood by considering the extent to which synaptic
potentials attenuate as they propagate from the dendrite to the
soma. Compartmental models constrained by simultaneous
Figure 53. CA1 passive properties and sag. A. Voltage responses to
recordings have been used to predict synaptic attenuation, and
step current injections from 200 to 200 pA in 50-pA increments.
the results are striking. Distal synaptic inputs are expected to
B. Voltagecurrent plot of the steady-state responses shown in A, as
well as similar responses following the addition of 5 mM CsCl to
attenuate many times 10-fold from the most distal sites to
the bath. Solid lines are the points used for linear ts to the data. the soma (Mainen et al., 1996; Andreasen and Lambert, 1998;
Dashed lines are extrapolations of the ts. C. Effect of 5 mM CsCl Golding et al., 2005) (Fig. 54). Most of this enormous atten-
on the hyperpolarizing voltage response to a current step of 150 uation occurs between the synapse and the primary apical
pA. Note the block of the sag in the voltage response. (Source: dendrite, but direct dendritic recordings indicate an addi-
Adapted from Staff et al., 2000.) tional three- to fourfold attenuation between dendritic sites at
about 300 m and the soma. The structure of the CA1 den-
dritic tree appears to maximize synaptic attenuation (Jaffe
ure of voltage attenuation. Modeling of neurons from which and Carnevale, 1999), a point that is considered in more detail
attenuation was measured with simultaneous somatic and below by way of comparison to dentate granule cells (see
dendritic patch-clamp electrodes yielded a value for Ri of Section 5.5.5).
about 180 cm for CA1 pyramidal neurons (Golding et al., The severe synaptic attenuation predicted on the basis of in
2005). This study also revealed that the membrane resistivity vitro studies and modeling might even be an underestimate of
of CA1 neurons is nonuniform, with the dendritic membrane the attenuation that could occur in passive dendrites in vivo.
increasingly leaky at greater distances from the soma. Because dendrites are constantly receiving synaptic inputs,
Similarly, Ih, which is activated at rest and therefore increases attenuation may be enhanced by the increased conductance
voltage attenuation in dendrites, also increases with distance produced at active synapses (Destexhe and Pare, 1999). The
from the soma (Magee, 1998; Golding et al., 2005) (see magnitude of this effect, however, has not been directly deter-
Section 5.2.13). mined in the hippocampus.

5.2.7 Implications for Voltage-Clamp 5.2.9 Mechanisms of Compensation

Experiments in CA1 Neurons for Synaptic Attenuation in CA1 Dendrites

Knowledge of the structure and membrane properties of CA1 Faced with this much attenuation, it seems almost pointless
neurons allows predictions regarding the functional coupling for a CA1 neuron to have synapses on its distal dendrites. Yet
of the soma and dendrites. Detailed compartmental modeling a large number of axons (primarily the perforant path input
of CA1 neurons suggests that functional coupling between the from the entorhinal cortex) do synapse on distal apical
 Hippocampal Neurons: Structure and Function 145

A B 1995) and that it is capable of discharging CA1 cells in vivo,

even after lesions that reduce or eliminate input from Schaffer
collaterals (McNaughton et al., 1989; Brun et al., 2002).
For perforant path synapses to have a signicant impact on
1 mV
CA1 output, some mechanism must compensate for dendritic
10 ms attenuation. One possibility is that distal synapses may result
in signicant somatic depolarization through amplication by
Vsynapse voltage-gated channels in dendrites. Alhough this may occur
in a graded fashion in the dendrites or the soma (Lipowsky et
al., 1996; Cook and Johnston, 1997, 1999; Andreasen and
Lambert, 1999), amplication may also occur through the
Vdendrite generation of dendritic spikes (see Sections 5.2.11 through
5.5.13). Another possibility is that distal synapses normally do
not exert a strong inuence over action-potential initiation on
Vsoma
their own but may interact effectively with more proximal
synaptic activation to trigger action potentials (Remondes and
Figure 54. CA1 dendritic attenuation of synaptic potentials. Schuman, 2002). Such interactions may occur in two ways.
Simulations of excitatory postsynaptic potential (EPSP) attenuation First, the distal, perforant path EPSP may sum with more
from the apical tuft to the soma. A model with nonuniform Rm proximal, Schaffer collateral EPSPs to reach threshold for an
and Gh was used in the simulation. The synaptic conductance was action potential in the soma or the apical dendrite (Levy et al.,
1 nS with a rise time constant of 0.5 ms and a decay time constant
1995; Jarsky et al., 2005; Kali and Freund, 2005). Second,
of 5 ms. A. The synapse location is indicated by the black dot in
strong activation of perforant-path synapses can lead to den-
the apical dendrite, 583 m from the soma. The somatic and den-
dritic recording sites are indicated by the electrode cartoons (den- dritic spikes, which only propagate effectively to the soma
dritic recording electrode at 365 m). B. Example of simulated when facilitated by coincident activation of the Schaffer col-
EPSPs in response to activation of the distal apical synapse (black lateral synapses (Jarsky et al., 2005).
dot in A) and measured at the synapse, the dendritic recording
site, and the soma. In this example the EPSP attenuates 26-fold 5.2.10 Pyramidal Neuron Function:
between the synapse and the soma. (Source: Adapted from Golding Passive Versus Active Dendrites
et al., 2005.)
Even without much information about membrane properties,
many neuroscientists predicted that synaptic potentials would
dendrites. Even Schaffer collaterals, the excitatory axons from attenuate severely as they propagate toward the soma if den-
CA3, form many synapses on relatively distal dendritic drites were truly passive. To counter the unlikely proposition
regions. Two mechanisms are likely to overcome this problem that distal dendritic synapses are irrelevant, some investigators
of dendritic disadvantage. First, there is now evidence that proposed that dendrites might contain voltage-gated conduc-
the average synaptic conductance in the stratum radiatum tances; others, however, took the view that voltage-gated
increases as a function of distance from the soma (Magee and channels were a unique property of the axon (reviewed in
Cook, 2000). The synapses at distal sites may have larger con- Llins, 1988; Adams, 1992). Although the debate over passive
ductances as a result of an increased abundance of large, per- versus active dendrites got started with studies of spinal
forated synapses and a higher density of AMPA-type motoneurons, the hippocampus soon took center stage, as its
glutamate receptors (Andrasfalvy and Magee, 2001; Smith et laminar structure and densely packed, uniformly oriented
al., 2003; Nicholson et al., 2006). However, both computer dendrites facilitated studies of dendritic function using extra-
simulations and electron microscopy indicate that this princi- cellular electrodes. Field potential recordings in the CA1
pal of synaptic scaling to compensate for dendritic distance region of the anesthetized rabbit indicated that spikes could
does not extend into the most distal apical dendrites, where be generated in the proximal dendrites of CA1 neurons and
the perforant-path bers form synapses (Kali and Freund, subsequently propagated actively away from this site (Cragg
2005; Nicholson et al., 2006). and Hamlyn, 1955; Andersen, 1960; Fujita and Sakata, 1962;
The perforant-path input to CA1 pyramidal neurons thus Andersen and Lomo, 1966; Andersen et al., 1966). These con-
appears to be disadvantaged because EPSPs from this input clusions were later conrmed and elaborated on by more
attenuate so much on their way to the soma, and they appar- detailed analyses using eld potentials, both in vitro and in
ently lack a conductance scaling mechanism to compensate. vivo (Turner et al., 1989, 1991; Herreras, 1990). Another
Exacerbating the problem is feedforward inhibition, which is important study was performed by Spencer and Kandel, who
so powerful, the excitatory nature of the perforant path inputs made intracellular recordings from CA1 neurons in the anes-
to CA1 was once in question (Buzski et al., 1995; Empson thetized cat (Spencer and Kandel, 1961). They observed fast,
and Heinemann, 1995a,b; Soltesz, 1995; Andreasen and spike-like events that were smaller than action potentials and
Lambert, 1998). It is now clear, however, that this input is exci- often preceded full-sized spikes. Based on the absence of such
tatory (Doller and Weight, 1982; Yeckel and Berger, 1990, events following antidromic stimulation of the axon, Spencer
146 The Hippocampus Book

and Kandel concluded that these fast prepotentials origi- neurons (Stuart et al., 1997). The reasons for this are not clear
nated in dendrites. This view has received some support but are likely related to the density and/or properties of Na
(Schwartzkroin, 1977; Wong and Stewart, 1992; Turner et al., channels in the axon (Colbert and Pan, 2002).
1993) but has remained controversial, as action potentials Following their initiation in the axon, action potentials
from CA1 neurons coupled by gap junctions have also been invade the dendritic tree of CA1 neurons (Spruston et al.,
implicated (MacVicar and Dudek, 1981; Schmitz et al., 2001). 1995; Golding et al., 2001) (Fig. 55A). Application of TTX to
Nevertheless, these studies collectively indicated that the den- the dendrites dramatically reduces the amplitude of these
drites of CA1 pyramidal neurons are capable of generating back-propagating action potentials, indicating that Na chan-
active responses owing to the presence of voltage-gated nels actively enhance action potential propagation in CA1
channels. dendrites (Spruston et al., 1995; Magee and Johnston, 1997;
The view that CA1 dendrites are active received consider- Golding et al., 2002). In keeping with this, Na channels have
able support from studies using microelectrodes to record been recorded directly in cell-attached patches in CA1 den-
from CA1 dendrites in slices. These recordings indicated that drites (Magee and Johnston, 1995a,b; Colbert et al., 1997; Jung
action potentials could be observed in dendritic recordings, et al., 1997; Mickus et al., 1999). The amplitude of the back-
even when dendrites were physically or pharmacologically propagating action potential, however, decreases with distance
isolated from the soma (Wong et al., 1979; Benardo et al., from the soma, indicating that back-propagation is not fully
1982; Poolos and Kocsis, 1990; Turner et al., 1991, 1993; Wong regenerative. At 300 m, the back-propagating action poten-
and Stewart, 1992; Colling and Wheal, 1994; Andreasen and tial amplitude is about half of its somatic amplitude (Golding
Lambert, 1995a). Calcium imaging studies also indicated that et al., 2001). Attenuation of back-propagating action poten-
dendrites could generate active responses capable of activating tials in the distal half of the apical dendrites is variable. Within
voltage-gated Ca2 channels and mediating signicant Ca2 a given cell, back-propagation is controlled by the membrane
entry into dendrites (Regehr et al., 1989; Jaffe et al., 1992; potential and the availability of Na and K channels
Regehr and Tank, 1992; Yuste and Denk, 1995; Yuste et al., (Bernard and Johnston, 2003). Between cells, two populations
1999). By the late 1980s and early 1990s, such a wealth of data of CA1 neurons have been identied (Golding et al., 2001). In
had been accumulated that it was undeniable that CA1 den- about half of the CA1 neurons, attenuation of the backpropa-
drites were active. Attention then shifted to answering more gating action potential from 300 to 400 m continues at
detailed questions about dendritic excitability, including what about the same rate as in more proximal dendrites. In the
types of channels are present, what properties they possess, other half of the CA1 neurons, attenuation in this distal region
how they are distributed, and how they contribute to the inte- is more pronounced. The relatively strong back-propagation
grative properties of CA1 neurons. may be promoted by somatic depolarization, as it is not
observed with antidromic stimulation (Bernard and
5.2.11 Dendritic Excitability Johnston, 2003). The dichotomy of action potential back-
and Voltage-Gated Channels propagation observed during somatic current injection sug-
in CA1 Neurons gests that there may be considerable cell-to-cell variability in
the densities of ion channels that control action potential
Studies of dendritic excitability in the hippocampus and else- back-propagation (Golding et al., 2001).
where have been greatly facilitated by the development of Although patch-clamp recordings from the basal dendrites
methods to routinely record from dendrites in brain slices of CA1 neurons have not yet been obtained, models based on
using patch pipettes (Stuart et al., 1993). The ability to record Na and K channel densities in the apical dendrites predict
simultaneously from the soma and the primary apical dendrite that action potential propagation into the shorter, basal den-
(up to about 400 m from the soma) has been particularly drites is much more reliable (Golding et al., 2001). This nd-
useful. One of the rst applications of this method was to ing is in keeping with imaging studies of the basal dendrites in
address the question of where action potentials are initiated. neocortical pyramidal cells, which indicate reliable back-
The classic view was that action potentials are initiated in the propagation in the basal dendrites (Schiller et al., 1995).
axon, but several of the studies discussed above had led to the Similarly, direct recordings from oblique branches of the api-
hypothesis that action potentials might actually be initiated in cal dendrites have not been obtained, but Ca2 imaging exper-
dendrites. Simultaneous somatic and dendritic recordings in iments suggest that action potentials actively invade these
CA1 neurons have indicated that fast, all-or-none action branches (Frick et al., 2003).
potentials begin near the soma (Spruston et al., 1995; Hoffman Attenuation of the back-propagating action potential
et al., 1997; Golding et al., 2001). Based on the sensitivity of the amplitude is even more severe during repetitive spiking (Fig.
action potential threshold to local application to tetrodotoxin 55A). Even at modest ring frequencies such as 20 Hz, the
(TTX), the site of action potential initiation has been further action potential amplitude, measured about 300 m from the
narrowed to a region near the rst node of Ranvier in the axon soma, attenuates to less than half of its amplitude at lower fre-
(Colbert et al., 1996). Similar results have been obtained in quencies, equivalent to less than one-fourth of the somatic
several other types of neurons, suggesting that the axon may action potential amplitude (Andreasen and Lambert, 1995a;
have a low threshold for action potential initiation in most Callaway and Ross, 1995; Spruston et al., 1995; Golding et al.,
 Hippocampal Neurons: Structure and Function 147

A B C

Figure 55. CA1 dendritic excitability. Left. Camera lucida drawing the s.l.m. On one trial the dendritic spike is associated with a
of a CA1 pyramidal neuron. Three types of dendritic excitability are somatic action potential, whereas on the other trial only an EPSP is
indicated. (Source: Adapted from Golding et al., 1999.) A. Trains of recorded at the soma (i.e., propagation failure). (Source: Adapted
back-propagating action potentials are initiated in the soma in from Golding and Spruston, 1998.) C. Dendritic Ca2 spikes were
response to step current injections and recorded in the dendrites elicited by a step current injection via a dendritic patch-clamp elec-
with a second patch-clamp electrode. The somatic and mid-proxi- trode 160 m from the soma. The second back-propagating action
mal recordings were obtained simultaneously from the same neu- potential shown is followed by a large, broad dendritic Ca2 spike.
ron. The more distal dendritic recording was obtained from a Simultaneous somatic recording shows that the dendritic Ca2 spike
different cell. (Source: Adapted from Golding et al., 2001.) B. caused an action potential burst in the soma. (Source: Adapted from
Dendritic Na spikes were elicited by strong synaptic stimulation in Golding et al., 1999.)

2001). This activity-dependent back-propagation is mediated conditions, be generated in the dendrites, as earlier studies
by a form of Na channel inactivation that recovers very had suggested (Spencer and Kandel, 1961; Schwartzkroin,
slowly (Colbert et al., 1997; Jung et al., 1997; Mickus et al., 1977; Schwartzkroin and Slawsky 1977; Wong et al., 1979;
1999). Patch-clamp recordings from CA1 somata and den- Benardo et al., 1982; Masukawa and Prince, 1984; Andreasen
drites have revealed that repetitive depolarizations lead to and Lambert, 1995a). Direct evidence for dendritically gener-
cumulative Na channel inactivation. Each brief depolariza- ated spikes soon followed from simultaneous somatic and
tion forces a fraction of the available Na channels into the dendritic patch-clamp recordings. When the synaptic stimu-
prolonged inactivated state. Recovery from this state back to lus intensity was increased well above the threshold for action
the closed (available) state has a time constant of about 1 sec- potential initiation in the axon, spikes were observed in apical
ond at the resting potential (Mickus et al., 1999). This kind of dendritic recordings before the action potential was recorded
inactivation, which is especially pronounced in the dendrites, in the soma (Golding and Spruston, 1998). With GABAA
results in fewer and fewer dendritic Na channels available to receptors blocked to increase the excitatory drive to dendrites,
support action potential back-propagation during repetitive two kinds of dendritic spike could be elicited by synaptic stim-
ring, resulting in progressive reduction in the amplitude of ulation: fast spikes mediated primarily by Na channels and
back-propagating action potentials. slower spikes mediated primarily by Ca2 channels (Golding
Although simultaneous somatic and dendritic patch- and Spruston, 1998; Golding et al., 1999) (Fig. 55B,C).
clamp recordings indicated that the threshold for action Intriguingly, these two types of dendritic spike do not propa-
potential initiation was lowest in the axon, these studies did gate reliably to the soma. Dendritic Na spikes were often
not address the question of whether spikes could, under some observed in dendritic recordings when no action potential
148 The Hippocampus Book

reached the soma, suggesting that these events could be Inhibition has been shown to inuence dendritic excitabil-
entirely restricted to the dendrites (Golding and Spruston, ity in two ways. First, in response to synaptic stimulation in
1998). As is the case for back-propagation, forward propaga- the slice, dendritic spikes occur much more frequently when
tion of dendritic Na spikes is sensitive to both distance and inhibition is blocked (Golding and Spruston, 1998; Golding et
membrane potential. Spikes initiated more distally are less al., 1999). This is consistent with the notion that dendritic
likely to propagate to the soma successfully, and membrane spikes are primarily observed during sharp waves, which
depolarization enhances the forward propagation of distally occur because of strong, synchronous synaptic excitation
generated spikes (Gasparini et al., 2004). Dendritic Ca2 (Kamondi et al., 1998). Second, inhibition has been shown to
spikes often trigger multiple fast action potentials in the axon limit action potential back-propagation (Buzski, 1996;
and soma but are also isolated to the dendrites in some cir- Tsubokawa and Ross, 1996). Both of these observations sug-
cumstances (Golding et al., 1999; Wei et al., 2001; Golding gest that inhibition limits dendritic excitability, but it should
et al., 2002). These ndings indicate that dendritic Ca2 spikes be noted that it could, if appropriately timed prior to an exci-
may function locally or may serve as a mechanism for tatory input, enhance dendritic excitability by removing inac-
producing bursts of action potentials in the soma and axon tivation from dendritic Na and Ca2 channels.
of CA1 neurons. Experiments combining imaging and elec- Activation of neuromodulatory receptors and second mes-
trophysiology suggest that spikes can also be generated senger systems has also been shown to inuence dendritic
locally in the basal dendrites of CA1 pyramidal neurons excitability in a variety of ways. Activation of mitogen-
(Ariav et al., 2003). activated protein (MAP) kinase, protein kinase C (PKC), or
An important question is whether the dendritic excitability protein kinase A (PKA) results in downregulation of A-type
observed in slice experiments also occurs under natural con- K current in CA1 dendrites and, accordingly, produces an
ditions in vivo. Only one study to date has addressed this increase in the amplitude of back-propagating action poten-
question directly. Microelectrode recordings from the den- tials (Hoffman and Johnston, 1998; Yuan et al., 2002). Similar
drites of CA1 neurons in anesthetized rats revealed a pattern results are also produced by activation of -adrenergic and
of dendritic excitability similar to that seen in vitro. Three muscarinic acetylcholine receptors, which activate the PKA
types of excitability were observed in the in vivo recordings and PKC systems, respectively. Activation of dopaminergic
that were putatively attributed to back-propagating action receptors produced a similar effect in a subset of CA1 neu-
potentials, dendritically generated Na spikes, and dendriti- rons, suggesting heterogeneity in the responses of CA1 neu-
cally generated Ca2 spikes (Kamondi et al., 1998). Putative rons to neuromodulators (Hoffman and Johnston, 1998).
back-propagating action potentials were identied by their Arachidonic acid produces a complementary effect, enhanc-
rapid time course and decline in amplitude as a function of ing the amplitude of back-propagating action potentials via a
distance of the recording from the soma. These events also reduction in dendritic A-type K current (Bittner and Mller,
decayed in amplitude during repetitive ring in a manner 1999; Colbert and Pan, 1999). Muscarinic receptor activation
similar to back-propagating action potentials in vitro reduces the activity dependence of action potential back-
(Kamondi et al., 1998; Golding et al., 2001). In addition to propagation, presumably by reducing prolonged inactivation
these events, which occurred both spontaneously and in of Na channels (Tsubokawa and Ross, 1997; Colbert and
response to current injection through the recording electrode, Johnston, 1998). A similar effect has also been observed fol-
putative dendritically generated spikes were observed during lowing rises in intracellular Ca2 evoked by repeated action
hippocampal population discharges (sharp waves). These potential ring (Tsubokawa et al., 2000).
events were either small in amplitude and rapid in time course Repeated synaptic activity can also modulate dendritic
(putative dendritic Na spikes) or large in amplitude and slow excitability. Frick and colleagues have shown that pairing of
in time course (putative dendritic Ca2 spikes). synaptic activity and action potential ring in a theta-burst
The dendritic excitability of CA1 pyramidal neurons is not pattern not only induces synaptic plasticity but also induces
likely to be static but probably changes depending on the long-term enhancement of dendritic excitability (Frick et al.,
behavioral state of the animal. The best direct evidence for 2004). This enhancement is caused by a leftward shift in the
this is that dendritic spikes are primarily observed in vivo dur- voltage dependence of inactivation for A-type K channels,
ing sharp waves, which occur only during awake immobility, which results in a reduction in the number of channels avail-
consummatory behaviors, and slow-wave sleep in rats able to be activated from the resting potential and causes an
(Kamondi et al., 1998) (see Chapter 11). Dendritic excitability increase in action potential back-propagation and dendritic
may also be heterogeneous within CA1, as is seen for action Ca2 entry (Frick et al., 2004).
potential invasion of the distal apical dendrites, which exhibits
a dichotomy of strong and weak back-propagation (Golding 5.2.12 Sources of Ca2 Elevation
et al., 2001). Several factors are likely to inuence dendritic in CA1 Pyramidal Neuron Dendrites
excitability in vivo: the balance and pattern of synaptic excita-
tion and inhibition, previous experience-dependent synaptic Calcium imaging experiments have provided considerable
and nonsynaptic plasticity, and the neuromodulatory state in insight into the function of CA1 dendrites. These experiments
the hippocampus. have identied three sources of Ca2 entry. First, dendritic
 Hippocampal Neurons: Structure and Function 149

depolarization can activate voltage-gated Ca2 channels. agating action potentials relieves the Mg2 block of NMDA
Second, neurotransmitters can activate ligand-gated channels receptors.
with signicant Ca2 permeability, such as the NMDA recep- Calcium release from intracellular stores has been observed
tor. Third, Ca2 can be released from intracellular stores in in CA1 dendrites in response to elevation of Ca2 via back-
dendrites. Signicant interactions have been identied among propagating action potentials and by activation of mGluRs
these mechanisms. coupled to inositol trisphosphate (IP3) receptors (Bianchi et
The large dendritic depolarizations produced by back- al., 1999; Nakamura et al., 1999, 2000; Sandler and Barbara,
propagating action potentials and dendritic spikes result in 1999). Again, these two means of activating Ca2 release are
activation of voltage-gated Ca2 channels, which produces not independent, as pairing of action potentials with EPSPs
Ca2 transients that are not uniform throughout the dendritic greatly enhances release (Nakamura et al., 1999, 2000).
tree (Regehr et al., 1989; Jaffe et al., 1992; Yuste and Denk, Although intracellular Ca2 release has as yet been observed
1995; Helmchen et al., 1996; Yuste et al., 1999; Golding et al., only in the proximal regions of the primary apical dendrite of
2001; Frick et al., 2003). Spatial gradients in Ca2 signals must CA1 neurons (Nakamura et al., 2000) and spine-restricted
be interpreted cautiously, as many factors may contribute. Ca2 transients are primarily coupled to activation of NMDA
Larger Ca2 transients may be produced by greater local depo- receptors (Yuste and Denk, 1995; Kovalchuk et al., 2000), the
larization, higher density of Ca2 channels, larger surface- presence of endoplasmic reticulum organelles in other regions
to-volume ratio (smaller dendrites), or weaker Ca2 buffering of the dendrites, including spines (Spacek and Harris, 1997) is
and extrusion (Volfovsky et al., 1999; Majewska et al., 2000; consistent with observations of spatially restricted Ca2 release
Murthy et al., 2000). Careful interpretation of the available in dendritic spines under some conditions (Emptage et al.,
data suggests, however, that depolarization gradients con- 1999).
tribute strongly to the gradients of Ca2 entry observed.
For example, back-propagating action potentials tend to 5.2.13 Distribution of Voltage-Gated
produce smaller Ca2 transients in distal dendrites than in Channels in the Dendrites of CA1 Neurons
proximal dendrites, largely due to the attenuation of the back-
propagating action potential with distance from the soma The subcellular distributions of various subtypes of voltage-
(Jaffe et al., 1992; Christie et al., 1995; Golding et al., 2001). gated ion channels have been investigated in CA1 neurons
This gradient is enhanced during repetitive ring due to the using physiological approaches, as well as antibody binding, in
activity dependence of back-propagation, which results in a some cases combined with electron microscopy. Physiological
steeper depolarization gradient along the dendrites (Golding studies fall into two categories: direct measurement of ion
et al., 2001). Neuromodulators coupled to G proteins reduce channel function using cell-attached or cell-excised patches
the dendritic calcium entry associated with back-propagating from the soma, axon, and dendrites and indirect assessment of
action potentials through membrane potential changes and ion channel distribution using whole-cell recording or cal-
a reduction in current through voltage-gated Ca2 channels cium imaging in combination with ion-channel pharmacol-
(Chen and Lambert, 1997; Sandler and Ross, 1999). ogy. Direct approaches offer a highly detailed picture of
Dendritically generated spikes produce a spatial gradient channel distribution and properties in various subcellular
opposite to that produced by back-propagating action poten- compartments, whereas indirect physiological observations
tials, with the largest Ca2 entry in more distal dendrites have provided a broader perspective and are potentially
(Ariav et al., 2003). informative about regions of the dendrites inaccessible to
Smaller depolarizations, such as those produced by EPSPs, patch-clamp recording.
tend to produce smaller Ca2 transients that are even more Early dendritic recordings strongly suggested the presence
spatially restricted (Regehr et al., 1989; Miyakawa et al., 1992; of voltage-gated Na channels in CA1 apical dendrites, and
Regehr and Tank, 1992; Yuste and Denk, 1995; Yuste et al., the reduction of back-propagating action potentials by TTX
1999; Kovalchuk et al., 2000; Murthy et al., 2000). These Ca2 conrmed this (see Section 5.2.11). Direct recordings of Na
transients may be produced by activation of NMDA receptors, channel activity in patches up to about 300 m from the soma
but in some cases activation of voltage-gated Ca2 channels on the primary apical dendrite in CA1 neurons have revealed
also contributes. In fact, it can be difficult to dissect the rela- that Na channels are distributed at an approximately con-
tive contribution of these two mechanisms, as blocking either stant density along this region of the dendrite (Magee and
one reduces dendritic depolarization, which may reduce Ca2 Johnston, 1995a) (Fig. 56A). Despite the relatively uniform
entry through either of these channel types. channel density, the properties of the channels change with
Pairing EPSPs with back-propagating action potentials has distance from the soma. Patches obtained at increasing dis-
been shown to produce Ca2 entry that exceeds the sum of the tances from the soma exhibited a greater degree of prolonged
two individual signals in dendritic shafts and spines (Yuste inactivation, a feature of the Na channel responsible for the
and Denk, 1995; Magee and Johnston, 1997; Yuste et al., 1999). activity dependence of action potential back-propagation
This effect is likely to occur because EPSPs can increase (Colbert et al., 1997; Mickus et al., 1999). The reasons for this
the amplitude of back-propagating action potentials and nonuniformity in channel function are not known but might
because the additional depolarization provided by back-prop- be mediated by a unique subunit composition (- and/or
150 The Hippocampus Book

-subunits), differential post-translational modication (e.g., al., 1997) (Fig. 56C). The farthest recordings, about 350 m
phosphorylation), or protein-protein interactions (e.g., with from the soma, indicate that the density of the A-type K cur-
the cytoskeleton). Little is known about the subcellular distri- rent is about ve times its somatic level. This high density of
bution of Na channel -subunits in the hippocampus, dendritic A-type K current contributes to a dramatic decline
though the available data suggest that Nav1.2 is most abun- in amplitude of back-propagating action potentials (Hoffman
dant during the rst few postnatal weeks, but that Nav1.6 is et al., 1997) (see Section 5.2.11).
the dominant subunit in adults (Schaller and Caldwell, 2000). Ca2 imaging experiments suggest that the density of A-
Numerous Ca2 imaging studies have demonstrated the type K channels may be especially high in oblique apical
presence of voltage-gated Ca2 channels in CA1 dendrites. dendrites (Frick et al.. 2003). Consistent with the high density
The presence of depolarization-induced Ca2 transients in of dendritic A-type K channels, application of 4-AP to the
dendritic spines suggests that Ca2 channels are also present bath or selectively to the apical dendrite results in an increased
in these postsynaptic specializations (Yuste and Denk, 1995; amplitude of back-propagating action potentials (Hoffman et
Sabatini et al., 2002). One imaging study used a pharmacolog- al., 1997; Magee and Carruth, 1999). 4-AP application also
ical approach to determine that high-threshold Ca2 channels increases the occurrence of dendritic Ca2 spikes, an effect
are the major contributors to Ca2 entry mediated by back- that is likely attributable to block of both A-type and D-type
propagating action potentials in the soma and proximal den- K channels in the dendrites (Andreasen and Lambert, 1995a;
drite, whereas low-threshold Ca2 channels contribute more Hoffman et al., 1997; Golding et al., 1999; Magee and Carruth,
to the Ca2 signal in the distal dendrites (Christie et al., 1995). 1999; Cai et al., 2004). The most likely candidates for the -
The most direct information regarding Ca2 channel distribu- subunits of the A-type K channel in CA1 neurons are mem-
tion was provided by single-channel recordings along the bers of the Kv4 family (Sheng et al., 1992; Maletic-Savatic et
length of the apical dendrite in CA1 neurons (Magee and al., 1995). Consistent with this, enhanced dendritic excitabil-
Johnston, 1995a). This study revealed that the density of Ca2 ity has been observed in a mouse expressing a Kv4.2 domi-
channels is approximately constant along the length of the nant-negative construct (Cai et al., 2004). Pharmacological
apical dendrite, although there was a trend for more high- studies have also revealed that the AHP of back-propagating
threshold channels to be encountered in the soma and proxi- action potentials is much less sensitive to blockers of BK-type
mal dendrites and more low-threshold channels to be Ca2-activated K channels than the AHP of somatic action
encountered in more distal dendrites (Fig. 56B), which is potentials suggesting that there may be a higher density of
consistent with the Ca2 imaging results. Such a distribution these channels in the soma than in the apical dendrites
is also consistent with an antibody-binding study for L-type (Andreasen and Lambert, 1995b; Poolos and Johnston, 1999).
(high-threshold) Ca2 channels, which indicated the highest These channels are present in dendrites, however, and they
density of these channels is in the soma and proximal apical have an important role in repolarizing dendritic Ca2 spikes
dendrites (Westenbroek et al., 1990). Unfortunately, more (Golding et al., 1999).
detailed information regarding the subcellular distribution of The hyperpolarization-activated, nonspecic cation cur-
Ca2 and Na channels in hippocampal neurons is not avail- rent (Ih) is also distributed nonuniformly along the length of
able. The availability of a number of high-quality Ca2 and the CA1 apical dendrite. Cell-attached patch recordings reveal
Na channel antibodies is required to provide this critical a linear increase in density along the apical dendrite, with cur-
information in the future. Such studies will have to be per- rents at 350 m approximately six times larger than in the
formed using electron microscopy, as Ca2 and Na channels soma (Magee, 1998) (Fig. 56D). In fact, Ih may reach an even
are abundant in both axons and dendrites, making it difficult higher density in more distal dendrites, as immunoreactivity
to determine the presynaptic versus postsynaptic localization for the channel is most intense in the distal dendrites (Santoro
of these channels using light microscopy. Although no data et al., 1997; Lorincz et al., 2002). This is also consistent with
are available regarding the subtypes of Ca2 channels in CA1 the dramatic increase in Ih observed in patch recordings from
axons, pharmacological studies of synaptic transmission and the distal dendrites of layer V pyramidal neurons (Williams
presynaptic Ca2 entry indicate that N-, P-, Q-, and R-type and Stuart, 2000; Berger et al., 2001). In addition to the direct
Ca2 channels are most abundant in other axon terminals of patch recordings, imaging studies reveal a large dendritic Na
the hippocampus (Wu and Saggau, 1994; Gasparini et al., entry activated by hyperpolarization in hippocampal neurons
2001; Qian and Noebels, 2001). (Tsubokawa et al., 1999). Modeling the effects of blocking Ih
Information regarding K channel distribution has also on voltage attenuation in CA1 dendrites is also consistent with
been obtained using both indirect and direct approaches. a high density of Ih in CA1 apical dendrites (Golding et al.,
Cell-attached patch-clamp recordings have provided the most 2005). Of the four Ih genes cloned so far, HCN 1, 2, and 4 are
direct information regarding K channel distribution in CA1 the most abundant in CA1 neurons (Moosmang et al., 1999;
dendrites. Patches obtained at different distances from the Santoro et al., 2000; Bender et al., 2001). The presence of a
soma revealed that the density of A-type K channels high density of Ih channels in CA1 dendrites has some inter-
increases approximately linearly along the primary apical den- esting functional consequences, which are discussed below.
drite, whereas the density of noninactivating (or slowly inac- These examples highlight the point that many channels
tivating) K channels is approximately constant (Hoffman et (e.g., A-type K and Ih channels) are distributed nonuni-
 Hippocampal Neurons: Structure and Function 151

A Na + channels

<4 weeks
>4 weeks

B Ca2+ channels

C K + current

D I h current

Figure 56. CA1 dendritic voltage-gated channels. A. Single- activated, medium conductance channels; hatched bars correspond
channel recordings of voltage-gated Na channels in dendritic to high-voltage activated large conductance channels. (Source:
cell-attached patches (left) and a plot of Na channel density as a Adapted from Magee and Johnston, 1995a.)C. Voltage-gated K
function of distance from the soma (right). Circles and triangles channels in cell-attached patches from dendrites and the soma,
are from rats younger and older than 4 weeks, respectively. (Source: separated into sustained and transient components (left). The two
Adapted from Magee and Johnston, 1995.) B. Single-channel components are plotted as a function of distance from the soma.
recordings of voltage-gated Ca2 channels in dendritic cell-attached The transient component is much larger in the dendrites. (Source:
patches [low-voltage activated (left) and high-voltage activated Adapted from Hofmann et al., 1997.) D. Hyperpolarization-
(middle); ensemble averages are also shown for each set of traces] activated current in cell-attached patches from dendrites and
and a plot of Ca2 channel density as a function of distance from the soma (left). Ih current density is plotted as a function of
the soma (right). Black portions of the bars correspond to low- distance from the soma (right). Ih is much larger in the dendrites.
voltage activated channels; open bars correspond to high-voltage (Source: Adapted from Magee, 1998.)
152 The Hippocampus Book

formly along the somato-dendritic axis. Even when channel The function of dendritic Na and Ca2 channels at mem-
density is relatively uniform (e.g., Na channels), the proper- brane potentials below threshold for the induction of regener-
ties of the channels may differ between the soma and the den- ative dendritic events is less clear. In principle, each of these
drites. Although the mechanisms for establishing and channel types (especially the low-threshold, T-type Ca2
maintaining these gradients are not known, more is known channels) could be activated during nonregenerative depolar-
about the functional implications of ion channels and their izations (e.g., moderate EPSPs) and generate additional depo-
nonuniform densities and properties in CA1 dendrites. larization and Ca2 entry. In fact, Na channels have been
shown to open during EPSPs, even when there are no obvious
5.2.14 Functional Implications of signs of regenerative activity (Magee and Johnston, 1995b).
Voltage-Gated Channels in CA1 Dendrites: This nding is consistent with experiments showing that Na
Synaptic Integration and Plasticity and Ca2 channel blockers can reduce postsynaptic EPSPs
without affecting the presynaptic input (Lipowsky et al., 1996;
Knowing that the dendrites of CA1 neurons contain many Gillessen and Alzheimer, 1997). The ability of Na and Ca2
voltage-gated channels, it is logical to ask: What for? Clearly, channels to amplify EPSPs effectively (without a dendritic
dendrites are able to generate a variety of active responses, spike), however, is compromised by coactivation of K chan-
including back-propagating action potentials and dendritic nels, which can cancel the effect of Na and Ca2 channel acti-
Na and Ca2 spikes. What is the purpose of these responses vation, resulting in linear summation of EPSPs (Cash and
in excitable dendrites? Yuste, 1998, 1999). It is likely, however, that neuromodulation,
One obvious answer is that the occurrence of a dendritic prior activity, and synaptic input pattern can alter the balance
spike generates output in the form of one or more axonal of Na, Ca2, and K channel activation during synaptic inte-
action potentials. This is correct for some dendritic spikes in gration, leading to nonlinearities in synaptic summation
CA1 neurons but not all. Large, broad Ca2 spikes can trigger under some conditions.
a burst of action potentials in the soma, indicating that these The presence of a high density of hyperpolarization-
events produce a reliable axonal output (Golding et al., 1999), activated, non-specic cation channels in CA1 dendrites has
and under some conditions the forward propagation of den- some interesting consequences for the temporal integration of
dritic spikes can be facilitated (Jarsky et al., 2005) (see EPSPs. Because some of Ih is active at the resting potential, it
Sections 5.2.10 and 5.2.11). Some dendritic Na and Ca2 contributes a signicant leak to the membrane. This has the
spikes, however, do not propagate well to the soma, so their effect of increasing attenuation of EPSPs as they propagate
inuence there can be quite small (Golding and Spruston, from the soma to the dendrite, as well as speeding the decay of
1998; Golding et al., 2002). This observation hints at the pos- EPSPs in the dendrite (Stuart and Spruston, 1998; Golding et
sibility of local functions for dendritic spikes. al., 2005). This effect on the amplitude of EPSPs may be com-
A likely function for back-propagating action potentials is pensated in part by increased synaptic conductance at more
to inform the dendrites and the synapses on them about the distal synapses. Dendritic Ih has also been shown to normalize
level of output in the axon. The most compelling evidence for the temporal ltering effects of dendrites (Magee, 1999).
supporting such a role for dendritic spikes comes from the In passive dendrites, the ltering effects result in EPSPs gener-
demonstration of their importance in the induction of long- ated at distal locations, producing slower EPSPs at the soma.
term potentiation (LTP). At Schaffer collateral synapses (from As a result, distal dendritic EPSPs summate more at the soma
CA3 axons), pairing of small EPSPs with somatic action than do size-matched EPSPs from more proximal locations.
potentials induces a form of LTP that disappears when action Dendritic Ih minimizes this spatial discrepancy of temporal
potentials are prevented from actively invading the dendrites summation by contributing a larger effective outward current
(Magee and Johnston, 1997).Thus, back-propagating action (due to the turning off of inward Ih) for distal synapses, thus
potentials seem to have an important role in providing the compensating for the additional temporal summation that
postsynaptic depolarization necessary for the induction of would otherwise occur at the soma (Magee, 1999).
some forms of Hebbian LTP (see Chapter 10). At more distal
synapses on CA1 neurons, in particular the perforant path 5.2.15 General Lessons Regarding
input from entorhinal cortex, LTP does not depend on back- Pyramidal Neuron Function
propagating action potentials (Golding et al., 2002). Rather,
dendritically generated Na and Ca2 spikes appear to pro- The CA1 pyramidal neuron is arguably the most extensively
vide an important component of the depolarization necessary studied neuron with respect to resting membrane properties,
to induce LTP at these synapses. A similar form of dendritic dendritic function, and synaptic integration. These studies
spike-induced LTP also occurs at the Schaffer collateral have produced a striking picture of the neuron in which the
synapses if the stimulus intensity is increased to produce voltage attenuation in dendrites, is predicted to be enormous
larger EPSPs during blockade of back-propagating action for the near-passive condition. CA1 dendrites impose power-
potentials (Golding et al., 2002). These ndings indicate that ful attenuation and ltering during both normal conditions
both back-propagating action potentials and dendritically and voltage clamp. The neuron appears to compensate for this
induced spikes can function to enhance the induction of LTP by at least two key mechanisms: synapse conductance scaling
at synapses on the CA1 dendritic tree. and excitable dendrites containing myriad voltage-gated
 Hippocampal Neurons: Structure and Function 153

channels. On the other hand, such specialized properties are dendritic excitability. A population approach to deter-
unlikely to exist exclusively to overcome the ltering proper- mining Na channel distributions does not reveal such
ties of dendrites. Rather, the structure and ion channel expres- potentially important differences across cells.
sion are likely central to the specialized function of CA1 5. Synaptic potentials attenuate dramatically between
pyramidal neurons. For instance, one emergent property of their site of origin in the dendrites and their nal site
the complexity of the CA1 dendritic tree appears to be coinci- of integration (with respect to action potential output)
dence detection for perforant-path and Schaffer-collateral in the axon. Large synaptic depolarizations may serve
synaptic inputs (Remondes and Schuman, 2002; Jarsky et al., local functions (e.g., release of retrograde messengers,
2005). initiation of dendritic spikes) independent of dendro-
As becomes evident in the sections that follow, it would be somatic attenuation. Because of the extensive ltering
inappropriate to generalize specic conclusions drawn from introduced by dendrites and the active responses that
studies of one cell type and assume that other cells behave the can be generated in dendrites, somatic voltage-clamp
same way. It is possible, however, to emphasize some conclu- recordings must be interpreted with great caution.
sions based on studies of CA1 pyramidal neurons that are 6. Different excitatory and inhibitory inputs can target
likely to direct studies of other neuronal cell types. different subcellular domains. Excitatory inputs from
different regions, such as the direct input from
1. Morphology is stereotypical but variable. Although all entorhinal cortex and the processed input from CA3
CA1 cells have basal dendrites and a main apical den- neurons in CA1, can contact distinct regions of the
drite that gives rise to oblique and tuft branches, den- dendritic tree. Inhibitory interneurons, such as basket
dritic morphology varies considerably from one CA1 cells and oriens-alveus/lacunosum-moleculare (O-LM)
neuron to the next. It is not known, but reasonable to neurons, can selectively innervate domains such as the
postulate, that such morphological variability may soma or the distal apical dendrites. This differential
have functional consequences. targeting suggests important differences in postsynap-
2. Physiological properties are also variable. As with mor- tic actions, even among synapses using the same trans-
phological variability, differences in physiological mitter.
properties may be functionally relevant. Distinguishing 7. The threshold for action-potential initiation is lower in
inconsequential variability about a mean from signi- the axon than in the dendrites. This nding generalizes
cant functional variability is challenging but impor- from CA1 neurons to many other neurons (Stuart et
tant. Physiological variability could arise as a result of al., 1997). One possible exception has been noted: O-
genetic determinism, history-dependent development LM interneurons in the hippocampus have a low
and plasticity, or both. threshold for action potential initiation in the den-
3. Channel and receptor distributions are often nonuni- drites (Martina et al., 2000) (see section 5.9.7).
form. In CA1 neurons, as in layer V neocortical 8. CA1 dendrites, like many other dendritic trees, contain
pyramidal neurons, distal dendrites contain a higher voltage-gated channels that support action-potential
density of hyperpolarization-activated cation channels back-propagation and dendritic spike initiation
(Magee, 1998; Williams and Stuart, 2000; Berger et al., (Husser et al., 2000). The presence of voltage-gated
2001; Lorincz et al., 2002) than the soma and more channels in dendrites is likely to have profound effects
proximal dendritic regions. These two cell types differ, on synaptic integration.
however, in that the high density of A-type K chan-
nels in CA1 dendrites is not observed in layer V den- These general conclusions share the common feature
drites (Hoffman et al., 1997; Bekkers, 2000; Korngreen that they all imply a tremendous degree of complexity and
and Sakmann, 2000). Even though Na and Ca2 specialization in neuronal function, even for a single cell type,
channel densities are uniform (on average) in CA1 such as the CA1 pyramidal neuron. As subsequent sections
dendrites, the properties of these channels vary sys- on the properties of other neurons in the hippocampus
tematically along the main apical dendrite. AMPA indicate, there are no shortcuts: The morphological and
receptors have a higher density in mushroom-shaped functional properties of each and every cell type must be
spines and perforated synapses, whereas dopamine explored in detail in the hippocampus as in the rest of the
receptors have the highest density in distal apical den- brain.
drites of CA1 neurons. Additional inhomogeneities of
channel and receptor density and function are likely to
be revealed by future studies.
4. Population studies inform only about average behavior 5.3 CA3 Pyramidal Neurons
but cannot reveal functional variability across cells. For
example, if some cells have increasing Na channel Following the CA1 pyramidal layer back toward the dentate
density gradients from the soma into the dendrites, gyrus, one encounters the CA3 pyramidal layer. (Little is
and others have decreasing gradients, they may have known about the physiology of neurons in the intervening
very different functional properties with respect to CA2 region.) The CA3 pyramidal neurons have been studied
154 The Hippocampus Book

extensively, in large part because of the unique functional spe- similar to that described above for CA1 pyramidal cells. CA3
cializations formed by the mossy ber inputs from the dentate pyramidal neurons, recovered from rat hippocampus in vitro
gyrus and because of the extensive axon collaterals between slices from the mid-transverse hippocampal axis, possess
CA3 neurons, which create a highly interconnected and extensive dendritic arborizations with four main features, (1)
excitable network. a basal arbor that extends throughout the stratum oriens, (2)
a short apical trunk in the stratum lucidum that branches into
5.3.1 Dendritic Morphology two or more secondary trunks, (3) oblique apical dendrites in
the stratum radiatum, and (4) an apical tuft that extends into
Pyramidal neurons in the CA3 region are structurally similar the stratum lacunosum-moleculare. On average, the total den-
to CA1 neurons; they consist of pyramid-shaped somata that dritic length for the apical and basal dendritic arbors is simi-
give rise to apical and basal dendritic trees (Fig. 57A). CA3 lar to that of CA1, but with a larger range (9.315.8 mm)
pyramidal neurons differ from their CA1 counterparts, how- (Ishizuka et al., 1995; Henze et al., 1996). This variability of
ever, in that the apical dendritic tree bifurcates closer to the dendritic structure is attributable to systematic differences in
soma. CA3 pyramidal neurons are typically somewhat larger the dendritic structure of pyramidal neurons in the various
than CA1 pyramidal cells, but their total dendritic length and subregions of CA3 (Ishizuka et al., 1995) (see Section 5.3.3).
organization is heterogeneous. The smallest CA3 pyramidal Of particular interest, despite the greater total length and sur-
cells are located in the limbs of the dentate gyrus, and the face area of the apical arbor, the number of terminal branches
largest cells are located in the distal portion of the CA3 sub- and number of dendritic segments are similar for both the
eld (Amaral et al., 1990). apical and basal arbors. This suggests that the distances
Detailed morphometric analysis of dye-injected cells has between successive branch points are shorter for the basal
allowed complete reconstruction of CA3 pyramidal neurons, dendrites (Henze et al., 1996). However, the basal dendrites

Figure 57. CA3 dendritic morphology, thorny excrescences, and 2001.) B. Photomicrograph of dendrites in a lled CA3 pyramidal
synaptic inputs and outputs. A. Computer-generated plot showing neuron providing clear examples of thorny excrescences. Particu-
dendritic morphology of a CA3 pyramidal neuron. Thorny excres- larly large spine clusters are indicated with arrows. Bar  25 m.
cences are apparent on the proximal apical dendrites in the stratum (Source: Adapted from Gonzales et al., 2001.) C. Three-dimensional
pyramidale (s.p.) and stratum lucidum (s.l.). The major excitatory reconstruction of a large branched spine with 12 heads. The recon-
synaptic inputs to each layer are indicated, as are the major synaptic structed thorn is light gray, and the PSDs are indicated in white.
outputs. Bar  50 m. (Source: Adapted from Gonzales et al., Bar  1 m. (Source: Adapted from Chicurel and Harris, 1992.)

A B
EC layer II
s.l-m.

commisural
associational
s.r.

mossy fibers s.l. C

s.p.
commisural
s.o.
associational

axon

CA1 Schaffer
commissural projection
collaterals fimbria lateral septal nucleus
 Hippocampal Neurons: Structure and Function 155

have approximately threefold fewer tips per primary dendrite shapes (Blackstad and Kjaerheim, 1961; Chicurel and Harris,
(8.5 vs. 29.1). This, together with the lower maximum branch 1992).
order for the basal dendrites (7.0 vs. 10.8), suggests that the The thorny excrescences (and hence the mossy ber
individual basal dendritic trees are signicantly less complex inputs) are concentrated in the rst 100 m or so of the api-
than the apical trees. Finally, in addition to being less complex, cal dendritic tree of CA3 pyramidal neurons (stratum
the mean distance to the basal dendritic tips is shorter than for lucidum). Across the remaining portions of both the apical
the apical tree (212 vs. 425 m) and suggests that synapses and basal dendrites, the shape and distribution of spines are
formed on the basal dendrites occupy a more restricted range somewhat similar to CA1 pyramidal neurons, described
of physical distances from the somata. The dendritic trees of above. Various calculations put the number of simple spines
CA3 pyramidal neurons also have a roughly symmetrical close to 30,000, which is similar to estimates of total
structure in their maximal transverse and septo-temporal spine number on CA1 pyramidal neurons (Jonas et al.,
extents (~ 300 and 270 m, respectively). 1993;Trommald et al., 1995).

5.3.2 Dendritic Spines and Synapses 5.3.3 Excitatory and Inhibitory Synaptic Inputs

Like their CA1 counterparts, CA3 pyramidal neuron dendrites The CA3 pyramidal neurons receive three prominent forms
are studded with thousands of spines. Even greater diversity of of excitatory synaptic input. The aforementioned mossy ber
spine morphology is apparent in CA3. In addition to the spine input from the dentate granule cells is the most extensively
shapes observed in CA1, the CA3 neurons also have another studied of these inputs. CA3 neurons also receive direct
major spine class, the thorny excrescences (Fig. 57B,C). input from layer II of the entorhinal cortex via the perforant
There are about 40 of these specialized spine clusters on each path. As in the CA1 region, this direct cortical input is limited
CA3 neuron (Blackstad and Kjaerheim, 1961; Chicurel and to the distal regions of the apical dendrites, in the stratum
Harris, 1992; Amaral and Witter, 1995; Gonzales et al., 2001). lacunosum-moleculare. The third prominent input to CA3
In the proximal regions of CA3 (closest to the dentate gyrus) neurons comes from the axons of other CA3 neurons. These
the thorny excrescences are distributed on apical and basal commissural/associational (C/A) inputs are numerous and
dendrites relatively close to the soma. In more distal CA3 neu- originate from CA3 neurons on both sides of the brain. This
rons (closer to CA1), the excrescences are found primarily on extensive network of recurrent collaterals has led to the pos-
the apical dendrite prior to the rst branch point located at tulate that the CA3 region may function as an autoassociative
distances of 10 to 120 m from the soma, where they form the network involved in memory storage and recall (Bennett et al.,
postsynaptic targets of the mossy ber synapses from granule 1994; Rolls, 1996) (see Chapter 14). A side effect of this exten-
cells of the dentate gyrus (Blackstad and Kjaerheim, 1961; sive interconnectivity, however, is that the CA3 network is
Hamlyn, 1962; Chicurel and Harris, 1992; Gonzales et al., highly excitable and prone to seizure activity when inhibition
2001). Three-dimensional reconstruction of thorny excres- is suppressed (Jung and Kornmuller, 1938; Ben-Ari, 1985) (see
cences on CA3 pyramidal neurons have revealed many Section 5.3.5). CA3 neurons also receive cholinergic input
unusual features that are peculiar to the mossy ber synapses from the medial septal nucleus and the nucleus of the diago-
and have not been described at other hippocampal synapses nal band of Broca, which terminate mostly in the stratum
or indeed other brain areas. Typically, postsynaptic thorny oriens (Amaral and Witter, 1995).
excrescences at mossy ber synapses comprise 1 to 16 The organization of the dendritic tree and its synaptic
branches that emerge from a single dendritic origin. These inputs are highly variable and strongly depend on the location
branched spines contain subcellular organelles typical of of the cell body within CA3. Cells located in the limbs of the
other spines (including ribosomes and multivesicular bodies), dentate gyrus have dendrites with a limited elaboration and
but they also contain organelles not found in other spines, do not extend dendrites into the stratum lacunosum-molecu-
such as mitochondria and microtubules. Multivesicular bod- lare (Amaral et al., 1990). In contrast, cells toward the more
ies occur most often in the spine heads, which also contain distal portion of the CA3 subeld project their dendrites
smooth endoplasmic reticulum; and ribosomes occur most throughout the stratum radiatum and stratum lacunosum-
often in spines that have spinules (nonsynaptic protruber- moleculare. Of particular importance regarding this heteroge-
ances that emerge from the spine head). Branched spines are neous anatomical organization, CA3 pyramidal cells at the
typically surrounded by a single mossy ber bouton, which most proximal and distal ends of the CA3 subeld are under
can establish synapses with multiple spine heads. The postsy- differential control by the various afferent pathways project-
naptic densities of these spines occupy ~ 10% to 15% of the ing into the CA3 hippocampus. Most notably, CA3 pyramidal
spine head membrane. Individual mossy ber boutons usually neurons closest to the dentate gyrus receive the proportionally
synapse with several branches belonging to more than one largest input from the dentate granule cells; these CA3 neu-
thorny spine excrescence but all originating from the same rons receive mossy ber axonal inputs onto both the apical
parent dendrite. Within a given branched spine, the dimen- and basal dendritic trees, whereas those located more distally
sion and volume of individual spine heads are comparable to in CA3 have mossy ber inputs primarily onto the proximal
those of the large mushroom spines found on CA1 pyramidal portion of their apical dendritic tree (Blackstad et al., 1970).
neurons, although the CA3 branches have more irregular This limited mossy ber projection to the basal CA3 pyrami-
156 The Hippocampus Book

dal cell dendrites is referred to as the infrapyramidal projec- at 150 to 300 mm. These collaterals cover a large area in the
tion and the granule cells in the infrapyramidal blade of the longitudinal direction, extending more than one-third the
dentate gyrus are the primary source of this projection. length of the septo-temporal axis (Ishizuka et al., 1990; Sik et
Although initial binding studies suggested that NMDA al., 1993; Li et al., 1994; Andersen et al., 2000). Pyramidal neu-
receptors were absent at mossy ber synapses (Monaghan et rons in the CA3a subeld (closest to CA1) arborize primarily
al., 1983), subsequent physiological studies revealed that acti- in the stratum oriens of CA1, whereas neurons in CA3c
vation of NMDA receptors indeed occurs during activation of (closer to the hilus) collateralize more extensively in the stra-
mossy ber synapses (Jonas et al., 1993). Dendritic recordings tum radiatum of CA1 (Ishizuka et al., 1990; Li et al., 1994). In
from the apical dendrites of CA3 pyramidal cells in areas close one complete reconstruction, the highest density of boutons
to the mossy-ber termination zone in stratum lucidum indi- in CA1 clustered about 600 to 800 m from the CA3 neuron
cate that the properties of dendritic ionotropic glutamate in both the septal and temporal directions (Sik et al., 1993).
receptors are remarkably similar to those expressed at the These axons contribute to the Schaffer collateral inputs to
soma of CA3 pyramidal cells (Spruston et al., 1995). AMPA CA1 pyramidal neurons. Another group of axon collaterals is
and NMDA-type glutamate receptors in CA3 neurons are also localized in the stratum oriens and stratum radiatum of CA3,
remarkably similar to these receptors in CA1 pyramidal neu- comprising the extensive associational connection common
rons. One notable difference, however, is the deactivation between CA3 pyramidal neurons. Boutons are located
kinetics of NMDA receptors, which were slower in CA1 than approximately every 4 m along the CA3 axon, but boutons
in CA3 pyramidal neurons. In addition, NMDA receptor- are distributed unevenly, with somewhat variable spacing
mediated Ca2 entry is limited in CA3 spines (Reid et al., (Shepherd et al., 2002). Estimates of the total number of
2001; but see Pozzo-Miller et al., 1996). A comparison of the synapses formed by a single axon in the ipsilateral hippocam-
properties of mossy ber synaptic transmission to AMPA pus range from 15,000 to 60,000 (Sik et al., 1993; Li et al.,
receptor properties in CA3 dendrites suggests that a typical 1994). A subset of the CA3 axon boutons contacts interneu-
quantal event would consist of 35 AMPA receptors being acti- rons. Each axon contacts each interneuron at a single location,
vated by a single quantum of glutamate (10 pS single- forming one synaptic release site. As in CA1, pyramidal cell-
channel conductance  350 pS quantal conductance). About to-interneuron synapses are powerful enough that a single
70 receptors are estimated to be present at each release site, axon is capable of producing an action potential in postsy-
but only half of these receptors are open at the peak of the naptic interneurons (Miles, 1990; Gulys et al., 1993a; Sik et
synaptic current (Jonas et al., 1993; Spruston et al., 1995). al., 1993).
Immunocytochemical analysis suggests that about four times Three-dimensional reconstruction of CA3 pyramidal neu-
as many AMPA receptors are present at each mossy ber ron axons in the CA1 region revealed that most of the vari-
synaptic specialization as at C/A synapses in CA3 or SC cosities or presynaptic boutons contained synaptic vesicles and
synapses in CA1 (Nusser et al., 1998). This high density of were opposed to a postsynaptic specialization that comprised
AMPA receptors and the large number of synaptic specializa- a single postsynaptic density (PSD). Approximately 20% of
tions on each thorny excrescence (Chicurel and Harris, 1992) presynaptic varicosities were apposed to postsynaptic special-
accounts for the large size and variability of unitary mossy izations comprising multiple PSDs (Shepherd and Harris,
ber synaptic currents, which have been estimated to com- 1998). CA3 pyramidal neuron axonal varicosities are typically
prise 2 to 16 quantal events (Jonas et al., 1993). oblong in shape and demonstrate considerable variation in
CA3 pyramidal neurons also receive substantial innerva- their length (~1 m) and volume (~0.13 m3). The intervari-
tion from inhibitory interneurons. In addition to the somatic cosity axonal shaft is narrow, tubular, and remarkably consis-
and axonal inhibition, which presumably limits action poten- tent in its diameter (~0.17 m). The narrow axonal shafts
tial initiation, a number of interneurons target CA3 dendrites. resemble dendritic spine necks, an arrangement that raises the
This dendritic inhibition has been shown to limit the initia- possibility of functional compartmentalization along individ-
tion and enhance termination of dendritic calcium spikes ual axons. Importantly, serial electron microscopy has revealed
(Miles et al., 1996). that a small number of CA3 axons (~20%) may make multiple
synapses on a single CA1 cell (Sorra and Harris, 1993).
5.3.4 Axon Morphology and Synaptic Targets
5.3.5 Resting Potential and Action
Each CA3 pyramidal neuron gives rise to a single axon, which Potential Firing Properties
projects bilaterally to the CA3, CA2, and CA1 regions, as well
as to the lateral septal nucleus (Ishizuka et al., 1990; Amaral Like CA1 pyramidal cells, CA3 pyramidal cells possess a reper-
and Witter, 1995). At the light microscopic level, individual toire of conductances that allow them to respond to
axons are thin and myelinated with abundant en passant bou- suprathreshold stimuli with either single action potentials or
tons or varicosities spaced, on average, 4.7 m apart (Ishizuka bursts. Although bursting also occurs in CA1 pyramidal neu-
et al., 1990; Shepherd and Harris, 1998; Shepherd et al., 2002). rons (see Section 5.2.5), bursting is more prominent in CA3,
CA3 axons project primarily to the CA1 region but also col- both in vivo and in the in vitro slice preparation (Kandel and
lateralize extensively within CA3. The total length of the CA3 Spencer, 1961; Spencer and Kandel, 1961; Wong and Prince,
collaterals in the ipsilateral hippocampus has been estimated 1978) and is therefore considered a hallmark feature of CA3
 Hippocampal Neurons: Structure and Function 157

pyramidal neurons. Bursts in CA3 neurons typically comprise A C

20 mV
several action potentials riding atop a depolarizing waveform,
often accompanied by smaller calcium-dependent spikes late 50 ms

in the waveform. Each burst is an all or nothing event last-

ing approximately 30 to 50 ms, with the frequency of action
potentials in the range of 100 to 300 Hz (Wong et al., 1979;
Wong and Prince, 1981; Traub and Miles, 1991). B1
Bursts can be triggered in several ways in CA3 neurons, and
the mechanism underlying bursting depends in large part on
the way the burst is evoked. One mechanism for burst gener-
ation is entirely intrinsic to the neuron and does not require
synaptic connectivity. A brief current injection via a somatic B2
recording electrode can produce an all-or-none burst of
20 mV 40 mV
action potentials in some CA3 neurons (Wong and Prince,
1981) (Fig. 58A), and in some cases these bursts occur spon- 100 ms 100 ms

taneously (Cohen and Miles, 2000). The burst is caused by an

ADP that follows a single action potential, which is likely to be
mediated by calcium current, as pharmacological blockade of Figure 58. CA3 pyramidal neuron bursting and epilepsy.
calcium current converts burst ring to single spiking (Wong A. Intrinsic burst ring in response to a short current injection in
and Prince, 1981). a CA3 pyramidal neuron. (Source: Adapted from Wong and Prince,
The CA3 neurons can also generate intrinsic bursts in 1981.) B. Similar appearance of intrinsic and network-driven bursts
response to brief synaptic activation or dendritic current in CA3 pyramidal neurons. B1 shows an endogenous burst in a
injection (Wong et al., 1979; Masukawa et al., 1982; Masukawa recording with the slice bathed in normal solution. B2 shows a
and Prince, 1984). Under these conditions the ADP may con- network-driven burst during a paroxysmal depolarization shift
(PDS), a network discharge induced by bathing the slice in a solu-
tribute to the bursting, but dendritically generated calcium
tion containing penicillin. Recordings are from two different cells.
spikes are likely to be a more important mechanism driving (Source: Adapted from Johnston and Brown, 1984.) C. The synaptic
the burst (Wong and Prince, 1978; Wong et al., 1979). nature of the PDS is shown by its reversal in polarity at a holding
Continuous current injection into the apical dendrites of CA3 potential of around 0 mV. The PDS was induced by bathing the
pyramidal neurons evokes rhythmic bursts at frequencies of at slice in a solution containing d-tubocurarine. (Source: Adapted
least 8 Hz (Traub and Miles, 1991). The burst pattern in the from Lebeda et al., 1982.)
dendrite differs markedly from that observed in the soma,
with dendritic fast spikes (presumed Na) usually having a
lower amplitude than slower spikes (presumed Ca2); this ing, each burst is terminated by a slow AHP (~1 second)
probably reects a greater distance of the dendrite from the mediated by a variety of voltage- and Ca2-activated K con-
axon initial segment, a low density of Na channels in the ductances (Alger and Nicoll, 1980; Hotson and Prince, 1980;
dendrite, or both (Traub et al., 1999). Schwartzkroin and Stafstrom, 1980; Lancaster and Adams,
The most robust bursts are generated in CA3 neurons when 1986; Lancaster and Nicoll, 1987; Storm, 1989). In contrast,
the entire network becomes synchronously active, usually in there is good evidence that network bursts are terminated by
response to suppression of synaptic inhibition or reduction of depletion of glutamate-containing synaptic vesicles during
K currents (Prince and Connors, 1986). Under these condi- the giant EPSPs that drive the PDS. Replenishment of these
tions, the network produces a paroxysmal depolarization shift vesicle pools is an important factor determining the rate at
(PDS), which is a hallmark of seizure activity in cortical net- which the PDS can recur during seizure activity (Staley et al.,
works. The PDS is a synchronous network depolarization 1998).
driven primarily by giant synaptic potentials (Johnston and Interestingly, not all CA3 pyramidal neurons respond to
Brown, 1981, 1984, 1986; Lebeda et al., 1982). Thus, intrinsic suprathreshold stimuli with intrinsic bursts. Intrinsic bursting
conductances can generate bursts that are likely to serve as is most typical of cells located at the distal portion of CA3,
important signals in a normally functional network, whereas closest to CA2. In guinea pig slices, cells located closer to the
massive, synchronous glutamate release contributes to syn- dentate gyrus typically respond with only a series of single
chronous bursting during hippocampal seizures (Fig. action potentials (Masukawa et al., 1982). Furthermore, CA3
58B,C). Because CA3 pyramidal neurons generate seizures, pyramidal neurons with somata located close to the stratum
they have been extensively studied as the possible pacemakers pyramidale/oriens border are more predisposed to burst gen-
of interictal epileptiform activity in a variety of animal mod- eration than those located closer to the stratum radiatum
els of electrographic seizures (see Schwartzkroin, 1993 for a (Bilkey and Schwartzkroin, 1990). One notable morphological
number of relevant reviews). distinction between these two cell groups is the length of the
Just as the mechanisms for generating normal and patho- initial portion of their apical dendrite; cells with somata closer
logical bursts are different, the mechanisms for terminating to the stratum pyramidale/oriens border have a signicantly
them are different as well. During normal intrinsic burst- longer initial apical dendrite. Consistent with the correlation
158 The Hippocampus Book

between anatomy and burst behaviors, L-type voltage-gated

Ca2 channels are clustered with high density on the base of
A
major dendrites of CA3 pyramidal neurons (Westenbroek et
al., 1990). Such an arrangement would permit large Ca2
uxes in response to summed excitatory inputs and perhaps
may account for the predisposition of certain deep CA3
pyramidal neurons to bursting activity.

5.3.6 Resting Membrane Properties

As discussed above for CA1 pyramidal neurons, accurate esti-

mation of the resting membrane properties of central neurons
has been problematic, so the conditions of any particular
measurement must be carefully considered. At resting mem-
brane potentials (
60 to
70 mV) and with Ih blocked, the
input resistance of CA3 neurons is slightly higher than in CA1
neurons, at about 160 M (Spruston and Johnston, 1992)
(Fig. 59). Estimates of the membrane time constant of CA3
B
neurons in slices from adult guinea pig yield a mean value of
90 ms, which is considerably slower than for CA1 neurons
under the same conditions (Spruston and Johnston, 1992).
Similar estimates in slices from young rats at room tempera-
ture resulted in even slower membrane time constants (112
ms), also without block of Ih (Jonas et al., 1993). Assuming a
Cm of 1 F/cm2 implies a membrane resistivity of 60,000 to
120,000 cm2 of CA3 neurons, which is more than twice as
high as Rm in CA1 neurons. It is not known whether Rm is
lower in dendrites of CA3 neurons than in the soma, as it is in
CA1 neurons (Spruston and Johnston, 1992; Stuart and
Spruston, 1998).
As with CA1 pyramidal neurons, estimates of m and RN
are larger at more depolarized potentials, consistent with the
notion that voltage-dependent conductances, active near the
resting membrane potential, contribute to the resting mem-
brane properties of CA3 neurons (Spruston and Johnston,
1992). Pharmacological elimination of Ih increases the magni-
C
tude of RN and m but did not alter their voltage dependence.
This is in contrast to CA1 pyramidal cells, where blocking Ih
reverses the voltage dependence of RN and m.
The high value of Rm for CA3 neurons implies that their
dendrites may be electrically compact for the steady-state con-
dition. As is the case for CA1 neurons, however, neuronal Figure 59. CA3 passive properties and sag. A. Voltage responses to
models indicate that attenuation of transient changes in step current injections from 80 to 40 pA in 20-pA increments
membrane potential is substantial, and voltage control of (20 pA not shown). B. Voltagecurrent plot of the steady-state
synaptic and voltage-gated currents using a somatic electrode responses shown in A. Solid lines are the points used for linear ts
would be nearly impossible for conductances located in CA3 to the data. Dashed lines are extrapolations of the ts. C. Effect of 3
dendrites, even at rather modest distances from the soma mM CsCl on the hyperpolarizing voltage response to a current step
(Spruston et al., 1993; Major et al., 1994). of 80 pA. Note the block of a small sag in the voltage response.
(Source: AC: Adapted from Spruston and Johnston, 1992.)
5.3.7 Active Properties of CA3 Dendrites
overcome, at least in part, by the presence of voltage-gated
Passive models of CA3 pyramidal neurons predict large atten- conductances in the dendrites. Although CA3 pyramidal neu-
uation of EPSPs between the dendrites and the soma rons possess most if not all of the same intrinsic conductances
(Carnevale et al., 1997; Jaffe and Carnevale, 1999). Although it described for the CA1 pyramidal cell, and in general these cur-
is not known whether the conductance of distal synapses are rents share many of the same properties and identities, the
larger to compensate for this attenuation (as shown in CA1) density of channels and/or the subcellular location often dif-
(Magee and Cook, 2000), dendritic attenuation is likely to be fer. For example, Ih is found in both CA1 and CA3 pyramidal
 Hippocampal Neurons: Structure and Function 159

neurons. However, the channels underlying the generation of sources and constitutes the major output of the hippocampus
this current in the two cell types appear to be different. Within (Naber et al., 2000) (see Chapter 3). Although neurons in this
the hippocampus two Ih subunit mRNA transcripts have been region have been studied much less than their neighbors in
detected: HCN1 and HCN2. Labeling for HCN1 is heavy in CA1 and CA3, the pronounced tendency of neurons in this
the CA1 pyramidal cells, whereas only moderate expression is region to re bursts of action potentials has led to consider-
observed in the CA3 pyramidal cell layer. The opposite distri- able interest in understanding the cell physiology of these
bution pattern is true for HCN2. Functionally, despite the dif- neurons.
fering expression patterns of HCN1 and HCN2, there appears
to be little overall difference between the biophysical proper- 5.4.1 Dendritic Morphology
ties of Ih in either cell type (Santoro et al., 2000; Fisahn et al.,
2002). Ih measured at the CA3 pyramidal neuron soma is con- Almost all principal neurons in the subiculum have a typical
siderably smaller than those observed in CA1 pyramidal neu- pyramidal morphology, with apical dendrites extending into
rons (Santoro et al., 2000), however, suggesting a different Ih the molecular layer and in many cases reaching the hip-
channel density in the two cell types. It is not known whether pocampal ssure (Harris et al., 2001) (Fig. 510). A quantita-
Ih is distributed at higher density in CA3 dendrites, as it is tive comparison of dendritic branching in the subiculum and
in CA1. CA1 revealed that subicular pyramidal neurons have slightly
Because CA3 pyramidal cells lack a large primary apical fewer branches in both the basal and apical dendritic trees,
dendrite such as that found in CA1, studies of dendritic with the largest difference in the proximal apical dendrites
excitability and ion channels in CA3 dendrites have lagged (Staff et al., 2000). No differences have been observed between
behind those in CA1. Consequently, little is known about the the dendritic morphology of regular-spiking and bursting
identities and properties of specic ion conductances in the neurons in the subiculum (Mason, 1993; Taube, 1993; Greene
dendrites of CA3 pyramidal neurons. A few studies have used and Totterdell, 1997; Staff et al., 2000; Harris et al., 2001).
sharp microelectrodes to record spikes from the dendrites of
CA3 pyramidal neurons (Wong et al., 1979; Miles et al., 1996). 5.4.2 Dendritic Spines and Synaptic Inputs
Calcium imaging experiments also indicate that voltage-gated
conductances are located in CA3 dendrites and spines (Pozzo- As in CA1, the dendrites of subicular pyramidal neurons are
Miller et al., 1993; Jaffe and Brown, 1997). Although much of studded with spines. Virtually no quantitative data are avail-
what we know about dendritic excitability in CA3 has been able regarding spine density, however, and little is known
inferred from a relatively small number of dendritic record- about the distribution of the numerous cortical and subcorti-
ings and calcium imaging, some interesting computer models cal inputs to the dendritic trees of subicular pyramidal neu-
have been developed for CA3 neurons. Roger Traub and col- rons. Two of the most prominent inputs to subiculum include
leagues used such computer models to demonstrate the topographical projections from CA1 and the entorhinal cor-
importance of dendritic Ca2 conductances in generating tex (see section 5.4.3 and Chapter 3 for details). The projec-
action potential bursts, as well as the importance of these tion from the entorhinal cortex originates in both layer II
bursts in the synaptic propagation of seizure activity through (which also projects to the dentate gyrus and CA3) and layer
a model of the CA3 network (Traub and Miles, 1991; Traub et III (which also projects to CA1) and forms synapses restricted
al., 1999). Another model of a CA3 pyramidal neuron demon- to the distal portions of the apical dendrites (Witter and
strated that bursting behavior does not require any special dis- Amaral, 1991; Lingenhhl and Finch, 1991; Tamamaki and
tribution of Ca2-dependent channels or mechanisms. Nojyo 1993; Amaral and Witter, 1995). These EC projections
Furthermore, a simple increase in the Ca2-independent K form functional excitatory synapses on subicular pyramidal
conductances was sufficient to convert the ring mode of neurons (van Groen and Lopes da Silva, 1986; Behr et al.,
modeled CA3 pyramidal neurons from bursting to nonburst- 1998; Gigg et al., 2000). Axons from CA1 pyramidal neurons
ing (Migliore et al., 1995). Hopefully, these models will stim- also form excitatory synapses on the dendrites of subicular
ulate additional experimental studies of the excitable pyramidal neurons (Finch and Babb, 1980, 1981; Tamamaki
properties of CA3 dendrites. and Nojyo 1990; Amaral et al., 1991; Taube, 1993; Gigg et al.,
2000).
Among the subcortical structures projecting to the subicu-
lum are the thalamic nucleus reuniens and weak cholinergic
5.4 Subicular Pyramidal Neurons projections from the septal nucleus and the nucleus of the
diagonal band. Modulatory inputs from the brain stem also
The subiculum lies on the opposite side of the CA1 region innervate the subiculum, including the locus coeruleus (nora-
than CA3, directly adjacent to the distal extent of area CA1. It drenergic), ventral tegmental area (dopaminergic), and raphe
begins where the Schaffer collaterals end, a point distinguish- nuclei (serotonergic). In addition to a direct input from the
able by the transition from the tightly packed CA1 pyramidal entorhinal cortex, the subiculum receives inputs from many of
layer to the more diffuse pyramidal layer of subiculum. The the same cortical areas that project to the entorhinal cortex.
subiculum plays an important role in the hippocampal for- Almost nothing is known about the inhibitory interneurons
mation, as it receives convergent input from numerous and their targeting of pyramidal neurons in the subiculum.
160 The Hippocampus Book

EC layers II & III

prefrontal cortex
anterior olfactory nucleus
retrosplenial cortex
CA1
septal complex
mammilary nuclei
nucleus accumbens
olfactory tubercle
CA1 s.o. thalamic nuclei
axon amygdaloid complex
bed nucleus of stria terminalis
EC layer V
presubiculum
parasubiculum
Figure 510. Subiculum dendritic morphology, spines, and synaptic inputs and outputs.
Camera lucida drawing shows the dendritic morphology of a subicular pyramidal neuron, with
its major inputs and outputs indicated. Bar  100 m. (Source: Adapted from Staff et al.,
2000.)

Synaptic stimulation in the CA1 region or entorhinal cor- the proximal subiculum (closest to CA1) projects to the most
tex results in apparently monosynaptic EPSPs that are blocked distal regions of the entorhinal cortex (the lateral entorhinal
by AMPA-type glutamate receptor antagonists and are only area), and the distal subiculum projects to the proximal
modestly affected by NMDA receptor blockers (Stewart and entorhinal cortex (the medial entorhinal area).
Wong, 1993; Stewart, 1997; Stanford et al., 1998; Staff et al., Among the other numerous targets of the subiculum are
2000). When inhibition is not blocked, synaptic stimulation in the medial prefrontal cortex, anterior olfactory nucleus, retro-
CA1 also produces a strong inhibitory postsynaptic potential splenial cortex, septal complex, mammilary nuclei, nucleus
(IPSP), presumably mediated in part by feedforward inhibi- accumbens, olfactory tubercle, several thalamic nuclei
tion resulting from activation of interneurons in the subicu- (including reuniens), amygdaloid complex, and the bed
lum (Menendez de la Prida, 2003). nucleus of the stria terminalis (Amaral and Witter, 1995) (see
Chapter 3 for more details). An interesting organizational fea-
5.4.3 Axon Morphology and Synaptic Targets ture of the axonal projections from the subiculum is that they
are organized along the septo-temporal (dorsal-ventral) axis
The axons of subicular pyramidal neurons collateralize exten- of the hippocampus as well as along the proximo-distal (CA1
sively in the subiculum as well as projecting out of the subicu- to presubiculum) axis. Neurons from four regions dened by
lum (Harris et al., 2001). The local collaterals are believed to these axes (septal-proximal, septal-distal, ventral-proximal,
form glutamatergic synapses contacts (Harris and Stewart, ventral-distal) project to different targets, and most subicular
2001). Pyramidal neurons with deeper cell bodies tend to neurons project to only a single cortical or subcortical target
form vertically oriented columns of local collaterals, whereas (Naber and Witter, 1998).
supercial pyramidal neurons exhibit a greater horizontal
spread (toward CA1 and EC) in their local axon collaterals 5.4.4 Resting and Active Properties
(Harris et al., 2001).
Projecting axon collaterals target a number of cortical and In awake, freely moving rats, subicular cells exhibit ring in
subcortical structures (Fig. 510). They do not project back to place elds, which are larger and noisier than those seen in
CA1 but do project to the entorhinal cortex and the pre- and CA1 (Sharp and Green, 1994; OMara et al., 2001) (see
parasubiculum (Amaral and Witter, 1995). The projection to Chapter 11). A prominent feature of the ring patterns
the entorhinal cortex terminates primarily in layer V. These observed in vivo is action potential bursting. Indeed, bursting
connections are organized such that subiculum projects back is much more prominent in the subiculum than in CA1. In
to the same regions of the entorhinal cortex from which it both in vitro and in vivo studies, pyramidal cells in the subicu-
receives input. This reciprocal projection is organized in an lum have been shown to fall into two broad physiological cat-
arrangement similar to the CA1-subiculum projection, in that egories: bursting cells and regular-spiking cells (Mason, 1993;
 Hippocampal Neurons: Structure and Function 161

Figure 511. Subiculum action potential bursting. A. Subicular bursting because of its response to a threshold-level step current
pyramidal neuron classied as regular spiking because of its injection (B1). A single two-spike burst occurs in response to a
response to a threshold-level step current injection (A1). A single threshold-level current injection in the shape of an EPSC (B2).
spike occurs in response to a threshold-level current injection in the (Source: A, B: Adapted from Cooper et al., 2003.)
shape of an EPSC (A2), but larger EPSC-like current injections pro-
duce bursting (A3). B. Subicular pyramidal neuron classied as

Mattia et al., 1993; Stewart and Wong, 1993; Taube, 1993; Staff from 69 to 64 mV in microelectrode and patch-clamp
et al., 2000; Menendez de la Prida et al., 2003) (Fig. 511). recordings (Mason, 1993; Mattia et al., 1993, 1997a; Stewart
Given the large number of brain areas innervated by subic- and Wong, 1993; Greene and Totterdell, 1997; Staff et al.,
ular neurons, it is tempting to speculate that strong bursting 2000). RN ranged from 24 to 42 M and m from 7 to 20 ms.
cells target different areas than weaker bursting or regular Most studies have revealed that these values are not signi-
spiking cells. Some evidence supports this hypothesis, as it has cantly different between bursting and regular-spiking neurons
been shown that bursting cells project preferentially to the in the subiculum (Taube, 1993; Staff et al., 2000). Compared
presubiculum and parasubiculum, whereas regular-spiking to CA1 pyramidal neurons, however, a recent patch-clamp
cells project preferentially to the entorhinal cortex (Stewart, study showed that subicular neurons have signicantly lower
1997; Funahashi et al., 1999). There is also some evidence for RN and faster m (Staff et al., 2000; but see Mason, 1993 and
preferential localization of bursting and regular-spiking neu- Taube, 1993). Like CA1 neurons, pyramidal neurons in the
rons in the subiculum, but the spatial gradients are weak. Two subiculum contain a substantial Ih current, which produces
studies showed that regular spiking cells are more prevalent in the characteristic sag toward the resting potential during
supercial layers (Greene and Totterdell, 1997; Harris et al., hyperpolarizing or depolarizing current pulses (Stewart and
2001). Evidence for a proximo-distal gradient is less consis- Wong, 1993; Staff et al., 2000; Menendez de la Prida et al.,
tent. A microelectrode study in slices showed a preference for 2003) (Fig. 512).
regular-spiking cells near the middle of the proximo-distal Action potentials were also found to be the same in burst-
axis of the subiculum (Greene and Totterdell, 1997), whereas ing and regular spiking pyramidal neurons in the subiculum
a patch-clamp study showed a slightly higher proportion of but were somewhat different from CA1 pyramidal neurons.
bursting cells in the distal subiculum (Staff et al., 2000). A Comparing single CA1 action potentials with subiculum
variety of technical differences could account for these differ- action potentials (either the rst in a burst or single action
ent conclusions, but the message emerging from these studies, potentials in regular-spiking cells), the subicular action poten-
taken together, is that proximo-distal gradients of bursting are tials were found to have slightly slower rise times and smaller
weak, if they exist at all. amplitudes but shorter half widths (Staff et al., 2000). The
The resting and active properties of subicular pyramidal dendritic morphology was also similar across all subicular
neurons have been examined in several studies. The measure- neurons but distinct from that of CA1 cells (Staff et al., 2000).
ments of resting potentials are remarkably consistent, ranging Taken together, these ndings suggest that the dichotomy
162 The Hippocampus Book

vated by Na-dependent action potentials (Jung et al., 2001).

A Ca2 channels are activated by each action potential, and cur-
rent ows through these channels during spike repolarization
as a result of an increased driving force and slow channel
deactivation. The resulting Ca2 tail current produces a spike
afterdepolarization (ADP), which triggers additional spikes
comprising the burst (Jung et al., 2001). This study also
showed that block of Ca2 channels can lead to an abnormal
form of bursting in part due to indirect block of Ca2-acti-
15 vated K channels. This nding reproduces and explains ear-
B lier demonstrations of the lack of effectiveness of some Ca2
V(mV)
10 channel blockers to block bursting despite the fact that Ca2
channels do play a pivotal role in producing the bursts under
I(pA) 5 normal conditions. The results also explain the effectiveness of

200
100 TTX as a blocker of bursting. Even at concentrations too low
to eliminate the rst spike, the threshold for the second spike
100 200 was raised to a point that the ADP no longer triggers addi-

5 tional spikes (Jung et al., 2001).

10
5.4.6 Membrane Potential Oscillations

15
Another important feature of pyramidal neurons in the
C subiculum is their ability to generate subthreshold membrane
potential oscillations in the theta frequency range of 4 to 9 Hz
(Mattia et al., 1997b). These oscillations increase in amplitude
as the membrane potential is depolarized above rest, reaching
Figure 512. Subiculum passive properties and sag. A. Voltage a peak amplitude of about 3 mV just below the action poten-
responses to step current injections from 200 to 300 pA in tial threshold. If the membrane is depolarized further, clusters
50-pA increments (250 pA response not shown). Bar  10 mV, of action potentials re at a rate of about 1 to 2 Hz.
100 ms. B. Voltagecurrent plot of the steady-state responses shown Subthreshold theta oscillations in subicular pyramidal cells
in A. Solid lines are the points used for linear ts to the data. are abolished by the Na-channel blocker TTX but not by
Dashed lines are extrapolations of the ts. C. Effect of 5 mM CsCl blockers of glutamate or GABA receptors or by Ca2 channels,
on the hyperpolarizing voltage response to a current step of 150 suggesting that oscillations are an intrinsic property that
pA. Note the block of sag in the voltage response. Bar  3 mV, 100
depends critically on voltage-gated Na channels (Mattia et
ms. (Source: Adapted from Staff et al., 2000.)
al., 1997b). Similar oscillations have also been observed in
CA1 pyramidal neurons, where they have been shown to arise
from interactions between voltage-gated Na current, K cur-
between bursting and regular-spiking neurons in the subicu-
rent (IM), and Ih (Leung and Yu, 1998; Hu et al., 2002).
lum may be somewhat articial. Another possibility is that
Subicular pyramidal cells also participate in synaptically
subicular pyramidal neurons all have similar basic properties
mediated oscillations in the gamma frequency range of 20 to
(yet distinct from those of CA1 cells), but that they vary
50 Hz. Bursting neurons in subiculum re doublets of action
along a continuum with respect to the strength of the burst-
potentials during these oscillations, which likely promote the
generating mechanism (Staff et al., 2000; Jung et al., 2001).
spread of network activity in the gamma range (Stanford et
al., 1998).
5.4.5 Mechanisms of Bursting

The mechanism responsible for bursting in the subiculum has

been a matter of some controversy. Two early studies identi-
ed a calcium-dependent spike that could contribute to burst- 5.5 Dentate Granule Neurons
ing, possibly as a result of dendritic spike initiation (Stewart
and Wong, 1993; Taube, 1993). Another study, published The main cell type of the dentate gyrus is the granule cell.
about the same time, reported that bursting persisted in the These neurons have been studied extensively, in part because
presence of Ca2 channel block but not Na channel block they were the rst to be shown to exhibit LTP and in part
and concluded that bursting is driven by Na currents (Mattia because they receive spatially segregated synaptic inputs on
et al., 1993; see also Mattia et al., 1997a). These seemingly con- different regions of their dendrites. Granule cells were also the
tradictory ndings were reconciled, to some extent, by a study rst cell type in which it was shown that mRNA could be
showing that Ca2 currents drive bursting, but they are acti- translated into proteins in dendrites, which has signicant
 Hippocampal Neurons: Structure and Function 163

implications for plasticity and learning (Jiang and Schuman, A

2002; Steward and Worley 2002).
EC layer II
5.5.1 Dendritic Morphology and Spines lateral

The principal neurons of the dentate gyrus are distinctly dif- EC layer II
ferent from all of the neurons discussed so far because they are medial s.m.
not pyramidal in morphology. Rather, they comprise a small,
mossy cells
ovoid cell body with a single, approximately conical (but
commisural
somewhat elliptical) dendritic tree (Fig. 513). The granule associational
cells lie in densely packed columnar stacks underneath the rel-
s.g.
atively cell-free molecular layer and the cell-rich polymorphic
layer (also known as the hilus). Typically, granule cell somata hilus
are about 10 m in diameter and about 18 m in length, with B
their long axis oriented perpendicular to the granule cell layer.
Interestingly, the packing density of dentate granule cells and
CA3 pyramidal cells vary along the septo-temporal axis of the
hippocampus but in opposite directions. Thus, the ratio of
granule cells to CA3 cells ranges from about 12:1 at the septal
pole to 2:3 at the temporal pole, implying substantial differ-
ences in connectivity along this axis (Amaral and Witter,
1995).
Granule cell dendrites all extend into the molecular layer
and terminate near the hippocampal ssure. The number of
dendritic branches is highly variable, with the calculated sum
of the dendritic lengths ranging from 2.3 to 4.6 mm. These
dendritic lengths are signicantly shorter than those of the
Figure 513. Dentate granule cell dendritic morphology, spines,
CA1 pyramidal neurons. Dendritic morphology also varies and synaptic inputs and outputs. A. Computer-generated plot of a
depending on the position in the dentate gyrus. Neurons reconstructed granule cell from the suprapyramidal blade of the
located in the suprapyramidal blade (closest to CA1) are gen- dentate gyrus showing the cell body in the stratum granulosum
erally larger (greater total dendritic length, more dendritic (s.g.) and the dendrites in the stratum moleculare (s.m.), with the
segments, and a greater transverse spread) than those in the major inputs indicated. (Source: Adapted from Claiborne et al. 1990,
infrapyramidal blade (Desmond and Levy 1982, 1985; with permission.) B. Camera lucida drawing of the dendritic mor-
Claiborne et al., 1990). phology and axonal arborization of a granule cell in the infrapyra-
Like pyramidal neurons, the dendrites of granule cells are midal blade of the dentate gyrus. The axons are in the molecular
heavily studded with spines. Estimates of spine density indi- layer (ML, same as s.m.). The major targets of the mossy ber axon
(thin structure extending right of the granule layer, GL) are the
cate that there are two to four spines per linear micrometer of
mossy cells and CA3 pyramidal neurons, as well as interneurons in
dendrite. The number of spines increases (within this range)
the hilus (H) and CA3 regions. (Source: Adapted from Lbke et al.,
in progressively more distal dendrites, and there are slightly 1998, with permission. A color version of this gure with the axon
more spines in the suprapyramidal blade than in the in red is available online.)
infrapyramidal blade (Desmond and Levy 1982; Hama et al.,
1989). The available data suggest that a granule cell located
in the suprapyramidal layer would have close to 5600 spines, third receives input from relatively medial regions of the
whereas a cell located in the infrapyramidal cell layer would entorhinal cortex, and the distal third receives input from the
have signicantly fewer spines, at about 3600. These values lateral entorhinal cortex (Blackstad 1958; Amaral and Witter,
are considerably lower than the spine densities and total 1995). Inputs to the dentate gyrus from layer II of the entorhi-
spine numbers calculated for both CA1 and CA3 pyramidal nal cortex arrive via the perforant path. The spatial separation
neurons. of lateral and medial perforant path inputs was elegantly
exploited in early studies that described the cooperative and
5.5.2 Excitatory and Inhibitory Synaptic Inputs associative properties of LTP (McNaughton et al., 1978; Levy
and Steward 1979) (see Chapter 10). Both inputs are primarily
Dentate granule cells receive synaptic input from several glutamatergic, although the medial perforant path inputs also
sources. Granule cell dendrites are generally divided into contain cholecystokinin, and the lateral perforant path con-
thirds according to the differential inputs they receive. The tains enkephalin (Amaral and Witter, 1995). Several differ-
proximal third (closest to the cell body) receives input from ences between synaptic transmission at the medial and lateral
the commissural/associational bers, which consist primarily perforant path have been reported, including differences in
of axons from the mossy cells (see Section 5.6). The middle NMDA receptor activation as well as modulation by choliner-
164 The Hippocampus Book

gic, adrenergic, and metabotropic glutamate receptor activa- mossy terminals along a single axon is about 10 in the hilar
tion (Abraham and McNaughton, 1984; Kahle and Cotman, region and about 12 (range 1018) in the CA3 region, with the
1989; Dahl et al., 1990; Pelletier et al., 1994; Macek et al., 1996). terminals being distributed more or less evenly along the
The dentate gyrus receives several neuromodulatory mossy ber axon. Mossy cells constitute the principal target of
inputs, including diffuse projections from the septal nuclei the large mossy terminals in the hilus (see Section 5.6), and
(acetylcholine), locus coeruleus (norepinephrine), and ventral the small lopodial and en passant terminals selectively inner-
tegmental area (dopamine). The supramammillary region of vate only inhibitory interneuron targets (Acsdy et al., 1998).
the hypothalamus (possibly containing substance P) projects These terminals form only single, often perforated, asymmet-
to the proximal third of the molecular layer, and the raphe rical synapses on the cell bodies, dendrites, and spines of
nuclei (serotonin) project primarily to the hilus (Amaral and GABAergic inhibitory interneurons. This anatomical arrange-
Witter, 1995). ment stands in marked contrast to the large mossy ber ter-
Granule cells also receive inputs from a variety of interneu- minals, which contain numerous release sites housing
rons, mostly lying in the hilus. Basket cells and chandelier cells thousands of both small, clear vesicles and large, dense-core
target primarily somata, whereas various interneurons target vesicles. The mossy bers are glutamatergic but also contain
the dendrites. In addition, both the commissural/associational GABA (Sandler and Smith, 1991; Sloviter et al., 1996; Walker
and perforant path inputs contain minor GABAergic compo- et al., 2001, 2002a) and the opiate peptide dynorphin
nents (Amaral and Witter, 1995). Of particular interest, (McGinty et al., 1983). Each mossy ber bouton contains
almost all of the serotonergic input to the hilus terminates on smooth endoplasmic reticulum and a large number of mito-
GABAergic interneurons that project to the distal dendrites of chondria (Chicurel and Harris, 1992). The main body of the
granule cells (Halasy et al., 1992). large mossy bouton envelops the thorny excrescences of the
CA3 pyramidal neuron, as described above.
5.5.3 Axon Morphology and Synaptic Targets Of particular interest, the lopodial extensions and small,
en passant synapses outnumber the large mossy ber termi-
The somata of granule cells taper at their base, where a single nals by about 10-fold, raising the intriguing idea that the pri-
unmyelinated axon emerges and is directed toward the hilus. mary targets of dentate gyrus granule cells are inhibitory
Ramon y Cajal termed the axons of granule cells mossy interneurons (Acsdy et al., 1998). This anatomical segrega-
bers for their anatomical similarity to the bers in the cere- tion of terminal types is unprecedented throughout the mam-
bellum. The large varicosities and lopodial extensions are malian CNS and suggests that it may provide functional
reminiscent of moss, thus providing the name for these path- specialization of synaptic output.
ways. Anatomically, as far as it is known these axons are Electrophysiological experiments have largely conrmed
unique among hippocampal and cortical neurons (and indeed the hypothesis that the large and small synaptic specializations
throughout the mammalian CNS) in that they form anatom- of the mossy bers are functionally distinct. Individual mossy
ically specialized synapses depending on the nature of their ber release sites onto principal cells have a low initial release
postsynaptic targets. probability, which allows a high degree of short-term and
Each granule cell gives rise to a single unmyelinated axonal frequency-dependent facilitation (Jonas et al., 1993; Salin et
ber about 0.2 m in diameter, with a total length (including al., 1996; Toth et al., 2000; Lawrence et al., 2004). In contrast,
collaterals) of more than 3 mm (Claiborne et al., 1986; mossy ber synapses onto interneurons in the stratum
Frotscher et al., 1991; Acsady et al., 1998). The axons form lucidum demonstrate either mild facilitation or depression in
numerous collaterals throughout the hilus that innervate the response to brief stimulus trains (Toth et al., 2000). In addi-
numerous cell types present in the hilar subeld before pro- tion, a form of NMDA receptor-independent LTP observed at
jecting to the apical dendrites and in some cases basal den- mossy ber synapses onto principal cells (which is thought to
drites of CA3 pyramidal neurons (Claiborne et al., 1986). The be expressed entirely in the presynaptic terminal) is absent at
mossy-ber projection to CA3 forms a tight bundle running mossy ber to interneuron synapses (Maccaferri et al., 1998)
parallel to the CA3 pyramidal cell body layer in a suprapyra- (see Chapter 10). Instead, paradigms that induce LTP at prin-
midal region termed the stratum lucidum because of its cipal cell synapses induce two forms of long-term depression
appearance at the light microscopic level. This region corre- at interneuron synapses (Lei and McBain, 2002, 2004).
sponds to approximately the rst 100 m of the CA3 pyram-
idal neuron apical dendrite. After leaving the hilar region, the 5.5.4 Resting Potential and Action
mossy ber axon contains no further branch points. Potential Firing Properties
Three basic types of mossy ber terminal exist along the
entire length of the main axon: large mossy terminals (410 The resting membrane potential of granule cells is typically
m), lopodial extensions that project from the large mossy more hyperpolarized than that of either CA1 or CA3 pyrami-
boutons (0.52.0. m), and small en passant varicosities dal neurons, being close to 84 mV, based on recordings made
(0.52.0 m) (Amaral, 1979; Claiborne et al., 1986; Frotscher in vivo (Penttonen et al., 1997) and in vitro brain slices
et al., 1991; Acsady et al., 1998). The latter two small terminals (Spruston and Johnston, 1992; Staley et al., 1992).
are somewhat larger than varicosities in axons of CA3 pyram- Measurements of the action potential threshold have provided
idal neurons (Acsady et al., 1998). The total number of large numbers similar to those observed for both CA1 and CA3
 Hippocampal Neurons: Structure and Function 165

pyramidal cells: 45 mV. Given the relatively negative resting

membrane potential of granule cells, larger depolarizations A
are required to bring single granule cells to threshold for
action potential generation. These properties likely underlie
the low spontaneous ring frequencies associated with gran-
ule cells in vivo (Penttonen et al., 1997). Surprisingly, little
accurate information exists regarding granule cell ring pat-
terns in vivo. This is due in part to the difficulty associated
with unambiguously identifying granule cell activity during 20 mV
extracellular recordings. During behavioral tasks, granule cell
ring frequencies are very low, except when the animal is in a 200 ms
place eld (Jung and McNaughton, 1993; Skaggs et al., 1996)
and during odor-guided tasks (Wiebe and Staubli, 1999) (see
Chapter 11).
Upon reaching threshold for action potential initiation, in V(mV)
B 20
response to long depolarizing pulses granule cells typically re
trains of action potentials that demonstrate marked accom-
modation (Staley et al., 1992; Penttonen et al., 1997). As with 10
pyramidal cells, spike accommodation is largely due to the
activation of Ca2-dependent, voltage-gated K conduc-
50
30
10 I(pA)
tances; addition of Ca2 buffers to intracellular recording
electrodes removes such accommodation (Staley et al., 1992). 10 30 50 70

5.5.5 Resting Membrane Properties

10

At the resting membrane potential granule cells have input

resistances considerably higher than those of both CA1 and
20
CA3 pyramidal cells, with estimates ranging from 230 to 450 Figure 514. Dentate granule cell passive and active properties. A.
M (Spruston and Johnston, 1992; Staley et al., 1992) (Fig. Voltage responses to current injections of 50 to 80 pA in 10-pA
514) and a m between 30 and 50 ms. These parameters are increments. B. Voltage-current plot of the data in A. The line repre-
also developmentally regulated, a factor that is especially per- sents an extrapolated linear t to the steady-state voltage responses
tinent because granule cells continue to be born and develop to 20 to 20 pA current injections. (Source: Adapted from Lbke
in adults (Spigelman et al., 1992; Gould and Cameron, 1996; et al., 1998, with permission.)
Liu et al., 1996). Lower values are obtained with sharp micro-
electrode recordings, but these values are likely to be affected
by the microelectrode-induced leak (Spruston and Johnston, ssure) as others in the same cell. Because of this dendritic
1992; Staley et al., 1992). Leak-induced effects are likely to be geometry, it is reasonable to consider an equivalent cylinder
larger in granule cells than in pyramidal cells because the model of granule cell dendrites (Durand et al., 1983; Turner
smaller cell body and thinner dendrites mean that a leak con- and Schwartzkroin, 1983; Desmond and Levy, 1984; Turner,
stitutes a larger fraction of the total conductance. 1984). Unfortunately, the methods for determining electro-
As in pyramidal neurons, estimates of passive membrane tonic length (L) from the time course of somatic voltage tran-
properties are highly nonlinear across small deviations from sients (applied to granule cells and others by several
the resting membrane potential (Spruston and Johnston, investigators during the 1980s) are limited by the existence of
1992), indicating that voltage-dependent conductances close a soma and further complicated by the presence of a micro-
to the resting membrane potential contribute to their passive electrode leak (Rall, 1969). Furthermore, anatomical data
membrane properties. Blockers of Ih have little effect on the alone cannot be used to determine L because an important
passive membrane properties of granule cells (Spruston et al., parameter, the intracellular resistivity (Ri) is not known for
1992), thus implicating other conductances as inuencing the granule cells. The effects of a microelectrode leak on RN and
resting membrane properties. the apparent m in granule cells, however, are strongly depend-
Granule cells have a relatively simple dendritic branching ent on L. Using this approach, the electrotonic length of gran-
pattern. Their dendrites branch rather symmetrically, with ule cells in the suprapyramidal blade of adult guinea pig slices
each of the daughter branches being smaller in diameter than has been estimated at 0.74 (Spruston and Johnston, 1992). In
the parent dendrite. In fact, the branch points in granule cell an equivalent cylinder of this length, a voltage imposed at the
dendrites come close to obeying the assumptions necessary to soma would be expected to attenuate only 22% by the end of
collapse the dendritic tree to an equivalent cylinder. In addi- the dendritic tree (Rall, 1959). Similar attenuation estimates
tion, each dendritic branch has similar geometry and termi- were obtained independently using computer models
nates at approximately the same distance (the hippocampal (Carnevale et al., 1997).
166 The Hippocampus Book

Based on these analyses, one might be tempted to conclude anemone toxin BDS-I, properties consistent with it being car-
that dentate granule cells are electrotonically compact. ried by channels containing the Kv3.4 subunit. In granule cells
However, steady-state voltage attenuation is much less than this current has its highest density in nucleated patches excised
the attenuation of transient potentials, such as EPSPs. from the basal end of the granule cell body, consistent with its
Furthermore, in the voltage-clamp condition, space-clamp immunohistochemical detection in the axons of granule cells.
errors are compounded because they are bidirectional In a technically superb series of experiments, Geiger and
(synapse to soma and back) and because escape of the den- Jonas (2000) provided direct evidence for voltage-gated ion
dritic potential from the command value introduces addi- channels in the presynaptic mossy ber terminals of granule
tional errors (Spruston et al., 1993). For these reasons, it may cells (Fig. 515). In whole-cell recordings made directly from
be argued that the descriptor electrotonically compact is at visually identied large mossy ber boutons, action potential
best misleading and at worst false for most neurons and waveforms were observed to be remarkably brief (half dura-
experimental conditions. The term is best used as a compara- tion 380 s) compared to action potentials measured at the
tor; indeed, computer models predict that granule cells are soma (half duration about 600 s). Action potentials meas-
more electrotonically compact than most pyramidal neurons ured at the soma were typically followed by a prominent after-
(Carnevale et al., 1997). depolarization, whereas those at the mossy ber terminal had
Despite the severity of EPSP attenuation, distal dendritic a brief after-hyperpolarization followed by an after-depolar-
synapses on granule cells are likely to be more effective than ization. Action potentials at mossy ber terminals were
synapses at similar distances on pyramidal cells. The reason is dynamically modulated during brief periods of high-
the absence of a large, main apical dendrite and the smaller frequency activity, with appreciable spike broadening occur-
soma in granule cells. A xed current entering a synapse on a ring as a result of a decrease in the rate of repolarization. In
granule cell dendrite is not subject to as much attenuation as contrast, only modest spike broadening was observed in
in pyramidal cell dendrites and can therefore generate a sig- similar experiments recorded at the granule cell soma, indi-
nicant somatic EPSP even after its attenuation in the den- cating differential expression of voltage-gated currents in the
dritic tree (Jaffe and Carnevale, 1999). In fact, because the soma and the axon. Consistent with this hypothesis, voltage-
diameter of granule-cell dendrites decreases with distance gated K currents in axon terminals inactivated rapidly
from the soma, the local depolarization from a xed somatic but recovered from inactivation very slowly, suggesting
current increases with distance from the soma in a way that that cumulative channel inactivation mediated the activity-
largely cancels the effect of the increased voltage attenuation. dependent spike broadening. Prolongation of the presynaptic
Thus, synapses are at least partially normalized for their posi- spike waveform directly increased the amount of calcium
tion on the dendritic tree. This effect has been called passive that entered the presynaptic terminal during action poten-
normalization of synaptic integration because the mecha- tial activity, resulting in a signicant increase in evoked
nism depends entirely on dendritic geometry and does not synaptic currents. Pharmacological identification of the
require any scaling of synaptic conductances with distance or responsible conductance implicated channels formed by
interaction with voltage-gated conductances in the dendritic Kv1.1/Kv1.4 or Kv1.1/ -subunit combinations (Geiger and
tree (Jaffe and Carnevale, 1999). Passive normalization has Jonas, 2000).
been shown to be plausible in other cell types that lack a large Propagation of action potentials into the granule cell den-
primary apical dendrite, including CA3 pyramidal neurons dritic structure has been much less studied than in pyramidal
and interneurons. In CA1 neurons, passive normalization neurons. This is primarily due to the small diameter of
does not occur in the apical dendritic tree because of the long granule-cell dendrites, which has made dendritic recording
primary dendrite, but it could occur (at least partially) in the extremely difficult. Dendritic eld potential recordings and
basal dendritic tree, which lacks a primary dendrite (Jaffe and current-source density analysis, however, have indicated that
Carnevale, 1999). These differences highlight the importance action potentials spread actively from an initiation zone near
of dendritic architecture in synaptic integration and the func- the soma into the dendrites (Jeffreys, 1979).
tional relevance of variability of dendritic structure across the
various regions of the hippocampus.

5.5.6 Active Properties of Granule Cells 5.6 Mossy Cells in the Hilus

Similar to pyramidal neurons, granule cells have a wide reper- Numerous cell types are present in the polymorphic layer
toire of voltage-gated currents. Granule cells express two (hilus) between the dentate granule cells and the CA3 pyram-
forms of the voltage-gated, transient K current (Riazanski et idal neurons (Amaral, 1978). The most prominent and well
al., 2001). One of these currents is indistinguishable from the studied principal neurons in this region are the mossy cells.
transient currents expressed in most neurons, which have a These cells have been studied quite extensively because of the
low threshold for inactivation and block by 4-AP. The second observation that they are among the rst neurons to degener-
current, however, activates at more depolarized potentials and ate during epilepsy and ischemia (Sloviter, 1987, 1991; Hsu
is sensitive to low micromolar concentrations of TEA and the and Buzski, 1993; but see Ratzliff et al., 2002).
 Hippocampal Neurons: Structure and Function 167

Ba Bb
A

half-duration
MFB 350 ms

-80 mV

100 mV
Ca Cb
GC soma half-duration
710 ms

-80 mV

20 ms 2 ms

Da Db
100 Hz

half-duration of APs
(normalized)
50 Hz
100th AP
30 Hz
50th AP
1st AP 10 Hz

Ea 1 ms
Eb
number of AP within train

100 pA
30 ms 100 pA 50 ms

Figure 515. Dentate granule cell mossy ber terminals. A. somata, which had a signicantly longer half duration and were
Ultrastructure of the mossy ber synaptic complex. Upper panel: followed by a more pronounced afterdepolarization. D. Action
Electron micrograph through a mossy ber synaptic complex. The potentials in presynaptic mossy ber boutons show marked
postsynaptic CA3 pyramidal cell dendrite is indicated (DEN) to the activity-dependent broadening. Trains of action potentials
left of the presynaptic mossy ber terminal. The presynaptic termi- were evoked at 50 Hz. Da. Every 50th action potential is shown
nal contains numerous small vesicles and mitochondria. The post- superimposed with the rst action potential in the train. Db. Plot
synaptic thorny excrescence is indicated (S) where it penetrates the of the action potential duration at half-maximal amplitude in whole
terminal. Several symmetrical junctions between the dendritic shaft cell recorded mossy ber boutons against the number of action
and the terminal are observed (arrowheads) and are representative potentials in the train for the frequencies indicated. E. Frequency-
of puncta adherentia. Note that these symmetrical junctions do not dependent spike broadening is explained in part by the biophysical
have any vesicles nearby. In contrast, asymmetrical junctions properties of voltage-gated outward potassium currents present in
(arrows) onto the thorn show a signicant clustering of vesicles. the presynaptic mossy ber bouton. Ea. K currents activated in a
Lower panel: Representation of a mossy ber synaptic complex mossy ber bouton outside-out patch by test pulses between 70
depicting the various features observed at the electron microscopic and 20mV. Eb. Cumulative inactivation of K channels induced
level. (Source: Adapted from Amaral and Dent, 1981; from Henze et by repetitive activation. Traces are K currents activated by a series
al., 2000.) B, C. Action potentials recorded from a presynaptic of test pulses to 20 mV (3 ms duration, 7 ms interpulse interval).
mossy ber bouton (MFB, panel B) and in a granule cell soma (GC, Such cumulative inactivation plays a signicant role in shaping the
panel C) evoked by a 1 ms current pulse and shown at different action potential trajectory in response to high frequency stimula-
time bases. The properties of action potentials in the presynaptic tion. (Source: BE: Geiger and Jonas, 2000.)
mossy ber bouton differ markedly from those in the granule cell
168 The Hippocampus Book

5.6.1 Dendritic Morphology and Spines form synapses in the hilus and the granule cell layer
(Buckmaster et al., 1996). The axon terminals in the inner
Mossy cells are located primarily in a C-shaped region of the molecular layer form asymmetrical synaptic contacts, prima-
hilus directly below the granule cell layer (zone 4 of Amaral, rily with granule cell dendrites. In the hilus, mossy cells form
1978). They consist of ovoid cell bodies that give rise to sev- asymmetrical synapses on GABA-positive dendritic shafts
eral primary dendrites (Fig. 516A,B). Each of these dendrites and GABA-negative dendritic spines (Wenzel et al., 1997).
branches three or four times while remaining roughly parallel Mossy cells stain positive for glutamate and negative for
to the granule cell layer and with almost all branches remain- GABA, suggesting an excitatory function (Soriano and
ing within the hilus (Amaral, 1978; Ribak et al., 1985; Frotscher, 1994; Wenzel et al., 1997), but may also contain
Buckmaster et al., 1992; but see Frotscher et al., 1991 and peptide neurotransmitters (Fredens et al., 1987; Freund et al.,
Scharfman, 1991). 1997).
The mossy cells are named for the mossy appearance cre- Two particularly important features of the mossy cell axon
ated by the numerous thorny excrescences covering their projection are that it is bilateral and it has an unusually exten-
proximal dendrites within 40 to 80 m from the soma sive longitudinal component (Soltesz et al., 1993; Buckmaster
(Amaral, 1978; Ribak et al., 1985; Frotscher et al., 1991; Lbke et al., 1996) (Fig. 516C). In fact, each mossy cell forms synap-
et al., 1998) (Fig. 516B). These excrescences are similar, tic contacts spanning more than half the length of the hip-
though not identical, to the thorny excrescences on CA3 pocampus, with most of the synaptic contacts formed by
pyramidal neurons. The major difference is a paucity of necks distal segments of the axon. These features imply that mossy
on the spines, with each excrescence formed by a cluster of cells allow granule cell activity to spread longitudinally and
spine heads (Amaral, 1978). The remaining portions of mossy bilaterally in the dentate gyrus. Like the recurrent collaterals
cell dendrites are covered in more conventional spines. of the CA3 network, the positive feedback network formed by
mossy cells has been suggested to contribute to the ability of
5.6.2 Excitatory and Inhibitory Synaptic Inputs the hippocampus to encode memories via an autoassociative
network (Buckmaster and Schwartzkroin, 1994).
Mossy cells receive direct excitatory input from the granule
cell mossy bers, which contact the thorny excrescences and at 5.6.4 Resting and Active Properties
least some of the conventional spines on more distal dendrites
(Frotscher et al., 1991; Buckmaster et al., 1992; Scharfman, Mossy cells in vitro have resting potentials of 61 to 65 mV,
1994). An additional excitatory input arises from a subset of RN of 67 to 200 M, and m of 16 to 41 ms (Fig. 517).
CA3 pyramidal neurons that send axon collaterals back These ranges represent the mean values taken from
into the hilus (Ishizuka et al., 1990; Scharfman, 1994). sharp-microelectrode recordings and patch-clamp recordings,
Intracellular recordings have revealed that mossy cells receive respectively (Scharfman and Schwartzkroin, 1988;
frequent, often large, spontaneous EPSPs (Scharfman and Buckmaster et al., 1993; Lbke et al., 1998). Mossy cells can
Schwartzkroin, 1988). Consistent with the large size of these re up to about 50 spikes during a 1-second current injection.
spontaneous EPSPs, kinetic analysis has revealed that sponta- Maximum spike rates can be higher, however, as mossy cells
neous EPSCs in mossy cells are slower than in aspiny hilar can exhibit both bursting and spike-frequency accommoda-
interneurons, a difference attributed to different glutamate tion. These physiological features readily distinguish mossy
receptor subtypes on these two classes of neurons (Livsey and cells from granule cells or basket cells of the dentate gyrus, but
Vicini, 1992; Livsey et al., 1993). they are not easily distinguished from other neurons in the
In addition to these powerful excitatory synaptic inputs, hilus (Scharfman, 1992; Lbke et al., 1998). Functionally,
mossy cells receive substantial inhibitory inputs. After block- the most relevant functional features of mossy cells are likely
ing glutamate receptors, a high frequency of spontaneous to be their frequent and large EPSPs, which can trigger bursts
IPSCs are observed in all hilar neurons, including mossy of action potentials (Scharfman and Schwartzkroin, 1988).
cells (Scharfman, 1992; Soltesz and Mody, 1994; Acsady et al., This unusual excitability of mossy cells has been proposed
2000). It is likely that inhibitory synaptic input arrives from to be relevant during normal function and pathology
a variety of sources, including dentate basket cells, aspiny (Sloviter, 1991; Buckmaster and Schwartzkroin, 1994; Ratzliff
hilar interneurons (ipsilaterally and contralaterally), and et al., 2002).
GABAergic input from the medial septum.
5.6.5 Other Spiny Neurons in the Hilus
5.6.3 Axon Morphology and Synaptic Targets
A population of spiny neurons in the hilus that differs from
The mossy cell axon specically targets the inner third of the mossy cells has also been identied. These neurons, fewer in
molecular layer of the dentate gyrus (Ribak et al., 1985; number than mossy cells, are distinguishable by a lack of
Buckmaster et al., 1992). More than two-thirds of mossy thorny excrescences and the presence of axon collaterals that
cell axon collaterals ramify in this region, and the remainder extend to the outer molecular layer of the dentate gyrus
 Hippocampal Neurons: Structure and Function 169

A C

Figure 516. Mossy cell dendritic morphology, spines, and synaptic covering a section of the dendrite. (Source: Adapted from Lbke et
inputs and outputs. A. Camera lucida drawing of a mossy cell with al., 1998, with permission.) C. Three sections showing axon collater-
the cell body in the hilus near the granule layer (GL) and dendrites als from a single labeled mossy cell innervating nearly the full septo-
extending farther into the hilus. (A color version of this gure with temporal extent of the hippocampus. Note that axon collaterals are
the axon in red is available online.) (Source: Adapted from Lbke et restricted to the hilus and inner molecular layer (ml, near the gran-
al. 1998, with permission.) B. Photomicrograph of a lled mossy ule cell layer, gcl). (Source: Adapted from Buckmaster et al., 1996,
cell showing the axon emerging from the cell body (arrow). The with permission.)
inset is a higher magnication view showing thorny excrescences

(Lbke et al., 1998). Although these neurons are spiny, it is not

clear if they form excitatory or inhibitory synapses on their 5.7 Pyramidal and Nonpyramidal Neurons
targets in the molecular layer. They have higher RN than mossy of Entorhinal Cortex
cells (despite similar m), presumably owing to more limited
dendritic arborization. They can also re at higher rates, on The entorhinal cortex is the major gateway between the hip-
average, although the overlap in ranges of these values makes pocampal formation and the rest of the brain. It receives both
it difficult to distinguish them conclusively from mossy cells intrinsic inputs from the hippocampus (primarily subiculum)
or aspiny hilar interneurons on the basis of physiological and pre/parasubiculum as well as extrinsic inputs from
properties alone (Lbke et al., 1998). numerous cortical areas. Various neurons in the entorhinal
170 The Hippocampus Book

layers (IV-VI) receive extensive input from the hippocampus

A (subiculum) and presubiculum. In addition to projections
back to other cortical areas, a projection from the deep layers
back to the supercial layers of entorhinal cortex completes
the closed-loop structure of the hippocampal formation
(Chrobak and Buzski, 1994; Chrobak et al., 2000) (see
Chapter 3 for more details). Here we consider three cell classes
about which detailed physiological and morphological infor-
mation is available: stellate cells in layer II, pyramidal cells in
layers II/III, and pyramidal neurons in layers V/VI.

5.7.1 Stellate Cells of Layer II

Stellate cells are the most abundant cell type in layer II and are
especially numerous in the medial entorhinal cortex. They
derive their name from their star-like appearance, with
numerous dendrites radiating out from the soma (Fig.
518A). In a typical stellate cell of rat medial entorhinal cor-
B tex, about ve thick (56 m diameter) primary dendrites
emerge from the cell body (Klink and Alonso, 1997a). In most
cases, one or two of the primary dendrites extend upward
toward layer I, and a few primary dendrites extend downward
toward layer III. Each of these two groups of dendrites
branches extensively, giving rise to bitriangular dendritic
arborization. On average, each primary dendrite gives rise to
15 terminal dendrites (Lingenhhl and Finch, 1991). The
upper dendritic tree (layer I) has a mediolateral expanse (up
to 700 m in the rat) that is greater than that of the lower den-
dritic tree (layer III) and exceeds that of most pyramidal neu-
rons with apical dendrites in layer I (Klink and Alonso,
1997a). The dendrites of stellate cells are uniformly studded
with spines, usually having a long, thin neck and a larger spine
head. Dendritic branches often terminate in bouquets of
about six dendritic spines (Klink and Alonso, 1997a). The
Figure 517. Mossy cell passive and active properties. A. Voltage
responses of a mossy cell (same as Figure 516) to step current dendritic spines presumably form the sites of contact for the
injections of 150 to 150 pA in 50-pA increments and 40 to 40 synaptic inputs reaching layers I to III from other cortical
pA in 10-pA increments. Input resistance is 181 M, and resting areas as well as from the pre/parasubiculum.
potential is 64 mV. B. Voltage-current relation of the steady-state The stellate cell axon usually emerges from a primary den-
() and peak () voltage responses shown in A. The line represents drite, close to the soma. The axon descends to the angular
an extrapolated linear t of the points between 50 and 50 pA. bundle, which eventually forms the perforant path. As
(Source: A, B: Adapted from Lbke et al., 1998, with permission.) described in Chapter 3, axons from stellate cells in the lateral
entorhinal cortex terminate on the distal third of granule cell
cortex provide the major input to the dentate gyrus and the dendrites, whereas those from the medial entorhinal cortex
hippocampus; they receive input back from the hippocampus terminate on the middle third of granule cell dendrites.
and in turn project to other cortical areas, often the same areas Detailed reconstructions have revealed that the axonal
that provide input to the entorhinal cortex. arborization from a single layer II stellate cell is extensive,
The entorhinal cortex is a conventional six-layered struc- encompassing both the suprapyramidal and infrapyramidal
ture comprising several excitatory cell types (Ramon y Cajal, blades of the dentate gyrus and encompassing more than two-
1904; Schwartz and Coleman, 1981; Carboni et al., 1990; thirds of the septo-temporal extent of the dentate gyrus
Germroth et al., 1991; Belichenko, 1993; Mikkonen et al., (Tamamaki and Nojyo, 1993; Tamamaki, 1997; see also Par
2000). Neurons in the supercial layers (II/III) receive most and Llins, 1994). The axons of layer II stellate cells also form
of their input from other cortical areas (largely olfactory synapses on CA3 pyramidal neurons. Importantly, many stel-
in the rodent) as well as from presubiculum and parasubicu- late cell axons do not cross the hippocampal ssure; rather,
lum. These layer II/III neurons then constitute the principal they project along one side of the ssure to CA3 and bend
inputs to the dentate gyrus and hippocampus, and they around the ssure toward the dentate gyrus. The presence of
provide feedback to other cortical areas. Neurons in the deep varicosities on axons in both CA3 and the dentate gyrus sug-
 Hippocampal Neurons: Structure and Function 171

A B

Figure 518. Stellate cells of layer II. A. Camera lucida drawing of a fast AHP, an ADP, and a medium AHP. (Source: Adapted from
spiny stellate cell in layer II of entorhinal cortex showing its den- Dickson et al., 2000a, with permission.) C. Membrane potential
dritic tree, which extends into layers I and III, and the axon emerg- oscillations (1 and 2) and spike clustering (3) in a layer II stellate
ing from the base of the soma (arrows). Bar  40 m. (Source: cell held at membrane potentials positive to 55 mV. Oscillations in
Adapted from Klink and Alonso, 1997a, with permission.) B. 13 are at increasingly depolarized membrane potentials. The auto-
Voltage responses to step current injections as shown. Note the correlation (inset in 1) demonstrates the rhythmicity of the sub-
prominent sag in hyperpolarizing voltage responses and rebound threshold membrane potential oscillations. (Source: Adapted from
ring at the offset of the current injection. The single action poten- Dickson et al., 2000a, with permission.)
tial shown on an expanded time scale at the right (from C2) shows a

gests that axons from individual stellate cells can project to inward rectication properties of stellate cells contribute to a
both areas (Witter, 1989). Axon collaterals also branch and hump at the onset of depolarizing current pulses. Hence,
form synaptic connections within the entorhinal cortex action potential ring occurs near the beginning of long cur-
(supercial and deep layers) and subiculum (Germroth et al., rent injections above threshold. Ih and Na current also con-
1991; Lingenhhl and Finch, 1991; Tamamaki and Nojyo, tribute to a depolarizing hump at the offset of hyperpolarizing
1993; Klink and Alonso, 1997a). current injections, which in some cases produce rebound
In microelectrode recordings from layer II of rat medial spike ring (Fig. 518B). Action potentials in stellate cells are
entorhinal cortex, stellate cells have resting potentials of about followed rst by a fast AHP of about 10 mV, then a depolariz-
62 mV, RN of about 36 M, and m of about 9 ms (Klink and ing after-potential of about 1 mV, and nally a medium AHP
Alonso, 1993; see also Jones, 1994). However, stellate cells also of about 11 mV. In response to long current injections stellate
exhibit active nonlinearities near the resting potential, includ- cells exhibit strong spike frequency accommodation; and a
ing TTX-sensitive amplication of subthreshold depolariza- large, slow AHP occurs following the offset of the current
tions and ZD-72688-sensitive sag of small hyperpolarizing injection (Klink and Alonso, 1993).
and depolarizing responses (Fig. 518B). These two proper- One of the most striking physiological features of layer II
ties, mediated by voltage-gated Na channels and Ih channels, stellate cells in entorhinal cortex is the theta frequency oscilla-
respectively, contribute to pronounced inward rectication tion that occurs during subthreshold depolarizations (Alonso
(the voltage response to current injection is larger for depo- and Llins, 1989). These 5- to 15-Hz oscillations begin upon
larization than for hyperpolarization) in stellate cells (Klink depolarization from rest and reach an amplitude of about 3
and Alonso, 1993). mV when Vm is around 55 mV (Fig. 518C). Further depo-
The action potential threshold is about 52 mV and the larization results in spike ring in a characteristic cluster pat-
action potential duration about 1.2 ms in stellate cells. The tern, with an interspike frequency of about 20 Hz and an
172 The Hippocampus Book

intercluster frequency of 1 to 3 Hz (Alonso and Klink, 1993). Many of the basic properties of layer II pyramidal neurons
Subthreshold theta oscillations in stellate cells are blocked by are similar to those of their stellate counterparts (Vrest
64
either TTX or ZD-72688, indicating that they result from mV, RN 40 M, m 12 ms, threshold
49 mV); both exhibit
interplay between voltage-gated Na current and Ih (Klink spike-frequency accommodation and a prominent TTX-
and Alonso, 1993; Dickson et al., 2000a,b). The Na depend- resistant, persistent Na current, which is also present in the
ence of the subthreshold oscillations in layer II stellate cells dendrites (White et al., 1993; Magistretti et al., 1999) (Fig.
has prompted careful analysis of the voltage-gated Na chan- 519C). However, several physiological properties clearly dis-
nels in these neurons. Na currents in neurons dissociated tinguish pyramidal cells from stellate cells in layer II (Alonso
from layer II have almost 10-fold larger Na conductance than and Klink, 1993). First, the action potential duration is signif-
is seen in cells dissociated from layer III (Fan et al., 1994). icantly longer in pyramidal cells than in stellate cells (1.8 vs.
Furthermore, compared to layer III neurons, Na currents in 1.2 ms at threshold). Second, pyramidal cells exhibit less sag
layer II activate at lower voltages and inactivate at higher volt- than stellate cells, especially in the depolarizing direction. As a
ages, resulting in a substantial window current, a steady-state result, layer II pyramidal neurons do not exhibit rebound
current in the voltage range where activation is signicant but spike ring, even following large hyperpolarizations; and they
inactivation is incomplete (Fan et al., 1994). Biophysical exhibit longer latencies to spiking than do stellate cells (Fig.
analysis has revealed that a high-conductance, persistent Na 519B). Third, pyramidal cells lack the outward rectication
current is responsible for most of the window current, which that is prominent in stellate cells in the presence of TTX.
is likely to produce subthreshold oscillations in layer II stellate Fourth, spike repolarization in pyramidal cells is severely
cells (Magistretti et al., 1999a; Magistretti and Alonso, 1999). impaired in the absence of Ca2 compared to the modest
These currents are present in the dendrites as well as the soma increase in spike duration observed under the same conditions
of layer II neurons (Magistretti et al., 1999b). Although the in stellate cells (Alonso and Klink, 1993; Klink and Alonso,
subthreshold oscillations are blocked by high concentrations 1993). Finally, the membrane potential oscillations generated
of TTX (Klink and Alonso, 1993), the Na channels thought below threshold and the spike clusters occurring above thresh-
to mediate the oscillations have been termed TTX resistant old in stellate cells are absent in layer II pyramidal neurons
because of the relatively high concentration of TTX required (Alonso and Klink, 1993). Activation of muscarinic receptors,
to block Na currents in layer II neurons (White et al., 1993). however, produces depolarization and low-frequency (0.20.5
The membrane properties of layer II stellate cells are also Hz) oscillations of action potential bursting in layer II pyram-
subject to modulation by muscarinic receptor activation idal neurons (Klink and Alonso, 1997b).
(Klink and Alonso, 1997b). Carbachol causes an increase in
input resistance and membrane depolarization, which pro- 5.7.3 Pyramidal Cells of Layer III
duces subthreshold voltage oscillations at a lower frequency
(about 6 Hz) than the depolarization-induced theta oscilla- Intracellular recordings followed by biocytin labeling have
tions that occur in the absence of the agonist. Because the revealed two types of pyramidal neurons in layer III, each with
effects of carbachol are potentiated by Ca2 inux, it has been distinct morphological and physiological properties (Gloveli
suggested that Ca2-activated, nonspecic-cation conduc- et al., 1997a; but see Dickson et al., 1997). These two types of
tance may be activated by muscarinic receptor activation pyramidal neuron were distinguished primarily on the basis
(Klink and Alonso, 1997c; Gloveli et al., 1999; Shalinsky et al., of dendritic spines, which are present on the dendrites of type
2002). 1 but not type 2 layer III pyramidal neurons (Gloveli et al.,
1997a). Both types have pyramidal morphology (Fig. 520A),
5.7.2 Pyramidal Cells of Layer II but the type 1 (spiny) cells cover broader expanses in both the
apical and basal dendritic trees. The apical dendrites of both
Most of the pyramidal neurons in layer II have their cell bod- cell types extend toward the cortical surface, whereas basal
ies in the deepest third of the layer. Their morphology is typ- dendrites remain largely restricted to layer III (Gloveli et al.,
ical of pyramidal neurons, with a somewhat triangular soma, 1997a; but see Dickson et al., 1997).
a thick apical dendrite extending into layer I, and thinner basal The axons of both types of layer III pyramidal neurons
dendrites extending into layer III (Fig. 519A). Both apical could be traced to the angular bundle, but only the axons of
and basal dendrites branch extensively and are studded with the type 1 (spiny) neurons could be followed to CA1 and the
spines at even higher density than in layer II stellate cells subiculum (Gloveli et al., 1997a). Thus, spiny pyramidal neu-
(Klink and Alonso, 1997a). rons give rise to the perforant-path projection to CA1 (or
The axon of layer II pyramidal neurons emerges from the temporo-ammonic path), but the targets of type 2 (aspiny)
base of the soma and extends through layers II and III toward neurons are less clear.
the angular bundle. Extensive axon collaterals are also located Layer III pyramidal neurons are easily distinguished from
within entorhinal cortex (Klink and Alonso, 1997a). The their layer II counterparts by the absence of sag (Dickson et
axons of layer II pyramidal cells, like those of the stellate cells al., 1997; Gloveli et al., 1997a; van der Linden and Lopes da
in the same layer, form the perforant path, which projects to Silva, 1998) (Fig. 520B). Type 1 (spiny) pyramidal neurons of
the dentate gyrus and CA3 (Schwartz and Coleman, 1981). layer III have relatively high RN (70 M) and long m (20 ms),
 Hippocampal Neurons: Structure and Function 173

Figure 519. Pyramidal cells of layer II. A. Camera lucida drawing with permission.) C. Voltage-gated Na channels in the dendrites of
of a spiny pyramidal neuron in layer II of entorhinal cortex showing layer II stellate cells showing transient (left) and persistent (right)
its dendritic morphology, which consists of an apical tree extending gating. Channel activity was elicited by consecutive depolarizing
to the limit of layer I and a basal tree in layers II and III. The axon voltage steps to 30 mV (beginning at arrow) in two cell-attached
(arrows) emerges from the base of the soma. Bar  40 m. patches from the dendrites of isolated layer II pyramidal cells. Bars
(Source: Adapted from Klink and Alonso, 1997a, with permission.)  2 pA, 5 ms (left); 2 pA, 50 ms (right). (Source: Adapted from
B. Voltage responses of a layer II pyramidal neuron to the current Magistretti et al. 1999b, with permission.)
injections shown. (Source: Adapted from Alonso and Klink, 1993,
174 The Hippocampus Book

A B

Figure 520. Pyramidal cells of layer III. A. Camera lucida drawing of a pyramidal neuron in
layer III of medial entorhinal. Bar  200 m. B. Hyperpolarizing and depolarizing voltage
responses to current injections of 300 and 300 pA. Bars  20 mV, 100 ms. (Source: A, B.
Adapted from van der Linden and Lopes da Silva, 1998, with permission.)

whereas type 2 (aspiny) pyramidal neurons have lower RN (30 large pyramidal neurons. Small pyramidal cells have basal
M) and shorter m (8 ms) (Gloveli et al., 1997a; see also dendrites in layers V and VI, whereas large pyramidal cells
Dickson et al., 1997). The resting potentials (about
70 mV) have dendrites that extend horizontally within layer V and
and action potential thresholds (about
55 mV) are similar have therefore been alternately described as horizontal cells
in the two cell types, but type 2 (aspiny) pyramidal cells (Hamam et al., 2002) (Fig. 521A).
showed more spike-frequency accommodation than type 1 In layer V of the lateral entorhinal cortex, both small and
(spiny) cells (Gloveli et al., 1997a). An interesting difference large pyramidal neurons have an average of six to seven pri-
between the pyramidal cells of layers II and III is that layer II mary dendrites: one apical dendrite (average diameter 4.2 m
neurons tend to be activated by synaptic activation at fre- in small and 5.2 m in large pyramidal cells) and three to
quencies above 5 Hz, whereas layer III neurons respond best eight primary basal dendrites (average diameter 1.7 m in
to frequencies below 10 Hz (Gloveli et al., 1997b). small and 3.3 m in large pyramidal cells) for a total dendritic
length of about 8 mm. Polymorphic cells also have an average
5.7.4 Pyramidal Cells of Deep Layers of six to seven primary dendrites with an average diameter of
3.4 m and a total dendritic length of about 10 mm (Hamam
Layer III of the entorhinal cortex is separated from the next et al., 2002; see also Lingenhhl and Finch, 1991; Hamam et
layer of cells by a cell-free zone called lamina dissecans. In al., 2000).
Cajals nomenclature, this acellular layer is synonymous with Spines are present on the dendrites of all deep-layer
layer IV, in which case there are neurons in layers V and VI. In pyramidal neurons and polymorphic cells (Lingenhhl and
some other studies, however, the acellular layer is considered Finch, 1991; Hamam et al., 2000, 2002). The synaptic inputs to
separately and the cellular layers are divided into layers IV to these principal neurons presumably originate from multiple
VI (see Chapter 3). At least two-thirds of the projecting neu- sources, including other cortical areas, presubiculum, para-
rons in the deep layers of entorhinal cortex are pyramidal neu- subiculum, and subiculum (Amaral and Witter, 1989; Lopes
rons, and the rest are polymorphic or multipolar neurons da Silva et al., 1990) (see Chapter 3).
(Lorente de No, 1933; Lingenhhl and Finch, 1991; Hamam et The axons of deep-layer pyramidal and nonpyramidal pro-
al., 2000, 2002; Egorov et al., 2002a). jecting neurons of entorhinal cortex extend into the angular
Although the pyramidal neurons and polymorphic neu- bundle, from which they project to other cortical areas,
rons comprise heterogeneous groups of neurons, these groups including many of the same regions that provide input to
are clearly distinguished by their dendritic architecture. the supercial layers of entorhinal cortex (see Chapter 3).
Pyramidal neurons consist of a large apical dendrite extending There is also considerable evidence, however, that deep-layer
to layer I or II and multiple basal dendrites with variable dis- neurons project back to the dentate gyrus and hippocampus
tributions, whereas polymorphic neurons consist of numer- as part of the perforant path (Swanson and Cowan, 1977;
ous dendrites radiating in all directions from the cell body. Khler, 1985; Deller et al., 1996b; Gloveli et al., 2001). In addi-
Pyramidal cells have been further subdivided into small and tion to entering the angular bundle, axons of both pyramidal
 Hippocampal Neurons: Structure and Function 175

and nonpyramidal neurons in the deep layers of the entorhi- Intracellular microelectrode recordings from deep-layer
nal cortex give off numerous branches, which are abundant in neurons in entorhinal cortex slices indicate that the basic elec-
layers V and VI but in some cases also extend into the super- trophysiological properties of small pyramidal cells, large
cial layers (Gloveli et al. 2001; Hamam et al., 2000, 2002). pyramidal cells (horizontal cells), and polymorphic cells are
Collaterals in layers V and VI form part of an associational heterogeneous but not statistically different among the mor-
system, which may contribute to the susceptibility of the phological groups (Hamam et al., 2000, 2002). Average resting
entorhinal cortex to epileptic activity (Jones and Heinemann, potentials were
62 to
65 mV, RN values were 50 to 80 M,
1988). Collaterals to the supercial layers complete a closed and m values were 11 to 15 ms. Similar values have been
loop from layers II and III to the hippocampus, to layers V and reported in other studies (Jones and Heinemann, 1988;
VI, and back to layers II and III (Chrobak and Buzski, 1994; Egorov et al., 2002a). Voltage-current plots revealed inward
Chrobak et al., 2000). rectication, in most cases consistent with a time-dependent

Figure 521. Pyramidal cells of layer VVI. A. Reconstructions of muscarinic AChR agonist carbachol (10 M). Note the progressive
three subtypes of neuron in layer V of lateral entorhinal cortex. Bar increase in persistent action potential ring that continues after
 100 m. (Source: Adapted from Hamam et al., 2002, with per- each current injection is terminated. Each depolarizing current step
mission.) B. Responses of a layer V pyramidal neuron to successive is 4 seconds in duration. (Source: Adapted from Egorov et al., 2002b,
current injections (300 pA, 4 seconds each) in the presence of the with permission.)

20 sec
176 The Hippocampus Book

voltage sag, presumably due to activation of Ih (Hamam et al., anatomical transition zone between the three-layered hip-
2000, 2002). In pyramidal neurons, sag was much smaller than pocampus and the six-layered parasubiculum. The second
reported for layer II stellate cells (Klink and Alonso, 1993), reason for considering the pre- and parasubiculum separately
and one study reports that sag is absent from nonpyramidal from the subiculum is that these regions have distinctly differ-
neurons in the deep layers of the entorhinal cortex (Egorov et ent physiological properties (see below).
al., 2002a). Action potentials were about 65 mV in amplitude The major projection from the presubiculum is to layer III
(threshold to peak) followed by an AHP of about 17 mV. of the medial entorhinal cortex, whereas the major projection
Polymorphic neurons exhibit a voltage-dependent delay to from the parasubiculum is to layer II of the medial and lateral
spike ring, which is mediated by a D-type K current entorhinal cortex. Both pre- and parasubiculum also project
(Egorov et al., 2002a). Repetitive ring was largely regular weakly back to the stratum lacunosum-moleculare of the hip-
spiking with some spike-frequency accommodation, but burst pocampus and to the supercial molecular layer of the dentate
ring was not observed (Hamam et al., 2000, 2002; Egorov et gyrus. Both regions also have numerous other cortical and
al., 2002a; but see Jones and Heinemann, 1988). subcortical inputs and outputs (see Chapter 3).
Another important characteristic of pyramidal neurons in Microelectrode recordings have been used to study the
the deep layers of the entorhinal cortex is that most exhibit electrophysiological properties of neurons in the pre- and
membrane-potential oscillations in the theta range when held parasubiculum (Funahashi and Stewart, 1997a). Despite the
just below threshold for action potential ring (Dickson et al., fact that pyramidal and nonpyramidal (stellate) neurons are
2000b). These oscillations have amplitudes of 2 to 7 mV and present in all layers of the cortex, the basic resting and active
frequencies of 5 to 15 Hz (Dickson et al., 2000b; Hamam et al., properties of neurons in the pre- and parasubiculum are
2000, 2002). Like the prominent oscillations in stellate cells of remarkably homogeneous, with resting potentials in the range
layer II, the oscillations in layer V neurons are TTX-sensitive of
60 to
68 mV, RN of 42 to 84 M, and m of 8 to 12 ms.
and therefore require voltage-gated Na current. An impor- A modest degree of membrane sag was observed in supercial
tant difference, however, is that the oscillations in layer V do neurons, while sag was absent in deep-layer pyramidal and
not seem to require Ih, as pyramidal cells lacking voltage sag stellate neurons (Fig. 522B). In this respect, the properties of
still exhibit prominent theta oscillations (Dickson et al., these neurons are functionally quite distinct from their neigh-
2000b). The relation between intracellular theta oscillations bors in the subiculum and entorhinal cortex.
and the network theta, gamma, and ripple (sharp-wave) oscil- Another major difference between neurons in the pre- and
lations observed in the entorhinal cortex during various parasubiculum and the subiculum is the absence of intrinsic
behavioral states remains unclear (Chrobak and Buzski, burst ring. In response to depolarizing current injection,
1994, 1996; Buzski, 1996). neurons in these regions were found to re only regular trains
Perhaps the most intriguing property of layer V pyramidal of action potentials (Funahashi and Stewart, 1997a). Despite
cells in the entorhinal cortex is their ability to integrate neural the absence of intrinsic burst ring, neurons in the deep lay-
activity over long periods of time. During activation of mus- ers of the pre- and parasubiculum exhibit synaptically driven
carinic acetylcholine receptors, stimuli lasting a few seconds burst discharges (Fig. 522C). In response to brief stimulation
can lead to sustained action potential ring for several min- of either the subiculum or the deep medial entorhinal cortex,
utes (Egorov et al., 2002b). Furthermore, repeated stimulation deep-layer neurons in the pre- and parasubiculum depolarize
during cholinergic activation leads to sustained ring at and re for hundreds of milliseconds (Funahashi and Stewart,
higher frequencies (Fig. 521B). This unique form of neural 1997b). These bursts are driven by giant, glutamatergic EPSPs
integration, which is mediated by a calcium-activated, non- and may be sustained by reciprocal connections with intrinsi-
specic cation current, has been proposed to contribute to the cally bursting neurons in the subiculum (Funahashi et al.,
function of the entorhinal cortex in working and long-term 1999). Neurons in the supercial layers of the pre- and para-
memory (Egorov et al., 2002b). subiculum respond to synaptic stimulation of the subiculum
or entorhinal cortex with EPSPs and can produce synaptically
driven bursts. The supercial neurons send axon collaterals to
the deep layers of the pre- and parasubiculum, but reciprocal
5.8 Pyramidal and Nonpyramidal Neurons connections are absent (Funahashi and Stewart, 1997b).
of Presubiculum and Parasubiculum

The presubiculum and parasubiculum are considered sepa-

rately from the subiculum for two important reasons. First, 5.9 Local Circuit Inhibitory Interneurons
these regions are distinct from the subiculum in that they are
multilaminate (Fig. 522A). The parasubiculum has a conven- Whereas most principal (i.e., glutamatergic) cell types of the
tional six-layered cortical structure. Presubiculum has also hippocampus have their cell bodies organized into highly
been described as a six-layered cortex, but the layers are not as structured lamina (e.g., CA3/CA1 stratum pyramidale), the
well dened, so it may be more appropriately considered an somata of local circuit GABAergic inhibitory interneurons
 Hippocampal Neurons: Structure and Function 177

m
lu
icu
b
su
ra
aP
m
iculu
rPesub

B layer II stellate cell layer II pyramidal cell

layer V stellate cell layer V pyramidal cell

Figure 522. Pyramidal and stellate cells of the presubiculum and various neurons to current injections (approximately 1.5 to 0.5
parasubiculum. A. Camera lucida drawings of several neurons in nA). Note the similarity of response characteristics of different
various layers of presubiculum and parasubiculum. A, pyramidal neurons within the same layers but differences in neurons between
cell in layer II of parasubiculum; B, stellate cell in layer II of pre- layers. C. Responses of a layer V neuron in the parasubiculum to
subiculum; C, pyramidal cell in layer III of presubiculum; D, synaptic stimulation of the entorhinal cortex or subiculum. Note
stellate cell in layer V of presubiculum; E, pyramidal cell in layer V also the burst-ring responses despite the absence of burst ring in
of presubiculum; F, pyramidal cell in layer V parasubiculum; G, response to current injections (B). (Source: AC. Adapted from
stellate cell in layer V of parasubiculum. B. Voltage responses of Funahashi and Stewart, 1997a, with permission.)
178 The Hippocampus Book

show no such apparent organization. In fact, the somata of inhibitory interneurons have traditionally been considered as
this highly diverse population of neurons are scattered little more than the regulators of principal neuron activity
throughout almost all subelds and strata of the hippocam- the yin to the excitatory yang. Evidence suggests, however, that
pus (Fig. 523). Moreover, despite representing only ~10% of in addition to that role their network connectivity and prop-
the total hippocampal neuronal population, inhibitory erties of intrinsic voltage-gated currents are nely tuned to
interneurons represent perhaps one of the most diverse cell permit inhibitory interneurons to generate and control the
populations, a claim based not only on their highly divergent rhythmic output of large populations of principal cells and
anatomical properties but also on their functional properties. other populations of inhibitory interneurons. The axons of
The axons of this diverse cell population make mainly these cells are often targeted to specic dendritic domains of
short-range projections and release the inhibitory neurotrans- pyramidal neurons as well as to other inhibitory interneuron
mitter GABA onto their targets. Consequently, hippocampal types; this has led to the suggestion that specic interneuron

Figure 523. Inhibitory interneuron diversity. A. Composite draw- pocampus. Filled circles mark the general cell body location of each
ing of characteristic hippocampal and dentate gyrus inhibitory interneuron type that gives rise to thick horizontal and/or vertical
interneuron types assembled from reconstructions of in vivo (5, lines, indicating the predominant orientation and laminar distribu-
1013) or in vitro (13, 69) intracellularly labeled or immunos- tion of the dendritic tree. Boxes represent the laminae, where the
tained (4, 14, 15) neurons in the rat (15, 1015) and guinea pig axon of each interneuron type typically arborizes. Shaded boxes
(69). Thick lines represent the dendritic trees; thin lines are the indicate that other interneurons, rather than principal cells, are the
axons. B. Summary diagram shows the laminar distribution of den- primary targets. The transverse extent of the dendrites or axon is
dritic and axonal arborizations of various types of calcium-binding not indicated. (Source: Modied from Freund and Buzski, 1996;
protein- and neuropeptide-containing interneurons in the hip- from McBain and Fisahn, 2001.)

11 12
A 10

CA1 13
9
CA3
1 2
6 14

5 15

7 DG
8 4
3

B
SLM

SR

SP
SO
yramidal Cell
P V
P P
V V
P CCK CB SO
M SO
M VIP CR
 Hippocampal Neurons: Structure and Function 179

types are tailor-made to play fundamental roles in control- Table 51.

ling the generation of Na- and Ca-mediated action poten- Morphological classication of hippocampal
tials, the regulation of synaptic transmission and plasticity, interneurons
and the generation and pacing of large-scale synchronous Chandelier or Axo-axonic Cells
oscillatory activity.
Basket Cells
Interneurons Innervating Principal Cell Dendrites
5.9.1 Understanding Interneuron Diversity
Dentate gyrus
By denition, all local circuit inhibitory interneurons synthe- Interneurons with hilar dendrites and ascending axons (HIPP cells)
size and release GABA as their primary neurotransmitter. Hilar border neurons with ascending axons and dendrites (HICAP
Anatomically, however, interneurons represent one of the cells)
most diverse populations in the mammalian CNS. To under- Neurons with axons and dendrites in stratum moleculare (MOPP
stand anything of inhibitory interneuron function one must cells)
rst consider the diversity of the population and the efforts Hilar neurons with a projection to the hippocampus and subiculum
made by numerous anatomical and physiological investigators Hilar neurons with unknown projections
alike to break this population down into subpopulations Hippocampus
based on anatomical, neurochemical, or functional properties.
Cells terminating in conjunction with entorhinal afferents (O-LM
A major problem that has invariably plagued any study of
cells)
interneuron function, however, has been the inability to clas- Interneurons innervating pyramidal cell dendrites in strata radia-
sify interneurons neatly into functional or anatomical sub- tum and oriens (bistratied and trilaminar cells)
populations. Even within a single region of the hippocampus, Interneurons with axon collaterals and dendrites in stratum radia-
classifying interneurons is difficult and leads to different tum
numbers of types depending on the classication scheme cho- Interneurons in stratum lacunosum-moleculare
sen. Because of this problem, we have chosen not to organize Interneurons in stratum lucidum
the discussion of hippocampal interneurons around an arbi- Interneurons projecting across subeld boundaries
trary classication scheme. Rather, this section is organized
Interneurons Specialized to Innervate Other Interneurons
around a variety of commonly accepted features observed in
IS-1 neurons
interneurons. Such features can include anatomical, neuro-
IS-2 neurons
chemical, or functional commonalities that at the most ele-
IS-3 neurons
mentary level allow identication of themes across neurons in
different regions of the hippocampal formation. Ultimately, Source: Adapted and extended from Freund and Buzsaki (1996).
however, when such features are cross-referenced, such rigid
classication schemes tend to break down and can often prove
unhelpful, as is discussed below. Although we have chosen this
approach, a detailed understanding of hippocampal function tion, shunt excitatory inputs on their way to the soma, or reg-
eventually requires that models be developed for several ulate dendritic spike initiation and/or propagation. Each of
classes of interneurons, and different models may require dif- these interneuron types exhibits both differences and similari-
ferent classication schemes. Thus, some discussion of the ties in their physiological properties. Even within a single layer
numerous attempts to classify these neurons into distinct sub- of the hippocampus, interneurons with unique axon targeting
populations is warranted. and physiological properties can be identied (Fig. 524).
Historically, initial attempts to classify interneurons prima- Perhaps one of the most useful characterizations of
rily used two approaches: classication based on anatomy or interneuron subtypes has been based on neurochemical con-
based on neurochemical markers (for review see Freund and tent (Table 52). Although the precise physiological roles
Buzski, 1996). Classications based solely on anatomy (Table played by this wide repertoire of peptide transmitters and
51) tell us little about interneuron function, but the recent Ca2-binding proteins contained in specific inhibitory
mapping of axonal arbors to specic domains across the den- interneuron subpopulations is largely unclear, their role in
dritic trees of their targets and the position of dendrites in spe- aiding anatomical and functional characterization of
cic hippocampal subelds have provided important clues to interneurons cannot be overemphasized. Much of our current
specic functional roles played by various interneuron sub- understanding of interneuron anatomy has its origins in stud-
types (Fig. 523). For example, interneurons with axons that ies utilizing antibodies raised against these markers.
innervate either pyramidal cell somata or axon initial segments Neurochemically identical cells can, however, have surpris-
(basket cells, chandelier cells, axo-axonic cells) almost certainly ingly different functional properties, somewhat undermining
regulate the local generation of Na-dependent action poten- this attempt to provide convenient interneuron subdivisions.
tials. In contrast, inhibition arriving at dendritic locations For example, fast-spiking parvalbumin (PV)-containing
likely has little direct inuence over somatic action potential interneurons were once considered to comprise only basket
generation but may strongly inuence local dendritic integra- and axo-axonic cells types, but this interneuron class has
180 The Hippocampus Book

A1 B1 C1

A2 B2 C2

Figure 524. Differential axon targeting of various inhibitory immunoreactive and target other interneurons (B1) (Gulys et al.,
interneurons. Three types of interneuron are shown, each with a 2003). In contrast to O-LM cells, this cell rapidly accommodated
cell body and dendrites in the stratum oriens of CA1 but having upon depolarization, ring only a single spike (B2). Furthermore,
distinct axonal projection patterns and physiological properties. upon repolarization, the cell red a spike with a slow component
A. An O-LM cell targets the distal dendrites of CA1 pyramidal suggestive of a T-type Ca2 current. C. Stratum oriens interneuron
cells (A1). Upon injection of a depolarizing current pulse, O-LM type targeting the proximal dendritic trees of CA1 pyramidal cells
interneurons exhibit an accommodating train of action potentials (C1). Also in contrast to O-LM interneurons, this cell exhibited a
followed by a slow AHP. A prominent sag indicative of Ih is present reduced fast AHP following each spike and a spike with a pro-
upon hyperpolarization (A2). B. Interneuron type exhibiting a dif- nounced Ca2 component upon repolarization (C2). (Source:
fuse axonal arborization pattern suggestive of a septally projecting Lawrence and McBain, unpublished.)
interneuron. These interneurons are typically somatostatin

recently seen new additions (Maccaferri et al., 2000). ing distinct neurochemical markers have been found to also
Interneurons with cell bodies in the stratum oriens and axons have different dendritic morphology (Gulys et al., 1999)
in the stratum lacunosum-moleculare (O-LM cells) or in both (Fig. 525).
stratum radiatum and stratum oriens (bistratied cells) also Characterization based on function has proven even more
contain PV. Given the dendritic location of their axonal problematic. The classic subdivisions were originally based
targeting, it is highly unlikely that the latter cell types perform solely on action potential ring patterns (e.g., fast-spiking ver-
roles similar to those of the basket or axoaxonic cells. Similar sus regular spiking). However, action potential generation
additions to the somatostatin- and cholecystokinin (CCK)- results from the combined activity of numerous voltage-gated
containing families have also been described, further compli- channels, with each channel type potentially having a unique
cating classication systems based solely on neurochemical expression pattern throughout the interneuron subpopula-
content (Oliva et al., 2000; Cope et al., 2002). Although these tions that imparts subtle characteristics to action potential r-
two methods of classication (morphology and neurochem- ing. Therefore, although this classification has been
istry) may be considered separately, substantial effort has been historically useful, it has only limited value given the ever-
made to relate the two criteria. For example, neurons express- expanding repertoire of voltage-gated channels being identi-
 Hippocampal Neurons: Structure and Function 181

Table 52. Scanziani, 2004). In this study, two groups of interneurons

Neurochemical markers used to classify were identied: one whose probability of ring was highest at
hippocampal interneurons the onset of a series of presynaptic stimuli and then fell rap-
Classical Neurotransmitters of Hippocampal Neurons
idly with subsequent stimuli (onset transient interneurons)
and a second where the probability of spike generation was
GABAergic
lowest after the rst stimuli but increased to a plateau in
Cholinergic?
response to later events in the stimulus train (late-persistent
Calcium-Binding Proteins interneurons). Of particular interest, the axons of onset tran-
Parvalbumin sient interneurons were largely distributed around the somata
Calbindin D28k of pyramidal cells (suggestive of basket, axo-axonic, or bis-
Calretinin tratied cells), whereas the axons of late-persistent interneu-
rons largely innervated the stratum radiatum-lacunosum
Neuropeptides moleculare (indicative of dendrite-targeting O-LM interneu-
Somatostatin rons). This anatomical and functional characterization sug-
Neuropeptide Y gests that two distinct inhibitory interneuron subclasses can
Cholecystokinin be sequentially recruited to produce inhibition that rst tar-
Vasoactive intestinal polypeptide gets the soma of pyramidal cells followed by inhibition tar-
Enkephalin
geted to pyramidal cell dendrites.
Neurokinin
This study, together with numerous others (Maccaferri et
Enzymes al., 1998; Markram et al., 1998; Scanziani et al., 1998), high-
Nitric oxide synthase lights an important feature of this type of classication
schemethat single axons may transmit information in a
Neurotransmitter Receptors target-specic manner. This target specicity of transmission
Glutamate receptors can be governed by both the transmitter release mechanisms
GABA receptors of individual presynaptic boutons and the type of ionotropic
Monoamine receptors or metabotropic receptors present at the postsynaptic face.
Acetylcholine receptors In addition, this can be further complicated by the observa-
Opiate receptors tion that whether a particular synapse demonstrates short-
Substance P receptor term depression or facilitation may be developmentally
regulated (for further discussion see McBain and Fisahn,
Source: Adapted from Freund and Buzski (1996).
2001; Maccaferri and Lacaille, 2003).

5.9.2 Dendritic Morphology

ed on inhibitory neurons. Subdivisions of interneurons
based on responses to neuromodulators, properties of intrin- Despite representing only ~10% of the total hippocampal
sic currents and conductances, or expression of ligand-gated neuron population, inhibitory interneurons have highly vari-
channels have all had limited success. Comparisons across able morphologies even within distinct subpopulations (Figs.
these groups, however, have underscored problems with such 523 to 525). For example, cells of a given neurochemical
classication systems and have threatened the very notion of subclass can have dendrites that are organized in a widely het-
homogeneous interneuron subpopulations. In fact, interneu- erogeneous orientation. Despite this marked heterogeneity,
rons are excited or inhibited by a bewildering number of com- however, a few cell types are highly recognizable based on
binations of modulators; and when these properties are their cell morphology. A good example of this is the O-LM
mapped onto their respective anatomical properties, a stag- cell, which typically has its soma and dendrites oriented par-
gering number of subpopulations may exist (Parra et al., allel to the pyramidal cell layer in the stratum oriens-alveus,.
1998). The O-LM cell typically projects its main axon across the stra-
Classications based on the nature of excitatory synaptic tum pyramidale to ramify predominantly in the stratum
transmission have become increasingly common. In hip- lacunosum-moleculare, with few collaterals projecting back to
pocampal circuits, repetitive activation of glutamatergic affer- the stratum oriens-alveus. This cell belongs to the subpopula-
ents progressively increases (augmentation/facilitation) or tion of somatostatin-containing interneurons; but other than
decreases (depression) the amplitudes of excitatory synaptic this neurochemical feature, it has little in common with other
events. The list of synapses that demonstrate facilitation or somatostatin-containing interneurons. As discussed above,
depression has become increasingly lengthy over the last few this presents a problem when attempting to dene the basic
years primarily from studies involving connected pairs of neu- anatomical features of inhibitory interneurons. Consequently,
rons. One of the best examples of this characterization comes we describe in detail only those interneuron subpopulations
from studying CA1 pyramidal cell-mediated feedback excita- whose anatomical features have been subject to rigorous mor-
tion of interneurons in the stratum oriens (Pouille and phometric analysis.
182 The Hippocampus Book

B 7000 0.70 C
6000 0.60

5000 0.50
micrometers

cubicmm

4000 0.40

3000 ** 0.30

2000 0.20
**
1000 0.10

0 0.00
PVCB
Averagelength Occupiedvolume Individualcell
Figure 525. Inhibitory interneuron neurochemistry and dendritic inhibitory inputs on medium-diameter dendrites of PV-, CB-, and
morphology. A. Reconstructed dendritic trees of parvalbumin CR-containing interneurons in CA1 stratum radiatum. The den-
(PV)-, calbindin (CB)-, and calretinin (CR)-containing interneu- drites and synapses were reconstructed from serial ultrathin sec-
rons from the CA1 region of the rat hippocampus. Two examples of tions immunostained for GABA. A large difference can be seen in
each cell type are shown, illustrating the characteristics of the the absolute and relative numbers of excitatory and inhibitory
branching patterns. The different types of dendritic segments synapses terminating on the three types of dendrite. The surface of
separated on the basis of their diameter are indicated with different the PV-positive dendrite is densely covered by synapses, in contrast
levels of gray. Note the variance in the total length of the dendrites to the sparse innervation of CB- and CR-positive dendrites. On the
of individual cells within a given group and the differential distribu- other hand, the proportion of inhibitory terminals compared to all
tion of dendrites in distinct layers for the three cell populations. synaptic inputs is lowest on the PV and highest on the CB den-
B. Parvalbumin cells have the largest dendritic tree and calretinin drites. Excitatory terminals are colored light gray (e.g., e1e9),
the smallest. The horizontal extent of the dendritic tree was widest GABAergic boutons are dark (e.g., i1i4). Note the large variance
for calbindin cells and is reected in the largest occupied volume in the size of the axon terminals. (Source: Data are from Gulys
obtained for this class of cells. C. Distribution of excitatory and et al., 1999.)

Arguably the largest of all the inhibitory interneuron sub- mary dendrites that arise from the somata and run radially
populations, the PV-containing, fast-spiking basket cells are into the stratum oriens or through the stratum radiatum into
found throughout the hippocampal formation (Fig. 525). the stratum lacunosum-moleculare making infrequent
Their cell bodies in general are restricted to the stratum pyra- branches. Although PV-containing interneurons have the
midale, and its border with both the stratum oriens and largest dendritic tree of all interneuron types, these cells
stratum radiatum. Occasionally, however, PV-containing demonstrate large variation in individual cell size, with a more
interneuron cell bodies are found in the stratum radiatum. than twofold variation in the range of dendritic length being
Among interneurons, PV-containing cells have the most elab- reported (Gulys et al., 1999). The distribution of dendrites
orate dendrites, which on average measure  4 mm in their among layers suggests that these cells are likely to collect most
full extent (Gulys et al., 1999). The surface area of PV-con- of their inputs in the strata radiatum and oriens and only a
taining interneuron somata has been estimated at ~1000 m2. smaller portion in the strata lacunosum-moleculare and pyra-
PV-containing interneurons have two to six (mean 5.5) pri- midale (Gulys et al., 1999).
 Hippocampal Neurons: Structure and Function 183

The cross-sectional diameters of PV-containing interneu- Similar to CB-containing interneurons, calretinin (CR)-
ron dendrites are highly variable (0.32.7 m) but are, in gen- containing interneurons can be subdivided into at least two
eral, among the thickest of all interneuron classes studied to classes based on morphology: a spiny type and a spine-free
date. The thickest dendrites typically arise from the somata type (Fig. 525). In the CA1 subeld only spine-free CR-con-
and then taper rapidly after the rst branch point. Apical den- taining interneurons are found, whereas both classes are
drites have about eight branch points, whereas basilar den- found in CA3, with spiny CR-containing interneurons being
drites have about ve. The number of terminal branches is highly enriched in the stratum lucidum, suggesting they
also highly variable but is, on average, ten and seven for apical receive innervation primarily from mossy ber axons of den-
and basilar dendrites, respectively. Although PV-containing tate gyrus granule cells (Gulys et al., 1992). In the CA1 sub-
interneurons can be subclassied into several functionally dis- field, spine-free CR-containing interneurons are found
tinct subgroups (i.e., basket and axo-axonic cells), there throughout all layers, from the alveus to the hippocampal s-
appears to be little anatomical difference between the meas- sure. The cell bodies of this class of interneurons have surface
ured somata and dendrites of these two subclasses. areas about one-half that of PV-containing interneurons, with
Similar morphological estimates have been derived from a mean surface area calculated at ~500 m2. The dendrites of
physiologically characterized fast-spiking basket cells of the CR-containing interneurons arise from highly diverse multi-
CA1 stratum pyramidale, which presumably represent either polar, bipolar, or fusiform cell bodies and primarily run radi-
PV- or CCK-containing interneurons, although their neuro- ally, traversing several layers. Of particular interest, dendrites
chemical content was not determined (Thurbon et al., 1994). of CR-containing cells often form extensive dendrodendritic
The mean total dendritic surface area of fast-spiking interneu- contacts with neighboring cells (Gulys et al., 1999).
rons was estimated at ~5000 m2 and ~2000 m2 for apical CR-containing interneurons have a comparatively small
and basilar dendrites, respectively. Thus, the total surface area total dendritic length (mean ~2500 m) and so are
of a CA1 fast-spiking basket cell is close to 7000 to 10,000 m2. significantly smaller than both PV- and CB-containing
By comparison the total surface area of CA1 pyramidal neu- interneurons. Unlike PV- and CB-containing interneurons,
rons has been estimated at around 24,400 m2 (Turner, 1984). CR-containing interneurons have rather homogeneous fea-
Calbindin (CB)-containing nonpyramidal neurons (some tures, with few primary dendrites (mean of three) that ascend
CA1 pyramidal neurons have high CB content) have the high- or descend radially to invade all layers of the CA1 subeld.
est density within the stratum radiatum, with the greatest The diameters of CR-containing interneurons are approxi-
numbers of cells located at the border with the stratum mately the same as those of CB-containing interneurons, with
lacunosum-moleculare. Smaller numbers of CB-containing the surface of the thickest dendritic branches often being
interneurons are found in the strata oriens and pyramidale, beaded.
with few cell bodies located in the stratum lacunosum-mole- Spiny CR-containing interneurons are present largely in
culare (Freund and Buzski, 1996). Two types of CB-contain- the hilus of the dentate gyrus and the CA3 stratum lucidum
ing interneuron can be distinguished based on morphology (i.e., areas where mossy ber axons of dentate gyrus granule
(Gulys et al., 1999) (Fig. 525). The so-called type I cells are cells are at their highest density). The dendrites of this cell
most numerous in the stratum radiatum, with lower numbers class often bear numerous long protrusions that penetrate
in the strata pyramidale and oriens. These cells have highly bundles of mossy bers (Freund and Buzski, 1996). In the
variable morphologies and are typically multipolar or bitufted dentate gyrus, their dendrites are restricted to the hilus, never
with three to six primary dendrites (mean of four) running in entering the granule cell layer. In CA3 stratum lucidum the
all directions, often descending to the stratum oriens but only dendrites run parallel with the pyramidal cell layer and typi-
rarely entering the stratum lacunosum-moleculare. Their cal- cally do not penetrate the stratum pyramidale or stratum
culated somatic surface area is ~800 m2. The primary den- radiatum. Detailed morphometric analysis of this cell type has
drites arising from the soma are thick and branch within 50 to not been performed.
100 m of the soma. The lengths of the dendrites of these cells
are highly variable but, on average, they are in excess of 3000 5.9.3 Dendritic Spines
m (mean 3441 m), with ~75% of the entire dendritic tree
located in the stratum radiatum. Generally the dendritic The conspicuous absence of dendritic spines on most
diameter of CB-containing interneurons is small, ranging interneuron types undoubtedly has a major impact on their
from 0.2 to 1.8 m. computational properties, in particular synapse specicity
Type II CB-containing interneurons represent a special and intracellular calcium handling. Some well characterized
class of inhibitory interneurons; they are atypical because interneurons are, however, covered densely by dendritic
they project out of the hippocampus to the medial septum spines. These are the hippocampal O-LM and the dentate
(i.e., axons are long-ranging and not local circuit) (see gyrus hilar perforant path-associated (HIPP) cells. Both cell
Chapter 3). These cells, typically found in the stratum oriens, types contain the neuropeptide somatostatin; and their den-
have large, fusiform cell bodies and several long horizontally drites, with prominent spines, can be visualized by immunos-
oriented dendrites. Accurate morphometric analysis of this taining against mGluR1, substance P receptors, or in some
cell type has not been reported (Gulys et al., 2003). cases calretinin (Freund and Buzski, 1996).
184 The Hippocampus Book

When present, spines on interneurons show profound dif- In addition to their inhibitory inputs onto principal cells,
ferences from their counterparts on pyramidal neurons. For interneurons provide inhibitory input to other interneurons.
example, pyramidal cell spines have a spine apparatus, an In contrast to our appreciation of inhibitory transmission
organelle not detected in the spines of interneurons, despite onto principal cells, much less is known regarding the
examining many immunocytochemically visualized spines in nature of inhibition between interneuron subpopulations.
serial sections (Gulys et al., 1992; Acsady et al., 1998). Although poorly appreciated, specicity appears to exist
Another important difference between interneuron and between the targets of identied populations of inhibitory
pyramidal cell spines is the number of synaptic boutons. The interneurons. Both CCK- and PV-containing interneurons
spines of interneurons are covered by numerous excitatory innervate the soma and proximal dendrites of other CCK- and
synaptic boutons (four to eight synapses per spine). By con- PV-containing cells (as well as contacting other interneuron
trast, most pyramidal cell spines have only one bouton, occa- populations), respectively. Of particular interest, some sub-
sionally two; typically, when a second synaptic bouton is populations of interneurons [e.g., vasoactive inhibitory
encountered on a pyramidal cell spine, it forms a symmetrical, peptide (VIP)-containing, calretinin-containing] have axons
presumably GABAergic synapse (Gulys et al., 1992; Acsady et that exclusively innervate only other interneurons. Moreover,
al., 1998). The large number of excitatory synapses on the targets of these subpopulations of interneuron-selective
interneuron spines raises the possibility that spines of interneurons are extremely specic; for example, the interneu-
interneurons serve to increase the synaptic surface area and do ron selective CR-positive cells innervate the CCK-containing
not function as a compartmentalization device as in pyrami- cells but not the PV-containing cells (Gulys et al., 1996). Like
dal cells (McBain et al., 1999). Spines on pyramidal cells have excitatory synaptic transmission onto interneurons, the time
traditionally been considered to be structures that provide course of the GABA-mediated inhibitory postsynaptic cur-
functional microdomains or compartments essential for rents (IPSCs) in cortical interneurons is markedly faster than
input-specific synaptic plasticity. Yuste and colleagues the kinetics of IPSCs in principal neurons of the same circuit
(Goldberg et al., 2003a) provided the rst direct observation (Bartos et al., 2001; Jonas et al., 2004). Another distinction
that functional microdomains restricted to less than 1 m of between inhibition on principal cells and inhibitory interneu-
dendritic space can exist on interneurons in the absence of rons is that the inhibitory synaptic reversal potential in
spines. This compartmentalization was determined by the interneurons is often more depolarized and close to the rest-
combination of rapid-kinetic, Ca2-permeable AMPA recep- ing membrane potential (Banke and McBain, 2005; Vida et al.,
tor-containing synapses located close to a fast extrusion 2006), such that inhibition in these cells is primarily shunting.
mechanism governed by the Na/Ca2 exchanger. In identied fast-spiking basket cells, this shunting inhibition
improves the robustness of gamma oscillatory activity (Vida
5.9.4 Excitatory and Inhibitory Synapses et al., 2006). In interneurons of the stratum lucidum, placing
the inhibitory synaptic reversal potential close to the resting
Within the hippocampal circuit, interneurons receive afferent membrane potential allows rapid reversal of the polarity of
excitatory input from a number of intrinsic and extrinsic synaptic inhibition during high-frequency transmission, such
sources. The nature of the afferents innervating specic pop- that GABAA receptor-mediated events depolarize other
ulations of interneurons depends of course on the location of interneurons and recruit coordinated inhibitory input onto
their soma and the extent of their dendritic tree. For example, principal cell targets (Banke and McBain, 2005).
interneurons with somata and dendrites located in the CA1 Precise estimates of synapse numbers and ratios of excita-
stratum radiatum primarily receive excitatory input from the tory to inhibitory synapses have been performed for only a
CA3 Schaffer collateral projections (i.e., are activated in a handful of interneuron subtypes, primarily the neurochemi-
feedforward manner) (Gulys et al., 1999; Pouille and cally dened PV-, CCK-, CB-, and CR-containing interneu-
Scanziani, 2001). In contrast, interneurons located in the stra- ron subtypes (Gulys et al., 1999, Mtys et al., 2004).
tum oriens/alveus receive little or no innervation from the Although the total number of synapses (excitatory plus
Schaffer collateral input and are primarily excited by the axons inhibitory) is several-fold higher on PV-containing cells
of CA1 pyramidal neurons to provide feedback (recurrent) (~16,000) than on CCK- (~8200), CB- (~4000), or CR-con-
inhibition to CA1 pyramidal neurons (Blasco-Ibanez and taining (~2200) interneurons, the percentage of GABAergic
Freund, 1995; Maccaferri and McBain, 1996). The myriad inputs is higher in CB- (~30%), CCK- (~35%), and CR-con-
roles that interneurons play in neuronal networks are often taining interneurons (~20%) than in PV-containing interneu-
crucially dependent on the rapid, temporally precise conver- rons (~6%). The pattern of excitatory innervation also varies
sion of an excitatory synaptic input to an inhibitory synaptic among the three cell types. Whereas the dendritic and synap-
output (Jonas et al., 2004). Consequently, glutamatergic tic arrangements suggest that both PV- and CR-containing
inputs to interneurons are often very powerful; combined interneurons receive afferent input in all layers of the hip-
with the typically more depolarized resting membrane poten- pocampus, CB-containing cells receive input largely from
tials and higher input resistances, these synapses tend to dis- Schaffer collateral afferents in the stratum radiatum (Gulys et
charge the postsynaptic cell reliably and rapidly (within 12 al., 1999). This high input specicity of CB-containing cells
ms) (Lawrence and McBain, 2003; Jonas et al., 2004). suggests that they are activated primarily in a feedforward
 Hippocampal Neurons: Structure and Function 185

manner, whereas PV- and CR-contining interneurons are acti- of both an M current and a slow calcium-activated potassium
vated in both a feedforward manner by CA3 (Schaffer collat- current and activation of a calcium-dependent nonselective
erals), entorhinal cortex (perforant path), and thalamic cationic current (ICAT). In contrast, stratum oriens cell types
(nucleus reuniens) afferents and a feedback manner by local with axon arborization patterns different from O-LM cells
CA1 recurrent collaterals. lacked this large muscarinic after-depolarization, suggesting
Activation of inhibitory interneurons by feedforward or that cholinergic specializations may be present in anatomi-
feedback afferent projections often permits individual cally distinct subpopulations of hippocampal interneurons.
interneurons to perform a dual role in the hippocampal cir-
cuit (Lei and McBain, 2002; Walker et al., 2002b). Indeed, glu- 5.9.5 Axon Morphology and Synaptic Targets
tamate receptors expressed at synapses formed between
various afferent projections onto single interneurons are often The axonal arborization of local-circuit inhibitory interneu-
comprised of AMPA and NMDA receptors of distinct molec- rons constitutes another diverse feature of their anatomy.
ular composition (Toth and McBain, 1998; Lei and McBain, Each cell type or subclass of inhibitory neuron innervates dis-
2002). AMPA receptors at synapses formed between the recur- tinct subcellular compartments of each of their targets. As
rent collaterals of CA3 pyramidal cells and interneurons of the described above, such an arrangement of axonal distribution
stratum lucidum have a highly edited GluR2 content, making predicts what role each interneuron plays in inuencing activ-
them essentially impermeable to calcium ions. In contrast, ity of the postsynaptic cell. A detailed discussion of the axonal
AMPA receptors at the synapses made by mossy bers and the arborization of each type of interneuron is provided else-
same interneuron lack, or have a low, GluR2 content, render- where (see Freund and Buzski, 1996). However, it is worth-
ing these receptors highly permeable to calcium (Toth and while discussing the axons of some of the most studied classes
McBain, 1998). It is particularly interesting that the time of inhibitory interneuron for comparison with the axons of
course of synaptic conductances associated with these two CA1 and CA3 pyramidal neurons described earlier in the
afferent projections are also markedly different despite the chapter.
synapses having overlapping electrotonic locations (Walker et Fast-spiking basket cells (presumably PV-containing) of
al., 2002b). Such an arrangement of synaptic receptors allows the CA1 hippocampus (Halasy et al., 1996) typically have
the output of the single interneurons to respond differentially axons that emerge from either the soma or a proximal den-
to different afferent inputs, thus increasing the computational drite. The axon makes collaterals that extend into the stratum
repertoire of the interneuron. radiatum but primarily arborize throughout the pyramidal
In addition to excitatory glutamatergic innervation, many cell layer (Halasy et al., 1996). Basket cell axons have a medio-
interneurons also express muscarinic and nicotinic receptors lateral extent of  700 m. The initial part of the primary
and receive cholinergic input from the medial septum. There axon and secondary branches are typically myelinated, and
is also evidence for noradrenergic input from the locus axonal branches wrap around the somata and proximal den-
coeruleus, serotonergic input from the raphe nucleus, and his- drites of pyramidal cells establishing numerous synaptic con-
taminergic input from the supramammillary nucleus, to name tacts. In contrast, the axon initial segments and spines of
a few. Although the roles of many of these neuromodulatory pyramidal cells are rarely contacted by basket cell axons. A sin-
systems are poorly understood, several studies have elucidated gle basket cell can make up to 15 varicosities (presumably
the roles of a few of these modulators in inuencing interneu- presynaptic specializations) on a single pyramidal neuron. In
ron activity. For example, histamine excitation of somato- the case of one well lled basket cell, the total number of pre-
statin- and PV-containing interneurons via H2 receptors sumed synaptic boutons was in excess of 10,000 and was esti-
modulates the activity of voltage-gated potassium channels mated to converge onto ~110 pyramidal neurons (Halasy et
formed by Kv3.2 subunits (Atzori et al., 2000). Following H2 al., 1996).
receptor activation, PKA phosphorylates channels containing Chandelier cells or axo-axonic cells of the dentate gyrus
Kv3.2 to decrease their outward current, consequently reduc- and hippocampus are so named because of the appearance of
ing the upper ring frequency of these fast-spiking interneu- their axonal arborization. The cell type was rst described by
rons. This reduction in interneuron ring frequency has a Szentagothai and Arbib (1974) in the neocortex and was sub-
major inuence on high-frequency oscillatory activity in the sequently found in the hippocampus (Kosaka, 1983; Somogyi
CA3 hippocampus. Similarly, muscarinic receptor activation et al., 1983a,b). The axons of the stratum pyramidale chande-
has a differential effect on interneuron function that depends lier cells originate from the soma or primary dendrite and
on the subclass of interneuron being activated and suggests form rows of 2 to 30 boutons throughout the principal cell
differential expression of muscarinic receptors (Hajos et al., layers and in the stratum oriens. These rows of boutons are
1998; McQuiston and Madison, 1999a,b). Activation of M1 aligned parallel to the axon initial segments of principal cells
and M3 muscarinic acetylcholine receptors on stratum (pyramidal cells in CA1 or CA3 and granule cells in the den-
oriens-lacunosum moleculare (O-LM) interneurons enhances tate gyrus) (Freund and Buzski, 1996) and drape the princi-
their ring frequency and produces large, sustained after- pal cell layer with a termination pattern reminiscent of a
depolarizations (Lawrence et al., 2006). The generation of chandelier. The axonal arbor occupies an elliptical area of
after-depolarization results from the coordinated inhibition ~600 to 850 m, elongated in the septo-temporal direction;
186 The Hippocampus Book

and a single chandelier cell is estimated to innervate ~1200 Inhibitory interneurons located in the CA1 pyramidal cell
postsynaptic CA1 cells (Li et al., 1992). Interestingly, the num- layer and morphologically identied as basket cells had highly
ber of cells targeted by a single chandelier cell in the CA3 sub- variable input resistances, but on average the RN, was ~180
eld may be almost double that in CA1 (Gulys et al., 1993b). M (Thurbon et al., 1994). Estimates of Rm in this cell popu-
Although both fast spiking (PV- or CCK-containing lation were also highly variable and ranged from 82,000 to
interneurons) and axo-axonic interneurons predominantly 281,000 cm2 (mean 66,200 cm2). Ri was also highly vari-
innervate the somata, proximal dendrites, and axon initial able and ranged from 60 to  500 cm (mean ~300 cm). By
segments of principal cells, respectively, other interneurons combining the physical lengths of the apical and basal den-
exclusively contact dendritic locations on principal cells. drites of these cells with their Rm and Ri, the electrotonic
Two examples of dendritic projecting interneurons are the lengths (L) were calculated to be ~1.0 and 0.5 for the apical
bistratied and the O-LM interneuron populations. These and basal dendrites, respectively. These values highlight an
two cell types form an axonal plexus largely complementary to interesting aspect of interneuron physiology. Although the
basket and axo-axonic cells. Bistratied cells (PV-positive, total surface area of the interneurons and the physical length
somatostatin-positive, CB-positive, neuropeptide Y-positive) of their dendrites are considerably shorter than for CA1
have axons that innervate the dendrites of pyramidal cells pyramidal neurons, their electrotonic proles are often simi-
in both the stratum radiatum and stratum oriens and largely lar. This suggests that neurons with small physical proles
spare the principal cell somata. In one conrmed CB-positive cannot simply be assumed to be more electrotonically com-
bistratied interneuron (Sik et al., 1995) the area occupied by pact than larger neurons.
the axon collaterals was 1860 m (septotemporal)  2090 m The membrane time constants of both spiny and aspiny
(mediolateral). The total axon length was 78,800 m, had stratum lucidum interneurons (m  ~40 and ~30 ms, respec-
16,600 boutons, and innervated as many as 2500 pyramidal tively) are slower than that of CA1 pyramidal neurons but
cells. O-LM cells (somatostatin-positive, mGluR1-positive, faster than that of CA3 pyramidal neurons, suggesting that
and to a lesser extent PV-positive) have horizontally oriented these cells have Rm intermediate to both pyramidal neuron
cell bodies residing deep within the stratum oriens and send types (Spruston et al., 1997). Input resistances (RN  ~350
out a largely nonbranching axon, often from a proximal den- and 290 M, respectively), however, are typically higher than
drite, which crosses the strata pyramidale and radiatum to that of both CA1 and CA3 pyramidal neurons, presumably
form an extensive axonal plexus in the stratum lacunosum owing to the smaller somata and dendritic trees in these
moleculare, making synapses onto the distal dendrites of neurons.
pyramidal cells. In one well-lled O-LM cell from an in vivo Estimates of passive membrane parameters for a variety of
study (Sik et al., 1995) the axon had a rather limited sep- interneuron types throughout all ve laminar substrata in
totemporal and mediolateral extent of its the termination eld CA3 (stratum oriens through stratum lacunosum-molecu-
(840 m by 500 m). The total axon length of this cell was lare) have also been determined (Chitwood et al., 1999).
~65,000 m, and the calculated total number of presynaptic Although these numbers were not accompanied by a detailed
boutons was 16,874. Thus, even though the spatial extent of analysis of the corresponding anatomy, the measured param-
the terminal eld was limited, the total number of putative eters provide worthwhile comparisons with other published
synapses was typically higher than that of basket cells meas- data. Input resistances were signicantly higher than those
ured in the same study. reported for basket cells but were similar to those obtained
from identied stratum lucidum interneurons; on average
5.9.6 Resting Membrane Properties they were ~440 M. The membrane time constants of CA3
interneurons were also highly variable (2769 ms) with a
Characterization of the passive membrane properties of mean of ~60 ms. Interestingly, no signicant differences in RN,
inhibitory interneurons has been limited, with only a handful m, Rm, or Cm were observed for interneurons located across
of studies providing reliable data. Given the numerous classes the ve substrata despite the marked heterogeneity in cell
and subpopulations of inhibitory interneurons in the mam- morphology across all subelds.
malian hippocampus, extrapolation of these numbers
between subpopulations is to be avoided. Here, we discuss 5.9.7 Voltage-Gated Channels
only those cells in which the anatomy is known and the pas- in Inhibitory Interneurons
sive membrane properties have been estimated using patch-
clamp recording. How an interneuron responds to afferent activity depends on
Where studied, inhibitory interneurons have resting mem- many factors, in particular the synaptic conductance time
brane potentials that are slightly more depolarized than CA1 course and the nature of the intrinsic channel proteins
and CA3 pyramidal neurons. Estimates of interneuron resting expressed on the cell surface. As was discussed in detail for
membrane potentials have provided numbers spanning a CA1 pyramidal neurons, the properties, density, and distribu-
wide range but typically lie 5 to 10 mV more depolarized than tion of these conductances determine how synaptic inputs are
pyramidal neurons. integrated and converted to unique patterns of action poten-
 Hippocampal Neurons: Structure and Function 187

tial ring. Like most CNS neurons, interneurons differentially tional roles played by voltage-gated potassium channels has
express a wide array of conductances with overlapping voltage arguably provided the best insight into determinants of
dependence (e.g., activation by depolarization or hyperpolar- interneuron function (for review see McBain and Fisahn,
ization) and kinetics. 2001). Although attempts to correlate interneuron K channel
The rst clue that interneurons might express voltage- function with properties of recombinant K channels has
gated ion channels distinct from their principal neuron coun- been problematic, major differences between principal cell
terparts came from inspection of action potential waveforms. and inhibitory interneuron voltage-gated potassium channels
Many interneurons re brief action potentials with rapid have emerged.
repolarization, followed by only a simple fast after-hyperpo- In recombinant systems, the K channel subunits Kv3.1b
larization. Responses of interneurons to sustained depolariza- and Kv3.2 confer currents that activate at highly depolarized
tion are often quite distinct from both CA1 and CA3 potentials, show little inactivation during sustained depolar-
pyramidal neurons, with many cell types showing high ring ization, and deactivate rapidly upon repolarization (Rudy and
rates, brief interspike intervals, and little action potential McBain, 2001). Homo-tetramers formed by recombinant
accommodation. The most parsimonious explanation for Kv3.1b and Kv3.2 have similar biophysical properties, with the
these distinct properties was the presence of voltage-gated notable difference that channels formed by Kv3.2 have a PKA-
channel types unique to interneurons. The demonstration phosphorylation site that reduces currents through these
that low-threshold voltage-activated calcium channels con- channels when phosphorylated. In the hippocampus, Kv3.1b
ferred burst-ring properties on interneurons of the stratum and Kv3.2 are expressed in all PV-containing interneurons,
lacunosum-moleculare provided the rst direct evidence that and Kv3.2 is also found in ~40% of somatostatin-containing
interneurons expressed voltage-gated channels with proper- interneurons; both are interneuron subpopulations with fast
ties distinct from principal cells (Fraser and MacVicar, 1991). spiking phenotypes. Outward currents through Kv3 channels
Sodium channels with properties distinct from those of the act to keep action potentials brief by activating at potentials
principal cells have also been described (Martina and Jonas, close to the action potential peak, rapidly repolarizing the
1997). Demonstration of the differential expression and func- membrane potential, and limiting the duration of the after-

Figure 526. Inhibitory interneuron dendritic excitability. A. B. Simultaneous current clamp (CC) recordings from a dendrite
Camera lucida reconstruction of a biocytin-lled oriens-alveus (gray) and soma (black). In response to long, threshold-level cur-
interneuron. Soma and dendrites (thick lines) are typically rent injections, the action potential occurs rst in the axon-bearing
restricted to the strata oriens and alveus. The axon (thin lines) proj- dendrite (top). With briefer, stronger-current injections, the action
ects across the pyramidal cell layer through the stratum radiatum to potential occurs rst at the site of current injection (bottom).
ramify extensively throughout the stratum lacunosum-moleculare. (Source: Data are from Martina et al., 2000.)

A B
188 The Hippocampus Book

hyperpolarization by rapidly deactivating. High ring fre- Adams PR (1992) The platonic neuron gets the hots. Curr Biol
quencies can be achieved by the ability of Kv3 channels to 2:625627.
facilitate the recovery of both Na channels and transient A- Alger BE, Nicoll RA (1980) Epileptiform burst afterhyperpolariza-
type potassium channels from inactivation. Of particular tion: calcium-dependent potassium potential in hippocampal
CA1 pyramidal cells. Science 210:11221124.
interest, PKA phosphorylation of Kv3.2 subunits is triggered
Ali AB, Deuchars J, Pawelzik H, Thomson AM (1998) CA1 pyramidal
by histamine via H2 receptor activation, which reduces out-
to basket and bistratied cell EPSPs: dual intracellular record-
ward currents, broadens action potentials, and reduces the ings in rat hippocampal slices. J Physiol (Lond) 507:201217.
maximal ring frequency of dentate gyrus interneurons by Alonso A, Klink R (1993) Differential electroresponsiveness of stel-
altering Kv3.2 function. late and pyramidal-like cells of medial entorhinal cortex layer II.
Although the type of channels expressed by the cells are J Neurophysiol 70:128143.
important determinants of interneuron function, the density Alonso A, Llins RR (1989) Subthreshold Na-dependent theta-like
and spatial distribution of those channels is also important. rhythmicity in stellate cells of entorhinal cortex layer II. Nature
The density of both Na and sustained K currents are signif- 342:175177.
icantly higher and more uniform in dendrites of somato- Amaral DG (1978) A Golgi study of cell types in the hilar region of
statin-containing interneurons of the CA1 stratum the hippocampus in the rat. J Comp Neurol 182:851914.
Amaral DG (1979) Synaptic extensions from the mossy bers of the
oriens-alveus than those in the dendrites of neighboring prin-
fascia dentata. Anat Embryol 155:241251.
cipal cells (Martina et al., 2000). Consequently, action poten-
Amaral DG, Dent JA (1981) Development of the mossy bers of
tials in interneurons are initiated at sites across the dendritic the dentate gyrus. I. A light and electron microscopic study
tree and propagate rapidly to and from the soma with little of the mossy bers and their expansions. J Comp Neurol
decrement even during high-frequency ring (Fig. 526). 195:5186.
Action potential initiation can switch between the soma and Amaral DG, Witter MP (1989) The three-dimensional organization
the axon-bearing dendrite, depending on the nature of the of the hippocampal formation: a review of anatomical data.
excitatory stimulus. Such organization of voltage-gated chan- Neuroscience 31:571591.
nels acts to ensure fast and reliable signaling in interneurons. Amaral DG, Witter MP (1995) Hippocampal formation. In: The rat
This channel organization is in marked contrast to channel nervous system (Paxinos G, ed). San Diego: Academic Press.
expression in principal cells, where dendritic action potentials Amaral DG, Ishizuka N, Claiborne B (1990) Neurons, numbers and
the hippocampal network. Prog Brain Res 93:111.
are attenuated by a gradient of dendritic A-type K channels
Amaral DG, Dolorfo C, Alvarez-Royo P (1991) Organization of CA1
(Hoffman et al., 1997; however, see Goldberg et al., 2003b)
projections to the subiculum: a PHA-L analysis in the rat.
(see Section 5.2.11 for further discussion). Whether such Hippocampus 1:415436.
organization of voltage-gated channels and dendritic initia- Andersen P (1960) Interhippocampal impulses. II. Apical dendritic
tion of action potentials exists in other interneurons remains activation of CA1 neurons. Acta Physiol Scand 48:178208.
to be seen, but it highlights another feature of interneuron Andersen P, Lmo T (1966) Mode of activation of hippocampal
design important for rapid signaling. Studies of the transcrip- pyramidal cells by excitatory synapses on dendrites. Exp Brain
tional regulation and selective targeting of both ligand- and Res 2:247260.
voltage-gated channels will ultimately provide valuable Andersen P, Blackstad TW, Lomo T (1966a) Location and identica-
insight into the molecular events regulating expression of tion of excitatory synapses on hippocampal pyramidal cells. Exp
these important membrane proteins. Brain Res 1:236248.
Andersen P, Holmqvist, Voorhoeve PE (1966b) Excitatory synapses
on hippocampal apical dendrites activated by entorhinal stimu-
lation Acta Physiol Scand 66:461472.
ACKNOWLEDGMENTS
Andersen P, Silfvenius H, Sundberg SH, Sveen O (1980) A com-
parison of distal and proximal dendritic synapses on CA1 pyra-
The authors thank members of their laboratories for helpful com- mids in guinea-pig hippocampal slices in vitro. J Physiol 307:
ments and discussion during the preparation of the chapter. 273299.
Andersen P, Soleng AF, Raastad M (2000) The hippocampal lamella
hypothesis revisited. Brain Res 886:165171.
REFERENCES Andrade R, Nicoll RA (1987) Pharmacologically distinct actions of
serotonin on single pyramidal neurones of the rat hippocampus
Abraham WC, McNaughton N (1984) Differences in synaptic trans- recorded in vitro. J Physiol 394:99124.
mission between medial and lateral components of the per- Andrasfalvy BK, Magee JC (2001) Distance-dependent increase in
forant path. Brain Res 303:251260. AMPA receptor number in the dendrites of adult hippocampal
Acsdy L, Kamondi A, Sik A, Martinez-Guijarro FJ, Buzski G, CA1 pyramidal neurons. J Neurosci 21:91519159.
Freund TF (1998) GABAergic cells are the major postsynaptic Andreasen M, Lambert JD (1995a) Regenerative properties of
targets of mossy bers in the rat hippocampus. J Neurosci pyramidal cell dendrites in area CA1 of the rat hippocampus. J
18:33863403. Physiol 483:421441.
Acsdy L, Kamondi A, Sik A, Freund T, Buzski G (2000) Unusual tar- Andreasen M, Lambert JD (1995b) The excitability of CA1 pyrami-
get selectivity of perisomatic inhibitory cells in the hilar region dal cell dendrites is modulated by a local Ca2-dependent K-
of the rat hippocampus. J Neurosci 20:69076919. conductance. Brain Res 698:193203.
 Hippocampal Neurons: Structure and Function 189

Andreasen M, Lambert JD (1998) Factors determining the efficacy Bender RA, Brewster A, Santoro B, Ludwig A, Hofmann F, Biel M,
of distal excitatory synapses in rat hippocampal CA1 pyrami- Baram TZ (2001) Differential and age-dependent expression of
dal neurones. J Physiol 507:441462. hyperpolarization-activated, cyclic nucleotide-gated cation
Andreasen M, Lambert JD (1999) Somatic amplication of distally channel isoforms 14 suggests evolving roles in the developing
generated subthreshold EPSPs in rat hippocampal pyramidal rat hippocampus. Neuroscience 106:689698.
neurones. J Physiol 519:85100. Benson DM, Blitzer RD, Landau EM (1988) An analysis of the depo-
Aniksztejn L, Demarque M, Morozov Y, Ben-Ari Y, Represa A (2001) larization produced in guinea-pig hippocampus by cholinergic
Recurrent CA1 collateral axons in developing rat hippocampus. receptor stimulation. J Physiol 404:479496.
Brain Res 913:195200. Bernard C, Johnston D (2003) Distance-dependent modiable
Ariav G, Polsky A, Schiller J (2003) Submillisecond precision of the threshold for action potential back-propagation in hippocam-
input-output transformation function mediated by fast sodium pal dendrites. J Neurophysiol 90:18071816.
dendritic spikes in basal dendrites of CA1 pyramidal neurons. J Bennett MR, Gibson WG, Robinson J (1994) Dynamics of the CA3
Neurosci 23:77507758. pyramidal neuron autoassociative memory network in the hip-
Atzori M, Lau D, Tansey EP, Chow A, Ozaita A, Rudy B, McBain CJ pocampus. Philos Trans R Soc Lond B Biol Sci 343:167187.
(2000) H2 histamine receptor-phosphorylation of Kv3.2 modu- Berger T, Larkum ME, Luscher HR (2001) High Ih channel density
lates interneuron fast spiking. Nat Neurosci 3:791798. in the distal apical dendrite of layer V pyramidal cells increases
Azouz R, Alroy G, Yaari Y (1997) Modulation of endogenous ring bidirectional attenuation of EPSPs. J Neurophysiol 85:855
patterns by osmolarity in rat hippocampal neurones. J Physiol 868.
502:175187. Bianchi R, Young SR, Wong RK (1999) Group I mGluR activation
Banke T, McBain CJ (2005) A frequency-dependent switch of synap- causes voltage-dependent and -independent Ca2 rises in hip-
tic inhibition polarity recruits a network of stratum lucidum pocampal pyramidal cells. J Neurophysiol 81:29032913.
interneurons [abstract 955.8]. Presented at the Annual Society Bilkey DK, Schwartzkroin PA (1990) Variation in electrophysiology
for Neuroscience Meeting, San Diego. and morphology of hippocampal CA3 pyramidal cells. Brain
Bannister NJ, Larkman AU (1995a) Dendritic morphology of CA1 Res 514:7783.
pyramidal neurones from the rat hippocampus. I. Branching Bittner K, Muller W (1999) Oxidative downmodulation of the tran-
patterns. J Comp Neurol 360:150160. sient K-current IA by intracellular arachidonic acid in rat hip-
Bannister NJ, Larkman AU (1995b) Dendritic morphology of CA1 pocampal neurons. J Neurophysiol 82:508511.
pyramidal neurones from the rat hippocampus. II. Spine distri- Blackstad TW (1956) Commissural connections of the hippocampal
butions. J Comp Neurol 360:161171. region in the rat, with special reference to their mode of termi-
Bartos M, Vida I, Frotscher M, Geiger JR, Jonas P (2001) Rapid sig- nation. J Comp Neurol 105:417537.
naling at inhibitory synapses in a dentate gyrus interneuron Blackstad TW (1958) On the termination of some afferents to the
network. J Neurosci 21:26872698. hippocampus and fascia dentata; an experimental study in the
Behr J, Gloveli T, Heinemann U (1998) The perforant path projection rat. Acta Anat (Basel) 35:202214.
from the medial entorhinal cortex layer III to the subiculum in Blackstad TW, Kjaerheim A (1961) Special axo-dendritic synapses in
the rat combined hippocampal-entorhinal cortex slice. Eur J the hippocampal cortex: electron and light microscopic studies
Neurosci 10:10111018. on the layer of mossy bers. J Comp Neurol 117:133159.
Bekkers JM (2000) Distribution and activation of voltage-gated Blackstad TW, Brink K, Hem J, Jeune B (1970) Distribution of hip-
potassium channels in cell-attached and outside-out patches pocampal mossy bers in the rat: an experimental study with
from large layer 5 cortical pyramidal neurons of the rat. J Physiol silver impregnation methods. J Comp Neurol 138:433450.
525:611620. Blasco-Ibanez JM, Freund TF (1995) Synaptic input of horizontal
Belichenko PV (1993) Neuronal cell types in entorhinal cortex and interneurons in stratum oriens of the hippocampal CA1 sub-
hippocampal formation of man and other mammalia: an inter- eld: structural basis of feed-back activation. Eur J Neurosci
species comparison. Hippocampus 3:310. 7:21702180.
Ben-Ari Y (1985) Limbic seizure and brain damage produced by Bonhoeffer T, Yuste R (2002) Spine motility: phenomenology, mech-
kainic acid mechanisms and relevance to human temporal lobe anisms, and function. Neuron 35:10191027.
epilepsy. Neuroscience 14:375403. Bowden SHE, Fletcher S, Loane DJ, Marrion NV (2001) Somatic colo-
Benardo LS, Prince DA (1982a) Ionic mechanisms of cholinergic calization of rat SK1 and D class (Cav1.2) L-type calcium chan-
excitation in mammalian hippocampal pyramidal cells. Brain nels in rat CA1 hippocampal neurons. J Neurosci 21:RC175 (16).
Res 249:333344. Brown TH, Fricke RA, Perkel DH (1981) Passive electrical constants
Benardo LS, Prince DA (1982b) Cholinergic excitation of mam- in three classes of hippocampal neurons. J Neurophysiol
malian hippocampal pyramidal cells. Brain Res 249:315331. 46:812827.
Benardo LS, Prince DA (1982c) Cholinergic pharmacology of mam- Brun VH, Otnass MK, Molden S, Steffenach HA, Witter MP, Moser
malian hippocampal pyramidal cells. Neuroscience 7:17031712. MB, Moser EI (2002) Place cells and place recognition main-
Benardo LS, Prince DA (1982d) Dopamine modulates a Ca2- tained by direct entorhinal-hippocampal circuitry. Science
activated potassium conductance in mammalian hippocampal 296:22432246.
pyramidal cells. Nature 297:7679. Buckmaster PS, Schwartzkroin PA (1994) Hippocampal mossy cell
Benardo LS, Prince DA (1982e) Dopamine action on hippocampal function: a speculative view. Hippocampus 4:393402.
pyramidal cells. J Neurosci 2:415423. Buckmaster PS, Strowbridge BW, Kunkel DD, Schmiege DL,
Benardo LS, Masukawa LM, Prince DA (1982) Electrophysiology of Schwartzkroin PA (1992) Mossy cell axonal projections to the
isolated hippocampal pyramidal dendrites. J Neurosci dentate gyrus molecular layer in the rat hippocampal slice.
2:16141622. Hippocampus 2:349362.
190 The Hippocampus Book

Buckmaster PS, Strowbridge BW, Schwartzkroin PA (1993) A com- dimensional analysis of granule cell dendrites in the rat dentate
parison of rat hippocampal mossy cells and CA3c pyramidal gyrus. J Comp Neurol 302:206219.
cells. J Neurophysiol 70:12811299. Cohen I, Miles M (2000) Contributions of intrinsic and synaptic
Buckmaster PS, Wenzel HJ, Kunkel DD, Schwartzkroin PA (1996) activities to the generation of neuronal discharges in in vitro
Axon arbors and synaptic connections of hippocampal mossy hippocampus. J Physiol 524:485502.
cells in the rat in vivo. J Comp Neurol 366:270292. Colbert CM, Johnston D (1996) Axonal action-potential initiation
Burnashev N, Monyer H, Seeburg PH, Sakmann B (1992) Divalent and Na channel densities in the soma and axon initial segment
ion permeability of AMPA receptor channels is dominated by of subicular pyramidal neurons. J Neurosci 16:66766686.
the edited form of a single subunit. Neuron 8:189198. Colbert CM, Johnston D (1998) Protein kinase C activation decreases
Buzski G (1996) The hippocampo-neocortical dialogue. Cereb activity-dependent attenuation of dendritic Na current in hip-
Cortex 6:8192. pocampal CA1 pyramidal neurons. J Neurophysiol 79:491495.
Buzski G, Penttonen M, Bragin A, Nadasdy Z, Chrobak JJ (1995) Colbert CM, Pan E (1999) Arachidonic acid reciprocally alters the
Possible physiological role of the perforant path-CA1 projec- availability of transient and sustained dendritic K channels in
tion. Hippocampus 5:141146. hippocampal CA1 pyramidal neurons. J Neurosci 19:81638171.
Cai X, Liang CW, Muralidharan S, Kao JP, Tang CM, Thompson SM Colbert CM, Pan E (2002) Ion channel properties underlying axonal
(2004) Unique roles of SK and Kv4.2 potassium channels in action potential initiation in pyramidal neurons. Nat Neurosci
dendritic integration. Neuron 44:351364. 5:533538.
Cajal SR (1911) Histology of the nervous system of man and vertebrates Colbert CM, Magee JC, Hoffman DA, Johnston D (1997) Slow recov-
(Swanson N, Swanson L, translators). Oxford: Oxford ery from inactivation of Na channels underlies the activity-
University Press, 1995. dependent attenuation of dendritic action potentials in
Callaway JC, Ross WN (1995) Frequency-dependent propagation of hippocampal CA1 pyramidal neurons. J Neurosci 17:65126521.
sodium action potentials in dendrites of hippocampal CA1 Colling SB, Wheal HV (1994) Fast sodium action potentials are gen-
pyramidal neurons. J Neurophysiol 74:13951403. erated in the distal apical dendrites of rat hippocampal CA1
Carboni AA, Lavelle WG, Barnes CL, Cipolloni PB (1990) Neurons of pyramidal cells. Neurosci Lett 172:7396.
the lateral entorhinal cortex of the rhesus monkey: a Golgi, his- Cook EP, Johnston D (1997) Active dendrites reduce location-
tochemical, and immunochemical characterization. J Comp dependent variability of synaptic input trains. J Neurophysiol
Neurol 291:583608. 78:21162128.
Carnevale NT, Tsai KY, Claiborne BJ, Brown TH (1997) Comparative Cook EP, Johnston D (1999) Voltage-dependent properties of den-
electronic analysis of three classes of rat hippocampal neurons. drites that eliminate location-dependent variability of synaptic
J Neurophysiol 78:703720. input. J Neurophysiol 81:535543.
Cash S, Yuste R (1998) Input summation by cultured pyramidal neu- Cooney JR, Hurlburt JL, Selig DK, Harris KM, Fiala JC (2002)
rons is linear and position-independent. J Neurosci 18:1015. Endosomal compartments serve multiple hippocampal den-
Cash S, Yuste R (1999) Linear summation of excitatory inputs by CA1 dritic spines from a widespread rather than a local store of recy-
pyramidal neurons. Neuron 22:383394. cling membrane. J Neurosci 22:22152224.
Chen H, Lambert NA (1997) Inhibition of dendritic calcium inux Cooper DC, Moore SJ, Staff NP, Spruston N (2003) Psychostimulant-
by activation of G-protein-coupled receptors in the hippocam- induced plasticity of intrinsic neuronal excitability in ventral
pus. J Neurophysiol 78:34843488. subiculum. J Neurosci 23:99379946.
Chicurel ME, Harris KM (1992) Three-dimensional analysis of the Cope DW, Maccaferri G, Marton LF, Roberts JD, Cobden PM,
structure and composition of CA3 branched dendritic spines Somogyi P (2002) Cholecystokinin-immunopositive basket and
and their synaptic relationships with mossy ber boutons in the Schaffer collateral-associated interneurones target different
rat hippocampus. J Comp Neurol 325:169182 domains of pyramidal cells in the CA1 area of the rat hip-
Chitwood RA, Hubbard A, Jaffe DB (1999) Passive electrotonic prop- pocampus. Neuroscience 109:6380.
erties of rat hippocampal CA3 interneurons. J Physiol (Lond) Cragg BG, Hamlyn LH (1955) Action potentials of the pyramidal
515:743756. neurones in the hippocampus of the rabbit. J Physiol 129:-
Christie BR, Eliot LS, Ito K, Miyakawa H, Johnston D (1995) 608627.
Different Ca2 channels in soma and dendrites of hippocampal Csicsvari J, Hirase H, Czurko A, Buzski G (1998) Reliability and
pyramidal neurons mediate spike-induced Ca2 influx. J state dependence of pyramidal cellinterneuron synapses in the
Neurophysiol 73:25532557. hippocampus: an ensemble approach in the behaving rat.
Chrobak JJ, Buzski G (1994) Selective activation of deep layer Neuron 21:179189.
(VVI) retrohippocampal cortical neurons during hippocampal Dahl D, Burgard EC, Sarvey JM (1990) NMDA receptor antagonists
sharp waves in the behaving rat. J Neurosci 14:61606170. reduce medial, but not lateral, perforant path-evoked EPSPs in
Chrobak JJ, Buzski G (1996) High-frequency oscillations in the out- dentate gyrus of rat hippocampal slice. Exp Brain Res
put networks of the hippocampal-entorhinal axis of the freely 83:172177.
behaving rat. J Neurosci 16:30563066. Dailey ME, Smith SJ (1996) The dynamics of dendritic structure in
Chrobak JJ, Lrincz A, Buzski G (2000) Physiological patterns in the developing hippocampal slices. J Neurosci 16:29832994.
hippocampal-entorhinal cortex system. Hippocampus 10:457 Deller T, Adelmann G, Nitsch R, Frotscher M (1996a) The alvear
465. pathway of the rat hippocampus. Cell Tissue Res 286:293303.
Claiborne BJ, Amaral DG, Cowan WM (1986) A light and electron Deller T, Martinez A, Nitsch R, Frotscher M (1996b) A novel entorhi-
microscopic analysis of the mossy bers of the rat dentate gyrus. nal projection to the rat dentate gyrus: direct innervation of
J Comp Neurol 246:435458. proximal dendrites and cell bodies of granule cells and
Claiborne BJ, Amaral DG, Cowan WM (1990) Quantitative, three- GABAergic neurons. J Neurosci 16:33223333.
 Hippocampal Neurons: Structure and Function 191

Desmond NL, Levy WB (1982) A quantitative anatomical study of rons of guinea pig entorhinal cortex. J Neurophysiol
the granule cell dendritic elds of the rat dentate gyrus using a 71:19861991.
novel probabilistic method. J Comp Neurol 212:131145. Fiala JC, Feinberg M, Popov V, Harris KM (1998) Synaptogenesis via
Desmond NL, Levy WB (1984) Dendritic caliber and the 3/2 power dendritic lopodia in developing hippocampal area CA1.
relationship of dentate granule cells. J Comp Neurol 227: J Neurosci 18:89008911.
589596. Fiala JC, Kirov SA, Feinberg MD, Petrak LJ, George P, Goddard CA,
Desmond NL, Levy WB (1985) Granule cell dendritic spine density Harris KM (2003) Timing of neuronal and glial ultrastructure
in the rat hippocampus varies with spine shape and location. disruption during brain slice preparation and recovery in vitro.
Neurosci Lett 54:219224. J Comp Neurol 465:90103.
Destexhe A, Par D (1999) Impact of network activity on the inte- Finch DM, Babb TL (1980) Neurophysiology of the caudally directed
grative properties of neocortical pyramidal neurons in vivo. J hippocampal efferent system in the rat: projections to the subic-
Neurophysiol 81:15311547. ular complex. Brain Res 197:1112.
Deuchars J, Thomson AM (1996) CA1 pyramid-pyramid connec- Finch DM, Babb TL (1981) Demonstration of caudally directed hip-
tions in rat hippocampus in vitro: dual intracellular recordings pocampal efferents in the rat by intracellular injection of horse-
with biocytin lling. Neuroscience 74:10091018. radish peroxidase. Brain Res 214:405410.
Dickson CT, Mena AR, Alonso A (1997) Electroresponsiveness of Fisahn A, Yamada M, Duttaroy A, Gan J-W, Deng C-X, McBain CJ,
medial entorhinal cortex layer III neurons in vitro. Neuroscience Wess J (2002) Muscarinic induction of hippocampal gamma
81:937950. oscillations requires coupling of the M1 receptor to two mixed
Dickson CT, Magistretti J, Shalinsky MH, Fransn E, Hasselmo ME, cation currents. Neuron 33:615624.
Alonso A (2000a) Properties and role of Ih in the pacing of sub- Fox S, Ranck JB (1975) Localization and anatomical identication of
threshold oscillations in entorhinal cortex layer II neurons. J theta and complex spike cells in dorsal hippocampal formation
Neurophysiol 83:25622579. of rats. Exp Neurol 49:299313.
Dickson CT, Magistretti J, Shalinsky M, Hamam B, Alonso A (2000b) Frank LM, Brown EN, Wilson MA (2001) A comparison of the
Oscillatory activity in entorhinal neurons and circuits. Ann NY ring properties of putative excitatory and inhibitory neurons
Acad Sci 911:127150. from CA1 and the entorhinal cortex. J Neurophysiol 86:
Dolleman-Van der Weel MJ, Witter MP (1996) Projections from the 20292040.
nucleus reuniens thalami to the entorhinal cortex, hippocampal Fraser DD, MacVicar BA (1991) Low-threshold transient calcium
eld CA1, and the subiculum in the rat arise from different pop- current in rat hippocampal lacunosum-moleculare interneu-
ulations of neurons. J Comp Neurol 364:637650. rons: kinetics and modulation by neurotransmitters. J Neurosci
Dolleman-Van der Weel MJ, Lopes da Silva FH, Witter MP (1997) 11:28122820.
Nucleus reuniens thalami modulates activity in hippocampal Fraser DD, MacVicar BA (1996) Cholinergic-dependent plateau
eld CA1 through excitatory and inhibitory mechanisms. J potential in hippocampal CA1 pyramidal neurons. J Neurosci
Neurosci 17:56405650. 16:41134128.
Doller HJ, Weight FF (1982) Perforant pathway activation of hip- Fredens K, Stengaard-Pedersen K, Wallace MN (1987) Localization of
pocampal CA1 stratum pyramidale neurons: electrophysiologi- cholecystokinin in the dentate commissural-associational sys-
cal evidence for a direct pathway. Brain Res 237:113. tem of the mouse and rat. Brain Res 401:6878.
Durand D, Carlen PL, Gurevich N, Ho A, Kunov H (1983) Freund TF, Buzski G (1996) Interneurons of the hippocampus.
Electrotonic parameters of rat dentate granule cells measured Hippocampus 6:347470.
using short current pulses and HRP staining. J Neurophysiol Freund TF, Hajos N, Acsady L, Gorcs TJ, Katona I (1997) Mossy cells
50:10801097. of the rat dentate gyrus are immunoreactive for calcitonin gene-
Egorov AV, Heinemann U, Mller W (2002a) Differential excitability related peptide (CGRP). Eur J Neurosci 9:18151830.
and voltage-dependent Ca2 signaling in two types of medial Frick A, Magee J, Koester HJ, Migliore M, Johnston D (2003)
entorhinal cortex layer V neurons. Eur J Neurosci 16:13051312. Normalization of Ca2 signals by small oblique dendrites of
Egorov AV, Hamam BN, Fransen E, Hasselmo ME, Alonso AA CA1 pyramidal neurons. J Neurosci 23:32433250.
(2002b) Graded persistent activity in entorhinal cortex neurons. Frick A, Magee J, Johnston D (2004) LTP is accompanied by an
Nature 420:173178. enhanced local excitability of pyramidal neuron dendrites. Nat
Empson RM, Heinemann U (1995a) The perforant path projection Neurosci 7:126135.
to hippocampal area CA1 in the rat hippocampal-entorhinal Fricker D, Verheugen JAH, Miles R (1999) Cell-attached measure-
cortex combined slice. J Physiol 484:707720. ments of the ring threshold of rat hippocampal neurons. J
Empson RM, Heinemann U (1995b) Perforant path connections to Physiol (Lond) 517:791804.
area CA1 are predominantly inhibitory in the rat hippocampal- Frotscher M, Seress L, Schwerdtfeger WK, Buhl E (1991) The mossy
entorhinal cortex combined slice preparation. Hippocampus cells of the fascia dentata: a comparative stuy of their ne struc-
5:104107. ture and synaptic connections in rodents and primates. J Comp
Emptage N, Bliss TV, Fine A (1999) Single synaptic events evoke Neurol 312:145163.
NMDA receptor-mediated release of calcium from internal Fujita Y, Sakata H (1962) Electrophysiological properties of CA1 and
stores in hippocampal dendritic spines. Neuron 22:115124. CA2 apical dendrites of rabbit hippocampus. J Neurophysiol
Engert F, Bonhoeffer T (1999) Dendritic spine changes associated 25:209222.
with hippocampal long-term synaptic plasticity. Nature Funahashi M, Stewart M (1997a) Presubicular and parasubicular cor-
399:6670. tical neurons of the rat: electrophysiological and morphological
Fan S, Stewart M, Wong RKS (1994) Differences in voltage-depend- properties. Hippocampus 7:117129.
ent sodium currents exhibited by supercial and deep layer neu- Funahashi M, Stewart M (1997b) Presubicular and parasubicular
192 The Hippocampus Book

cortical neurons of the rat: functional separation of deep and Golding NL, Kath WL, Spruston N (2001) Dichotomy of action
supercial neurons in vitro. J Physiol (Lond) 501:387403. potential backpropagation in CA1 pyramidal neurons.
Funahashi M, Harris E, Stewart M (1999) Re-entrant activity in pre- J Neurophysiol 86:29983010.
subiculum-subiculum circuit generates epileptiform activity in Golding NL, Jung H, Mickus TJ, Spruston N (1999) Dendritic cal-
vitro. Brain Res 849:139146. cium spike initiation and repolarization are controlled by dis-
Ganeshina O, Berry RW, Petralia RS, Nicholson DA, Geinisman Y tinct potassium channel subtypes in CA1 pyramidal neurons.
(2004a) Differences in the expression of AMPA and NMDA J Neurosci 19:87898798.
receptors between axospinous perforated and nonperforated Golding N, Staff NP, Spruston N (2002) Dendritic spikes as a mech-
synapses are related to the conguration and size of postsynap- anism for cooperative long-term potentiation. Nature
tic densities. J Comp Neurol 468:8695. 418:326331.
Ganeshina O, Berry RW, Petralia RS, Nicholson DA, Geinisman Y Golding N, Mickus T, Katz Y, Kath WL, Spruston N (2005) Factors
(2004b) Synapses with a segmented, completely partitioned mediating powerful voltage attenuation along CA1 dendrites.
postsynaptic density express more AMPA receptors than other J Physiol 568:6982.
axospinous synaptic junctions. Neuroscience 125:615623. Goldsmith SK, Joyce JN (1994) Dopamine D2 receptor expression in
Gasparini S, Kasyanov AM, Pietrobon D, Voronin LL, Cherubini E hippocampus and parahippocampal cortex of rat, cat, and
(2001) Presynaptic R-type calcium channels contribute to fast human in relation to tyrosine hydroxylase-immunoreactive
excitatory synaptic transmission in the rat hippocampus. bers. Hippocampus 4:354373.
J Neurosci 21:87158721. Gonzales RB, DeLeon Galvan CJ, Rangel YM, Claiborne BJ (2001)
Gasparini S, Migliore M, Magee JC (2004) On the initiation and Distribution of thorny excrescences on CA3 pyramidal neurons
propagation of dendritic spikes in CA1 pyramidal neurons. in the rat hippocampus. J Comp Neurol 430:357368.
J Neurosci 24:1104611056. Gould E, Cameron HA (1996) Regulation of neuronal birth,
Gentet LJ, Stuart GJ, Clements JD (2000) Direct measurement of spe- migration and death in the rat dentate gyrus. Dev Neurosci
cic membrane capacitance in neurons. Biophys J 79:314320. 18:2235.
Geiger JRP, Jonas P (2000) Dynamic control of presynaptic Ca2 Gray EG (1959) Electron microscopy of synaptic contacts on den-
inow by fast inactivating K channels in hippocampal mossy drite spines of the cerebral cortex. Nature 183:15921593.
ber boutons. Neuron 28:927939. Greene JRT, Totterdell S (1997) Morphology and distribution of elec-
Geinisman Y (2000) Structural synaptic modications associated trophysiologically dened classes of pyramidal and nonpyrami-
with hippocampal LTP and behavioral learning. Cereb Cortex dal neurons in rat ventral subiculum in vitro. J Comp Neurol
10:952962. 380:395408.
Germroth P, Schwerdtfeger WK, Buhl EH (1991) Ultrastructure and Grutzendler J, Kasthuri N, Gan WB (2002) Long-term dendritic
aspects of functional organization of pyramidal and nonpyra- spine stability in the adult cortex. Nature 420:812816.
midal entorhinal projection neurons contributing to the per- Gulys AI, Miettinen R, Jacobowitz DM, Freund TF (1992) Calretinin
forant path. J Comp Neurol 305:215231. is present in non-pyramidal cells of the rat hippocampus. I. A
Gigg J, Finch DM, OMara SM (2000) Responses of rat subicular neu- new type of neuron specically associated with the mossy bre
rons to convergent stimulation of lateral entorhinal cortex and system. Neuroscience 48:127.
CA1 in vivo. Brain Res 884:3550. Gulyas AI, Miles R, Sik A, Toth K, Tamamaki N, Freund T (1993a)
Gillessen T, Alzheimer C (1997) Amplication of EPSPs by low Ni2- Hippocampal pyramidal cells excite inhibitory neurons through
and amiloride-sensitive Ca2 channels in apical dendrites of rat a single release site. Nature 366:683687.
CA1 pyramidal neurons. J Neurophysiol 77:16391643. Gulys AI, Miles R, Hajos N, Freund TF (1993b) Precision and vari-
Gloveli T, Schmitz D, Empson RM, Dugladze T, Heinemann U ability in postsynaptic target selection of inhibitory cells in the
(1997a) Morphological and electrophysiological characteriza- hippocampal CA3 region. Eur J Neurosci 5:17291751.
tion of layer III cells of the medial entorhinal cortex of the rat. Gulys AI, Hajos N, Freund TF (1996) Interneurons containing cal-
Neuroscience 77:629648. retinin are specialized to control other interneurons in the rat
Gloveli T, Schmitz D, Empson RM, Heinemann U (1997b) hippocampus. J Neurosci 16:33973411.
Frequency-dependent information ow from the entorhinal Gulys AI, Toth K, McBain CJ, Freund TF (1998) Stratum radiatum
cortex to the hippocampus. J Neurophysiol 78:34443449. giant cells: a type of principal cell in the rat hippocampus. Eur
Gloveli T, Egorov AV, Schmitz D, Heinemann U, Mller W (1999) J Neurosci 10:38133822.
Carbachol-induced changes in excitability and [Ca2]i signaling Gulys A, Megias M, Emri Z, Freund TF (1999) Total number and
in projection cells of medial entorhinal cortex layers II and III. ratio of excitatory and inhibitory synapses converging onto sin-
Eur J Neurosci 11:36263636. gle interneurons of different types in the CA1 area of the rat
Gloveli T, Dugladze T, Schmitz D, Heinemann U (2001) Properties of hippocampus. J Neurosci 19:1008210097.
entorhinal cortex deep layer neurons projecting to the rat den- Gulys AI, Hajos N, Katona I, Freund TF (2003) Interneurons are the
tate gyrus. Eur J Neurosci 13:413420. local targets of hippocampal inhibitory cells which project to
Goldberg JH, Tamas G, Aronov D, Yuste R (2003a) Calcium the medial septum. Eur J Neurosci 17:18611872.
microdomains in aspiny dendrites. Neuron 40:807821. Hajos N, Papp EC, Acsady L, Levey AI, Freund TF (1998) Distinct
Goldberg JH, Tamas G, Yuste R (2003b) Ca2 imaging of mouse neo- interneuron types express m2 muscarinic receptor immunore-
cortical interneurone dendrites: Ia-type K channels control activity on their dendrites or axon terminals in the hippocam-
action potential backpropagation. J Physiol 551:4965. pus. Neuroscience 82:355376.
Golding NL, Spruston N (1998) Dendritic spikes are variable triggers Halasy K, Miettinen R, Szabat E, Freund TF (1992) GABAergic inter-
of axonal action potentials in hippocampal CA1 pyramidal neu- neurons are the major postsynaptic targets of median raphe
rons. Neuron 21:11891200. afferents in the rat dentate gyrus. Eur J Neurosci 4:144153.
 Hippocampal Neurons: Structure and Function 193

Halasy K, Buhl EH, Lorinczi Z, Tamas G, Somogyi P (1996) Synaptic Hoffman DA, Johnston D (1998) Downregulation of transient K
target selectivity and input of GABAergic basket and bistratied channels in dendrites of hippocampal CA1 pyramidal neurons
interneurons in the area CA1 of the rat hippocampus. by activation of PKA and PKC. J Neurosci 18:35213528.
Hippocampus 6:306329. Hoffman DA, Magee JC, Colbert CM, Johnston D (1997) K channel
Halliwell JV, Adams PR (1982) Voltage-clamp analysis of muscarinic regulation of signal propagation in dendrites of hippocampal
excitation in hippocampal neurons. Brain Res 250:7192. pyramidal neurons. Nature 387:869875.
Hama K, Arii T, Kosaka T (1989) Three-dimensional morphometri- Hosokawa T, Rusakov DA, Bliss TV, Fine A (1995) Repeated confocal
cal study of dendritic spines of the granule cell in the rat dentate imaging of individual dendritic spines in the living hippocam-
gyrus with HVEM stereo images. J Electron Microsc Tech pal slice: evidence for changes in length and orientation associ-
12:8087. ated with chemically induced LTP. J Neurosci 15:55605573.
Hamam BN, Kennedy TE, Alonso A, Amaral DG (2000) Morphologi- Hotson JR, Prince DA (1980) A calcium-activated hyperpolarization
cal and electrophysiological characteristics of layer V neurons of follows repetitive ring in hippocampal neurons. J Neurophysiol
the rat medial entorhinal cortex. J Comp Neurol 418:457472. 43:409419.
Hamam BN, Amaral DG, Alonso A (2002) Morphological and elec- Hotson JR, Prince DA, Schwartzkroin PA (1979) Anomalous inward
trophysiological characteristics of layer V neurons of the rat lat- rectification in hippocampal neurons. J Neurophysiol
eral entorhinal cortex. J Comp Neurol 451:4561. 42:889895.
Hamlyn LH (1962) The ne structure of the mossy ber endings in Hsu M, Buzski G (1993) Vulnerability of mossy ber targets in the
the hippocampus of the rabbit. J Anat 96:112120. rat hippocampus to forebrain ischemia. J Neurosci
Harris E, Stewart M (2001) Intrinsic connectivity of the rat subicu- 13:39643979.
lum. II. Properties of synchronous spontaneous activity and a Hu H, Vervaeke K, Storm JF (2002) Two forms of electrical resonance
demonstration of multiple generator rhythms. J Comp Neurol at theta frequencies, generated by M-current, h-current and
436:506518. persistent Na current in rat hippocampal pyramidal cells.
Harris E, Witter MP, Weinstein G, Stewart M (2001) Intrinsic con- J Physiol 545:783805.
nectivity of the rat subiculum. I. Dendritic morphology and Ishizuka N, Weber J, Amaral DG (1990) Organization of intrahip-
patterns of axonal arborization by pyramidal neurons. J Comp pocampal projections originating from CA3 pyramidal cells in
Neurol 436:490505. the rat. J Comp Neurol 295:580623.
Harris KM, Kater SB (1994) Dendritic spines: cellular specializations Ishizuka N, Cowan WM, Amaral DG (1995) A quantitative analysis
imparting both stability and exibility to synaptic function. of the dendritic organization of pyramidal cells in the rat hip-
Annu Rev Neurosci 17:341371. pocampus. J Comp Neurol 362:1745.
Harris KM, Stevens JK (1989) Dendritic spines of CA1 pyramidal Isaac JT (2003) Postsynaptic silent synapses: evidence and mecha-
cells in the rat hippocampus: serial electron microscopy with nisms. Neuropharmacology 45:450460.
reference to their biophysical characteristics. J Neurosci Isaac JT, Nicoll RA, Malenka RC (1995) Evidence for silent synapses:
9:29822997. implications for the expression of LTP. Neuron 15:427434.
Harris KM, Jensen FE, Tsao B (1992) Three-dimensional structure of Jaffe DB, Brown TH (1997) Calcium dynamics in thorny excrescences
dendritic spines and synapses in rat hippocampus (CA1) at of CA3 pyramidal neurons. J Neurophysiol 78:1018.
postnatal day 15 and adult ages: implications for the maturation Jaffe DB, Carnevale NT (1999) Passive normalization of synaptic
of synaptic physiology and long-term potentiation. J Neurosci integration inuenced by dendritic architecture. J Neurophysiol
12:26852705. 82:32683285.
Hausser M, Spruston N, Stuart GJ (2000) Diversity and dynamics of Jaffe DB, Johnston D, Lasser-Ross N, Lisman JE, Miyakawa H, Ross
dendritic signaling. Science 290:739744. WN (1992) The spread of Na spikes determines the pattern of
Helmchen F, Imoto K, Sakmann B (1996) Ca2 buffering and action dendritic Ca2 entry into hippocampal neurons. Nature
potential-evoked Ca2 signaling in dendrites of pyramidal neu- 357:244246.
rons. Biophys J 70:10691081. Jarsky T, Roxin A, Kath WL, Spruston N (2005) Conditional dendritic
Henze DA, Buzski G (2001) Action potential threshold of hip- spike propagation following distal synaptic activation of hip-
pocampal pyramidal cells in vivo is increased by recent spiking pocampal CA1 pyramidal neurons. Nat Neurosci 8:16671676.
activity. Neuroscience 105:121130. Jay TM, Glowinski J, Thierry AM (1989) Selectivity of the hip-
Henze DA, Cameron WE, Barrionuevo GJ (1996) Dendritic mor- pocampal projection to the prelimbic area of the prefrontal cor-
phology and its effects on the amplitude and rise-time of synap- tex in the rat. Brain Res 505:337340.
tic signals in hippocampal CA3 pyramidal cells. J Comp Neurol Jefferys JGR (1979) Initiation and spread of action potentials in gran-
369:331344. ule cells maintained in vitro in slices of guinea-pig hippocam-
Henze DA, Urban NN, Barrionuevo G (2000) The multifarious hip- pus. J Physiol (Lond) 289:375388.
pocampal mossy fiber pathway: a review. Neuroscience Jensen MS, Azouz R, Yaari Y (1994) Variant ring patterns in rat hip-
98:407427. pocampal pyramidal cells modulated by extracellular potas-
Hering H, Sheng M (2001) Dendritic spines: structure, dynamics and sium. J Neurophysiol 71:831839.
regulation. Nat Rev Neurosci 2:880888. Jiang C, Schuman EM (2002) Regulation and function of local pro-
Herreras O (1990) Propagating dendritic action potential mediates tein synthesis in neuronal dendrites. Trends Biochem Sci
synaptic transmission in CA1 pyramidal cells in situ. J 27:506513.
Neurophysiol 64:14291441. Johnston D (1981) Passive cable properties of hippocampal CA3
Hjorth-Simonsen A (1973) Some intrinsic connections of the hip- pyramidal neurons. Cell Mol Neurobiol 1:4145.
pocampus in the rat: an experimental analysis. J Comp Neurol Johnston D, Brown TH (1981) Giant synaptic potential hypothesis
147:145161. for epileptiform activity. Science 211:294297.
194 The Hippocampus Book

Johnston D, Brown TH (1984) The synaptic nature of the paroxysmal Koch C, Zador A (1993) The function of dendritic spines: devices
depolarizing shift in hippocampal neurons. Ann Neurol subserving biochemical rather than electrical compartmental-
16(Suppl):S65S71. ization. J Neurosci 13:413422.
Johnston D, Brown TH (1986) Control theory applied to neural net- Khler C (1985) A projection from the deep layers of the entorhinal
works illuminates synaptic basis of interictal epileptiform activ- area to the hippocampal formation in the rat brain. Neurosci
ity. Adv Neurol 44:263274. Lett 56:1319.
Jonas P, Major G, Sakmann B (1993) Quantal components of unitary Korngreen A, Sakmann B (2000) Voltage-gated K channels in layer
EPSCs at the mossy bre synapse on CA3 pyramidal cells of rat 5 neocortical pyramidal neurones from young rats: subtypes
hippocampus. J Physiol (Lond) 472:615663. and gradients. J Physiol 525:621639.
Jonas P, Bischofberger J, Fricker D, Miles R (2004) Interneuron diver- Kosaka T (1983) Axon initial segments of the granule cell in the rat
sity series: fast in, fast outtemporal and spatial signal process- dentate gyrus: synaptic contacts on bundles of axon initial seg-
ing in hippocampal interneurons. Trends Neurosci 27:3040. ments. Brain Res 274:129134.
Jones R (1994) Synaptic and intrinsic properties of neurons of ori- Kovalchuk Y, Eilers J, Lisman J, Konnerth A (2000) NMDA receptor-
gin of the perforant path in layer II of the rat entorhinal cortex mediated subthreshold Ca2 signals in spines of hippocampal
in vitro. Hippocampus 4:335353. neurons. J Neurosci 20:17911799.
Jones RSG, Heinemann U (1988) Synaptic and intrinsic responses of Krettek JE, Price JL (1977) The cortical projections of the mediodor-
medial entorhinal cortical cells in normal and magnesium-free sal nucleus and adjacent thalamic nuclei in the rat. J Comp
medium in vitro. J Neurophysiol 59:14761496. Neurol 171:157191.
Jung H, Mickus T, Spruston N (1997) Prolonged sodium channel Laatsch RH, Cowan WM (1966) Electron microscopic studies of the
inactivation contributes to dendritic action potential attenua- dentate gyrus of the rat. I. Normal structure with special refer-
tion in hippocampal pyramidal neurons. J Neurosci ence to synaptic organization. J Comp Neurol 128:359395.
17:66396646. Lancaster B, Adams PR (1986) Calcium-dependent current generat-
Jung H, Staff NP, Spruston N (2001) Action potential bursting in ing the afterhyperpolarization of hippocampal neurons.
subicular pyramidal neurons is driven by a calcium tail current. J Neurophysiol 55:12681282.
J Neurosci 21:33123321. Lancaster B, Nicoll RA (1987) Properties of two calcium-activated
Jung MW, McNaughton BL (1993) Spatial selectivity of unit hyperpolarizations in rat hippocampal neurones. J Physiol
activity in the hippocampal granular layer. Hippocampus 389:187203.
3:165182. Lawrence JJ, McBain CJ (2003) Interneuron diversity series: contain-
Jung R, Kornmueller AE (1938) Eine Metodik der Ableitung ing the detonationfeedforward inhibition in the CA3 hip-
lokalisierter Potentialschwankungen aus subcorticalen pocampus. Trends Neurosci 26:631640.
Hirngebieten. Arch Psychiatr Nervenkr 109:130. Lawrence JJ, Grinspan ZM, McBain CJ (2004) Quantal transmission
Kahle JS, Cotman CW (1989) Carbachol depresses synaptic responses at mossy bre targets in the CA3 region of the rat hippocampus.
in the medial but not the lateral perforant path. Brain Res J Physiol 554:175193.
482:159163. Lawrence JJ, Statland JM, Grinspan ZM, McBain CJ (2006) Cell
Kali S, Freund TF (2005) Distinct properties of two major excitatory type-specic dependence of muscarinic signaling in mouse
inputs to hippocampal pyramidal cells: a computational study. hippocampal stratum oriens interneurons. J Physiol 570:
Eur J Neurosci 22:20272048. 595610.
Kamondi A, Acsady L, Buzsaki G (1998) Dendritic spikes are Lebeda FJ, Hablitz JJ, Johnston D (1982) Antagonism of GABA-
enhanced by cooperative network activity in the intact hip- mediated responses by d-tubocurarine in hippocampal neu-
pocampus. J Neurosci 18:39193928. rons. J Neurophysiol 48:622632.
Kandel ER, Spencer WA (1961) Electrophysiology of hippocampal Lei S, McBain CJ (2002) Distinct NMDA receptors provide differen-
neurons. II. Afterpotentials and repetitive ring. J Neurophysiol tial modes of transmission at mossy ber-interneuron synapses.
24:243259. Neuron 33:921933.
Kemppainen S, Jolkkonen E, Pitkanen A (2002) Projections from the Lei S, McBain CJ (2004) Two loci of expression for long-term depres-
posterior cortical nucleus of the amygdala to the hippocampal sion at hippocampal mossy fiber-interneuron synapses. J
formation and parahippocampal region in rat. Hippocampus Neurosci 24:21122121.
12:735755. Leung LS, Yu HW (1998) Theta-frequency resonance in hippocampal
Klink R, Alonso A (1993) Ionic mechanisms for the subthreshold CA1 neurons in vitro demonstrated by sinusoidal current injec-
oscillations and differential electroresponsiveness of medial tion. J Neurophysiol 79:15921596.
entorhinal cortex layer II neurons. J Neurophysiol 70:144157. Levy WB, Steward O (1979) Synapses as associative memory elements
Klink R, Alonso A (1997a) Morphological characteristics of layer II in the hippocampal formation. Brain Res 175:233245.
projection neurons in the rat medial entorhinal cortex. Levy WB, Colbert CM, Desmond NL (1995) Another network model
Hippocampus 7:571583. bites the dust: entorhinal inputs are no more than weakly exci-
Klink R, Alonso A (1997b) Muscarinic modulation of the oscillatory tatory in the hippocampal CA1 region. Hippocampus
and repetitive ring properties of entorhinal cortex layer II neu- 5:137140.
rons. J Neurophysiol 77:18131828. Li XG, Somogyi P, Tepper JM, Buzsaki G (1992) Axonal and dendritic
Klink R, Alonso A (1997c) Ionic mechanisms of muscarinic depolar- arborization of an intracellularly labeled chandelier cell in the
ization in entorhinal cortex layer II neurons. J Neurophysiol CA1 region of rat hippocampus. Exp Brain Res 90:519525.
77:18291843. Li XG, Somogyi P, Ylinen A, Buzsaki G (1994) The hippocampal CA3
Knowles WD, Schwartzkroin PA (1981) Axonal ramications of hip- network: an in vivo intracellular labeling study. J Comp Neurol
pocampal Ca1 pyramidal cells. J Neurosci 1:12361241. 339:181208.
 Hippocampal Neurons: Structure and Function 195

Liao D, Hessler NA, Malinow R (1995) Activation of postsynaptically Madison D, Nicoll RA (1984) Control of the repetitive discharge of
silent synapses during pairing-induced LTP in CA1 region of rat CA1 pyramidal neurons in vitro. J Physiol 354:319331.
hippocampal slice. Nature 375:400404. Madison DV, Lancaster B, Nicoll RA (1987) Voltage clamp analysis of
Lingenhhl K, Finch DM (1991) Morphological characterization of cholinergic action in the hippocampus. J Neurosci 7:733741.
rat entorhinal neurons in vivo: soma-dendritic structure and Magee JC (1998) Dendritic hyperpolarization-activated currents
axonal domains. Exp Brain Res 84:5774. modify the integrative properties of hippocampal CA1 pyrami-
Lipowsky R, Gillessen T, Alzheimer C (1996) Dendritic Na channels dal neurons. J Neurosci 18:76137624.
amplify EPSPs in hippocampal CA1 pyramidal cells. J Magee JC (1999) Dendritic Ih normalizes temporal summation in
Neurophysiol 76:21812191. hippocampal CA1 neurons. Nat Neurosci 2:508514.
Liu YB, Lio PA, Pasternak JF, Trommer BL (1996) Developmental Magee JC, Carruth M (1999) Dendritic voltage-gated ion channels
changes in membrane properties and postsynaptic currents of regulate the action potential ring mode of hippocampal CA1
granule cells in rat dentate gyrus. J Neurophysiol 76:10741088. pyramidal neurons. J Neurophysiol 82:18951901.
Livsey CT, Vicini S (1992) Slower spontaneous excitatory postsynap- Magee JC, Cook EP (2000) Somatic EPSP amplitude is independent
tic currents in spiny versus aspiny hilar neurons. Neuron of synapse location in hippocampal pyramidal neurons. Nat
8:745755. Neurosci 3:895903.
Livsey CT, Costa E, Vicini S (1993) Glutamate-activated currents in Magee JC, Johnston D (1995a) Characterization of single voltage-
outside-out patches from spiny versus aspiny hilar neurons of gated Na and Ca2 channels in apical dendrites of rat CA1
rat hippocampal slices. J Neurosci 13:53245333. pyramidal neurons. J Physiol 487:6790.
Llinas RR (1988) The intrinsic electrophysiological properties of Magee JC, Johnston D (1995b) Synaptic activation of voltage-gated
mammalian neurons: insights into central nervous system func- channels in the dendrites of hippocampal pyramidal neurons.
tion. Science 242:16541664. Science 68:301304.
Lopes da Silva FH, Witter MP, Boeijinga PH, Lohman AH (1990) Magee JC, Johnston D (1997) A synaptically controlled, associative
Anatomic organization and physiology of the limbic cortex. signal for Hebbian plasticity in hippocampal neurons. Science
Physiol Rev 76:453511. 275:209213.
Lorente de No R (1933) Studies on the structure of the cerebral cor- Magistretti J, Alonso A (1999) Biophysical properties and slow volt-
tex. I. The area entorhinalis. J Psychol Neurol 45:381438. age-dependent inactivation of a sustained sodium current in
Lorente de No R (1934) Studies on the structure of the cerebral cor- entorhinal cortex layer-II principal neurons: a whole-cell and
tex. II. Continuation of the study of ammonic system. J Psychol single-channel study. J Gen Physiol 115:491509.
Neurol 46:113177. Magistretti J, Ragsdale DS, Alonso A (1999a) High conductance sus-
Lorincz A, Notomi T, Tamas G, Shigemoto R, Nusser Z (2002) tained single-channel activity responsible for the low-threshold
Polarized and compartment-dependent distribution of HCN1 persistent Na current in entorhinal cortex neurons. J Neurosci
in pyramidal cell dendrites. Nat Neurosci 5:11851193. 19:73347341.
Lbke J, Frotscher M, Spruston N (1998) Specialized electrophysio- Magistretti J, Ragsdale DS, Alonso A (1999b) Direct demonstration
logical properties of anatomically identied neurons in the hilar of persistent Na channel activity in dendritic processes of
region of the rat fascia dentata. J Neurophysiol 79:15181534. mammalian cortical neurones. J Physiol 521:629636.
Lujan R, Nusser Z, Roberts JD, Shigemoto R, Somogyi P (1996) Mainen ZF, Carnevale NT, Zador AM, Claiborne BJ, Brown TH
Perisynaptic location of metabotropic glutamate receptors (1996) Electrotonic architecture of hippocampal CA1 pyrami-
mGluR1 and mGluR5 on dendrites and dendritic spines in the dal neurons based on three-dimensional reconstructions.
rat hippocampus. Eur J Neurosci 8:14881500. J Neurophysiol 76:19041923.
Maccaferri G, McBain CJ (1996) Long-term potentiation in distinct Majewska A, Brown E, Ross J, Yuste R (2000) Mechanisms of cal-
subtypes of hippocampal nonpyramidal neurons. J Neurosci cium decay kinetics in hippocampal spines: role of spine
16:53345343. calcium pumps and calcium diffusion through the spine
Maccaferri G, Lacaille JC (2003) Interneuron diversity series: hip- neck in biochemical compartmentalization. J Neurosci
pocampal interneuron classicationsmaking things as simple 20:17221734.
as possible, not simpler. Trends Neurosci 26:564571. Major G, Larkman AU, Jonas P, Sakmann B, Jack JJ (1994) Detailed
Maccaferri G, Toth K, McBain CJ (1998) Target-specic expression of passive cable models of whole-cell recorded CA3 pyramidal
presynaptic mossy ber plasticity. Science 279:13681370. neurons in rat hippocampal slices. J Neurosci 14:46134638.
Maccaferri G, Roberts JD, Szucs P, Cottingham CA, Somogyi P Maletic-Savatic M, Lenn NJ, Trimmer JS (1995) Differential spa-
(2000) Cell surface domain specic postsynaptic currents tiotemporal expression of K channel polypeptides in rat hip-
evoked by identied GABAergic neurones in rat hippocampus pocampal neurons developing in situ and in vitro. J Neurosci
in vitro. J Physiol 524:91116. 15:38403851.
Macek TA, Winder DG, Gereau RW 4th, Ladd CO, Conn PJ (1996) Maletic-Savatic M, Malinow R, Svoboda K (1999) Rapid dendritic
Differential involvement of group II and group III mGluRs as morphogenesis in CA1 hippocampal dendrites induced by
autoreceptors at lateral and medial perforant path synapses. J synaptic activity. Science 283:19231927.
Neurophysiol 76:37983806. Markram H, Wang Y, Tsodyks M (1998) Differential signaling via the
MacVicar BA, Dudek FE (1981) Electrotonic coupling between same axon of neocortical pyramidal neurons. Proc Natl Acad Sci
pyramidal cells: a direct demonstration in rat hippocampal USA 95:53235328.
slices. Science 213:782785. Marrion NV, Tavalin SJ (1998) Selective activation of Ca2-activated
Madison DV, Nicoll RA (1982) Noradrenaline blocks accommoda- K channels by co-localized Ca2 channels in hippocampal
tion of pyramidal cell discharge in the hippocampus. Nature neurons. Nature 395:900905.
299:636638. Marshall L, Henze DA, Hirase H, Leinekugel X, Dragoi G, Buzsaki G
196 The Hippocampus Book

(2002) Hippocampal pyramidal cell-interneuron spike trans- distribution of inhibitory and excitatory synapses on hip-
mission is frequency dependent and responsible for place mod- pocampal CA1 pyramidal cells. Neuroscience 102:527540.
ulation of interneuron discharge. J Neurosci 22:RC197(15). Menendez de la Prida L (2003) Control of bursting by local inhibi-
Martina M, Jonas P (1997) Functional differences in Na channel tion in the rat subiculum in vitro. J Physiol 549:219230.
gating between fast-spiking interneurones and principal neu- Menendez de la Prida L, Suarez F, Pozo MA (2003) Electro-
rones of rat hippocampus. J Physiol 505:593603. physiological and morphological diversity of neurons from the
Martina M, Vida I, Jonas P (2000) Distal initiation and active propa- rat subicular complex in vitro. Hippocampus 13:728744.
gation of action potentials in interneuron dendrites. Science Metz A, Jarsky T, Martina M, Spruston N (2005) R-type calcium
287:295300. channels produce an afterdepolarization and bursting in
Mason A (1993) Electrophysiology and burst-ring of rat subicular hippocampal CA1 pyramidal neurons. J Neurosci 25:5763
pyramidal neurons in vitro: a comparison with area CA1. Brain 5773.
Res 600:174178. Mickus T, Jung H, Spruston N (1999) Properties of slow, cumulative
Masukawa LM, Prince DA (1984) Synaptic control of excitability in sodium channel inactivation in rat hippocampal CA1 pyrami-
isolated dendrites of hippocampal neurons. J Neurosci dal cells. Biophys J 76:846860.
4:217227. Migliore M, Cook EP, Jaffe DB, Turner DA, Johnston D (1995)
Masukawa LM, Benardo LS, Prince DA (1982) Variations in electro- Computer simulations of morphologically reconstructed CA3
physiological properties of hippocampal neurons in different hippocampal neurons. J Neurophysiol 73:11571168.
subelds. Brain Res 242:341344. Mikkonen M, Pitknen A, Soininen H, Alafuzoff I, Miettinen R
Matsuzaki M, Ellis-Davies GC, Nemoto T, Miyashita Y, Iino M, Kasai (2000) Morphology of spiny neurons in the human entorhinal
H (2001) Dendritic spine geometry is critical for AMPA recep- cortex: intracellular lling with lucifer yellow. Neuroscience
tor expression in hippocampal CA1 pyramidal neurons. Nat 96:515522.
Neurosci 4:10861092. Miles R (1990) Variation in strength of inhibitory synapses in the
Matsuzaki M, Honkura N, Ellis-Davies GC, Kasai H (2004) CA3 region of guinea-pig hippocampus in vitro. J Physiol
Structural basis of long-term potentiation in single dendritic (Lond) 431:659676.
spines. Nature 429:761766. Miles R, Toth K, Gulyas AI, Hajos N, Freund TF (1996) Differences
Mattia D, Hwa GGC, Avoli M (1993) Membrane properties of rat between somatic and dendritic inhibition in the hippocampus.
subicular neurons in vitro. J Neurophysiol 70:12441248. Neuron 16:815823.
Mattia D, Kawasaki H, Avoli M (1997a) In vitro electrophysiology of Miyakawa H, Ross WN, Jaffe D, Callaway JC, Lasser-Ross N, Lisman
rat subicular bursting neurons. Hippocampus 7:4857. JE, Johnston D (1992) Synaptically activated increases in Ca2
Mattia D, Kawasaki H, Avoli M (1997b) Repetitive ring and oscilla- concentration in hippocampal CA1 pyramidal cells are prima-
tory activity of pyramidal-like bursitng neurons in the rat rily due to voltage-gated Ca2 channels. Neuron 9:11631173.
subiculum. Exp Brain Res 114:507517. Monaghan DT, Holets VR, Toy DW, Cotman CW (1983) Anatomical
Mtys F., Freund TF, Gulys AI (2004) Convergence of excita- distributions of four pharmacologically distinct 3H-L-glutamate
tory and inhibitory inputs onto CCK-containing basket cells binding sites. Nature 306:176179.
in the CA1 area of the rat hippocampus. Eur J Neurosci 19: Moosmang S, Biel M, Hofmann F, Ludwig A (1999) Differential dis-
1243. tribution of four hyperpolarization-activated cation channels in
McBain CJ, Fisahn A (2001) Interneurons unbound. Nat Rev Neurosci mouse brain. Biol Chem 380:975980.
2:1123. Murthy VN, Sejnowski TJ, Stevens CF (2000) Dynamics of dendritic
McBain CJ, Freund T, Mody I (1999) Glutamatergic synapses onto calcium transients evoked by quantal release at excitatory hip-
hippocampal interneurons: precision timing without lasting pocampal synapses. Proc Natl Acad Sci USA 97:901906.
plasticity. Trends Neurosci 22:228235. Naber PA, Witter MP (1998) Subicular efferents are organized mostly
McGinty JF, Henriksen SJ, Goldstein A, Terenius L, Bloom FE (1983) as parallel projections: a double labeling, retrograde-tracing
Dynorphin is contained within hippocampal mossy bers: study in the rat. J Comp Neurol 393:284297.
immunochemical alterations after kainic acid administration Naber PA, Witter MP, Lopes da Silva FH (2000) Networks of hip-
and colchicine-induced neurotoxicity. Proc Natl Acad Sci USA pocampal memory system of the rat. Ann NY Acad Sci
80:589593. 911:392403.
McNaughton BL, Douglas RM, Goddard GV (1978) Synaptic Nafstad PHJ, Blackstad TW (1966) Distribution of mitochondria in
enhancement in fascia dentata: cooperativity among coactive pyramidal cells and boutons in hippocampal cortex. Z Zellforsch
afferents. Brain Res 157:277293. Mikrosk Anat 3:234245.
McNaughton BL, Barnes CA, Meltzer J, Sutherland RJ (1989) Nakamura T, Barbara JG, Nakamura K, Ross WN (1999) Synergistic
Hippocampal granule cells are necessary for normal spatial release of Ca2 from IP3sensitive stores evoked by synaptic
learning but not for spatially-selective pyramidal cell discharge. activation of mGluRs paired with backpropagating action
Exp Brain Res 76:485496. potentials. Neuron 24:727737.
McQuiston AR, Madison DV (1999a) Muscarinic receptor activity Nakamura T, Nakamura K, Lasser-Ross N, Barbara JG, Sandler VM,
induces an afterdepolarization in a subpopulation of hip- Ross WN (2000) Inositol 1,4,5trisphosphate (IP3)-mediated
pocampal CA1 interneurons. J Neurosci 19:57035710. Ca2 release evoked by metabotropic agonists and backpropa-
McQuiston AR, Madison DV (1999b) Muscarinic receptor activity gating action potentials in hippocampal CA1 pyramidal neu-
has multiple effects on the resting membrane potentials of CA1 rons. J Neurosci 20:83658376.
hippocampal interneurons. J Neurosci 19:56935702. Nicholson D, Katz Y, Trana R, Kath WL, Spruston N, Geinisman Y
Megias M, Emri Z, Freund TF, Gulyas AI (2001) Total number and (2006) Distance-dependent differences in synapse number and
 Hippocampal Neurons: Structure and Function 197

AMPA receptor expression in hippocampal CA1 pyramidal NMDA receptor-mediated EPSPs in CA1 hippocampal pyrami-
neurons. Neuron 50:431442. dal cells. Brain Res 524:342346.
Nikonenko I, Jourdain P, Alberi S, Toni N, Muller D (2002) Activity- Poolos NP, Johnston D (1999) Calcium-activated potassium conduc-
induced changes of spine morphology. Hippocampus 12: tances contribute to action potential repolarization at the soma
585591. but not the dendrites of hippocampal CA1 pyramidal neurons.
Nimchinsky EA, Sabatini BL, Svoboda K (2002) Structure and func- J Neurosci 19:52055212.
tion of dendritic spines. Annu Rev Physiol 64:313353. Popov VI, Davies HA, Rogachevsky VV, Patrushev IV, Errington ML,
Nimchinsky EA, Yasuda R, Oertner TG, Svoboda K (2004) The num- Gabbott PL, Bliss TV, Stewart MG (2004) Remodelling of
ber of glutamate receptors opened by synaptic stimulation in synaptic morphology but unchanged synaptic density during
single hippocampal spines. J Neurosci 24:20542064. late phase long-term potentiation (LTP): a serial section elec-
Nusser Z, Lujan R, Laube G, Roberts JD, Molnar E, Somogyi P (1998) tron micrograph study in the dentate gyrus in the anaesthetised
Cell type and pathway dependence of synaptic AMPA receptor rat. Neuroscience 128:251262.
number and variability in the hippocampus. Neuron Pouille F, Scanziani M (2001) Enforcement of temporal delity in
21:545559. pyramidal cells by somatic feed-forward inhibition. Science
Oertner TG, Sabatini BL, Nimchinsky EA, Svoboda K (2002) 293:11591163.
Facilitation at single synapses probed with optical quantal Pouille F, Scanziani M (2004) Routing of spike series by dynamic cir-
analysis. Nat Neurosci 5:657664. cuits in the hippocampus. Nature 429:717723.
Oliva AA Jr, Jiang M, Lam T, Smith KL, Swann JW (2000). Novel hip- Pozzo-Miller LD, Petrozzino JJ, Mahanty NK, Connor JA (1993)
pocampal interneuronal subtypes identied using transgenic Optical imaging of cytosolic calcium, electrophysiology, and
mice that express green uorescent protein in GABAergic ultrastructure in pyramidal neurons of organotypic slice cul-
interneurons. J Neurosci 20:33543368. tures from rat hippocampus. Neuroimage 1:109120.
OMara SM, Commins S, Anderson M, Gigg J (2001) The subiculum: Pozzo-Miller LD, Petrozzino JJ, Golarai G, Connor JA (1996) Ca2
a review of form, physiology and function. Prog Neurobiol release from intracellular stores induced by afferent stimulation
64:129155. of CA3 pyramidal neurons in hippocampal slices. J Neurophysiol
Otmakhova NA, Lisman JE (1999) Dopamine selectively inhibits the 76:554562.
direct cortical pathway to the CA1 hippocampal region. J Prince DA, Connors BW (1986) Mechansims of interictal epileptoge-
Neurosci 19:14371445. nesis. Adv Neurol 44:275299.
Otmakhova NA, Lisman JE (2000) Dopamine, serotonin, and nora- Pyapali GK, Sik A, Penttonen M, Buzsaki G, Turner DA (1998)
drenaline strongly inhibit the direct perforant path-CA1 synap- Dendritic properties of hippocampal CA1 pyramidal neurons in
tic input, but have little effect on the Schaffer collateral input. the rat: intracellular staining in vivo and in vitro. J Comp Neurol
Ann NY Acad Sci 911:462464. 391:335352.
Par D, Llins R (1994) Non-lamellar propagation of entorhinal Qian J, Noebels JL (2001) Presynaptic Ca2 channels and neuro-
inuences in the hippocampal formation: multiple electrode transmitter release at the terminal of a mouse cortical neuron. J
recordings in the isolated guinea pig brain in vitro. Neurosci 21:37213728.
Hippocampus 4:403409. Racca C, Stephenson FA, Streit P, Roberts JD, Somogyi P (2000)
Parnass Z, Tashiro A, Yuste R (2000) Analysis of spine morphological NMDA receptor content of synapses in stratum radiatum of the
plasticity in developing hippocampal neurons. Hippocampus hippocampal CA1 area. J Neurosci 20:2512222.
10:561568. Rall W (1959) Branching dendritic trees and motoneuron membrane
Parra P, Gulyas AI, Miles R (1998) How many subtypes of inhibitory resistivity. Exp Neurol 1:491527.
cells in the hippocampus? Neuron 20:983993. Rall W (1964) Theoretical signicance of dendritic trees for neuronal
Pasquier DA, Reinoso-Suarez F (1978) The topographic organization input-output relations. In: Neural theory and modeling (Reiss
of hypothalamic and brain stem projections to the hippocam- RF, ed). Palo Alto: Stanford University Press.
pus. Brain Res Bull 3:373389. Rall W (1969) Time constants and electrotonic length of membrane
Pearce RA (1993) Physiological evidence for two distinct GABAA cylinders and neurons. Biophys J 9:14831508.
responses in rat hippocampus. Neuron 10:189200. Ramon y Cajal S (1904, 1995) Histology of the nervous system, vol 2
Pedarzani P, Storm JF (1996) Evidence that Ca/calmodulin-depend- (Swanson N, Swanson L, translators). New York: Oxford
ent protein kinase mediates the modulation of the Ca2- University Press.
dependent K current, IAHP, by acetylcholine, but not by Ranck JB (1973) Studies on single neurons in dorsal hippocampal
glutamate, in hippocampal neurons. Pugers Arch 431:723728. formation and septum in unrestrained rats. Part I. Behavioral
Pelletier MR, Kirkby RD, Jones SJ, Corcoran ME (1994) Pathway correlates and ring repertoires. Exp Neurol 41:461531.
specicity of noradrenergic plasticity in the dentate gyrus. Ratzliff AdH, Santhakumar V, Howard A, Soltesz I (2002) Mossy cells
Hippocampus 4:181188. in epilepsy: rigor mortis or vigor mortis? Trends Neurosci
Penttonen M, Kamondi A, Sik A, Acsady L, Buzsaki G (1997) Feed- 25:140144.
forward and feed-back activation of the dentate gyrus in vivo Regehr WG, Tank DW (1992) Calcium concentration dynamics pro-
during dentate spikes and sharp wave bursts. Hippocampus duced by synaptic activation of CA1 hippocampal pyramidal
7:437450. cells. J Neurosci 12:42024223.
Peters A, Kaiserman-Abramof IR (1970) The small pyramidal neuron Regehr WG, Connor JA, Tank DW (1989) Optical imaging of calcium
of the rat cerebral cortex: the perikaryon, dendrites and spines. accumulation in hippocampal pyramidal cells during synaptic
Am J Anat 127:321355. activation. Nature 341:533536.
Poolos NP, Kocsis JD (1990) Dendritic action potentials activated by Reid CA, Fabian-Fine R, Fine A (2001) Postsynaptic calcium tran-
198 The Hippocampus Book

sients evoked by activation of individual hippocampal mossy Scharfman HE, Schwartzkroin PA (1988) Electrophysiology of mor-
ber synapses. J Neurosci 21:22062214. phologically identied mossy cells of the dentate hilus recorded
Remondes M, Schuman EM (2002) Direct cortical input modulates in guinea pig hippocampal slices. J Neurosci 8:38123821.
plasticity and spiking in CA1 pyramidal neurons. Nature Schiller J, Helmchen F, Sakmann B (1995) Spatial prole of dendritic
416:736740. calcium transients evoked by action potentials in rat neocortical
Riazanski V, Becker A, Chen J, Sochivko D, Lie A, Wiestler OD, Elger pyramidal neurones. J Physiol 487:583600.
CE, Beck H (2001) Functional and molecular analysis of tran- Schmitz D, Schuchmann S, Fisahn A, Draguhn A, Buhl EH, Petrasch-
sient voltage-dependent K currents in rat hippocampal gran- Parwez E, Dermietzel R, Heinemann U, Traub RD (2001) Axo-
ule cells J Physiol 537:391406. axonal coupling: a novel mechanism for ultrafast neuronal
Ribak CE, Seress L, Amaral DG (1985) The development, ultrastruc- communication. Neuron 31:831840.
ture and synaptic connections of the mossy cells of the dentate Schwartz SP, Coleman PD (1981) Neurons of origin of the perforant
gyrus. J Neurocytol 14:835857. path. Exp Neurol 74:305312.
Rolls ET (1996) A theory of hippocampal function in memory. Schwartzkroin PA (1977) Further characteristics of hippocampal
Hippocampus 6:601620. CA1 cells in vitro. Brain Res 128:5368.
Ropert N (1988) Inhibitory action of serotonin in CA1 hippocampal Schwartzkroin PA (1993) Epilepsy; models, mechanisms and concepts.
neurons in vitro. Neuroscience 26:6981. Cambridge: Cambridge University Press.
Rudy B, McBain CJ (2001) Kv3 channels: voltage-gated K channels Schwartzkroin PA, Slawsky M (1977) Probable calcium spikes in hip-
designed for high-frequency repetitive ring. Trends Neurosci pocampal neurons. Brain Res 135:157161.
24:517526. Schwartzkroin PA, Stafstrom CE (1980) Effects of EGTA on the cal-
Sabatini BL, Oertner TG, Svoboda K (2002) The life cycle of Ca2 cium-activated afterhyperpolarization in hippocampal CA3
ions in dendritic spines. Neuron 33:439452. cells. Science 210:11251126.
Salin PA, Scanziani M, Malenka RC, Nicoll RA (1996) Distinct short- Segev I, Rinzel J, Shepherd GM (1995) The theoretical foundation of
term plasticity at two excitatory synapses in the hippocampus. dendritic function: selected papers of Wilfrid Rall with commen-
Proc Natl Acad Sci USA 93:1330413309. taries. Cambridge, MA: MIT Press.
Sandler R, Smith AD (1991) Coexistence of GABA and glutamate in Shalinsky MH, Magistretti J, Ma L, Alonso A (2002) Muscarinic acti-
mossy ber terminals of the primate hippocampus: an ultra- vation of a cation current and associated current noise in
structural study. J Comp Neurol 303:177192. entorhinal-cortex layer-II neurons. J Neurophysiol
Sandler VM, Barbara JG (1999) Calcium-induced calcium release 88:11971211.
contributes to action potential-evoked calcium transients in Sharp PE, Green C (1994) Spatial correlates of ring patterns of sin-
hippocampal CA1 pyramidal neurons. J Neurosci 19:4325 gle cells in the subiculum of the freely moving rat. J Neurosci
4336. 14:23392356.
Sandler VM, Ross WN (1999) Serotonin modulates spike backprop- Sheng M (2001) Molecular organization of the postsynaptic special-
agation and associated [Ca2]i changes in the apical dendrites of ization. Proc Natl Acad Sci USA 98:70587061.
hippocampal CA1 pyramidal neurons. J Neurophysiol 81: Sheng M, Tsaur M-L, Jan YN, Jan LY (1992) Subcellular segregation
216224. of two A-type K channel proteins in rat central neurons.
Santoro B, Grant SG, Bartsch D, Kandel ER (1997) Interactive cloning Neuron 9:271284.
with the SH3 domain of N-src identies a new brain specic ion Shepherd GMG, Harris KM (1998) Three dimensional structure and
channel protein, with homology to eag and cyclic nucleotide- composition of CA3-CA1 axons in rat hippocampal slices:
gated channels. Proc Natl Acad Sci USA 94:1481514820. implications for presynaptic connectivity and compartmental-
Santoro B, Chen S, Luthi A, Pavlidis P, Shumyatsky GP, Tibbs GR, ization. J Neurosci 18:83008310.
Siegelbaum SA (2000) Molecular and functional heterogeneity Shepherd GMG, Raastad M, Andersen P (2002) General and variable
of hyperpolarization-activated pacemaker channels in the features of varicosity spacing along unmyelinated axons in the
mouse CNS. J Neurosci 20:52645275. hippocampus and cerebellum. Proc Natl Acad Sci USA
Scanziani M, Gahwilar BH, Charpak S (1998) Target cell-specic 99:63406345.
modulation of transmitter release at terminals from a single Sik A, Tamamaki N, Freund TF (1993) Complete axon arborization
axon. Proc Natl Acad Sci USA 95:1200412009. of a single CA3 pyramidal cell in the rat hippocampus, and its
Schaffer K (1892) Beitrag zur Histologie der Ammons Horn-forma- relationship with postsynaptic parvalbumin-containing
tion. Arch Mikrosc Anat 39:611632. interneurons. Eur J Neurosci 5:17191728.
Schaller KL, Caldwell JH (2000) Developmental and regional expres- Sik A, Penttonen M, Ylinen A, Buzsaki G (1995) Hippocampal CA1
sion of sodium channel isoform NaCh6 in the rat central nerv- interneurons: an in vivo intracellular labeling study. J Neurosci
ous system. J Comp Neurol 420:8497. 15:66516665.
Scharfman HE (1991) Dentate hilar cells with dendrites in the molec- Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta
ular layer have lower thresholds for synaptic activation by per- phase precession in hippocampal neuronal populations and
forant path than granule cells. J Neurosci 11:16601673. the compression of temporal sequences. Hippocampus
Scharfman HE (1992) Differentiation of rat dentate neurons by mor- 6:149172.
phology and electrophysiology in hippocampal slices: granule Sloviter RS (1987) Decreased hippocampal inhibition and a selective
cells, spiny hilar cells and aspiny fast-spiking cells. Epilepsy Res loss of interneurons in experimental epilepsy. Science 235:7376.
Suppl 7:93109. Sloviter RS Somogyi P (1991) Permanently altered hippocampal
Scharfman HE (1994) Evidence from simultaneous intracellular structure, excitability, and inhibition after experimental status
recordings in rat hippocampal slices that area CA3 pyramidal epilepticus in the rat: the dormant basket cell hypothesis and
cells innervate dentate hilar mossy cells. J Neurophysiol its possible relevance to temporal lobe epilepsy. Hippocampus
72:21672180. 1:4166.
 Hippocampal Neurons: Structure and Function 199

Sloviter RS, Dichter MA, Rachinsky TL, Dean E, Goodman JH, Sollas lucidum of the rat hippocampus: an anatomical and electro-
AL, Martin DL (1996) Basal expression and induction of gluta- physiological characterization. J Comp Neurol 385: 427440.
mate decarboxylase and GABA in excitatory granule cells of the Staff NP, Jung H, Thiagarajan T, Yao M, Spruston N (2000) Resting
rat and monkey hippocampal dentate gyrus. J Comp Neurol and active properties of pyramidal neurons in subiculum and
373:593618. CA1 of rat hippocampus. J Neurophysiol 84:23982408.
Smith MA, Ellis-Davies GC, Magee JC (2003) Mechanism of the dis- Staley KJ, Otis TS, Mody I (1992) Membrane properties of dentate
tance-dependent scaling of Schaffer collateral synapses in rat gyrus granule cells: comparison of sharp microelectrode and
CA1 pyramidal neurons. J Physiol 548:245258. whole cell recordings. J Neurophysiol 67:13461358.
Soltesz I (1995) Brief history of cortico-hippocampal time with a Staley KJ, Longacher M, Bains JS, Yee A (1998) Presynaptic modula-
special reference to the direct entorhinal input to CA1. tion of CA3 network activity. Nat Neurosci 1:201209.
Hippocampus 5:120124. Stanford IM, Traub RD, Jefferys JGR (1998) Limbic gamma
Soltesz I, Mody I (1994) Patch-clamp recordings reveal powerful rhythms. II. Synaptic and intrinsic mechanisms underlying
GABAergic inhibition in dentate hilar neurons. J Neurosci spike doublets in oscillating subicular neurons. J Neurophysiol
14:23652376. 80:162171.
Soltesz I, Bourassa J, Deschenes M (1993) The behavior of mossy cells Stanzione P, Calabresi P, Mercuri N, Bernardi G (1984) Dopamine
of the rat dentate gyrus during theta oscillations in vivo. modulates CA1 hippocampal neurons by elevating the thresh-
Neuroscience 57:555564. old for spike generation: an in vitro study. Neuroscience
Somogyi P, Nunzi MG, Gorio A, Smith AD (1983a) A new type of 13:11051116.
specic interneuron in the monkey hippocampus forming Steward O, Worley P (2002) Local synthesis of proteins at synaptic
synapses exclusively with the axon initial segments of pyramidal sites on dendrites: role in synaptic plasticity and memory con-
cells. Brain Res 259:137142. solidation? Neurobiol Learn Mem 78:508527.
Somogyi P, Smith AD, Nunzi MG, Gorio A, Takagi H, Wu JY (1983b) Steward O, Falk PM, Torre ER (1996) Ultrastructural basis for gene
Glutamate decarboxylase immunoreactivity in the hippocam- expression at the synapse: synapse-associated polyribosome
pus of the cat: distribution of immunoreactive synaptic termi- complexes. J Neurocytol 25:717734.
nals with special reference to the axon initial segment of Stewart M (1997) Antidromic and orthodromic responses by subicu-
pyramidal neurons. J Neurosci 3:14501468. lar neurons in rat brain slices. Brain Res 769:7185.
Soriano E, Frotscher M (1993) Spiny nonpyramidal neurons in the Stewart M, Wong RKS (1993) Intrinsic properties and evoked
CA3 region of the rat hippocampus are gluatmate-like responses of guinea pig subicular neurons in vitro. J
immunoreactive and receive convergent mossy ber input. J Neurophysiol 70:232245.
Comp Neurol 332:435448. Storm JF (1987) Action potential repolarization and a fast after-
Soriano E, Frotscher M (1994) Mossy cells of the rat fascia dentata are hyperpolarization in rat hippocampal pyramidal cells. J Physiol
glutamate-immunoreactive. Hippocampus 4:6569. 385:733759.
Sorra KE, Harris KM (1993) Occurrence and three-dimensional Storm JF (1989) An afterhyperpolarization of medium duration
structure of multiple synapses between individual radiatum in rat hippocampal pyramidal cells. J Neurophysiol 409:171190.
axons and their target pyramidal cells in hippocampal area CA1. Storm JF (1990) Potassium currents in hippocampal pyramidal cells.
J Neurosci 13:37363748. Prog Brain Res 83:161187.
Sorra KE, Harris KM (2000) Overview on the structure, composition, Storm-Mathisen J (1977) Localization of transmitter candidates in
function, development, and plasticity of hippocampal dendritic the brain: the hippocampal formation as a model. Prog
spines. Hippocampus 10:501511. Neurobiol 8:119181.
Spacek J, Harris KM (1997) Three-dimensional organization of Storm-Mathisen J, Fonnum F (1972) Localization of transmitter can-
smooth endoplasmic reticulum in hippocampal CA1 dendrites didates in the hippocampal region. Prog Brain Res 36:4158.
and dendritic spines of the immature and mature rat. J Neurosci Stuart G, Spruston N (1998) Determinants of voltage attenuation in
17:190203. neocortical pyramidal neuron dendrites. J Neurosci 18:3501
Spencer WA, Kandel ER (1961) Electrophysiology of hippocampal 3510.
neurons IV. Fast prepotentials. J Neurophysiol 24:272285. Stuart GJ, Dodt HU, Sakmann B (1993) Patch-clamp recordings from
Spigelman I, Zhang L, Carlen PL (1992) Patch-clamp study of post- the soma and dendrites of neurons in brain slices using infrared
natal development of CA1 neurons in rat hippocampal slices: video microscopy. Pugers Arch 423:511518.
membrane excitability and K currents. J Neurophysiol 68:5569. Stuart G, Spruston N, Sakmann B, Husser M (1997) Action poten-
Spruston N, Johnston D (1992) Perforated patch-clamp analysis of tial initiation and backpropagation in neurons of the mam-
the passive membrane properties of three classes of hippocam- malian central nervous system. Trends Neurosci 20:125131.
pal neurons. J Neurophysiol 67:508529. Su H, Alroy G, Kirson ED, Yaari Y (2001) Extracellular calcium mod-
Spruston N, Jaffe DB, Williams SH, Johnston D (1993) Voltage- and ulates persistent sodium current-dependent burst-ring in hip-
space-clamp errors associated with the measurement of pocampal pyramidal neurons. J Neurosci 21:41734182.
electrotonically remote synaptic events. J Neurophysiol 70: Suzuki SS, Smith GK (1985) Burst characteristics of hippocampal
781802. complex spike cells in the awake rat. Exp Neurol 89:9095.
Spruston N, Jaffe DB, Williams SH, Johnston D (1994) Dendritic Swanson LW, Cowan WM (1977) An autoradiographic study of the
attenuation of synaptic potentials and currents: the role of pas- organization of the efferent connections of the hippocampal
sive membrane properties. Trends Neurosci 17:161166. formation in the rat. J Comp Neurol 172:4984.
Spruston N, Jonas P, Sakmann B (1995) Dendritic glutamate receptor Swanson LW, Khler C, Bjrklund A (1987) The limbic region. I. The
channels in rat hippocampal CA3 and CA1 pyramidal neurons. septo-hippocampal system. In: Integrated systems of the CNS
J Physiol (Lond) 482:325352. (Bjrklund A, Hkfelt T, Swanson L, eds), pp. 125269. New
Spruston N, Lbke J, Frotscher M (1997) Interneurons in the stratum York: Elsevier.
200 The Hippocampus Book

Szentagothai J, Arbib MA (1974) Conceptual models of neural in hippocampal neurons (CA1, CA3 and dentate) Biophys J
organization. Neurosci Res Prog Bull 12:307510. 46:7384.
Takumi Y, Ramirez-Leon V, Laake P, Rinvik E, Ottersen OP (1999) Turner DA, Schwartzkroin PA (1983) Electrical characteristics of
Different modes of expression of AMPA and NMDA receptors dendrites and dendritic spines in intracellularly stained CA3
in hippocampal synapse. Nat Neurosci 2:618624. and dentate hippocampal neurons. J Neurosci 3:23812394.
Tamamaki N (1997) Organization of the entorhinal projection to the Turner RW, Meyers DE, Barker JL (1989) Localization of
rat dentate gyrus revealed by DiI anterograde labeling. Exp tetrodotoxin-sensitive eld potentials of CA1 pyramidal cells in
Brain Res 116:250258. the rat hippocampus. J Neurophysiol 62:13751387.
Tamamaki N, Nojyo Y (1990) Disposition of slab-like modules Turner RW, Meyers DE, Richardson TL, Barker JL (1991) The site for
formed by axon branches originating from single CA1 pyrami- initiation of action potential discharge over the somatodendritic
dal neurons in the rat hippocampus. J Comp Neurol 291: axis of rat hippocampal CA1 pyramidal neurons. J Neurosci
509519. 11:22702280.
Tamamaki N, Nojyo Y (1993) Projection of the entorhinal layer II Turner RW, Meyers DE, Barker JL (1993) Fast pre-potential genera-
neurons in the rat as revealed by intracellular pressure-injection tion in rat hippocampal CA1 pyramidal neurons. Neuroscience
of neurobiotin. Hippocampus 3:471480. 53:949959.
Tamamaki N, Abe K, Nojyo Y (1987) Columnar organization in the Van der Linden S, Lopes da Silva FH (1998) Comparison of the elec-
subiculum formed by axon branches originating from single trophysiology and morphology of layers III and II neurons of
CA1 pyramidal neurons in the rat hippocampus. Brain Res the rat medial entorhinal cortex in vitro. Eur J Neurosci
412:156160. 10:14791489.
Taube JS (1993) Electrophysiological properties of neurons in the rat Van Groen T, Wyss JM (1990) Extrinsic projections from area CA1 of
subiculum. Exp Brain Res 96:304318. the rat hippocampus: olfactory, cortical, subcortical, and bilat-
Thurbon D, Field A, Redman S (1994) Electronic proles of interneu- eral hippocampal formation projections. J Comp Neurol
rons in stratum pyramidale of the CA1 region in the rat hip- 302:515528.
pocampus. J Neurophysiol 71:19481958. Van Groen T, Lopes da Silva FH (1986) Organization of the recipro-
Toth K, McBain CJ (1998) Afferent-specic innervation of two dis- cal connections between the subiculum and the entorhinal cor-
tinct AMPA receptor subtypes on single hippocampal interneu- tex in the cat. II. An electrophysiological study. J Comp Neurol
rons. Nat Neurosci 1:572578. 251:111120.
Toth K, Suares G, Lawrence JJ, Philips-Tansey E, McBain CJ (2000) Verdoorn TA, Burnashev N, Monyer H, Seeburg PH, Sakmann B
Differential mechanisms of transmission at three types of mossy (1991) Structural determinants of ion ow through recombi-
ber synapse. J Neurosci 20:82798289. nant glutamate receptor channels. Science 252:17151718.
Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E, Vida I, Bartos M, Jonas P (2006) Shunting inhibition improves
Svoboda K (2002) Long-term in vivo imaging of experience- robustness of gamma oscillations in hippocampal interneu-
dependent synaptic plasticity in adult cortex. Nature ron networks by homogenizing firing rates. Neuron 49:
420:788794. 107117.
Traub RD, Miles R (1991) Neuronal networks of the hippocampus. Vissavajjhala P, Janssen WG, Hu Y, Gazzaley AH, Moran T, Hof
Cambridge, UK: Cambridge University Press. PR, Morrison JH (1996) Synaptic distribution of the AMPA-
Traub RD, Jefferys JGR, Whittington MA (1999) Fast oscillations in GluR2 subunit and its colocalization with calcium-binding pro-
cortical circuits. Cambridge, MA: MIT Press. teins in rat cerebral cortex: an immunohistochemical study
Trommald M, Hulleberg G (1997) Dimensions and density of den- using a GluR2-specic monoclonal antibody. Exp Neurol 142:
dritic spines from rat dentate granule cells based on reconstruc- 296312.
tions from serial electron micrographs. J Comp Neurol Volfovsky N, Parnas H, Segal M, Korkotian E (1999) Geometry of
377:1528. dendritic spines affects calcium dynamics in hippocampal neu-
Trommald M, Jensen V, Andersen P (1995) Analysis of dendritic rons: theory and experiments. J Neurophysiol 82:450462.
spines in rat CA1 pyramidal cells intracellularly lled with a u- Walker MC, Ruiz A, Kullmann DM (2001) Monosynaptic GABAergic
orescent dye. J Comp Neurol 353:260274. signaling from dentate to CA3 with a pharmacological and
Tsubokawa H, Ross WN (1996) IPSPs modulate spike backpropaga- physiological prole typical of mossy ber synapses. Neuron
tion and associated [Ca2]i changes in the dendrites of 29:703715.
hippocampal CA1 pyramidal neurons. J Neurophysiol 76:2896 Walker MC, Ruiz A, Kullmann DM (2002a) Do mossy bers release
2906. GABA? Epilepsia 43:196202.
Tsubokawa H, Ross WN (1997) Muscarinic modulation of spike Walker HC, Lawrence JJ, McBain CJ (2002b) Activation of kinetically
backpropagation in the apical dendrites of hippocampal CA1 distinct synaptic conductances on inhibitory interneurons by
pyramidal neurons. J Neurosci 17:57825791. electrotonically overlapping afferents. Neuron 35:161171.
Tsubokawa H, Miura M, Kano M (1999) Elevation of intracellu- Wei DS, Mei YA, Bagal A, Kao JP, Thompson SM, Tang CM (2001)
lar Na induced by hyperpolarization at the dendrites of Compartmentalized and binary behavior of terminal den-
pyramidal neurones of mouse hippocampus. J Physiol 517: drites in hippocampal pyramidal neurons. Science 293:2272
135142. 2275.
Tsubokawa H, Offermanns S, Simon M, Kano M (2000) Calcium- Wenthold RJ, Petralia RS, Blahos J II, Niedzielski AS (1996) Evidence
dependent persistent facilitation of spike backpropagation in for multiple AMPA receptor complexes in hippocampal
the CA1 pyramidal neurons. J Neurosci 20:48784884. CA1/CA2 neurons. J Neurosci 16:19821989.
Turner DA (1984) Segmental cable evaluation of somatic transients Wenzel HJ, Buckmaster PS, Anderson NL, Wenzel ME, Schwartzkroin
 Hippocampal Neurons: Structure and Function 201

PA (1997) Ultrastructural localization of neurotransmitter Wouterlood FG, Saldana E, Witter MP (1990) Projection from the
immunoreactivity in mossy cell axons and their synaptic targets nucleus reuniens thalami to the hippocampal region: light and
in the rat dentate gyrus. Hippocampus 7:559570. electron microscopic tracing study in the rat with the antero-
Westenbroek RE, Ahlijanian MK, Catterall WA (1990) Clustering of grade tracer Phaseolus vulgaris-leucoagglutinin. J Comp Neurol
L-type Ca2 channels at the base of major dendrites in hip- 296:179203.
pocampal pyramidal neurons. Nature 347:281284. Wu LG, Saggau P (1994) Pharmacological identication of two types
White JA, Alonso A, Kay AR (1993) A heart-like Na current in the of presynaptic voltage-dependent calcium channels at CA3-
medial entorhinal cortex. Neuron 11:10371047. CA1 synapses of hippocampus. J Neurosci 14:56135622.
Wickens J (1988) Electrically coupled but chemically isolated Xiao B, Tu JC, Worley PF (2000) Homer: a link between neural activ-
synapses: dendritic spines and calcium in a rule for synaptic ity and glutamate receptor function. Curr Opin Neurobiol
modication. Prog Neurobiol 31:507528. 10:370374.
Wiebe SP, Staubli UV (1999) Dynamic ltering of recognition mem- Yasuda R, Sabatini BL, Svoboda K (2003) Plasticity of calcium chan-
ory codes in the hippocampus. J Neurosci 19:1056210574. nels in dendritic spines. Nat Neurosci 6:948955.
Williams SR, Stuart GJ (2000) Site independence of EPSP time course Yeckel MF, Berger TW (1990) Feedforward excitation of the hip-
is mediated by dendritic I(h) in neocortical pyramidal neurons. pocampus by afferents from the entorhinal cortex: redenition
J Neurophysiol 83:31773182. of the role of the trisynaptic pathway. Proc Natl Acad Sci USA
Witter MP (1989) Connectivity of the rat hippocampus. In: The hip- 87:58326583.
pocampus: new vistas, pp 5369. New York: Alan R. Liss. Yeckel MF, Berger TW (1995) Monosynaptic excitation of hippocam-
Witter MP, Amaral DG (1991) Entorhinal cortex of the monkey. V. pal CA1 pyramidal cells by afferents from the entorhinal cortex.
Projections to the dentate gyrus, hippocampus, and subicular Hippocampus 5:108114.
complex. J Comp Neurol 307:437459. Yuan LL, Adams JP, Swank M, Sweatt JD, Johnston D (2002) Protein
Wong RKS, Prince DA (1978) Participation of calcium spikes during kinase modulation of dendritic K channels in hippocampus
intrinsic burst ring in hippocampal neurons. Brain Res involves a mitogen-activated protein kinase pathway. J Neurosci
159:385390. 22:48604868.
Wong RKS, Prince DA, Basbaum AI (1979) Intradendritic recordings Yue C, Remy S, Su H, Beck H, Yaari Y (2005) Proximal persistent Na
from hippocampal neurons. Proc Natl Acad Sci USA 76: channels drive spike afterdepolarizations and associated burst-
986990. ing in adult CA1 pyramidal cells. J Neurosci 25:97049720.
Wong RKS, Prince DA (1981) Afterpotential generation in hip- Yuste R, Denk W (1995) Dendritic spines as basic functional units of
pocampal pyramidal cells. J Neurophysiol 45:8697. neuronal integration. Nature 375:682684.
Wong RKS, Stewart M (1992) Different ring patterns generated in Yuste R, Majewska A, Cash SS, Denk W (1999) Mechanisms of cal-
dendrites and somata of CA1 pyramidal neurones in guinea-pig cium inux into hippocampal spines: heterogeneity among
hippocampus. J Physiol 457:675687. spines, coincidence detection by NMDA receptors, and optical
Wong RKS, Traub RD, Miles R (1986) Cellular basis of neuronal syn- quantal analysis. J Neurosci 19:19761987.
chrony in epilepsy. Adv Neurol 44:583592.
This page intentionally left blank
6 Dimitri Kullmann

Synaptic Function

For the purpose of this chapter a synapse is taken to repre-

6.1 Overview sent the specialized structural and functional unit that is com-
posed of one or more contiguous synaptic specializations
Ever since Sherrington coined the term synapse (from the occurring at the interface between two neurons (the presy-
Greek hold together), this organelle has attracted intense naptic and postsynaptic partners). Thus, a single synapse can
interest for several reasons. First, the information passing contain more than one presynaptic active zone (where neuro-
through the synapses supplied by an individual neuron transmitter molecules are released) and postsynaptic special-
reects the outcome of its integrative functions. Second, with ization (where receptors detect them). However, a pair of
some notable exceptions it acts as a one-way valve, transmit- pre- and postsynaptic neurons can also be connected via more
ting information from the pre- to the postsynaptic neuron. than one synapse. Although this denition of a synapse is
Third, plasticity of synaptic transmission is the leading candi- morphological, it should be borne in mind that major insights
date for the cellular substrate of memory formation and stor- into synaptic function have relied principally on electrophysi-
age. This chapter addresses some of the ways that individual ological, pharmacological, and biochemical methods, often in
synapses in the hippocampus are specialized to transmit sig- the absence of histological reconstruction of the underlying
nals and encode a history of their recent activity. It also pro- elements.
vides the physiological context to Chapter 7, which examines Not all communication between neurons is via synapses:
the function of many of the molecules underlying synaptic Some neurotransmitters are released from axonal varicosities
transmission. The mechanisms of induction and expression of in the absence of postsynaptic specialization. The fact that
long-term changes in synaptic strength are addressed in receptors are sometimes relatively remote from the site of
Chapter 10. release of their endogenous ligands (e.g., monoamines, pep-
Although the hippocampus has provided extensive infor- tides, acetylcholine) has been taken as evidence that two dis-
mation on the mechanisms by which neurons communicate tinct forms of signaling exist: (1) point-to-point or wiring
with one another, many important principles underlying transmission at synapses and (2) a more diffuse spillover,
synaptic transmission were originally elucidated at the mam- extrasynaptic, or volume transmission mediated by diffu-
malian or arthropod neuromuscular junction. Indeed, many sion of neuroactive substances through the extracellular space
of the mechanisms underlying the trafficking of presynaptic (Agnati et al., 1995). As is argued below, the distinction
vesicles are shared with other intracellular membrane-bound between these concepts is blurred by the existence of numer-
compartments and are even present in prokaryotes. We there- ous types of receptor at variable distances from the release
fore start by summarizing some general features of inter- sites of their endogenous agonists. Neurons can also exert
neuronal communication that are shared with other systems effects on their neighbors via mechanisms that do not involve
without going into detail into the experimental basis for these neurotransmitters at all, for instance via eld effects (Jefferys,
generally agreed upon principles. A detailed treatment of this 1995) or gap junctions (Galarreta and Hestrin, 1999; Gibson
subject is given in Synapses (Cowan et al., 2003). We then et al., 1999). These act as relatively low-impedance pathways
examine the various types of hippocampal synapse (classied that permit the passive ow of electrical signals among con-
anatomically) to see how their functional properties differ and nected neurons. They can also permit ions to ow down their
how these distinct characteristics shed light on their roles. electrochemical gradients, thereby dissipating their local accu-

203
204 The Hippocampus Book

mulation and depletion. Preliminary data also imply that described in detail in Chapter 7, are thought to contribute to
axons of neighboring pyramidal neurons may be coupled the formation, stabilization, and plasticity of synapses (see
through electrical contacts (Schmitz et al., 2001b). Moreover, also Chapter 10).
not all hippocampal synapses are formed between neurons: The presynaptic element is readily identied by the pres-
Morphological and electrophysiological evidence supports ence of vesicles containing neurotransmitter molecules. These
the existence of excitatory synapses between hippocampal vesicles are generally of relatively uniform size and are aggre-
neurons and a subset of oligodendrocyte precursors (Bergles gated near a membrane specialization that is often identied
et al., 2000). ultrastructurally as a thickening. This is thought to represent
the active zone, which is the site of exocytosis; and the thick-
ening reects the presence of membrane proteins necessary
for exocytosis. Among them are proteins that interact with
6.2 General Features of Synaptic vesicular partners (SNARE proteins, see below), as well as
Transmission: Structure voltage-dependent channels that mediate the inux of Ca2,
which triggers exocytosis upon action potential invasion. It is
Synaptic transmission can be broken down into neurotrans- almost universally agreed that vesicles are the morphological
mitter release from a presynaptic specialization, diffusion counterpart of the quanta, that make up synaptic signals (see
of the neurotransmitter across the synaptic cleft, and activa- below). In addition, mitochondria are frequently present,
tion of postsynaptic receptors. These phenomena must be reecting the energetic demands of transmitter release (prin-
understood in the context of the detailed ultrastructure of the cipally represented by the work required for transmitter syn-
synapse. thesis, vesicular packaging, exocytosis, and reuptake)
The electron micrograph shown in Figure 61, see color (Shepherd and Harris, 1998).
insert illustrates some features of hippocampal synapses. The On the postsynaptic side, there is also an increased density
pre- and postsynaptic elements are separated by a synaptic of the membrane. Synapses with obvious postsynaptic thick-
cleft. This cleft is often narrower than the gap separating non- ening are known as asymmetrical, or type I, synapses (Gray,
synaptic membranes and appears relatively dense, reecting 1959) and were shown to be excitatory by Andersen and col-
the presence of a basal substance containing intercellular leagues (1966), who examined the ultrastructural conse-
adhesion molecules. These molecules, some of which are quences of lesioning monosynaptic excitatory pathways. The

Figure 61. Fine structure of the hip-

pocampal neuropil. Arrows point to indi-
vidual asymmetrical (glutamatergic)
synapses in a section taken from the stra-
tum radiatum of CA1 in the rat. The
presynaptic boutons contain numerous
vesicles, which are frequently clustered
close to membrane specializations (the
active zone). A proportion of docked
vesicles are in contact with the presynap-
tic membrane and are thought to be
ready for exocytosis. The electron-dense
area is the postsynaptic density, which
distinguishes these synapses from numer-
ically less abundant symmetrical
(inhibitory) synapses (see Fig. 69).
Astrocyte proles, colored in blue, lack-
ing vesicles and postsynaptic densities,
contact the synaptic perimeter in some
cases (arrowheads). Bar  1 m. (Source:
Ventura and Harris, 1999, with permis-
sion.)
 Synaptic Function 205

postsynaptic density (PSD) is a relatively detergent-resistant mate anion is driven into vesicles by the vesicle potential gra-
structure containing glutamate receptors and associated dient formed by the high intraluminal H concentration. A
macromolecules (Kennedy, 1997) (discussed in Chapter 7). family of proteins that mediate glutamate uptake (VGLUT1-3)
Some synapses lack this specialization and are therefore has recently been identied (Bellocchio et al., 2000; Takamori
known as symmetrical, or type II, synapses (Gray, 1959). et al., 2000). In the case of GABA (-aminobutyric acid), in
They are thought to be principally GABAergic inhibitory addition to uptake driven by the potential gradient, another
synapses (see Chapter 8). Although they represent a small mechanism exists: Cl
is driven in by the potential gradient,
number of synapses in the hippocampus as a whole, they play and then exchanged for the GABA anion (Fykse and Fonnum,
numerous important roles, including setting the overall 1996). The vesicular transporter VGAT appears to be respon-
excitability of the structure and its rhythms. GABAergic sible for GABA uptake (McIntire et al., 1997).
synapses are concentrated around the soma and axonal initial Vesicles are reversibly tethered by synapsin molecules to a
segment of many neurons but are also found on dendrites. presynaptic actin scaffold. A subset of vesicles are especially
In addition to the pre- and postsynaptic elements that closely apposed to the presynaptic active zone and have been
make up a synapse, astrocyte processes commonly occur in described as beging docked, presumably ready for exocytosis
close proximity to the synaptic cleft. As is shown below, they (Harris and Sultan, 1995). These vesicles may correspond to a
play a critical role in clearing neurotransmitter from the extra- population of quanta that are available for release on a rela-
cellular space. They also play an important role in buffering tively rapid time scale (dened physiologically as the readily
extracellular ion transients and in providing for the energetic releasable pool) (Stevens and Tsujimoto, 1995; Rosenmund
demands of synaptic transmission. Although astrocyte and Stevens, 1996). Evidence obtained in isolated prepara-
processes are often thought of as an obligatory partner, mak- tions suggests that vesicle docking may be reversible (Steyer et
ing up the third element of a tripartite synapse, their pres- al., 1997; Murthy and Stevens, 1999), and there may be a fur-
ence at excitatory and inhibitory synapses is much more ther stage of preparation for exocytosis that is not resolved
haphazard in the hippocampus than in other regions of the ultrastructurally (named priming).
brain, such as the cerebellar cortex (Spacek, 1985; Lehre et al., Considerable effort has gone into identifying the molecu-
1995; Ventura and Harris, 1999). Indeed, many synapses in the lar identity of the proteins that mediate and detect the inux
hippocampus are not contacted by an obvious astrocytic of Ca2 and trigger the fusion of the vesicle membrane with
process at all (Ventura and Harris, 1999). the presynaptic neuronal membrane (Augustine, 2001).
Voltage-dependent Ca2 channels clearly play a critical
6.2.1 Transmitter Release and Diffusion role, and it is important to understand their properties to
explain use-dependent plasticity and pharmacological modu-
Anatomical, biochemical, and electrophysiological methods lation of transmitter release. Of the various pharmacologically
have converged on the following general account of transmit- and biophysically dened classes, P/Q
and N-type channels
ter release (Zucker et al., 1998; Chen and Scheller, 2001; (also known as CaV2.1 and CaV2.2, respectively) account for
Sudhof, 2004) (Fig. 62). most of the Ca2 inux that triggers exocytosis (Dunlap et al.,
On the presynaptic side, neurotransmitter molecules are 1995). A third class of Ca2 channels, R type (CaV2.3), play a
translocated from the cytoplasm into vesicles by specic trans- relatively small role at most synapses. Because of their high
porters, which exploit a proton gradient formed by vesicular depolarization threshold, all three types are normally kept
ATPases. The two principal vesicular amino acid neurotrans- shut at resting potentials. They open rapidly upon depolariza-
mitters differ slightly in their uptake mechanisms. The gluta- tion; but because their closure lags slightly behind action

Figure 62. Stages of the vesicle

cycle. 1, Vesicle lling. 2, Trafficking
to the release site. 3, Docking/prim- 9
ing. 4, Fusion/release. 5, Membrane ATP
attening. 6, Clathrin coating. 7,
10
Endocytosis. 8, Clathrin uncoating. ADP H
9, Merging with endosome. 10,
Recycling. Some evidence suggests 1 8
that docking/priming and
fusion/release may be reversible
(indicated by broken line arrows). 4 6
2 3
7
Ca2 5

Synaptic
cleft
206 The Hippocampus Book

potential repolarization, they are still able to carry Ca2 ux

when the presynaptic membrane potential has returned close VAMP
to the resting value. The bulk of the Ca2 inux triggering
Nucleation
exocytosis therefore probably occurs during repolarization,
because the inward driving force for Ca2 ions is maximal at
this time point. Interestingly, whereas both N- and P/Q-type
channels contribute to triggering glutamate release at individ- SNAP-25 syntaxin
Ca2
ual synapses, many GABAergic terminals appear to fall into
two classes: those that rely mainly on N-type channels and Zippering
those that predominantly use P/Q-type channels (Poncer et
al., 1997). It is not known if the Ca2 channel identity is some-
how tailored to the ring patterns of the presynaptic neurons. Fusion
Measurement of the intraterminal spatiotemporal Ca2
concentration prole that occurs following action potential
invasion is technically difficult in relatively small hippocampal
synapses. Nevertheless, it has been estimated that the average
concentration of free Ca2 in a presynaptic varicosity rises to Figure 63. Proposed mechanism of SNARE-mediated exocytosis.
about 500 nM (Koester and Sakmann, 2000). This underesti- Assembly of the core complex (VAMP/synaptobrevin, syntaxin,
mates the peak concentration seen by the Ca2 sensors SNAP-25) is seen as a nucleation event necessary to prepare the
responsible for triggering exocytosis. Based on studies at other vesicle for exocytosis. Ca2 inux induces a conformational change
synapses, the concentration at the site of exocytosis is thought of the core complex (zippering), which brings the vesicle and
to rise transiently approximately 100-fold, to approach 10 M plasma membranes into contact, followed by transmitter release.
(Bollmann et al., 2000; Schneggenburger and Neher, 2000). (Source: Lin and Scheller, 2000, with permission.)
Estimates of the size of the Ca2 ion ux through an individ-
ual Ca2 channel and the distance separating them from cleaved by at least one clostridial toxin (botulinum toxins AG
docked vesicles are consistent with the view that relatively few and tetanus toxin), which results in blockade of neurotrans-
voltage-sensitive Ca2 channels (possibly fewer than 100) mitter release.
must open to trigger exocytosis of a single vesicle (Borst and A compelling suggestion for the role of the core complex
Sakmann, 1996). formed by VAMP/synaptobrevin, syntaxin, and SNAP-25 is
Transmitter release shows a steep dependence on the extra- that it forces together the vesicular and target membranes in
cellular Ca2 concentration. A fourth-power relation has been preparation for their fusion. Nevertheless, the precise
observed at some synapses (Dodge and Rahamimoff, 1967), sequence of events that leads up to the conformational change
although this saturates at supraphysiological concentrations underlying exocytosis is incompletely understood. In particu-
so the curve relating transmitter release to Ca 2 concentra- lar, it remains unclear precisely how synaptotagmin interacts
tion is actually sigmoid. Several possible explanations exist for with the SNARE complex and how the cascade overcomes the
the steep Ca2 dependence, including summation of Ca2 steep energy barrier to fusion of the vesicular and target mem-
inux via multiple channels, saturation of buffers, and coop- brane bilayers. Ultimately, the transmitter is released into the
erativity of Ca2 sensors, in particular the vesicle-associated synaptic cleft. An outline of this cascade is summarized in
protein synaptotagmin I (Fernandez-Chacon et al., 2001). Figure 63.
Synaptotagmin I contains two Ca2-binding domains, and Before considering the consequence of transmitter release
genetic deletion of this protein severely disrupts fast Ca2- into the synaptic cleft, it is worth examining the fate of Ca2
dependent exocytosis, making it a strong candidate for the ions. Powerful presynaptic Ca2 buffers exist in the form of
Ca2 sensor responsible for triggering transmitter release proteins and mitochondria (Klingauf and Neher, 1997;
(Geppert et al., 1994). Ca2 binding is thought to trigger a Tang and Zucker, 1997). These, together with Ca2 extrusion
conformational change in a macromolecular complex consist- mechanisms, contribute to terminating the Ca2 transient in
ing of three groups of proteins: vesicular, target membrane- such a way that the system is returned to the resting state.
associated, and soluble proteins (Chen and Scheller, 2001). Nevertheless, it may take several minutes for this to be
The membrane-associated proteins are known as the SNAREs, achieved, during which time successive action potentials may
from SNAP-receptors (SNAP: synaptosomal-associated pro- become increasingly effective in triggering transmitter release
teins). The vesicular SNAREs (v-SNAREs) include VAMP (Kamiya and Zucker, 1994; Zucker and Regehr, 2002). Genetic
(vesicle-associated membrane protein, also known as synap- manipulation of Ca2-buffering proteins can alter the
tobrevin); and the target membrane SNAREs (t-SNAREs) dependence of synaptic strength on the recent history of
include SNAP-25 and syntaxin. These key proteins form a activity, underlining the importance of this phenomenon
complex consisting of four helices (one  helix contributed by (Caillard et al., 2000). However, this is only one of several
each of VAMP/synaptobrevin and syntaxin, and two  helices mechanisms that contribute to short-term use-dependent
contributed by SNAP-25). Each of these proteins can be plasticity of synaptic transmission (see below).
 Synaptic Function 207

It remains unclear whether exocytosis always involves com- It is known to interact with two other proteins, rabphilin and
plete fusion of the vesicle, f lattening and merging with the RIM (Rab-interacting molecule), that contain putative
presynaptic membrane, or can occur through the reversible Ca2-binding domains. There may also be a mechanism that
opening of a fusion pore, allowing the escape of all or part of regulates vesicle trafficking as a function of the neurotrans-
the vesicle contents. Early reports of incomplete exocytosis mitter contents: Interfering with vesicular ATPase has been
came from ultrastructural studies at the neuromuscular junc- reported not to reduce the quantal amplitude but only to
tion and from two types of electrophysiological measurement reduce the number of quanta available for release (Zhou et al.,
that offer very good temporal resolution: voltammetric 2000). This observation suggests that a form of quality control
recordings of monoamine release (Chow et al., 1992; Bruns exists to prevent underlled vesicles from docking and/or
and Jahn, 1995) and capacitance measurements of the release releasing.
of secretory granules (Breckenridge and Almers, 1987). None Following release, neurotransmitter molecules diffuse
of these measurements was obtained at hippocampal across the cleft to activate postsynaptic receptors. Because the
synapses, so their relevance to this structure remains indirect. synaptic cleft spans only approximately 20 nm, the time taken
However, recent optical measurements of dyes taken up by by neurotransmitter molecules to reach the nearest possible
presynaptic vesicles in hippocampal neurons in culture have postsynaptic receptors is negligible: Assuming that neuro-
suggested that individual vesicles can retain their identity transmitter molecules diffuse readily in the extracellular
through the cycle of release and reuptake (Murthy and space, they should reach most of the receptors in the synaptic
Stevens, 1998), consistent with the proposal that they dis- cleft within 100 s. However, although many receptors occur
charge their contents without completely fusing with the in the postsynaptic membrane opposite the active zone, oth-
presynaptic membrane. This behavior raises the possibility ers are farther away. Several classes of receptors (considered in
that a vesicle may release less than its entire contents (so- detail below) show distinct patterns of localization: in the
called kiss-and-run release) before separating from the presy- presynaptic membrane within the synaptic cleft or in a perisy-
naptic membrane and mixing with a pool of vesicles in the naptic halo around the synapse, either pre- or postsynaptically
terminal. Electrophysiological evidence has also been reported (Takumi et al., 1998). Some receptors are even found in the
from hippocampal neurons in culture suggesting that gluta- axonal membrane relatively far from the synaptic cleft (Yokoi
mate can be released through two interchangeable modes, et al., 1996). To assess the functional role of these types of
corresponding to full exocytosis or escape through a fusion receptor, it is important to understand how the endogenous
pore (Renger et al., 2001). ligands reach them following release from presynaptic vari-
Once fusion has occurred, the SNARE proteins must cosities. The factors that determine the activation of these
be dissociated and made available for another round of receptors are as follows.
exocytosis. This step involves the soluble ATPase NSF (N-
Amount of neurotransmitter released
ethylmaleimide-sensitive factor) and -SNAP (soluble NSF
Temporal prole of release
attachment protein). The requirement for ATP hydrolysis
Distance from release site to receptors
probably reects the energetically unfavorable reversal of
Diffusion coefficient (or diffusivity) of the neurotrans-
the formation of the core complex (Chen and Scheller, 2001;
mitter in the cleft and in the extracellular medium
Jahn et al., 2003). The vesicle membrane, together with the
Diffusional obstacles
membrane-associated proteins (including the vesicular
Neurotransmitter transporters, their affinity and kinet-
ATPase and transporters, v-SNAREs, and synaptotagmin)
ics, and other binding sites
must also be endocytosed if full fusion has occurred. This
Kinetics of the receptors
retrieval proceeds by formation of a lattice of clathrin mole-
cules on the cytoplasmic face of the vesicle membrane. This Astrocytes play an important role in determining receptor
lattice imposes a concavity on the membrane that is eventu- activation because they frequently extend processes that con-
ally pinched off by the action of dynamin, following which the tact the perimeter of the synaptic cleft and because they
recycled vesicle reenters the presynaptic pool (Cremona and express transporters and receptors on their surface, which
De Camilli, 1997). Whether the vesicle merges immediately buffer and/or sequester the neurotransmitter molecules.
with the pool available to be released or is targeted specically A further role played by astrocytes in the case of glutamate, is
to an intermediate endosomal compartment is unclear. to convert it enzymatically to glutamine (see below).
Numerous mechanisms regulate the availability of vesicles
for release. Among them is ATP-dependent mobilization of 6.2.2 Receptors and Receptor Activation
vesicles tethered to presynaptic microlaments by synapsins
(Rosahl et al., 1995). Rab3A is a GTP-binding protein that Receptors fall into two distinct classes. Ionotropic receptors,
reversibly associates with vesicles as they go through the exo- also known as ligand-gated ion channels, contain binding sites
cytosisendocytosis cycle. Genetic deletion of Rab3A inter- for neurotransmitters. Binding of the transmitter molecules
feres with the expression of long-term potentiation at mossy causes the channel to open, allowing charged ions to travel
bers (see below) (Castillo et al., 1997a), suggesting a regula- down their chemical and electrical gradients. The biophysical
tory role; but its precise function in the vesicle cycle is unclear. principles and molecular mechanisms underlying ion selectiv-
208 The Hippocampus Book

ity and permeation are described in detail in Ion Channels of

Excitable Membranes (Hille, 2001). With the other type, Box 61
Voltage-Clamp Techniques
known as metabotropic receptors, ligand binding triggers an
intracellular biochemical cascade beginning with dissociation Early studies of ion channels in neurons relied on measuring
of GTP-binding proteins (G proteins). the effects of drugs and stimuli on the membrane potential.
Ionotropic receptors generally occur postsynaptically, However, an ion ux through a population of channels can
although they can also occur presynaptically in some areas of itself alter the membrane potential. This has two conse-
quences: It alters the driving force for further ion ux and
the brain (MacDermott et al., 1999), and mediate depolariz-
can cause voltage-sensitive channels to open or close, there-
ing or hyperpolarizing synaptic signals that activate within by potentially setting off regenerative currents. Thus, it is
milliseconds. Depolarizing signals mediated by glutamate difficult to characterize the properties of different types
receptors take the form principally of an inux of Na ions, of receptors on the basis of voltage recordings alone. The
accompanied by variable amounts of Ca2 inux. Other exci- voltage-clamp method circumvents this problem by keeping
tatory ionotropic receptors (purinergic P2x receptors, nico- the membrane potential at a xed value and estimating the
tinic acetylcholine receptors, and serotonergic 5HT3 current owing through the membrane. The method relies
on a feedback circuit that pumps enough current back into
receptors) are present in far smaller numbers. The ion ux
the neuron to clamp its voltage at the desired level. The
constitutes the excitatory postsynaptic current (EPSC), whose amplitude and time course of this current in relation to
size is determined by the membrane potential, the reversal various stimuli (for instance, synaptic release or exogenous
potential for each ion species, its permeability through the application of neurotransmitters or changes in the command
receptor, and the mean number of receptor channels opening. voltage itself) then give an indication of the properties of
The EPSC depolarizes the postsynaptic membrane, and gives the ion channels. As originally developed to study muscle
rise to an excitatory postsynaptic potential (EPSP), which if it bers and axons, the method required separate electrodes to
detect the membrane potential and to pass the current. This
is sufficiently large and/or sums with other EPSPs can activate
was not feasible in any but the largest cells. The patch-clamp
regenerative currents. Such currents can reach threshold to method overcomes this difficulty by using the same low-
trigger an action potential in the axon initial segment, resistance electrode to detect the membrane potential and
although under some conditions an action potential can also pass the current. It relies on the nding that a glass pipette
arise in the parent dendrite (see Chapter 5). Box 61 summa- applied to the surface of a neuron can make a very high
rizes some principles underlying voltage-clamp recording resistance seal with the plasmalemma. The method can work
techniques, which have shed much light on the mechanisms in various congurations.
underlying the generation of synaptic signals. 1. Currents passing through individual ion channels
Conversely, inhibitory ionotropic signals are mediated under the membrane can be recorded in the cell
principally by GABAA receptors. Although glycine receptors attached mode.
also contribute to inhibition in spinal and brain stem neurons, 2. The membrane under the pipette can be ruptured to
allow the entire neuron to be voltage-clamped (whole
they are not known to play an important role in the hip-
cell modethe most versatile method for recording
pocampus. Inhibitory postsynaptic currents (IPSCs) are
synaptic currents).
carried by Cl
and HCO3-ions owing down their electro- 3. Another way to access the entire cell electrically while
chemical gradients. Because the permeability of GABAA minimizing the perturbation of its biochemical milieu
receptors for Cl
is greater than that for HCO3
and the equi- is to add an ionophore to the pipette solution while
librium potential for Cl
is relatively negative, the IPSC gen- remaining in the cell attached mode. This is the perfo-
erally hyperpolarizes the neuron, giving rise to an inhibitory rated patch technique.
postsynaptic potential (IPSP). Synaptic inhibition has two 4. In the whole cell conguration, slowly withdrawing the
effects on neuronal excitability. First, it drives the membrane pipette causes a patch of membrane to seal over the
potential farther away from the threshold for activating regen- pipette mouth (outside-out conguration). This
erative depolarizing currents; and, second, by lowering the allows the effects of brief pulses of agonists on small
numbers of receptors to be studied in great detail.
membrane resistivity it shunts EPSCs that would normally
5. It is also possible to rupture a vesicle of membrane in
depolarize the neuron.
such a way that the cytoplasmic face is exposed to the
Although IPSPs and EPSPs contribute to the excitability bath (inside-out conguration). This is useful for
of the neuron, they do not sum linearly because of several studying the regulation of ion channels by second mes-
factors addressed in Chapter 5. Briey, voltage- and time- sengers.
dependent electrical properties of dendrites cause synaptic
currents to dissipate or amplify as they propagate, the results
of which depend strongly on the initial spatial prole of the logical proles, have specic patterns of distribution. Most
membrane voltage, on the geometric distribution of active such receptors exert their actions over a slower time scale, last-
synapses, and on the density and kinetics of a set of voltage- ing up to seconds. As a general principle, presynaptic
gated ion channels. metabotropic receptors depress transmitter release (Pin and
Metabotropic receptors occur both pre- and postsynapti- Duvoisin, 1995; Schoepp, 2002). This effect is mediated partly
cally, although different types, which have distinct pharmaco- through G protein regulation of presynaptic Ca2 channels.
 Synaptic Function 209

N-type Ca2 channels in particular are directly modulated by presynaptic action potentials give rise to postsynaptic signals
the G protein subunit G  (Zhang et al., 1996; Dolphin, 1998); that uctuate from trial to trial among discrete levels (Katz,
but in addition, some metabotropic receptors affect transmit- 1969; Zucker et al., 1998). These discrete levels occur at inte-
ter release indirectly via inactivation of adenylate cyclase gral multiples (0, 1, 2, ) of an underlying unit, the quan-
(Tzounopoulos et al., 1998). Some metabotropic receptors tum. This model of transmission has received only partial
may decrease transmitter release separately from an effect on support from studies in the central nervous system (CNS). As
Ca2 inux (Scanziani et al., 1995; Blackmer et al., 2001). already hinted above, if exocytosis is incomplete or the neuro-
On the postsynaptic side, metabotropic receptors have transmitter is discharged slowly, a single vesicle may give a
more diverse actions, although again most of these effects graded postsynaptic effect, even if it originally contained a
involve G proteins. The effector mechanisms include opening xed amount of neurotransmitter. Moreover, nonuniformities
of K channels, giving rise to a slow hyperpolarization (Pin in the distribution of receptors available to detect transmitter
and Duvoisin, 1995; Luscher et al., 1997). Some metabotropic released at different synapses or in the degree of distortion
glutamate receptors in contrast trigger the closure of specic that the postsynaptic signals undergo as they propagate to
K channel subtypes, leading to depolarization or a change in the cell body are likely to conceal any quantal structure
the temporal prole of repetitive action potentials. Activation in the amplitude uctuations of postsynaptic signal recorded
of yet other metabotropic receptors can lead to a depolarizing at the soma. Finally, unlike the neuromuscular junction,
current by opening a cation channel with high permeability to where a single presynaptic end-plate has a large number
Na (Heuss et al., 1999). Other effects of metabotropic recep- of relatively uniform, well separated release sites, most hip-
tor activation include activation of phospholipase C, generat- pocampal synapses probably contain only one or a few release
ing inositol trisphosphate (IP3) and diacylglycerol (DAG). IP3 sites (Schikorski and Stevens, 1997), which probably interact
is intimately involved in regulating the release of Ca2 from nonlinearly. Nevertheless, recordings of synaptic signals have
internal stores, which may lead to triggering regenerative Ca2 reported large trial-to-trial uctuations in the size of the
release and more remote effects such as modulation of gene postsynaptic response to activation of a single or a small num-
transcription. ber of presynaptic axons. In some cases, it has even been
Receptor activation must eventually terminate. The most possible to discern clustering of postsynaptic response ampli-
important mechanism that underlies this is rapid dissipation tudes at integral multiples of the unit value, perhaps reecting
of the neurotransmitter pulse by diffusion into the large extra- the fortuitous case where underlying release events gave simi-
cellular space, which can drop the concentration 100-fold lar responses, and where they summed relatively linearly
within a few milliseconds. This phenomenon, on its own, thus across the various active synapses or release sites (e.g.,
causes a rapid decrease in ligand concentration to levels that Kullmann and Nicoll, 1992). These cases probably represent
may fall below those required to activate receptors. only a small number of recordings but have helped shed
Nevertheless, some receptors remain bound for several hun- light on the parameters that underlie transmission. Box 62
dred milliseconds because they have very high affinity (Lester summarizes the principles of quantal analysis (Zucker et al.,
et al., 1990). Uptake also plays a major role in the clearance of 1998).
neurotransmitter through both the rapid quenching of neuro- The more general nding is that although transmission is
transmitter molecules binding to unoccupied transporters stochastic and can even fail on occasion EPSCs, EPSPs, IPSCs,
and their translocation into the cytoplasm. A third mecha- and IPSPs tend not to cluster systematically at preferred
nism that can underlie the termination of receptor activation amplitudes (Raastad, 1995). It has nevertheless been possible
is desensitization: In the continued presence of agonist mole- to extract some information on the mechanisms of modula-
cules, ionotropic receptors generally enter a state that is no tion of synaptic strength from examining the trial-to-trial
longer able to open. (Much less is known about the desensiti- variability of such signals. This application of quantal analysis
zation mechanisms of metabotropic receptors.) The relative relies on making a few assumptions, but some of them have
importance of diffusion, uptake, and desensitization in termi- been found to be reasonably robust. For instance, if it is
nating activation may vary among receptors and even among assumed that failures of transmission reect cases where
synapses. However, even if uptake plays little role in the rapid the presynaptic action potential did not trigger exocytosis, it
termination of receptor activation, the neurotransmitters may be possible to use the frequency of such failures as an
must eventually be cleared from the extracellular space by indirect witness of the state of the presynaptic terminal.
active transport because, apart from acetylcholine, ATP, and That is, if a physiological or pharmacological event decreases
some neuroactive peptides, there are no extracellular mecha- the rate at which presynaptic action potentials elicit postsy-
nisms to break them down. naptic responses, it may be inferred that the presynaptic
release probability has decreased. Other approaches rely on
6.2.3 Quantal Transmission estimating the coefficient of variation of the postsynaptic sig-
nal (see Box 62) . These approaches, however, must be used
The conventional view of synaptic functionoriginating with caution. They have been at the center of a controversy
from studies by Katz and coworkers at the neuromuscular surrounding the mechanisms of expression of long-term
junctionis that transmission is quantized; that is, repeated potentiation (see Chapter 10), and some of the disagreements
210 The Hippocampus Book

Box 62
Quantal Analysis

Chemical synaptic transmission occurs via the presynaptic release of packages of neurotrans-
mitter, which diffuse to and act upon postsynaptic receptors. This phenomenon was originally
elucidated at the neuromuscular junction on the basis of electrophysiological recordings from
the postsynaptic muscle cells. Under certain conditions, the size of the postsynaptic response to
presynaptic stimulation uctuates from stimulus to stimulus among discrete levels, reecting 0,
1, 2 quanta released. These levels correspond to integral multiples of an underlying unit,
which coincide with the size of a spontaneous event arising from the spontaneous release of a
single package of neurotransmitter. From a wealth of other data, it has been established that the
quantum corresponds to the postsynaptic response to the neurotransmitter packaged in an
individual vesicle. The quantal description of neurotransmission not only gives a powerful
insight into the biophysics of transmitter release and receptor activation, it provides a shortcut
for investigating the mechanisms of synaptic signaling modulation. Thus, factors that reduce or
increase the presynaptic release of transmitter almost universally affect the average number of
quanta released across a population of release sites (the quantal content m, which is the prod-
uct of the number of available quanta n, and their average release probability p). Factors that
modulate the state of the postsynaptic receptors alter the size of the response to an individual
package of neurotransmitter (the quantal amplitude Q).
Extrapolation of the quantal model to the central nervous system (CNS) has uncovered
some subtle but important differences. Unlike the neuromuscular junctionwhich is often a
long ribbon-like structure with numerous vesicles available to be releasedactive zones in the
CNS tend to be small structures with only a few vesicles in close apposition to the presynaptic
membrane (docked vesicles). Correspondingly, postsynaptic signals often uctuate between
failures and an amplitude that may reect the release of only one vesicle (0 or 1 quantum).
Moreover, because the peak occupancy of postsynaptic receptors in response to release of a sin-
gle vesicle may be high ( 50%), release of two vesicles may not yield a signal whose amplitude
is twice as large. Nevertheless, some synapses (such as mossy ber synapses on CA3 pyramidal
neurons) have numerous active zones and postsynaptic densities; and at these synapses, multi-
quantal release can be detected. Another difference between quantal transmission in the CNS
and neuromuscular juction is that different synapses are probably not electrically equivalent:
Differences in the number of receptors and degrees of electrotonic attenuation at distinct
synapses may conceal any clear quantal increments in the distribution of amplitudes of the
postsynaptic signal obtained by repeatedly sampling the connections between pre- and postsy-
naptic neurons.
Despite the differences in quantal neurotransmission between the CNS and neuromuscular
junction, measuring the trial-to-trial uctuations in the amplitudes of postsynaptic currents or
potentials yields a powerful insight into the loci of alterations in synaptic strength. In particu-
lar, a change in the fraction of presynaptic stimuli evoking failure of transmission implies an
alteration in p or n. A more indirect method relies on measuring the trial-to-trial variability of
the signal, as estimated by the coefficient of variation (CV: standard deviation/mean). Although
the evidence is more circumstantial, changes in CV are again generally associated with alter-
ations in p or n. Conversely, if the average size of the postsynaptic response is not associated
with a change in either the failure rate or the CV, the simplest inference is that it is mediated by
a change in Q. An important application of quantal analysis is for understanding the mecha-
nisms of the expression of synaptic plasticity, and some of the results and pitfalls of this
approach are discussed in Chapter 10.

can be attributed to unjustied assumptions about the under- esting forms of synaptic plasticity (long-term potentiation
lying mechanisms. and depression, LTP and LTD, respectively) persist for hours if
not longer and are considered in detail in Chapter 10. There
6.2.4 Short-term Plasticity are, however, other forms of use-dependent plasticity that last
up to a few minutes and that may play an equally important
Implicit in the outline of quantal transmission given above is role in the second-to-second traffic of information through
the assumption that action potentials repeatedly invading a synapses. Phenomenologically, they are divided into increases
presynaptic terminal trigger neurotransmitter release with a and decreases in transmission: facilitation, augmentation, and
constant probability. In fact, synapses show numerous forms potentiation on the one hand and depression on the other
of memory of their activation history. Some of the most inter- (Zucker and Regehr, 2002). There is some redundancy in these
 Synaptic Function 211

terms. Facilitation describes the enhancement of transmission variables, such as their temporal relation and the distribution
frequently seen following a preceding action potential. If the of voltage-gated channels (Johnston et al., 1996; Larkum et al.,
synapse is stimulated twice in rapid succession and the second 1999; Hausser et al., 2001; Stuart and Hausser, 2001).
response is larger than the rst, the phenomenon is often Nonlinearities in the secondary message cascade may also
referred to as paired pulse facilitation. Augmentation refers amplify metabotropic receptor-mediated signals. Another
to a gradual increase in synaptic strength with repeated stim- postsynaptic mechanism contributing to the facilitation of
ulation. Potentiation (often referred to as post-tetanic poten- some glutamatergic signals is relief from voltage-dependent
tiation to distinguish it from long-term potentiation) requires block of the ion channel by polyamines (Rozov and
a high-frequency train of stimuli for its induction and persists Burnashev, 1999). Conversely, at some synapses desensitiza-
for up to a few minutes after the end of the train. Many tion of receptors can contribute to use-dependent depression
synapses do not show these phenomena and, instead, depress of neurotransmission (Brenowitz and Trussell, 2001).
when repeatedly stimulated. At some hippocampal synapses Because principal neurons frequently discharge in bursts of
early facilitation gives way to later depression with prolonged action potentials, the degree of postsynaptic facilitation or
trains of stimulation. depression during such bursts may contain much of the infor-
These distinct patterns of short-term plasticity result from mation transmitted through the network. Indeed, it has been
a large number of processes occurring principally in the argued that lasting changes in short-term plasticity are of
presynaptic terminals (Thomson, 2000; Zucker and Regehr, greater importance for information encoding than synaptic
2002). The following is an incomplete list of events that con- strength per se (dened as the size of an EPSP or IPSP evoked
tribute to short-term plasticity. after an interval sufficiently long for short-term plasticity phe-
nomena to have dissipated) (Markram and Tsodyks, 1996).
1. Facilitation/augmentation/potentiation
However, because short-term plasticity is mediated predomi-
If the presynaptic free Ca2 concentration has not
nantly through use-dependent alteration in release probabil-
returned to baseline levels or if the occupancy of
ity, the information contained in a facilitating or depressing
Ca2 buffers is high, successive action potentials may
burst is inevitably corrupted by the stochastic nature of trans-
give rise to larger increments of Ca2, thereby
mitter release. That is, the degree of facilitation or depression
enhancing exocytosis.
at a synapse can be reliably detected by the network only by
High-affinity Ca2 sensors detect a slow build-up of
averaging out the response to a stereotyped sequence of action
Ca2 in the presynaptic varicosity and contribute to
potentials repeatedly delivered to the presynaptic neuron. This
triggering exocytosis.
has led to an almost diametrically opposite view of the impor-
Some presynaptic protein kinases activated with
tance of bursts: Because high-probability synapses often
repeated Ca2 inux may phosphorylate proteins
depress and low-probability synapses often facilitate, the func-
involved in making vesicles available for exocytosis
tion of bursts is to cancel out the effects of the initial release
or in triggering exocytosis itself.
probability and minimize the sampling error arising from sto-
2. Depression
chastic release. Thus, it is argued that the most important
The readily releasable pool of quanta (possibly
modiable parameter relevant to information transmission is
equivalent to docked vesicles) may become depleted.
the quantal amplitude (Lisman, 1997). Against this back-
Repeated depolarizing pulses inactivate presynaptic
ground, there has recently been considerable interest in the
voltage-dependent Ca2 channels, reducing the
long-term consequences of particular timing patterns of pre-
amount of Ca2 entry following successive action
and postsynaptic action potentials. Notably, it has been
potentials.
reported at several cortical synapses that if presynaptic gluta-
Depletion of extracellular Ca2 from the restricted
mate release occurs shortly before a postsynaptic action
synaptic cleft may reduce its availability for the
potential it can lead to persistent enhancement of transmis-
presynaptic terminal.
sion. If, on the other hand, the presynaptic stimulus lags
Presynaptic autoreceptors activated by released neu-
behind the postsynaptic action potential, it can be followed by
rotransmitters decrease further transmitter release
depression of transmission (Bi and Poo, 2001). The mecha-
during subsequent action potentials.
nisms underlying these phenomena and their relevance to
The above list focuses on presynaptic mechanisms that learning are further considered in Chapter 10.
contribute to short-term plasticity. However, postsynaptic
mechanisms also contribute. In particular, depolarizing
synaptic potentials can summate, leading to activation of
regenerative currents in the postsynaptic dendrite. 6.3 Glutamatergic Synaptic Transmission
Interactions with dendritic action potentials further compli-
cate the relation between voltage signals measured at the soma The main excitatory transmitter in the hippocampus, as else-
and presynaptic transmitter release: Orthodromically or where in the mammalian CNS, is glutamate. Although the ear-
antidromically propagating action potentials can either liest evidence for this was principally obtained in the spinal
amplify or shunt synaptic potentials, depending on many cord, many important insights have come from work on
212 The Hippocampus Book

rodent hippocampal tissue: Glutamate depolarizes hippocam- cles, so it is unclear how it could be packaged appropriately to
pal neurons (Biscoe and Straughan, 1966), and tritiated gluta- fulll a role as a fast neurotransmitter.
mate is avidly taken up by areas enriched in synapses Following release, glutamate is taken up into neurons and
(Storm-Mathisen and Iversen, 1979). Moreover, electrical glia, although this process takes place at a much slower rate
stimulation evokes glutamate release (Dolphin et al., 1982; than passive diffusion away from the release site (see below).
Walker et al., 1995), and antibodies raised against glutamate Of the cloned glutamate transporters, the most abundant in
stain presynaptic terminals at asymmetrical synapses the rodent hippocampus is GLT (also known as EAAT2), fol-
(Ottersen et al., 1990). Importantly, glutamate activates the lowed by GLAST (EAAT1) and EAAC1 (EAAT3) (Danbolt,
three principal types of receptors that mediate ionotropic 2001). In the hippocampus, GLT and GLAST are almost exclu-
excitatory transmission: -amino-3-hydroxy-5-methyl-isoxa- sively expressed in astrocytes. EAAC1 is principally expressed
zole-propionic acid (AMPA), kainate, and N-methyl-D-aspar- in neurons although at a relatively low level, so it probably
tate (NMDA). Finally, a vesicular glutamate transporter has contributes relatively little to clearing glutamate following
recently been identied (Bellocchio et al., 2000; Takamori et exocytosis. However, it has been proposed to supply glutamate
al., 2000). for GABA synthesis in inhibitory boutons (Sepkuty et al.,
Although there is abundant evidence that glutamate medi- 2002; Mathews and Diamond, 2003).
ates fast excitatory transmission at glutamatergic synapses, Glutamate taken up into astrocytes is decarboxylated by
there are persistent reports that aspartate also can be released glutamine synthetase (Fig. 64). The intracellular concentra-
from hippocampal slices in an activity-dependent manner tion of glutamate in astrocytes is thereby kept low, which is
(Szerb, 1988; Nadler et al., 1990). Aspartate immunoreactivity necessary to make continued uptake of glutamate energeti-
has also been found in certain presynaptic glutamate-releasing cally favorable. The energy required for uptake of glutamate
terminals, and the abundance of this signal decreases follow- comes from the co-transport of three Na ions down their
ing intense synaptic activity (Gundersen et al., 1998). Because electrochemical gradient. In addition, ux measurements
aspartate is a weak agonist at NMDA receptors (although have shown that one H ion accompanies glutamate, and one
inactive at AMPA or kainate receptors), these observations K ion is countertransported (Zerangue and Kavanaugh,
raise the possibility that it may be co-released with glutamate. 1996). Because the net movement of glutamate is dictated by
However, aspartate is not taken up effectively by isolated vesi- the electrochemical gradients for these other species, it is pos-

Figure 64. Glutamateglutamine cycle. The excitatory neurotrans- also shows the arrangement of several types of glutamate
mitter glutamate is synthesized from -ketoglutarate, which is pro- receptors: AMPA and NMDA receptors tend to be located
duced by the Krebs cycle, and recycled from glutamine. Following opposite release sites, and group I metabotropic receptors
release, glutamate is taken up by glial cells, where it is rapidly [mGluR (I)] tend to be located in a perisynaptic distribution.
decarboxylated to form glutamine. Glutamine can diffuse across Group II and III metabotropic receptors tend to be located
the extracellular space back to presynaptic varicosities. The diagram presynaptically.

Presynaptic axonal
ATP H ADP varicosity
-ketoglutarate

H
Aminotransferase
mGluR (II/III)

Glutamine Glu
Glu
Glutaminase
Glutamine
Transporters
Glutamine Glu
Glu
, Na, H
synthetase Na, Ca2 Na
 K

Glial process

mGluR (I) NMDAR AMPAR

Postsynaptic neuron
 Synaptic Function 213

sible to reverse uptake; moreover, evidence exists that such the properties of native, although not necessarily synaptic,
reversed uptake can take place during ischemia, when extra- receptors) or from cells expressing recombinant receptors.
cellular K builds up and cells become depolarized because of
ATP depletion (Rossi et al., 2000). Because the extracellular 6.3.1 AMPA Receptors
accumulation of glutamate can exacerbate depolarization of
neurons by acting at glutamate receptors, there is the possibil- AMPA receptors are composed of different combinations of
ity of triggering a positive feedback loop. This cascade may four subunits (GluR1-4, also known as GluRA-B) (Hollmann
underlie part of the excitotoxic role of glutamate in stroke. et al., 1989; Keinanen et al., 1990; Hollmann and Heinemann,
Astrocytic glutamine, formed from decarboxylation of glu- 1994; Nakanishi et al., 1998; Ozawa et al., 1998). They occur at
tamate, can enter metabolic pathways in astrocytes; but it can almost all excitatory synapses in the hippocampus and gate a
also diffuse passively down its concentration gradient into the cation-selective channel. At resting membrane potentials, Na
extracellular space, a phenomenon facilitated by several trans- inux accounts for most of the current, but the channel is also
porters. From here it is taken up into presynaptic terminals, permeant to other small monovalent cations, so K efflux can
where it can be converted back to glutamate and repackaged also occur at depolarized potentials. Most AMPA receptors in
into vesicles. pyramidal neurons of the adult hippocampus (at least in
The three major classes of ionotropic glutamate receptors rodents) are thought to be GluR1-2 or GluR2-3 tetramers
(Table 61) take their names from agonists that activate them (Wenthold et al., 1996).
in a relatively selective fashion (Watkins and Evans, 1981; AMPA receptors require a rapid pulse of glutamate in
Nakanishi et al., 1998; Ozawa et al., 1998). These agonists do excess of approximately 100 M to open. When a membrane
not normally exist in the brain. At most excitatory hippocam- patch taken from the soma or proximal dendrite of a hip-
pal synapses, EPSCs are mediated by AMPA and NMDA pocampal neuron is exposed to a pulse of 1 mM glutamate
receptors, which have strikingly different biophysical and (roughly corresponding to the synaptic glutamate transient;
pharmacological properties. Kainate receptors play a relatively see below) a current is generated with a rapid rise time
poorly understood role in synaptic transmission. All three (100600 s at physiological temperature). This reects both
receptor types are heteromultimeric structures probably made very fast binding kinetics and a high opening probability: the
up of four pore-forming subunits. These subunits are coded peak probability that a bound receptor is in the open state has
for by several genes that show considerable homology and been estimated at approximately 0.6 (Jonas et al., 1993).
exist in several splice variants, which contribute to dening Native AMPA receptors deactivate rapidly following clear-
their kinetic properties. The importance of the molecular ance of synaptic glutamate (with a time constant of 2.33.0
variability of the subunits making up each class of receptors is ms) (Colquhoun et al., 1992). Deactivation is probably suffi-
considered further in Chapter 7. Here, we concentrate on the cient to explain the termination of AMPA receptor-mediated
major pharmacological and biophysical differences that exist EPSCs on the grounds that glutamate is cleared from the
among the various receptors. Much of the information sum- synaptic cleft faster than this. If glutamate is not cleared,
marized here has been obtained by studying the behavior of however, AMPA receptors close rapidly and enter a desensi-
small numbers of receptors present in membrane patches tized state from which they recover relatively slowly. The
exposed to brief pulses of glutamate and studied with voltage- time course of desensitization depends on the subunit com-
clamp methods. These patches are taken from the somata of position of the receptors and is affected by alternative splicing
neurons in acute hippocampal slices or in cultures (reecting of the subunit mRNA (so-called ip and op variants; see

Table 61.
Ionotropic Glutamate Receptors

Receptor type Agonists Subunits Permeant ions Examples of antagonists

AMPA (-amino-3- Glutamate (endogenous), GluR1-4 Na, K Kynurenic acid

hydroxy-5-methyl-4- AMPA, kainate (GluRA-D) (Ca2) 6-Cyano-7-nitroquinoxaline-2,3-dione
isoxazolepropionic acid) (CNQX)
GYKI53655
Kainate Glutamate (endogenous), Glur5-7, KA1,2 Na, K Kynurenic acid
kainate (Ca2) CNQX
LY382884 (GluR5-specic)
NMDA (N-methyl-D- Glutamate, glycine, NR1, Na, K, Kynurenic acid
aspartic acid) D-serine (endogenous co- NR2A-D Ca2 D-Aminophosphonovalerate (D-AP5)
agonists), NMDA 3-[(RS)-2-carboxypiperazine-4-yl]-propyl-
1-phosphonate (CPP)
214 The Hippocampus Book

Chapter 7), but it usually proceeds with a decay time constant form of GluR2 is present in the receptor, its conductance is
of the order of 510 ms) (Mosbacher et al., 1994). AMPA relatively small and is independent of the transmembrane
receptors can even desensitize in the presence of glutamate voltage. If the edited GluR2 is absent from the receptor, the
concentrations that are insufficient to open them or if the glu- currentvoltage relation becomes highly nonlinear (Fig. 65).
tamate concentration rises sufficiently slowly. This form of That is, the receptor functions as a rectier, with a conduc-
desensitization may be an adaptation that prevents excessive tance that increases with the transmembrane potential differ-
receptor activation under pathological conditions where ence. This transition between low- and high-conductance
extracellular glutamate accumulates. Desensitization may also states is due to blockade of the receptor by polyamine mole-
occur with extremely brief (millisecond) exposure to the ago- cules, which enter the ion channel but can be expelled in a
nist, so if a second pulse is applied a smaller response is voltage-dependent manner (Bowie and Mayer, 1995; Donevan
obtained. However, this phenomenon does not contribute and Rogawski, 1995; Kamboj et al., 1995; Koh et al., 1995).
appreciably to frequency-dependent plasticity of transmission The single-channel conductance of native receptors varies
in the hippocampus (Hjelmstad et al., 1999). broadly from  1 pS to approximately 30 pS, depending not
Depending on their subunit composition, AMPA receptors only the presence of GluR2 but also on the identity of other
can also show signicant permeability to Ca2 ions. This subunits (Swanson et al., 1997). Bound receptors do not
permeability is determined by the presence or absence of a crit- remain open continuously. In common with other ligand-
ical amino acid (arginine, R) in a pore-lining segment of the gated ion channels, they icker between open and closed
GluR2 subunit. This subunit undergoes post-transcriptional states; and even in their open state they uctuate among dis-
RNA editing resulting in a change of the amino acid tinct preferred conductance levels. Although most kinetic
at this position from glutamine (Q), encoded by the genomic schemes assume that two glutamate molecules must bind for
sequence, to arginine (Sommer et al., 1991). The presence the receptor to open (Jonas et al., 1993; Diamond and Jahr,
of the edited form of GluR2 ensures that the receptor is imper- 1997), there is little direct experimental data to support this
meable to Ca2, which is the case for most of the glutamate assumption. Indeed, it has been proposed that four binding
receptors in principal cells. If the GluR2 subunit is absent, the steps to a tetrameric receptor take place, and that the conduc-
receptor has signicant Ca2 permeability. Such receptors are tance level of an individual channel increases with the number
present in some hippocampal interneurons (Geiger et al., of glutamate molecules bound (Rosenmund et al., 1998). Both
1995). In addition, some fetal GluR2 subunits are not the opening probability and the conductance of the channel
edited. can be modulated by phosphorylation (Derkach et al., 1999;
Q/R editing has several other consequences for AMPA Banke et al., 2000), phenomena that may play an important
receptor function (Swanson et al., 1997). If the edited (R) role in synaptic plasticity (see Chapter 10).

Figure 65. Rectication of unedited GluR2-containing AMPA recording with an intracellular solution devoid of polyamines
receptors. A, Current (I)voltage (V) relationships of glutamate- (lled symbols) the IV relation becomes more linear, explained
evoked currents obtained in a cell expressing Ca2-permeable by washout of endogenous polyamines from the cytoplasm.
unedited GluR2 (GluRB) subunits. The open circles show a nonlin- B, IV relation measured with a pipette solution containing
ear, rectifying, IV relation 1 minute after beginning whole-cell 25 M spermine, an endogenous polyamine. The rectication
recording. This is explained by a blockade of the ion pore by is more marked and persists despite prolonged recording.
endogenous polyamines, which is relieved by increasing the voltage (Source: Rozov et al., 1998, with permission.)
gradient across the membrane. Fifteen minutes after beginning the

1 minute
15 minutes

I norm 1.0 I norm

0.2

80
60
40
20
0.5
V (mV) 20 40

0.2

80
60
40
20

0.4
V (mV) 20 40 60

0.6

0.5

0.8

1.0
1.0

A B
 Synaptic Function 215

6.3.2 Kainate Receptors kainate receptors at various synapses, activated by different

concentrations of agonists, either enhance or depress trans-
Kainate receptors share many features with AMPA receptors mitter release. The mechanisms underlying these phenomena
(Lerma et al., 1997, 2001; Bleakman, 1999; Chittajallu et al., are incompletely understood (Kullmann, 2001) and may
1999; Ben-Ari and Cossart, 2000; Frerking and Nicoll, 2000; include depolarization, Ca2 inux via permeable receptors,
Kullmann, 2001). They too are heteromultimeric, made up of and coupling to a metabotropic cascade (Rodriguez-Moreno
different combinations of ve subunits: GluR5-7 and KA1-2. and Lerma, 1998). Controversy also surrounds the subunit
However, not all of these subunits have the same status composition of the kainate receptors mediating postsynaptic
because receptors made up of KA1 or KA2 alone are nonfunc- and presynaptic kainate receptor-mediated signaling. Some
tional. GluR5-6 undergo Q/R editing with similar conse- pharmacological data suggest a major role for GluR5-contain-
quences as for GluR2, although the proportion of edited ing receptors at mossy bers (Bortolotto et al., 1999; Lauri et
subunits falls considerably below 100%. This means that the al., 2001). However, the results of genetic deletion experi-
Ca2 permeability, single-channel conductance, and rectica- ments are more consistent with a role for GluR6-containing
tion of the receptor cannot be inferred easily from the subunit receptors at mossy bers and GluR5-containing receptors on
composition. interneurons (Contractor et al., 2000, 2001; Mulle et al., 2000).
The biophysical properties of recombinant kainate recep- Although presynaptic kainate receptors exert a powerful inu-
tors are similar to those of AMPA receptors. That is, they open ence on synaptic function, the adaptive signicance of the
and desensitize rapidly, and they have single-channel conduc- enhancement and depression of transmitter release remains a
tances, rectication properties, and Ca2 permeabilities that subject of speculation.
depend on Q/R editing (Bowie and Mayer, 1995; Kamboj et
al., 1995). However, there are some unexplained discrepancies 6.3.3 NMDA Receptors
between the relatively low affinity and rapid kinetics of
kainate receptors studied in isolated cells and the nding that NMDA receptors consist of subunits belonging to two rela-
synaptic kainate receptor-mediated signals exhibit very slow tively distinct subtypes (McBain and Mayer, 1994); they exist
kinetics (Castillo et al., 1997b; Lerma et al., 1997, 2001; Vignes as heteromultimers of NR1 and NR2A-D subunits. (A recently
and Collingridge, 1997; Cossart et al., 1998; Frerking et al., described NR3 subunit does not appear to play an important
1998). Some kainate receptor-mediated EPSCs can last more role in the hippocampus.) The NR1 subunit is encoded by one
than 100 ms, at which time they would have been expected to gene but exists in several alternatively spliced isoforms. It does
have deactivated following clearance of glutamate or desensi- not bind glutamate but, instead, contains an important bind-
tized if glutamate persisted. Several explanations have been ing site for another amino acid, glycine or D-serine, which acts
proposed for this discrepancy, including the possibility that as a co-agonist (see below). The NR2A-D subunits, on the
synaptic kainate receptors differ from nonsynaptic receptors other hand, contain the glutamate-binding site (Laube et al.,
in their subunit composition because of the actions of acces- 1997). They are encoded by four genes and are variably
sory proteins or because of site-specic phosphorylation expressed in different regions of the brain and at different
(Garcia et al., 1998). Among other possible explanations for stages of development (see Chapter 7).
the slow kinetics of kainate receptor-mediated EPSCs is that NMDA receptors show many striking properties that mark
glutamate needs to diffuse a long distance to reach the recep- them out as quite different from AMPA and kainate receptors.
tors, perhaps because they are relatively remote from the site Notably, they have very slow kinetics and can continue to
of exocytosis. Finally, synaptic co-release of a modulatory sub- mediate an ion ux for several hundreds of milliseconds after
stance might alter the response of kainate receptors to gluta- the glutamate pulse has terminated (activation time constant
mate. Another puzzle is that, although kainate receptors are is approximately 7 ms; deactivation time constants are
abundant in the hippocampus, synaptic responses mediated approximately 200 ms and 13 s). The slow kinetics are
by them are very small and generally require trains of high- explained by an extremely slow receptor unbinding rate
frequency stimuli to be detected. Recently, however, relatively (Lester et al., 1990). That is, once glutamate molecules have
fast, apparently monoquantal kainate receptor-mediated become bound to NMDA receptors, they remain bound for a
EPSCs have been described in interneurons and CA3 pyrami- long time, during which time the ionophore can undergo
dal neurons (Cossart et al., 2002). Why such currents have repeated opening. The maximal probability that a bound
been overlooked previously is unclear. receptor is in the open state has been estimated at between
Evidence has emerged for a major role of presynaptic 0.05 and 0.3, which is lower than the corresponding probabil-
kainate receptors in modulating transmitter release ity for AMPA receptors (Jahr, 1992; Rosenmund et al., 1995).
(Chittajallu et al., 1996; Rodriguez-Moreno et al., 1997; Vignes The range of estimates reects alterations in NMDA receptor
et al., 1998; Kamiya and Ozawa, 2000; Cossart et al., 2001; kinetics as a function of intracellular factors, including Ca2
Jiang et al., 2001; Schmitz et al., 2001a), axon excitability ions. Reecting the very slow dissociation rate, the apparent
(Semyanov and Kullmann, 2001), and synaptic plasticity affinity of NMDA receptors under steady-state conditions is
(Contractor et al., 2001; Lauri et al., 2001). Surprisingly, much higher than that of AMPA receptors (the EC50 is in the
216 The Hippocampus Book

range of 110 M, instead of 100500 M) (Patneau and brane potentials, the EPSC has fast kinetics and is blocked by
Mayer, 1990). This distinction may explain several striking the AMPA/kainate receptor blocker 6-cyano-7-nitroquinoxa-
properties of glutamatergic EPSCs (discussed further below). line-2,3-dione (CNQX). Subsequent studies, taking advantage
In addition to their slow kinetics, NMDA receptors have three of more selective AMPA receptor antagonists (in particular
other important features. the 2,3-benzodiazepine GYKI53655), have established that
First, a second agonist-binding site (the strychnine-insen- AMPA receptors account for the fast EPSC at negative poten-
sitive glycine site) must be occupied before glutamate is able tials. This EPSC is analogous to the fast synaptic current car-
to activate them (Johnson and Ascher, 1987; Kleckner and ried by acetylcholine receptors at the neuromuscular junction,
Dingledine, 1988). However, some estimates of the tonic which short-circuits the transmembrane potential differ-
extracellular glycine concentration in the brain suggest that ence. At depolarized membrane potentials, a slower compo-
the glycine-binding site is normally occupied. Alternatively, nent of the EPSC emerges that is blocked by the NMDA
D-serine can substitute for glycine, and it has been proposed receptor antagonist D-amino-5-phosphonovalerate (APV).
that this amino acid plays a physiological role in regulating This NMDA receptor-mediated current accounts for most of
NMDA receptor function (Schell et al., 1995; Baranano et al., the charge transfer because of the very slow receptor kinetics.
2001). Both AMPA and NMDA receptor-mediated components dis-
Second, they are highly permeable to Ca2 ions and mono- appear near 0 mV because this represents the reversal poten-
valent cations (Ascher and Nowak, 1988). Ca2 inux via tial for the mixture of monovalent cations that account for the
NMDA receptors plays a central role in several forms of long- bulk of the current. However, a small Ca2 inux still occurs
term synaptic plasticity (see Chapter 10). Ca2 ions are actu- at this membrane potential because the reversal potential for
ally ubiquitous secondary messengers, and NMDA receptor Ca2 is positive. If the neuron is held at a positive membrane
activation has been shown to trigger further release of potential, the dual-component EPSC appears again, although
Ca2 from intracellular stores (Emptage et al., 1999). it is now an outward current (Fig. 66).
Accompanying the high Ca2 permeability of NMDA recep- Kainate receptors generally do not contribute appreciably
tors is a relatively high single-channel conductance (4050 to the fast EPSC recorded in principal neurons. However, a
pS), which is greater than that of most AMPA receptors (Jahr low-amplitude kainate receptor-mediated current can be
and Stevens, 1987; Gibb and Colquhoun, 1992). detected in some neurons in response to high-frequency
Third, Mg2 ions block the ionophore in a voltage-depend- trains of action potentials (especially in CA3 pyramidal neu-
ent manner (Mayer et al., 1984; Nowak et al., 1984). Thus, at rons in response to mossy ber stimulation and in some
resting membrane potentials (more negative than approxi- interneurons) (Castillo et al., 1997a; Vignes and Collingridge,
mately 50 mV), NMDA receptors are unable to mediate an 1997; Cossart et al., 1998, 2002). Because it has very slow
EPSC even if glutamate and glycine (or D-serine) are present. kinetics, it may account for a sizable fraction of the total
They mediate an ion ux only when the membrane is depo- charge injected. Although kainate receptor-mediated EPSCs
larized. have been described as synaptic, it remains to be determined
The Ca2 permeability and Mg2 blockade of NMDA whether they arise from the activation of kainate receptors in
receptors explain their role as synaptic coincidence detectors: synapses rather than by spillover of glutamate onto more
Ca2 inux occurs only if there is a conjunction of presynap- remote extrasynaptic receptors.
tic glutamate release and postsynaptic depolarization, a situa-
tion that arises when pre- and postsynaptic activity occur 6.3.5 Metabotropic Glutamate Receptors
together (Wigstrom and Gustafsson, 1986). The signicance
of this phenomenon for long-term modication of synaptic Metabotropic receptors contain seven transmembrane seg-
strength is addressed in Chapter 10. ments and are coupled to nucleotide-binding G proteins,
which mediate most of their actions. They assemble as dimers,
6.3.4 Co-localization of Glutamate Receptors which are thought to be in a dynamic equilibrium between
two conformations: Binding of a glutamate molecule stabi-
Both AMPA and NMDA receptors are present at a roughly lizes them in the active state (Kunishima et al., 2000).
100-fold higher density at synapses than in extrasynaptic Metabotropic glutamate receptors fall into three classes,
membranes (Bekkers and Stevens, 1989; Nusser et al., 1998; although eight genes have been identied (Pin and Duvoisin,
Bolton et al., 2000). However, some synapses may be devoid of 1995; Ozawa et al., 1998; Schoepp, 2002) (Table 62).
AMPA receptors, especially early in postnatal development Group I receptors (mGluR1 and 5) are generally localized
(Nusser et al., 1998; Takumi et al., 1998; Petralia et al., 1999). to postsynaptic membranes and tend to occur in a halo
Less is known about the subcellular location of kainate recep- around the synaptic cleft (Baude et al., 1993). They are cou-
tors because adequate antibodies have not been developed. pled via G proteins to phospholipase C, and their activation
Reecting the co-localization of AMPA and NMDA (and leads to an increase in both inositol trisphosphate and diacyl-
possibly kainate) receptors, EPSCs generally show several glycerol (Fagni et al., 2000). Activation of these receptors at
pharmacologically, electrophysiologically, and kinetically dis- mossy ber synapses can elicit a slow EPSP mediated by
tinct components (Forsythe and Westbrook, 1988; Hestrin et cation-selective conductance and can trigger Ca2 release
al., 1990). At most hippocampal synapses, at negative mem- from intracellular stores (Heuss et al., 1999; Yeckel et al.,
 Synaptic Function 217

A B

 20 mV pA 100

APV
150
100
50 50

40
mV

100

80 APV

200

100 pA

300

50 ms

Figure 66. Dual-component glutamatergic transmission at hip- tion of the APV-sensitive component (circles, measured at a time
pocampal synapses. A, EPSCs recorded in a pyramidal neuron at shown by the dotted line in A) shows a region of negative slope
various voltages. As the postsynaptic cell is depolarized, a slow com- (that is, increasing conductance as the cell is depolarized) character-
ponent appears. This slow current is abolished by the NMDA recep- istic of voltage-dependent relief of blockade of NMDA receptors by
tor blocker traces marked APV. B, Current-voltage (IV) relation Mg2 ions. In contrast, the peak of the EPSC (triangles) has an
is plotted for the peak of the EPSCs (triangles) and for a later time approximately linear IV relation because the AMPA receptors in
point indicated by the vertical dotted line in A. Filled symbols, con- pyramidal neurons mainly contain edited GluR2 subunits. (Source:
trol conditions; open symbols, in the presence of APV. The IV rela- Hestrin et al., 1990 with permission).

1999). Some responses appear not to involve G proteins, coupled to adenylate cyclase via G proteins. The distinction
implying that group I receptors may be coupled to other sec- between group II and group III receptors is pharmacological:
ond messenger cascades (Heuss et al., 1999). L-AP4 is a selective agonist at group III, but not group II,
Group II (mGluR 2, 3) and group III (mGluR4, 6, 7, 8) receptors. However, the EC50 for L-AP4 differs markedly
receptors tend to be located in presynaptic membranes. At between the individual receptors in group III, so this classi-
mossy bers, mGluR2 receptors are located relatively far from cation is somewhat arbitrary (Wu et al., 1998).
glutamate release sitesin axonal membranesimplying The physiological roles of metabotropic receptors are not
that they detect only glutamate molecules that have escaped fully understood. The perisynaptic postsynaptic group I
from the synaptic cleft (Yokoi et al., 1996). receptors may preferentially respond to trains of action poten-
Several group III receptors, on the other hand, tend to be tials that result in the prolonged presence of glutamate in their
located in synapses, that is, very close to or even within active vicinity. Indeed, such stimulus patterns have been used to
zones (Shigemoto et al., 1996, 1997). Both types are negatively evoke postsynaptic currents and Ca2 signals mediated by

Table 62.
Metabotropic Glutamate Receptors

Group Subtypes Transduction mechanisms Agonists Examples of antagonists

I mGluR1,5 Phospholipase C Glutamate, (S)-3,5-dihydroxyphenylglycine 3,5-Methyl-4-carboxyphenylglycine

(DHPG) (MCPG)
1S,3R-1-aminocyclopentane-1, 3-dicarboxylic
acid (1S,3R-ACPD)
II mGluR2,3 Adenylyl cyclase Glutamate MCPG
1S,3R-ACPD LY341495
(2S,2R,3R)-2-(2,3-dicarboxycyclopropyl)
glycine (DCG-IV)
III mGluR4,6-8 Adenylyl cyclase Glutamate Methylserine-O-phosphate (MSOP)
L-Amino-phosphonobutyrate (L-AP4)
218 The Hippocampus Book

group I receptors (Heuss et al., 1999; Yeckel et al., 1999). As for terminals of GABAergic neurons. Its expression varies in
group II receptors, their predominantly extrasynaptic presy- response to changes in neuronal activity, implying that it
naptic location implies that they detect the extracellular build- may play a regulatory role in GABAergic transmission
up of glutamate and that they therefore act as autoreceptors (Soghomonian and Martin, 1998). Uptake into vesicles is
that regulate neurotransmitter release as a function of the facilitated by the vesicular GABA transporter VGAT, which
volume-averaged excitatory traffic (Scanziani et al., 1997). can also mediate vesicular glycine transport, although this
The intrasynaptic presynaptic location of some group III enzyme may not be essential for vesicular GABA accumula-
receptors prompts the speculation that they act as autorecep- tion because it is absent from a subset of GABAergic terminals
tors on a smaller spatial scale. However, they are also present (Chaudhry et al., 1998). Following exocytosis, GABA diffuses
at some GABAergic terminals (Shigemoto et al., 1997), which out of the synaptic cleft and is taken up by a family of four
are not known to release glutamate. There is evidence that transporters: GAT1-4 (Schousboe, 2000) (Fig. 67). These
they detect glutamate released from neighboring synapses transporters are located not only in astrocytes but also in
(Semyanov and Kullmann, 2000), so their role may be akin to interneurons and principal cells. GABA uptake is electrogenic;
that of group II receptors. that is, the amino acid is transported together with two Na
ions (and possibly one Cl
ion) (Kavanaugh et al., 1992;
6.3.6 Receptor Targeting and Anchoring Cammack et al., 1994). This stoichiometry normally ensures
that GABA is taken up into cells because the driving force pro-
A major challenge is to understand how receptors are targeted vided by the Na electrochemical gradient is sufficient to
and anchored to postsynaptic densities and how they are traf- overcome the gradient for GABA. However, because the trans-
cked to and from the synapse. These issues are considered port stoichiometry is less electrogenic than for glutamate, it
from a molecular perspective in Chapter 7. AMPA receptor has been suggested that GABA uptake can be reversed when
insertion plays a major role in synapse maturation and long- cells are depolarized (Richerson and Wu, 2003). The various
term potentiation (further considered in Chapter 10). Briey, GABA transporters show some differences in localization and
considerable evidence is emerging for two types of interaction substrate specicity: For instance, GAT1 is relatively abundant
between AMPA receptor subunits and intracellular proteins in neurons and is relatively insensitive to -alanine compared
involved in trafficking (Shi et al., 2001). One process is medi- to the other major transporter GAT3. Although GABA trans-
ated by an interaction between the cytoplasmic C-termini of port shows many similarities with glutamate uptake, there are
GluR1 (and possibly GluR4) subunits with protein(s) con- suggestions that it may be less efficient, either because indi-
taining a consensus motif known as a PDZ domain (see vidual transporters function more slowly or because they are
Chapter 7). This process is triggered by activation of Ca2- present in a lower concentration. In particular, synaptically
calmodulin-dependent protein kinase II (CaMKII) and con- released GABA appears to exert actions at high-affinity,
tributes to long-term potentiation. Another process involves remotely located receptors; and GABA escaping from the
an interaction between the C-terminus of GluR2 subunits and synaptic cleft can even be detected by recording from a mem-
both NSF (the ATPase involved in dissociating the SNARE brane patch containing GABAA receptors positioned on the
core complex; see above) and PDZ proteins. surface of a slice (Isaacson et al., 1993). In contrast, although
NMDA receptors are anchored to a large number of scaf- glutamate also exerts extrasynaptic actions, they are spatially
folding, signaling, and transduction proteins (Husi et al., much more restricted (see below).
2000; Sheng and Sala, 2001). This anchoring is mediated via GABA receptors are divided into ionotropic (GABAA) and
an interaction between the C-termini of NR2 subunits and metabotropic (GABAB) receptors (Table 6-3). GABAC recep-
several PDZ domain proteins, in particular PSD-95 and tors are ionotropic receptors with an unusual pharmacologi-
SAP97. Some group I metabotropic glutamate receptors are cal prole and are made up of  subunits. They are arguably a
also anchored via a PDZ domain protein, Homer (Brakeman variant of GABAA receptors (Barnard et al., 1998; Bormann,
et al., 1997). Relatively less is known of the mechanisms that 2000). Because they are not known to play a major role in the
target and/or anchor other metabotropic receptors: One splice hippocampus, they are not considered here.
variant of mGluR7 interacts with another intracellular pro-
tein, PICK1 (Perroy et al., 2002), but how other receptors are 6.4.1 GABAA Receptors
localized to presynaptic terminals is unclear.
GABAA receptors are heteropentameric, and consist of sub-
units drawn from at least 7 different families: 1-6, 1-3, 1-3, ,
, , and  (Mehta and Ticku, 1999). Of these, 6, , , and 
6.4 GABAergic Synaptic Transmission appear to be excluded from the rodent hippocampus or to
occur at very low levels. The remaining subunits show differ-
The major inhibitory transmitter GABA is synthesized by ential distributions at a macroscopic level. Thus, 1, 2, 4, 3,
decarboxylation of glutamate. Two isoforms of glutamic acid 2, and  are present in the dentate gyrus. In the hippocampus
decarboxylase exist. GAD67 is widespread in the cytoplasm, proper, the principal subunits are 1, 2, 5, 3, and 2, with
whereas GAD65 is more closely associated with presynaptic relatively lower levels of 4, 1, 2, 3, and ; 3 and 1 are
 Synaptic Function 219

Presynaptic axonal
ATP H ADP varicosity

Glutamate H

Glutamate GABAB
decarboxylase GABA GABA
Succinate
GABA Transporters
transaminase GABA GABA, Na Cl

GABAA
Postsynaptic neuron GABAB
Figure 67. GABA cycle. The inhibitory neurotransmitter GABA is gram also shows the arrangement of GABAA and GABAB receptors.
synthesized from glutamate via the action of two enzymes, GAD-65 Ionotropic GABAA receptors tend to be located opposite release
and GAD-67. Following release, GABA is taken up by both neurons sites, although extrasynaptic receptors also occur in dentate granule
and glial cells. GABA can be converted back to glutamate or to suc- cells. GABAB receptors occur both pre- and postsynaptically.
cinate by the mitochondrial enzyme GABA transaminase. The dia-

present only at very low levels (Sperk et al., 1997). Most hip- receptors tend not to be conned to synapses but have a high
pocampal GABAA receptors probably contain two  subunits affinity for GABA and are relatively insensitive to benzodi-
and two subunits, together with either a  subunit or the  azepines. These -containing receptors are strong candidates
subunit but not both (Chang et al., 1996; Farrar et al., 1999; for mediating a GABAA receptor-dependent tonic inhibition
Whiting et al., 1999). The  subunits in particular play impor- in dentate granule cells (Overstreet and Westbrook, 2001;
tant roles in determining the affinity for GABA and the sensi- Nusser and Mody, 2002). The 4 2 receptors have recently
tivity to numerous modulatory agents such as Zn2 ions, been shown to be strongly potentiated by ethanol at low con-
steroids, ethanol, and exogenous pharmacological agents such centrations, making them a candidate for some of the psy-
as benzodiazepines, barbiturates, and general anesthetics chotropic properties of this drug (Sundstrom-Poromaa et al.,
(Barnard et al., 1998). The  subunits also affect several of 2002).
these parameters and, in addition, mediate anchoring of GABAA receptor-mediated IPSCs have a fast onset,
GABAA receptors to synapses via an indirect interaction with although their decay is generally slower than that of AMPA
gephyrin, a scaffolding protein that plays an important role in receptor-mediated EPSCs. Their single-channel conductance
the formation and stabilization of both GABAergic and is highly variable, ranging from < 1 to > 30 pS, depending on
glycinergic synapses (Essrich et al., 1998). The -containing their subunit composition.

Table 63.
GABA receptors

Receptor type Subunits Ions Agonists Antagonists Modulators

GABAA 1-6, 1-3, 1-3, , , ,  Cl

, HCO3
GABA Picrotoxin Benzodiazepines, bar-
(ion channel) Muscimol Bicuculline biturates, Zn2,
SR 95531 ethanol, neurosteroids
GABAB GBR1 G protein-gated inward recti- GABA Saclofen
GBR2 fying K (GIRK) channels Baclofen CGP35348
Ca2 channels
220 The Hippocampus Book

Depending on the subunit composition, benzodiazepines 6.4.2 GABAB Receptors

enhance the affinity of GABAA receptors for the agonist. This
causes prolongation of the IPSC and also usually increases the Metabotropic GABAB receptors occur as heterodimers com-
peak amplitude, implying that the occupancy of the receptors posed of GBR1 and GBR2 (Jones et al., 1998), both of which
is normally incomplete (Perrais and Ropert, 1999; Hajos et al., can undergo alternative splicing (Kuner et al., 1999). GABAB
2000). However, because benzodiazepines can, in addition, receptors are widespread at both pre- and postsynaptic ele-
increase single-channel conductance (Eghbali et al., 1997), ments of synapses and also in extrasynaptic membranes
estimating the occupancy of receptors from their effects on (Fritschy et al., 1999). The GABAB receptor agonist baclofen
IPSC amplitude is problematic. Barbiturates, on the other powerfully depresses the synaptic release of both glutamate
hand, enhance and prolong the IPSC by increasing the pro- and GABA, so the presynaptic targeting of GABAB receptors
portion of the time that GABA-liganded channels remain and/or their linkage to effector mechanisms is presumably not
open (Study and Barker, 1981). Both these classes of drugs confined to GABAergic terminals. Conversely, some
have the effect of potentiating GABAergic transmission, pos- GABAergic terminals appear to be insensitive to baclofen
sibly accounting for their usefulness as sedative, anxiolytic, (Lambert and Wilson, 1993). Among possible explanations for
and antiepileptic agents. Combined genetic and pharmaco- this nding are the absence of receptors at some terminals and
logical manipulation of the receptors has suggested that 1- heterogeneity in the identity of the presynaptic Ca2 channels
containing receptors preferentially mediate the sedative, but (see above).
not the anxiolytic, effects of systemic benzodiazepines GABAB receptors in the presynaptic membrane are nega-
(Rudolph et al., 1999), although whether and how the hip- tively coupled to N- and P/Q- type Ca2 channels via a G-pro-
pocampal formation contributes to these phenomena remains tein cascade (Anwyl, 1991; Mintz and Bean, 1993), although
to be determined. this may not account entirely for their effect on transmitter
GABAA receptors are permeable only to anions. The cur- release (Capogna et al., 1996). On the postsynaptic side, the
rent is carried overwhelmingly by Cl ions, but receptors are GABAB-triggered cascade leads to the opening of G protein-
also permeable to HCO3 ions (Kaila, 1994). Because neurons gated inward-rectifying K (GIRK) channels (Andrade et al.,
are normally depolarized relative to the Cl reversal potential, 1986; Misgeld et al., 1995). Activation of these channels causes
GABAA receptor-mediated IPSCs are generally hyperpolariz- a slow IPSP, lasting several hundred milliseconds. This is eas-
ing. The situation is different in neurons at early stages of ily distinguished from the GABAA receptor-mediated IPSP,
development (Ben-Ari et al., 1989). The intracellular Cl con- not only because of its slow kinetics but also because it is inde-
centration is relatively high because the principal extrusion pendent of the Cl reversal potential.
mechanism (the K/Cl co-transporter KCC2) is not In contrast to GABAA receptor-mediated IPSCs, it has
expressed abundantly (Rivera et al., 1999). The electrochemi- proved difficult to elicit GABAB receptor-mediated synaptic
cal gradient then dictates that Cl ions ow out of the cell, responses by activating individual presynaptic neurons. This
making the current depolarizing. Somewhat confusingly, this observation is difficult to reconcile with data that GABAB
is sometimes referred to as a depolarizing IPSC. Such depo- receptors generally have a higher affinity for GABA than do
larizing GABAergic signals can bring neurons to ring thresh- GABAA receptors. A possible resolution of the paradox is that
old and trigger Ca2 inux, and they may play an important the receptors are principally extrasynaptic, and that GABA
role in the early stages of neural circuit formation (Ben-Ari et released from multiple presynaptic terminals must accumu-
al., 1997). An analogous depolarizing response to GABA has late to activate them (Destexhe and Sejnowski, 1995;
also been observed in response to exogenous agonist applica- Scanziani, 2000). This argument suggests that uptake mecha-
tion to the dendrites of pyramidal neurons in acute slices nisms are insufficient to isolate GABAergic terminals func-
taken from adult brains (Andersen et al., 1980a). A major tionally from one another, at least as far as high affinity
mechanism underlying this phenomenon is that prolonged GABAB receptors are concerned.
and intense activation of GABAA receptors results in accumu-
lation of intracellular Cl, which eventually alters the reversal
potential for the receptors (Staley et al., 1995). In addition,
because GABAA receptors have a signicant permeability to 6.5 Other Neurotransmitters
HCO3, alterations in the electrochemical gradient for this
anion can contribute to the phenomenon. Extracellular K In comparison with glutamate and GABA, other neurotrans-
also accumulates through the action of KCC2, further con- mitters are present at far fewer synapses. Some of them appear
tributing to neuronal depolarization (Kaila et al., 1997). to carry a diffuse signal in the extracellular space and may
Although depolarizing responses can be achieved by ion- therefore play extensive roles in determining the overall state
tophoretic application of GABA and possibly during seizures of the hippocampal circuitry. Relatively little is known about
(Bracci et al., 2001), it is not clear that they occur with synap- the mechanisms underlying their release and postsynaptic
tic activity under physiological conditions in adult animals, actions.
which generally activates only a small proportion of the recep- ATP is thought to be co-released with other neurotrans-
tors on a neuron and for a short time. mitters, and activates ionotropic P2X receptors linked to a
 Synaptic Function 221

cation-selective channel. In the presence of broad-spectrum However, whether this is relevant to the role of endogenous
antagonists of amino acid receptors, a fast depolarizing dopamine depends on how and where it is released, which
purinergic postsynaptic current has been reported in hip- remains to be determined. An alternative possibility is that the
pocampal neurons (Pankratov et al., 1998; Mori et al., 2001). main role of dopamine is to modulate synaptic transmission.
However, to detect such a current it is often necessary to stim- Such a possibility is prompted by analogy with the effects of
ulate many presynaptic bers synchronously, implying that dopamine in the striatum, which receives a much stronger
the receptors are present in very low density or that little ATP projection from the substantia nigra. In this system, dopamin-
is released. The physiological or pathological role of this class ergic projections terminate on the necks of dendritic spines
of receptors is not known. ATP undergoes extracellular enzy- that receive glutamatergic terminals (Kotter, 1994). Activation
matic degradation, which eventually results in the formation of such dopamine receptors results in inhibition of EPSCs
of adenosine. Adenosine is probably also released by neurons (Calabresi et al., 1997), an action that has also been reported
directly, and extracellular adenosine accumulation occurs at some hippocampal synapses (Otmakhova and Lisman,
under conditions of metabolic stress. Presynaptic adenosine 1999).
receptors (mainly A1 in the hippocampus) are widespread on Acetylcholine is also released from extrinsic afferents,
excitatory terminals, where they depress glutamate release via although in this case their cell bodies lie in diencephalic struc-
a G protein-dependent mechanism (Wu and Saggau, 1994). tures, in particular the medial septal nuclei and the nucleus of
This phenomenon contributes to the potent depressant effect the diagonal band (see Chapter 3). Acetylcholine acts on
of hypoxia on excitatory synaptic transmission. GABA release both ionotropic (nicotinic) and metabotropic (muscarinic)
is relatively less sensitive to adenosine (Yoon and Rothman, receptors.
1991; Lambert and Wilson, 1993). Nicotinic receptors are pentameric, and two main types
Among other small molecules that act as neurotransmit- occur in the hippocampus: homomeric 7 and heteromeric
ters are the monoamines noradrenaline (norepinephrine), 4 2. These receptors can be distinguished by their affinity for
dopamine (DA), serotonin (5-HT), and histamine (Nicoll et acetylcholine, by the desensitization kinetics, and by selective
al., 1990) Norepinephrine acts at metabotropic -adrenergic agonists and antagonists. Homopentameric 7 receptors have
receptors, serotonin activates both ionotropic depolarizing 5- a lower affinity for the endogenous agonist than do 4 2
HT3 receptors and a large number of metabotropic receptors, receptors, desensitize fast, and are inhibited by methyllaca-
and histamine acts at metabotropic H1-3 receptors. The conitine (MLA) and -bungarotoxin. Activation of these
monoamines tend to be released from varicosities of extrinsic receptors by exogenous agonist application has been reported
afferent bers, mainly originating in hindbrain structures. to enhance evoked glutamate and GABA release (McGehee et
Norepinephrine, acting via receptors, increases the action al., 1995; Albuquerque et al., 1997; Dani, 2001). This obser-
potential frequency and reduces spike frequency adaptation vation implies that they are located presynaptically
during continued depolarizing current injection (Madison (MacDermott et al., 1999), a conclusion that has recently been
and Nicoll, 1982, 1984). This effect is mediated by a relatively supported by immunohistochemical evidence for abundant
selective inhibition of Ca2-gated K channels responsible -bungarotoxin staining co-localized with both glutamatergic
for the slow after-hyperpolarization following an action and GABAergic terminals in the hippocampus (Fabian-Fine et
potential (Haas and Konnerth, 1983) and requires a signaling al., 2001). However, postsynaptic staining was also seen in this
cascade involving protein kinase A (Pedarzani and Storm, study. Heteromeric 4 2 receptors have a relatively higher
1993). Serotonin and histamine have effects similar to those affinity for acetylcholine and nicotine, desensitize slowly, and
of norepinephrine but, in addition, exert other actions. are antagonized by dihydro- -erythroidine and mecamy-
Serotonin, in particular, activates GIRK channels (Andrade lamine. Activation of these receptors has been reported to
et al., 1986). The role of the ionotropic 5-HT3 receptors is depolarize a subset of neocortical interneurons (Porter et al.,
poorly understood. Histamine has been reported to enhance 1999). A major limitation of these studies is that they have
NMDA receptor-mediated currents (Bekkers, 1993). mainly relied on application of exogenous agonists. Thus,
The hippocampus receives dopaminergic projections from although they shed light on possible mechanisms of action of
both the substantia nigra pars compacta and the ventral nicotine, it is not known under what conditions endogenously
tegmental area. Dopamine acts at two groups of metabotropic released acetylcholine is able to activate these receptors.
receptors (Sealfon and Olanow, 2000). D1-like receptors (D1 Four muscarinic receptors (M1-4) are found in the hip-
and D5) are positively coupled to adenylate cyclase, which pocampus. Activation of muscarinic receptors has several
activates protein kinase A and leads to phosphorylation of the effects. The so-called M current is active at resting potentials
protein phosphatase inhibitor DARPP-32, whereas D2-like (Brown and Adams, 1980) and is inhibited by muscarinic
receptors (D2, D3, D4) are negatively coupled to this cascade. receptor agonists, thereby facilitating burst-ring of neurons.
Transcripts for all of these receptors can be detected in the It is mediated at least in part by heteromultimeric K channels
hippocampus. Relatively little is known about the synaptic containing KCNQ2 and KCNQ3 subunits (Wang et al., 1998;
function of this transmitter. Exogenous dopamine hyperpo- Selyanko et al., 1999; Shapiro et al., 2000). Muscarinic recep-
larizes a proportion of pyramidal neurons, an action that is tors have recently been reported to inuence the M current by
principally mediated by D1 receptors (Berretta et al., 1990). activation of phospholipase C, which leads to the breakdown
222 The Hippocampus Book

of phosphatidylinositol-4,5-bisphosphate (Suh and Hille, ally exert only indirect control over membrane potentials via
2002). The depolarizing actions of muscarinic agonists are G proteins. Before BDNF is elevated to the rank of neuro-
especially prominent in some interneurons (Pitler and Alger, transmitters, it is important to determine whether endoge-
1992; Behrends and ten Bruggencate, 1993). This phenome- nous BDNF can evoke the same effect.
non may also play an important role in generating the Some molecules have an uncertain status as neurotrans-
hippocampal  rhythm (Fischer et al., 1999), although the mitters in the hippocampus. Glycine receptors occur in the
precise synaptic and nonsynaptic mechanisms underlying this hippocampus, where they open Cl
conductance. However,
rhythm in vivo are far from clear (see Chapter 8). Other con- glycinergic synaptic transmission has not been demonstrated
sequences of muscarinic receptor activation include prolonga- (in contrast to the spinal cord). The amino acids taurine and
tion of action potential duration (Figenschou et al., 1996) and -alanine have recently been proposed to act as endogenous
activation of GIRK channels (Seeger and Alzheimer, 2001). agonists at glycine receptors, although it is unclear whether
The inhibition of Ca2-sensitive K channels by muscarinic they can be released in an activity-dependent manner (Mori et
receptors is reminiscent of the effect of receptor activation al., 2002).
by norepinephrine (see above). However, this action of mus- The Zn2 ion is associated with synaptic vesicles of gluta-
carinic receptors is mediated not by protein kinase A but by matergic synapses throughout the hippocampus but is espe-
Ca2/calmodulin-dependent protein kinase II (Pedarzani and cially prominent in mossy bers. Zn2 inhibits GABAA and
Storm, 1996). Muscarinic receptors also exist presynaptically, NMDA receptors, as well as glutamate uptake, although it
where they depress transmitter release (Behrends and ten is unclear whether this occurs during synaptic transmission
Bruggencate, 1993). This action is likely to be mediated by G under physiological conditions (Draguhn et al., 1990; Smart
protein-mediated inhibition of Ca2 channels. Again, the et al., 1994; Spiridon et al., 1998; Frederickson et al., 2000;
physiological role of this effect is unknown. Vogt et al., 2000). The dipeptide N-acetylaspartate may also
In contrast to the amino acids glutamate and GABA, sev- be released from presynaptic terminals and acts as an agonist
eral of the transmitters considered in this section tend not to at metabotropic glutamate receptors (Wroblewska et al.,
subserve conventional point-to-point (or wiring) transmis- 1993).
sion. Instead, they probably act principally by diffusing a rela- Nitric oxide (NO) and carbon monoxide (CO) deserve
tively long distance from the axonal varicosities where they are mention, although they do not appear to function as synaptic
released. Indeed, many such varicosities exist in the absence of neurotransmitters in the conventional sense. NO is synthe-
identiable postsynaptic structures. Although this principle sized in neurons expressing nitric oxide synthase (principally
would argue against an important role in the fast transmission a subset of interneurons) in a Ca2-dependent manner and
of information through the hippocampal formation, these diffuses through membranes to its targets, which include sol-
transmitters actually play crucial roles in regulating the uble guanylyl cyclase (Baranano et al., 2001). This agent
excitability of the circuitry. The involvement of the choliner- potentially acts as a retrograde messenger, but evidence to
gic projection to the hippocampus in setting the theta rhythm support thislet alone to indicate a physiological role for the
is an important example of such a phenomenon (see Chapters phenomenonis scanty.
8 and 11). Products of lipid metabolism may also play a role in inter-
A large number of peptides have also been shown to exist cellular communication. Notable among these products are
in axonal varicosities and to bind to specic receptors in the arachidonic acid (AA) (Piomelli, 1994), and the endocannabi-
hippocampus (Hkfelt et al., 1980). They include the opioid noids anandamide and 2-arachidonylglycerol (Davies et al.,
peptides, somatostatin, neuropeptide Y, galanin, and cholecys- 2002). Both AA and endocannabinoids are synthesized in a
tokinin. Of these, the opioid peptides (enkephalins, endor- Ca2-dependent manner and are potential retrograde mes-
phins, dynorphin, endomorphins) have been studied most sengers. The role of endocannabinoid signaling in modulation
extensively. They act at receptors that fall into three classes of inhibition is considered further below.
with distinct pharmacological proles (, , ), although
there is evidence that these receptors can heterodimerize
(Jordan and Devi, 1999).
Peptides are thought to be contained in dense-core vesicles, 6.6 Special Features of Individual
which occur in GABAergic synapses and at mossy ber Hippocampal Synapses
synapses. Many peptides probably act principally via spillover
to modulate either pre- or postsynaptic metabotropic recep- In contrast to the neuromuscular junction, hippocampal
tors in the vicinity of the site where they are released synapses usually occur at axonal varicosities (the widely used
(Weisskopf et al., 1993; Simmons and Chavkin, 1996). term terminal is thus misleading). These varicosities occur
Recently, brain-derived neurotrophic factor (BDNF) at regular intervals along many axons. Their spacing is
applied to hippocampal neurons was shown to evoke fast approximately 4 m for Schaffer collaterals and associational-
depolarizing currents mediated by an interaction between the commissural bers and approximately 5 m for mossy bers
TrkB receptor and NaV1.9 Na channels (Blum et al., 2002). (Shepherd et al., 2002) (see Chapter 3). They often contain
This nding is unexpected, not least because peptides gener- mitochondria, reecting the energetic demands of transmitter
 Synaptic Function 223

synthesis, vesicular packaging, release, and re-endocytosis postsynaptic kainate receptors (Castillo et al., 1997b; Vignes
(Shepherd and Harris, 1998). and Collingridge, 1997). The AMPA receptors at small spine
Only one type of axo-axonic synapse is known in the hip- synapses on principal neurons have a low permeability to
pocampus: the GABAergic synapse made by chandelier cells Ca2 and are nonrectifying, reecting the presence of edited
on axon initial segments (see Chapters 3 and 8). In contrast to GluR2 subunits in almost all receptors (Jonas et al., 1994).
the spinal cord, presynaptic specializations have not been Most AMPA receptors are thought to be GluR1/2 and GluR2/3
reported in contact with distal axons or axonal varicosities. It heteromers (Wenthold et al., 1996). Ca2 inux, however,
can thus be inferred that heterosynaptic modulation of trans- occurs as an indirect consequence of AMPA receptor activa-
mitter release by activation of metabotropic receptors occurs tion because the consequent depolarization allows Ca2-per-
via spillover transmission (see below). meable NMDA receptors to open (Yuste et al., 1999). In
addition, the depolarization allows voltage-gated Ca2 chan-
6.6.1 Small Excitatory Spine Synapses nels in the dendritic spine and nearby region of the dendrite
to open. A nal source of Ca2 is from intracellular stores,
The most abundant type of synapse in the hippocampus is the triggered by second messengers including inositol trisphos-
small glutamatergic synapse made on dendritic spines. This phate and Ca2 itself (Emptage et al., 1999). However, the lat-
subserves transmission from the perforant path to dentate ter is controversial as some studies have not found a
granule cells, as well as associational/commissural projections substantial role for Ca2 release from stores at these synapses
to CA3 pyramidal neurons and Schaffer collateral/commis- (Kovalchuk et al., 2000). Some of the disagreement about the
sural transmission to CA1 pyramidal neurons. The monosy- relative importance of the various sources of Ca2 probably
naptic projection from the perforant path to the distal reects differences in recording methods: Although whole-cell
dendrites of CA1 and CA3 pyramidal neurons has received recordings allow voltage control of the dendrite, they may
less attention but probably has broadly similar biophysical interfere with intracellular Ca2 homeostasis. Intracellular
properties (Berzhanskaya et al., 1998; Otmakhova et al., 2002). microelectrode recordings, though minimizing this source of
Similar synapses relay excitatory signals from CA1 pyramidal error, sacrice voltage control.
neurons to subicular pyramidal cells and from the latter to the Although AMPA and NMDA receptors both contribute
entorhinal cortex. Just over half of the presynaptic varicosities to EPSCs evoked by simultaneous stimulation of numerous
in CA1 appear to contain mitochondria (Shepherd and presynaptic bers, EPSCs evoked by action potentials in single
Harris, 1998). The postsynaptic densities range widely in area: presynaptic axons sometimes lack an AMPA receptor-medi-
from less than 0.07 m2 at synapses on thin spines to at least ated component (Kullmann, 1994; Isaac et al., 1995; Liao et al.,
0.42 m2 (Harris and Sultan, 1995). Dendritic spines in the 1995). Such EPSCs are evoked only under conditions where
dentate gyrus measure up to 0.8 m in diameter and approx- the postsynaptic neuron is depolarized or Mg2 ions are omit-
imately 1 m in length (Trommald and Hulleberg, 1997), and ted from the extracellular solution to reveal the NMDA
generally receive only one presynaptic bouton. However, the receptor-mediated component. This observation has led to
presynaptic varicosity is sometimes in synaptic contact with the hypothesis that AMPA receptors are absent or nonfunc-
more than one spine (Westrum and Blackstad, 1962; Sorra tional at a proportion of synapses, which are consequently
and Harris, 1993). Spines are thought to represent principally often referred to as silent synapses. Immunohistochemical
biochemical compartments rather than electrical barriers or studies provide some indirect support for this hypothesis:
ampliers for the propagation of synaptic currents to the den- Some small and/or immature synapses appear not to stain for
drite (Yuste et al., 2000). AMPA receptors (Nusser et al., 1998; Petralia et al., 1999;
The synaptic cleft is often closely apposed by an astrocytic Takumi et al., 1999). However, given the nite sensitivity of
process, although this is highly variable: In one study as few as the antibodies, inferring that the receptors are genuinely
57% of glutamatergic synapses in CA1 were contacted, and of absent relies on extrapolation. In contrast, NMDA receptor
these only approximately 43% of the perimeter of the synap- density appears to be more uniform across synapses of differ-
tic cleft was found to be bounded by an astrocytic process ent size (Racca et al., 2000). Taken together with the observa-
(Ventura and Harris, 1999). Because astrocytes express abun- tion that the proportion of silent synapses diminishes with
dant glutamate transporters, this implies that there may be age, it is tempting to conclude that synaptic maturation devel-
considerable variability in the spatiotemporal prole of gluta- ops by an initial expression of NMDA receptors, followed later
mate in the vicinity of the synapse following exocytosis. by insertion of AMPA receptors (Durand et al., 1996). This
A combination of immunohistochemical and ultrastruc- phenomenon may share mechanisms in common with long-
tural techniques has shown that both AMPA and NMDA term potentiation.
receptors are localized to the postsynaptic density (Nusser et Alternative explanations for silent synapses exist: Because
al., 1998; Takumi et al., 1998, 1999; Racca et al., 2000). NMDA receptors have a much higher steady-state affinity for
Whether kainate receptors are also present is not known. glutamate than AMPA receptors (Patneau and Mayer, 1990),
However, glutamatergic EPSCs at Schaffer collateralCA1 they might respond to a relatively low and prolonged pulse of
synapses are entirely blocked by antagonists of AMPA and glutamate that is insufficient to activate AMPA receptors. Such
NMDA receptors, leaving little room for the involvement of a glutamate prole could arise by diffusion from neighboring
224 The Hippocampus Book

synapses (spillover) (Asztely et al., 1997; Kullmann and and their interaction with receptors and transporters
Asztely, 1998). Thus, the NMDA receptors of the recorded cell around the synapse (Wahl et al., 1996; Rusakov and
might be acting as bystanders, eavesdropping on the excita- Kullmann, 1998; Barbour, 2001). There are considerable
tory traffic through many synapses formed on neighboring uncertainties in several critical parameters, including
cells. In favor of this model is the observation that raising the the diffusion coefficient of glutamate in the extracellular
recording temperature from room temperature (thereby space and the fraction of the local volume accounted for
enhancing glutamate uptake) reduces the discrepancy by the extracellular space. Nevertheless, some studies
between the number of quanta detected by NMDA and AMPA applying this approach have arrived at the conclusion
receptors. Conversely, blocking glutamate uptake pharmaco- that the glutamate concentration in the synaptic cleft
logically increases the ratio of NMDA to AMPA receptor- reaches a peak of approximately 1 to 3 mM almost
mediated signaling, consistent with an exacerbation of instantaneously and then drops rapidly with a multiex-
cross-talk among synapses. Further evidence in favor of gluta- ponential time course. Taking into account the kinetic
mate spillover comes from examining the effects of manipula- properties of NMDA receptors, this glutamate prole
tions of release probabilities and competitive glutamate implies that the NMDA receptors opposite a release site
antagonists on the time course of NMDA receptor-mediated should be almost saturated by a single vesicle. The occu-
EPSCs (Lozovaya et al., 1999; Diamond, 2001; Arnth-Jensen et pancy of AMPA receptors is generally found to be
al., 2002). slightly lower than that of NMDA receptors, reecting
Another possible explanation for selective activation of their lower affinity. Another result from numerical sim-
NMDA receptors is that the glutamate released from the ulations is that glutamate spreads outside the synaptic
presynaptic terminal to activate AMPA receptors is insuffi- cleft, although the extent of this extrasynaptic glutamate
cient. This could occur if release took place via a fusion pore. escape is sharply curtailed by transporters. Some studies
That is, glutamate could slowly escape from a vesicle via a have suggested that a signicant proportion of NMDA
small, possibly ickering, opening to the synaptic cleft. receptors within a radius of approximately 500 nm are
Evidence for this phenomenon comes in part from the obser- activated by such glutamate spillover (Rusakov and
vation that enhancing the affinity of AMPA receptors for glu- Kullmann, 1998). This roughly corresponds to the typi-
tamate with pharmacological tools can unsilence such a cal distance separating one synapse in the CA1 neuropil
whispering synapse (Choi et al., 2000; Renger et al., 2001). from its nearest neighbor, taking into account the fact
Moreover, manipulation of the presynaptic release probability that synapses are not arranged as a regular lattice
can differentially affect AMPA and NMDA receptor-mediated (Rusakov et al., 1998; Kirov et al., 1999). One report
components of EPSCs (Gasparini et al., 2000). arrived at lower estimates for the glutamate concentra-
The various explanations for silent synapses (absence of tions in both the synaptic cleft and the perisynaptic
functional AMPA receptors, spillover, fusion pore release) are space (Barbour, 2001); however, the discrepancy is
not mutually exclusive. Their relative importance is also unre- mainly explained by assuming a larger extracellular
solved. Indeed, a silent synapse could be deaf (that is, devoid space fraction, which gives rise to a greater degree of
of functional AMPA receptors) early in its development but be dilution of neurotransmitter following release.
mute (presynaptically silent but potentially susceptible to The second approach to estimate receptor occupancy is
spillover of glutamate from a neighbor) at a later stage. to compare the effects of receptor antagonists with dif-
Because these issues have profound implications for the ferent properties on the amplitude of the postsynaptic
interpretation of changes in quantal parameters following response (Clements et al., 1992). Low-affinity antago-
long-term potentiation or depression, they will no doubt con- nists should be more easily displaced from the receptors
tinue to attract considerable attention (Kullmann, 2003) (see by a pulse of glutamate than high-affinity antagonists.
Chapter 10). One critical question that bears on the interpre- By titrating the concentration of the antagonists and
tation of silent synapses is the degree of occupancy of AMPA establishing the kinetic properties of the receptors in
and NMDA receptors following presynaptic glutamate release. patches of membranes exposed to known concentra-
Despite much effort, there is still considerable uncertainty tions of agonists and antagonists, this approach can
over the extent to which AMPA and NMDA receptors are yield information about the glutamate concentration
bound by glutamate released from the presynaptic terminal. It prole in the synapse. This method relies on the
is also unclear whether presynaptic terminals are able to assumption that receptors in somatic membrane
release more than one vesicle of glutamate. Three approaches patches are representative of synaptic receptors.
have been used to estimate the occupancy of receptors follow- Nevertheless, it leads to the conclusion that glutamate
ing a presynaptic action potential. reaches a concentration in excess of 1 mM for 100 s,
subsequently decaying with an exponential time con-
First, by assuming that a single vesicle contains between stant of approximately 1 to 2 ms. These results have
2000 and 5000 molecules of glutamate, and that all of again argued for a relatively high occupancy of NMDA
them are released through a rapidly expanding pore, receptors (> 95% bound). Extending this method to
one can simulate diffusion of the glutamate molecules AMPA receptors has been difficult because of the lack
 Synaptic Function 225

of suitable antagonists; but driving kinetic models of naptic terminals. Equating mEPSCs with vesicles, however,
AMPA receptors with a similar concentration prole relies on two assumptions: rst, that exocytosis proceeds in
leads to the prediction that they should be only approx- the same way in the presence and absence of presynaptic
imately 60% bound at the peak of the response. action potentials; and second, that spontaneous presynaptic
The third approach has given a different picture of Ca2 uctuations do not trigger the simultaneous release of
receptor occupancy. High-resolution optical methods multiple vesicles. Nevertheless, mEPSCs recorded in individ-
have shown Ca2 transients in single dendritic spines, ual pyramidal neurons show a broad, skewed distribution,
which reects NMDA receptor opening in response to ranging from detection threshold ( 5 pA) up to approxi-
presynaptic glutamate release and which can be distin- mately 50 pA (Manabe et al., 1992). Part of this variability
guished from transmission failure (Mainen et al., 1999). may be due to different degrees of cable ltering of EPSCs
When two presynaptic stimuli were given in rapid suc- originating in different parts of the dendritic tree.
cession (with an interval too short to allow appreciable Nevertheless, the median amplitude probably corresponds to
NMDA receptor unbinding), the Ca2 transient was fre- about 100 AMPA receptor-gated channels opening.
quently larger than when only one stimulus was given. It has been suggested that quantal parameters are positively
This implies that the receptors were not saturated by correlated across different synapses (Schikorski and Stevens,
the rst transient. This approach again depends on a 1997). Thus, synapses with large presynaptic elements tend to
number of assumptions, not least that voltage-sensitive have many docked vesicles (as well as more vesicles in total),
Ca2 channels and postsynaptic Ca2 stores do not large active zones and PSDs, and large postsynaptic spines
contribute to the signals and that the visualized (Harris and Sultan, 1995). Thus, if multivesicular release
synapses are representative of the overall population. occurs, it should do so preferentially at such synapses and give
Notwithstanding these caveats, it implies that NMDA rise to large postsynaptic signals. Conversely, small synapses,
receptors are less than 60% bound following a single some of which are possibly devoid of AMPA receptors, may
release event. Again, it is difficult to extend this not contribute appreciably to the excitatory traffic impinging
approach to AMPA receptors because normally they are on the postsynaptic neuron.
not sufficiently permeable to Ca2 for a postsynaptic Even though a pyramidal neuron has 10,000 to 30,000
signal to be detected directly. Nevertheless, the conclu- spine synapses (see Chapters 3 and 5), it has been estimated
sion that NMDA receptor occupancy is less than 60% that only 16 to 26 need to re synchronously to bring it to
does not automatically imply that AMPA receptor occu- action potential threshold (Otmakhov et al., 1993). Clearly,
pancy should be less than this: The 100-fold difference fewer large quanta are required than small ones, and physio-
in affinities of the two receptors applies to steady-state logical patterns of discharges further complicate this equa-
agonist application. With a very brief pulse of transmit- tion. Because the dendrites of CA1 pyramidal neurons can
ter, the occupancy is determined principally by the extend up to 500 m from the soma, an important question is
binding rate, not by the balance of binding and unbind- whether distal synapses are as effective as proximal ones
ing rates (Kullmann, 1999). Because the binding rates of (Andersen et al., 1980b). If the dendrites acted as passive
the two receptor types are similar, it is not impossible cables, and the same synaptic current was injected irrespective
that both are equally and incompletely occupied by glu- of position, distal synapses would give rise to a somatic EPSP
tamate following a single release event. whose peak amplitude was considerably smaller than that
generated by proximal synapses. However, if the total charge
Overall, the average release probability at Schaffer collat- transfer is estimated, the difference between proximal and dis-
eral-CA1 pyramidal neuron synapses is probably less than tal synapses decreases. It has recently been reported that
50% (Raastad, 1995), although it is highly variable from EPSCs generated at different positions on the dendritic tree
synapse to synapse (Hessler et al., 1993; Rosenmund et al., show a location-related scaling that further compensates for
1993). In fact, it is difficult to estimate the mean probability the effect of cable attenuation (Magee and Cook, 2000). This
because there may be a large number of release sites with a nding remains to be conrmed independently, and the bio-
release probability close to 0. Moreover, the release probability physical mechanisms underlying such a scaling have yet to be
at neocortical synapses has been reported to increase with the resolved. Finally, voltage-dependent conductances may mod-
temperature of the preparation (Hardingham and Larkman, ulate the anterograde and retrograde propagation of electrical
1998), so estimates obtained at room temperature may be signals in the dendrites. These issues are considered further in
biased. Recent electrophysiological and optical approaches Chapter 5.
have suggested that more than one vesicle can be simultane- Dentate granule cells are much smaller and have fewer total
ously released at small glutamatergic synapses (Bolshakov et spines than pyramidal neurons, although they have the same
al., 1997; Oertner et al., 2002). Spontaneous miniature spine density per unit length of dendrite (Trommald and
EPSCs (mEPSCs), recorded in the presence of tetrodotoxin to Hulleberg, 1997). The two extrinsic excitatory projections to
block Na-dependent action potentials, have conventionally granule cells show a striking difference in their short-term
been taken to reect the postsynaptic effect of individual plasticity when two presynaptic stimuli are delivered in rapid
quanta, corresponding to single vesicles released from presy- succession. Lateral perforant path synapses exhibit marked
226 The Hippocampus Book

facilitation of the second EPSC compared to the rst EPSC Another difference between the two pathways is that LPP
(paired pulse facilitation). This behavior is essentially the but not MPP synapses exhibit the silent synapse phenomenon
same as at Schaffer collateral synapses on CA1 pyramidal neu- seen at Schaffer collateral synapses (Min et al., 1998a). The
rons. In contrast, medial perforant path synapses show less pharmacological and/or morphological differences that
pronounced paired-pulse facilitation or even depression underlie this difference are unknown.
(McNaughton, 1980). This phenomenon is seen under condi-
tions where inhibitory transmission is blocked, and it appears 6.6.2 Mossy Fiber Synapses
to reect intrinsic differences between the glutamatergic
synapses in the two pathways. Indeed, medial perforant path Mossy bers are the thin, unmyelinated axons of dentate gran-
synapses appear to have a higher initial release probability ule cells. They form several distinct types of synapse (Henze et
than lateral perforant path synapses (Min et al., 1998). al., 2000). Those formed on CA3 pyramidal neurons are in
Medial and lateral perforant path synapses also show sev- many ways unique (Fig. 68). First, they are much larger than
eral differences in sensitivity to neurotransmitters acting on any other synapses in the hippocampus, with a diameter up to
presynaptic receptors. Notably, metabotropic glutamate 10 m. Second, the large presynaptic element (volume
receptor agonists depress transmission in the two pathways approximately 50 m3) contains numerous dense-core vesi-
via distinct receptors (Macek et al., 1996), in approximate cles in addition to the small clear vesicles that predominate in
agreement with the distribution of subtypes resolved with other glutamatergic terminals. These dense-core vesicles are
immunohistochemical methods (Shigemoto et al., 1997). thought to contain dynorphin, Zn2, and other molecules,
Muscarinic receptor agonists depress the medial perforant many of which are released in an activity-dependent manner.
path input relatively selectively (Kahle and Cotman, 1989), Although mossy bers are excitatory, they contain abundant
and norepinephrine has also been reported to have pathway- GABA (Sandler and Smith, 1991; Sloviter et al., 1996), raising
specic actions (Dahl and Sarvey, 1989). It remains to be the possibility that they may also package this neurotransmit-
determined whether these differences have functional conse- ter in some vesicles and release it following action potential
quences. invasion. Indeed, this possibility has been given indirect

Figure 68. Mossy ber structure. The mossy ber

terminal (MF) contains numerous small (approxi-
mately 40 nm diameter) vesicles as well as mito-
chondria and larger vesicles, some of which are
dense-cored. Some vesicles are found close to mem-
brane specializations (arrows) that represent release
sites in contact with a postsynaptic complex spine
(S) belonging to a CA3 pyramidal neuron. DEN,
dendrite. Arrowheads point to puncta adhaerentia,
symmetrical membrane specializations often seen
at synapses. (Source: Amaral and Dent, 1981, with
permission.)
 Synaptic Function 227

experimental support (Gutierrez, 2000; Walker et al., 2001). other afferent bers) remains unclear. The functional role of
Third, the postsynaptic element is a highly branched structure presynaptic modulation of mossy ber transmission is also
(thorny excrescence) akin to a large, complex spine, occur- unclear: It may act as a brake on further transmitter release or
ring in the proximal part of the apical dendrite of the CA3 subserve a form of center-surround inhibition to sharpen the
pyramidal neuron. These structures are also found in rela- projection from the dentate gyrus. Such a role has been pro-
tively small numbers in the proximal basal dendrites, although posed for the release of dynorphin, which acts at  receptors
their incidence increases following intense excitatory traffic (Weisskopf et al., 1993; Simmons and Chavkin, 1996).
such as occurs during a prolonged seizure (see Chapter 15). Although mossy ber synapses on CA3 pyramidal neurons
The postsynaptic excrescence invaginates the presynaptic ter- are large and complex, this projection is sparse: Each mossy
minal. Mossy ber synapses on CA3 pyramidal neurons can ber makes only about 14 synapses in CA3 (see Chapter 3).
contain between 3 and 80 active zones, each of which is Moreover, at low frequencies unitary mossy ber EPSCs
apposed to a PSD (Henze et al., 2000) evoked in individual CA3 pyramidal neurons have a peak
Mirroring their anatomical complexity, mossy berCA3 amplitude corresponding to a synaptic conductance of
synapses show many physiological and pharmacological pecu- approximately 1 nS and consist of approximately 5 quanta
liarities. EPSCs contain a kainate receptor-mediated compo- (Jonas et al., 1993). This conductance is only approximately
nent in addition to the AMPA receptor-mediated component vefold larger than at the far more abundant glutamatergic
(Castillo et al., 1997b; Vignes and Collingridge, 1997), whereas input to CA3 pyramidal neurons, the associational-com-
the NMDA component is relatively small (Jonas et al., 1993). misural bers. However, because mossy ber EPCS can
These pharmacological features are in agreement with the increase up to 10-fold with high-frequency activity (Regehr et
relatively high kainate and low NMDA binding densities al., 1994), the projection sometimes becomes powerful for
observed in the stratum lucidum (Monaghan et al., 1983; depolarizing CA3 neurons. A brief train of action potentials in
Benke et al., 1993). The kainate component is generally a single mossy ber can bring a CA3 pyramidal neuron to r-
thought to be carried by GluR6-containing receptors (Mulle ing threshold in vivo (Henze et al., 2002). The sparse connec-
et al., 1998), although it remains unclear why this compo- tivity of granule cells to CA3 pyramidal neurons has impeded
nent of the EPSC is slow compared to the kinetics of recom- detailed biophysical study of this pathway.
binant GluR6-containing receptors. High-frequency stimuli A second type of mossy fiber synapse is formed on gluta-
also evoke a postsynaptic metabotropic glutamate receptor- matergic mossy cells of the dentate hilus. The presynaptic
mediated depolarizing current (Heuss et al., 1999; Yeckel et al., varicosities are also large and contain pleomorphic vesicles.
1999). Numerically far more important are small synapses on other
Mossy ber EPSCs exhibit pronounced facilitation with hilar interneurons (Acsady et al., 1998). They have a very dif-
very modest increases in discharge frequency, and part of this ferent shape in that the presynaptic element is a lopodial
facilitation is mediated by presynaptic kainate autoreceptors extension off the shaft of the mossy ber itself, which makes a
(Schmitz et al., 2001c). Following a high-frequency burst of single contact with the target interneuron. These synapses are
action potentials in the presynaptic axons, transmission at much smaller than the synapses on CA3 pyramidal neurons
mossy bers can remain elevated for hours (Henze et al., and probably contain only a single release site. Several physi-
2000). This form of long-term potentiation is distinct from ological studies of mossy berinterneuron transmission have
that seen at other spine synapses in the hippocampus in that not systematically identified the postsynaptic neurons
it is not dependent on NMDA receptors. Its underlying mech- (Maccaferri et al., 1998; Toth et al., 2000; Lei and McBain,
anisms are considered further in Chapter 10. 2002). However, because small lopodial synapses far out-
Transmission at mossy ber synapses can be profoundly number large synapses on mossy cells, they probably
depressed by activation of several classes of presynaptic recep- principally reect the properties of the former. Although
tors, including group II (and in some species group III) they are highly susceptible to presynaptic modulation via
metabotropic glutamate receptors (Yamamoto et al., 1983; metabotropic glutamate receptors, many mossy ber synapses
Lanthorn et al., 1984; Kamiya et al., 1996), GABAB receptors on interneurons appear not to exhibit marked frequency-
(Min et al., 1998b; Vogt and Nicoll, 1999) and opioid receptors dependent facilitation or NMDA receptor-independent LTP
(Weisskopf et al., 1993). Presynaptic kainate receptors have a (Maccaferri et al., 1998). However, this conclusion may not
paradoxical function: With low concentrations of agonists, apply universally to the interneuronal targets of mossy bers:
probably corresponding to that achieved with synaptic gluta- Short- and long-term potentiation have recently been
mate release, they facilitate transmission (Schmitz et al., reported in identied basket cells in response to high-fre-
2001c), whereas with higher concentrations transmission is quency trains of stimuli delivered to single presynaptic gran-
depressed (Chittajallu et al., 1996; Vignes et al., 1998; Kamiya ule cells (Alle et al., 2001) (see Chapter 10).
and Ozawa, 2000; Kullmann, 2001; Schmitz et al., 2001c). The Another peculiarity of mossy ber transmission to some
degree to which presynaptic receptors on mossy bers are interneurons is that Ca2-permeable AMPA receptors have
activated by neurotransmitters released from their own termi- been reported to mediate transmission (Toth et al., 2000).
nals, in contrast to neighboring axons (whether mossy ber or This property is shared with other excitatory synapses on
228 The Hippocampus Book

interneurons and probably depends on whether the interneu-

ron expresses GluR2 receptors (see below). However, there Box 63
Methods Used to Study Transmission
may be a further level of renement because it has been pro-
at Excitatory Hippocampal Synapses
posed that a given interneuron targets Ca2-permeable and
Ca2-impermeable receptors to different synapses (Toth and The hippocampus has yielded many of the most important
McBain, 1998). insights into synaptic transmission in the CNS. This is
The fourth type of synapse formed by mossy bers is on because of two principal factors. First, the hippocampal for-
mation has a remarkable laminar organization: Many of the
the proximal dendrites of other granule cells. Although this
bers making excitatory synapses run parallel to one another,
projection is normally sparse, it increases following intense and target neurons and their projections are tightly packed
seizure activity, in common with the infrapyramidal projec- (see Chapter 3). Second, development of the slice prepara-
tion to CA3 (mossy ber sprouting) (see Chapter 16). tion allowed neurons and their connections to be studied in
vitro, where the chemical and physical environment of the
6.6.3 Other Glutamatergic Synapses synapses could be manipulated experimentally. An important
consequence of the laminar structure of the hippocampus is
on Interneurons
that the currents evoked by synchronous activity in many
synapses can be detected via extracellular electrodes. Because
Excitatory synapses on interneurons made by pyramidal neu- synapses represent a current sink (current moves from the
rons are generally small and are made directly on the soma extracellular medium into neurons), eld EPSPs (fEPSPs)
and dendrites (Box 63). From limited ultrastructural and are detected as negative voltage deections in the dendritic
electrophysiological data obtained in parallel, they contain a regions where most excitatory synapses occur (the stratum
single release site (Gulyas et al., 1993). Two important phar- radiatum or stratum oriens of the hippocampus proper or
macological features distinguish them from small spine the stratum moleculare of the dentate gyrus). In the cell
body layer, there is a corresponding positive wave (represent-
synapses on principal neurons. First, at a subset of synapses
ing current owing out of the principal neurons) that com-
the AMPA receptors are permeable to Ca2 (Jonas et al., pletes the local circuit. With stimuli of sufficient amplitude,
1994), which can be explained by the absence of the GluR2 however, a sharp negative deection is superimposed on the
subunit that confers Ca2 impermeability at other receptors. positive fEPSP in the cell body layer. This population spike
Interestingly, many of the same interneurons express neuronal reects the synchronous discharge of principal cells that are
nitric oxide synthase (NOS) (Catania et al., 1995). It is tempt- brought to ring threshold by the EPSP propagating to the
ing to speculate that Ca2 inux via AMPA synaptic receptors cell body (Andersen et al., 1971).
may trigger the release of NO from these neurons, which Intracellular records from hippocampal neurons in vivo
have lent direct support for the identication of population
could act as a diffuse messenger, affecting several target neu-
spikes with somatic action potentials (Kandel and Spencer,
rons within a small distance. This property is, however, not
1961). With the development of the hippocampal slice,
ubiquitous among interneurons; and many appear neither Schwartzkroin (1975) showed that the dendritic negative
to express NOS nor to have Ca2-permeable AMPA receptors. fEPSP was indeed due to activation of synaptic potentials
Another highly variable property of glutamatergic synapses generated at the dendritic level, recorded as an intracellular
at interneurons is that some have a kainate receptor-mediated depolarizing EPSP in the soma. More recently, patch-clamp
component (Cossart et al., 1998; Frerking et al., 1998), recordings have been widely used, either from somata or
although it does not occur ubiquitously (Semyanov and from dendrites close to the excitatory synapses. They have
Kullmann, 2001). In contrast to mossy berCA3 synapses, shown even more directly that EPSCs result from the local
interneurons are enriched in GluR5. It is not known whether activation of afferent bers (Hestrin et al., 1990).
the presence or absence of a kainate receptor-mediated
component to EPSCs in interneurons correlates with the post-
synaptic interneurons morphological or immunocytochemi- Another difference between the synapses is the expression
cal identity (see Chapter 7). of presynaptic group III metabotropic glutamate receptors. As
The kinetics of AMPA receptor-mediated EPSCs vary already mentioned, these receptors are present at mossy ber
extensively among the various excitatory synapses in the hip- synapses in some species (for instance in guinea pigs) but not
pocampus. In general, the rise times and decay time constants at Schaffer collateral synapses in CA1 pyramidal neurons.
are slow in pyramidal neurons, intermediate in granule cells, However, excitatory transmission to interneurons is highly
and fast in interneurons (Geiger et al., 1997). Although this sensitive to agonists of group III metabotropic receptors
matches the passive membrane time constants of the various (Scanziani et al., 1998). Small synapses on pyramidal neurons,
neurons, the difference persists once the cable properties of in contrast, are insensitive. This differential sensitivity to
the neurons have been taken into account. That is, the differ- group III agonists parallels the distribution of some subtypes
ent kinetics appear to be an intrinsic property of the neurons of group III receptors imaged with high-resolution immuno-
and/or synapses (Colquhoun et al., 1992) and, indeed, appear histochemical methods. For instance, mGluR7 appears to be
to reect the subunit composition and splice variants of expressed in presynaptic active zones apposed to interneurons
AMPA receptors expressed in each neuron (Lambolez et al., and not at active zones apposed to pyramidal neurons
1996). (Shigemoto et al., 1996, 1997). This target cell dependence in
 Synaptic Function 229

the distribution of presynaptic receptors implies that a retro-

grade signal instructs the presynaptic membrane whether to Box 64
Early Insights into Inhibitory Synaptic
anchor these receptors.
Signaling in the Hippocampus
Many glutamatergic synapses on interneurons show simi-
lar patterns of short-term use-dependent plasticity as Schaffer Early intracellular recordings from principal neurons in the
collateral synapses on CA1 pyramidal neurons. That is, they hippocampus showed that following a strong afferent stimu-
show frequency-dependent facilitation with brief trains of lus the initial discharge was often followed by a pause in
spontaneous activity (Albe-Fessard and Buser, 1955). fEPSP
action potentials. However, a subset of basket and bistratied
recordings also showed that intense activation was followed
cells receive glutamatergic synapses that depress prominently by a temporarily reduced ability to excite CA1 pyramidal
(Thomson, 2000). The molecular basis of this distinction has cells and dentate granule cells synaptically. These observa-
not been elucidated in the hippocampus but may again imply tions led to the hypothesis that inhibitory processes are
a retrograde signal from the target neuron that species some entrained by the stimulus. The synaptic basis for this
of the properties of the presynaptic terminal. Such a principle inhibition was not immediately established because intracel-
is supported by the nding that different synapses supplied by lular recordings revealed long-lasting depolarizing waves.
However, it was subsequently discovered that they were IPSPs
the same neocortical neurons can show either depression or
whose polarity was articially reversed by the leakage of Cl

facilitation with repeated stimulation, depending on the iden-

ions into the neurons from the recording microelectrodes
tity of the postsynaptic neurons (Reyes et al., 1998).
(Spencer and Kandel, 1961). This inhibition was shown to be
Finally, glutamatergic synapses to interneurons frequently generated across the soma membrane (Andersen et al., 1964),
appear not to express NMDA receptor-dependent LTP, as would be expected if it was mediated by the perisomatic
although a form of LTD has been described (McMahon and terminals of basket cells. Extracellular stimuli not only
Kauer, 1997). One report implicates a coincidence of Ca2 recruited basket cells via axon collaterals of principal neurons
inux via Ca2-permeable AMPA receptors and activation of (disynaptic feedback inhibition) but could also directly
presynaptic metabotropic receptors in the induction of this activate inhibitory interneurons, yielding a monosynaptic
IPSP (feedforward inhibition). Both types of inhibition
phenomenon (Laezza et al., 1999). This depression of inhibi-
play an important role in regulating the excitability of the
tion potentially leads to a disynaptic disinhibition of principal circuitry and in synchronizing the activity of populations of
cells (Lu et al., 2000). neurons (see Chapter 8). A potential role for feedforward
inhibition in sharpening the temporal resolution of pattern
6.6.4 Inhibitory Synapses discrimination has recently been proposed (Pouille and
Scanziani, 2001).
GABAergic inhibition of principal neurons plays an essential
role in regulating the transmission of information through the
hippocampal formation (see Box 64). The wide range of postsynaptic neurons, these synapses appear to yield large-
interneuron morphologies is summarized in Chapter 3, and amplitude IPSCs with relatively fast kinetics and activity-
their connectivities are considered in Chapter 8. Not all dependent depression (Pawelzik et al., 1999; Maccaferri et al.,
interneurons are GABAergic: Mossy cells in the dentate hilus 2000).
and giant spiny interneurons in the stratum radiatum are glu- Axo-somatic synapses are made by basket cells, whose axon
tamatergic and more akin to pyramidal neurons. Because rel- collaterals form clusters of boutons surrounding the soma of
atively little is known about synapses made by glutamatergic the target neurons (see Chapter 3). Basket cells mediate pow-
interneurons, this section restricts its attention to GABAergic erful inhibition of principal neurons by synchronous release
signaling (Fig. 69). Complementing the variability of of GABA from the multiple synapses. These synapses also tend
interneuron morphology, GABAergic synapses fall into vari- to depress with repeated activation (Hefft et al., 2002).
ous types depending on their location on the target neuron. Interestingly, two subtypes of basket cells appear to signal
Three main types are recognized: axo-axonic, axo-somatic, via different postsynaptic GABAA receptors: Cholecystokinin-
and axo-dendritic. This distinction appears to be valid positive cells make synapses with a high 2 content, and
whether the postsynaptic target is a pyramidal neuron or parvalbumin-containing basket cells signal preferentially via
another interneuron. synapses containing 1 subunits (Klausberger et al., 2002).
Axo-axonic synapses are made by chandelier cells and How this complementary distribution of receptors ts with
occur 50 to 70 m from the soma of the target neuron. In the physiological roles of these two subsets of interneurons is
pyramidal neurons, postsynaptic 2 GABAA subunits are unknown.
restricted principally to these synapses (Nusser et al., 1996). Axo-dendritic synapses on pyramidal cells are made by a
They are also distinguished from other GABAergic synapses variety of interneurons whose cell bodies are found in the
by the presence of a postsynaptic Ca2-sequestering organelle strata oriens, radiatum, and lacunosum-moleculare. Synapses
(Benedeczky et al., 1994). Situated at the strategic point for formed on the distal dendrites are thought to control the
impulse initiation, these synapses probably exert extremely excitability of the postsynaptic neuron relatively indirectly.
powerful control over the discharge of target cells. From the They are unlikely to hyperpolarize the cell body effectively
small number of published paired recordings of pre- and because they are electrically remote, but they may be highly
230 The Hippocampus Book

Figure 69. Inhibitory synapse structure. The

arrows point to two symmetrical synapses on
neighboring CA1 pyramidal cell bodies. Bar  0.5
m. (Source: Harris and Landis, 1986, with
permission).

effective in shunting locally evoked EPSCs (Staley and Mody, free Ca2 with repeated action potential invasion (Caillard et
1992). al., 2000). Another phenomenon that may underlie differences
Possibly reecting the different sites at which GABAergic in use-dependent plasticity is that some GABAergic synapses
synapses contact postsynaptic neurons, GABAergic IPSCs rely exclusively on either N- or P/Q-type Ca2 channels to
in pyramidal neurons have bimodal kinetics. So-called trigger transmitter release (Poncer et al., 1997; Wilson et al.,
GABAA-Fast IPSCs have a faster rise-time ( 2 ms) and a cor- 2001) in contrast to glutamatergic synapses, which tend to use
respondingly faster decay time constant ( 20 ms) than a mixture of both types. Because N-type channels tend to
GABAA-Slow IPSCs (Pearce, 1993). This distinction is greater inactivate more rapidly and are a more sensitive to G protein-
than can be explained entirely by cable ltering by the den- mediated inhibition (Zhang et al., 1996), it is tempting to
dritic tree and suggests, instead, that different GABAA recep- speculate that synapses relying on this Ca2 channel subtype
tor subunits with differing kinetics are segregated to different depress more readily.
synapses. Indeed, GABAA-Fast IPSCs are sensitive to furosemide Although GABAergic IPSCs evoked by a single basket cell
(Pearce, 1993), which is consistent with presence of the 4 are generally multiquantal (Edwards et al., 1990), this proba-
subunit (Wafford et al., 1996). Circumstantial evidence sug- bly reects the numerous synapses made by a single neuron
gests that they originate from basket cells and may be prima- (Kraushaar and Jonas, 2000). The number of vesicles released
rily responsible for generating relatively high-frequency at a single synapse is not known with certainty. Evidence exists
(gamma range, approximately 40 Hz) oscillations in the hip- for multivesicular release at some cerebellar GABAergic
pocampus (Banks et al., 2000). In contrast, GABAA-Slow IPSCs synapses (Auger et al., 1998), but whether this also holds for
are insensitive to furosemide, are more likely to arise from hippocampal synapses is not known. Because the occupancy
dendritic synapses equipped with 1-containing receptors, of postsynaptic GABAA receptors following release of a single
and may play a more important role in generating theta- vesicle may be greater than 50% (Perrais and Ropert, 1999), it
rhythm (approximately 8 Hz) oscillations. cannot be inferred that release of two vesicles gives rise to a
Although GABA is the principal transmitter released at all response twice as large.
of the synapses considered above, some are also thought to GABAergic synapses in the hippocampus show a different
release peptides from dense-core vesicles. Basket cells contain prole of modulation by presynaptic receptors than most glu-
either cholecystokinin (CCK) or parvalbumin but not both. tamatergic synapses. Thus,  opioid receptors depress GABA
Although little is known about differences in the physiology of release relatively selectively (Simmons and Chavkin, 1996).
the synapses supplied by these two cell types, only cholecys- Cannabinoid receptors have a similar effect at some basket cell
tokinin-positive neurons appear to express CB1 cannabinoid synapses (Katona et al., 1999; Hoffman and Lupica, 2000) and
receptors (Marsicano and Lutz, 1999; Tsou et al., 1999). The mediate an unusual form of short-term depression triggered
evidence for a synaptic role for cholecystokinin and other by Ca2 transients in postsynaptic pyramidal neurons (Ohno-
neuro-active peptides such as somatostatin, CCK, and sub- Shosaku et al., 2001; Wilson and Nicoll, 2001). Because the
stance P is scanty. CB1 cannabinoid receptors mediating this phenomenon are
In contrast to most glutamatergic synapses, most apparently restricted to CCK-positive basket cells, this form of
GABAergic synapses show marked depression with repetitive retrograde signaling could exert a powerful disinhibitory
activity at moderate frequencies (Thomson, 2000). This may effect. Moreover, it could relatively selectively interfere with
reect a high baseline release probability, but the presence of rhythmogenesis, possibly contributing to the cognitive and
Ca2-buffering proteins also attenuates the accumulation of hallucinatory effects of exogenous cannabinoid agonists.
 Synaptic Function 231

In contrast to glutamatergic synapses, GABA release is rel- tering of glutamate receptors at synapses, have attracted
atively resistant to adenosine (Yoon and Rothman, 1991; intense interest because of their potential role in the develop-
Lambert and Wilson, 1993). Surprisingly, GABAergic synapses ment and plasticity of excitatory transmission. They are taken
on interneurons are as sensitive to group III metabotropic up in greater detail in Chapter 7. Other areas, such as the
glutamate agonists as glutamatergic synapses (Semyanov and detailed postsynaptic signaling cascades, leading to effects as
Kullmann, 2000) even though only the latter actually release remote as nuclear transcription, are discussed, in the context
the endogenous agonist. GABAergic transmission to pyrami- of LTP, in Chapter 10.
dal neurons, however, is relatively insensitive to group III ago- As in so many areas of neuroscience, insights at a molecu-
nists, mirroring the lack of sensitivity of Schaffer collateral lar level are outpacing progress at a more macroscopic level. It
synapses on pyramidal neurons. Thus, the expression of is some of these areas that ultimately may require major break-
presynaptic group III receptors appears to be determined in throughs to relate the mechanisms of synaptic function to the
large part by the target cell identity. The presence of these function of the network as a whole. For instance, although the
receptors on GABA-releasing synapses is difficult to explain involvement of cholinergic mechanisms in theta rhythms is on
unless they have evolved to detect glutamate released from a relatively secure footing, this understanding remains at a
neighboring synapses. Evidence that glutamate spillover mod- qualitative level. Many critical parameters remain far from
ulates inhibition via these receptors has been reported established: What is the temporal prole of acetylcholine
(Semyanov and Kullmann, 2000). release? Where are the target receptors in relation to the sites
There is relatively little evidence for long-term use- of release? How many individual terminals cooperate to acti-
dependent plasticity at GABAergic synapses akin to long-term vate muscarinic and nicotinic receptors? Which of the numer-
potentiation (see Chapter 10). However, some manipulations ous actions of these receptors actually take place in vivo?
in neuronal cultures can enhance the surface expression of Similar questions occur for each of the other neurotrans-
GABAA receptors (Wan et al., 1997), underlining the impor- mitters described here. Most of this chapter has concentrated
tance of this phenomenon for the development of synapses, if on the fast ionotropic signaling by glutamate and GABA; this
not their regulation in adults. is because, to a great extent, electrophysiological methods are
GABAA receptors in cerebellar granule cells can mediate a so powerful in detecting and quantifying EPSP/Cs and
continuous current, a phenomenon known as tonic inhibi- IPSP/Cs. More diffuse types of signaling mediated by
tion (to distinguish it from phasic inhibition represented by metabotropic receptors or even by extrasynaptic actions at
the occurrence of IPSCs) (Brickley et al., 1996; Wall and ionotropic receptors may be of similar or even greater impor-
Usowicz, 1997). Tonic GABAA receptor-mediated inhibition tance for understanding the roles of these transmitters in
also occurs in hippocampal granule cells (Overstreet and intercellular signaling. As for the monoamines and peptides,
Westbrook, 2001; Stell and Mody, 2002). In the hippocampus they are considered less important only on the grounds that
proper, interneurons appear to express tonic inhibition rela- the releasing axons are relatively sparse. However, not until
tively selectively, possibly reecting a homeostatic role in reg- their actions have been put on a sound quantitative footing
ulating circuit excitability (Semyanov et al., 2003). similar to that of the amino acid neurotransmitters will it be
Inhibitory synapses are also made among interneurons. possible to know if this ranking is correct.
These synapses appear to show use-dependent depression
similar to that seen at most GABAergic synapses on principal
cells, although the kinetics of the IPSCs are generally faster ACKNOWLEDGMENTS
(Bartos et al., 2001). Some parvalbumin-positive basket cells
My thanks to the editors, in particular Per Andersen, for helpful input
are also coupled by gap junctions (Fukuda and Kosaka, 2000).
to this chapter.
Genetic deletion of connexion 36 perturbs  frequency oscil-
lations, possibly because of impaired electrical coupling
among these interneurons (Hormuzdi et al., 2001). REFERENCES

Acsady L, Kamondi A, Sik A, Freund T, Buzsaki G (1998) GABAergic

cells are the major postsynaptic targets of mossy bers in the rat
6.7 Unresolved Issues hippocampus. J Neurosci 18:33863403.
Agnati LF, Zoli M, Stromberg I, Fuxe K (1995) Intercellular commu-
There are numerous unresolved issues surrounding synaptic nication in the brain: wiring versus volume transmission.
function that cannot be listed completely here. Many are fun- Neuroscience 69:711726.
Albe-Fessard D, Buser P (1955) Intracellular activities collected in the
damental questions concerning transmitter release, such as
sigmoid cortex of the cat; participation of the pyramidal neu-
the mechanisms by which synaptotagmin transduces the Ca2
rons in the somesthetic evoked potential. J Physiol (Paris)
surge to trigger exocytosis, how the degree of lling of a vesi- 47:6769.
cle affects its readiness for release, and whether and under Albuquerque EX, Alkondon M, Pereira EF, Castro NG, Schrattenholz
what conditions exocytosis can take place via a fusion pore. A, Barbosa CT, Bonfante-Cabarcas R, Aracava Y, Eisenberg
Other important questions concern postsynaptic mecha- HM, Maelicke A (1997) Properties of neuronal nicotinic
nisms. Some of these questions, such as the targeting and clus- acetylcholine receptors: pharmacological characterization and
232 The Hippocampus Book

modulation of synaptic function. J Pharmacol Exp Ther P (1993) The metabotropic glutamate receptor (mGluR1 alpha)
280:11171136. is concentrated at perisynaptic membrane of neuronal subpop-
Alle H, Jonas P, Geiger JR (2001) PTP and LTP at a hippocampal ulations as detected by immunogold reaction. Neuron
mossy berinterneuron synapse. Proc Natl Acad Sci USA 11:771787.
98:1470814713. Behrends JC, ten Bruggencate G (1993) Cholinergic modulation of
Amaral DG, Dent JA (1981) Development of the mossy bers of synaptic inhibition in the guinea pig hippocampus in vitro:
the dentate gyrus. I. A light and electron microscopic study excitation of GABAergic interneurons and inhibition of GABA-
of the mossy bers and their expansions. J Comp Neurol release. J Neurophysiol 69:626629.
195:5186. Bekkers JM (1993) Enhancement by histamine of NMDA-mediated
Andersen P, Eccles JC, Loyning Y (1964) Pathway of postsynaptic synaptic transmission in the hippocampus. Science
inhibition in the hippocampus. J Neurophysiol 27:608619. 261:104106.
Andersen P, Blackstad TW, Lomo T (1966) Location and identica- Bekkers JM, Stevens CF (1989) NMDA and non-NMDA receptors are
tion of excitatory synapses on hippocampal pyramidal cells. Exp co-localized at individual excitatory synapses in cultured rat
Brain Res 1:236248. hippocampus. Nature 341:230233.
Andersen P, Bliss TV, Skrede KK (1971) Unit analysis of hippocampal Bellocchio EE, Reimer RJ, Fremeau RT Jr, Edwards RH (2000) Uptake
polulation spikes. Exp Brain Res 13:208221. of glutamate into synaptic vesicles by an inorganic phosphate
Andersen P, Dingledine R, Gjerstad L, Langmoen IA, Laursen AM transporter. Science 289:957960.
(1980a) Two different responses of hippocampal pyramidal cells Ben-Ari Y, Cossart R (2000) Kainate, a double agent that generates
to application of gamma-amino butyric acid. J Physiol seizures: two decades of progress. Trends Neurosci 23:580587.
305:279296. Ben-Ari Y, Cherubini E, Corradetti R, Gaiarsa JL (1989) Giant synap-
Andersen P, Silfvenius H, Sundberg SH, Sveen O (1980b) A compar- tic potentials in immature rat CA3 hippocampal neurones.
ison of distal and proximal dendritic synapses on CA1 pyramids J Physiol 416:303325.
in guinea-pig hippocampal slices in vitro. J Physiol 307:273299. Ben-Ari Y, Khazipov R, Leinekugel X, Caillard O, Gaiarsa JL (1997)
Andrade R, Malenka RC, Nicoll RA (1986) A G protein couples sero- GABAA, NMDA and AMPA receptors: a developmentally regu-
tonin and GABAB receptors to the same channels in hippocam- lated menage a trois. Trends Neurosci 20:523529.
pus. Science 234:12611265. Benedeczky I, Molnar E, Somogyi P (1994) The cisternal organelle as
Anwyl R (1991) Modulation of vertebrate neuronal calcium channels a Ca(2)-storing compartment associated with GABAergic
by transmitters. Brain Res Rev 16:265281. synapses in the axon initial segment of hippocampal pyramidal
Arnth-Jensen N, Jabaudon D, Scanziani M (2002) Cooperation neurones. Exp Brain Res 101:216230.
between independent hippocampal synapses is controlled by Benke TA, Jones OT, Collingridge GL, Angelides KJ (1993) N-Methyl-
glutamate uptake. Nat Neurosci 5:325331. D-aspartate receptors are clustered and immobilized on den-
Ascher P, Nowak L (1988) The role of divalent cations in the N- drites of living cortical neurons. Proc Natl Acad Sci USA
methyl-D-aspartate responses of mouse central neurones in 90:78197823.
culture. J Physiol 399:247266. Bergles DE, Roberts JD, Somogyi P, Jahr CE (2000) Glutamatergic
Asztely F, Erdemli G, Kullmann DM (1997) Extrasynaptic glutamate synapses on oligodendrocyte precursor cells in the hippocam-
spillover in the hippocampus: dependence on temperature and pus. Nature 405:187191.
the role of active glutamate uptake. Neuron 18:281293. Berretta N, Berton F, Bianchi R, Capogna M, Francesconi W, Brunelli
Auger C, Kondo S, Marty A (1998) Multivesicular release at single M (1990) Effects of dopamine, D-1 and D-2 dopaminergic ago-
functional synaptic sites in cerebellar stellate and basket cells. nists on the excitability of hippocampal CA1 pyramidal cells in
J Neurosci 18:45324547. guinea pig. Exp Brain Res 83:124130.
Augustine GJ (2001) How does calcium trigger neurotransmitter Berzhanskaya J, Urban NN, Barrionuevo G (1998) Electrophysiolog-
release? Curr Opin Neurobiol 11:320326. ical and pharmacological characterization of the direct per-
Banke TG, Bowie D, Lee H, Huganir RL, Schousboe A, Traynelis forant path input to hippocampal area CA3. J Neurophysiol
SF (2000) Control of GluR1 AMPA receptor function by cAMP- 79:21112118.
dependent protein kinase. J Neurosci 20:89102. Bi G, Poo M (2001) Synaptic modication by correlated activity:
Banks MI, White JA, Pearce RA (2000) Interactions between distinct Hebbs postulate revisited. Annu Rev Neurosci 24:139166.
GABA(A) circuits in hippocampus. Neuron 25:449457. Biscoe TJ, Straughan DW (1966) Micro-electrophoretic studies of
Baranano DE, Ferris CD, Snyder SH (2001) Atypical neural messen- neurones in the cat hippocampus. J Physiol 183:341359.
gers. Trends Neurosci 24:99106. Blackmer T, Larsen EC, Takahashi M, Martin TF, Alford S, Hamm HE
Barbour B (2001) An evaluation of synapse independence. J Neurosci (2001) G protein betagamma subunit-mediated presynaptic
21:79697984. inhibition: regulation of eocytotic fusion downstream of Ca2
Barnard EA, Skolnick P, Olsen RW, Mohler H, Sieghart W, Biggio G, entry. Science 292:293297.
Braestrup C, Bateson AN, Langer SZ (1998) International Bleakman D (1999) Kainate receptor pharmacology and physiology.
Union of Pharmacology. XV. Subtypes of gamma-aminobutyric Cell Mol Life Sci 56:558566.
acid A receptors: classication on the basis of subunit structure Blum R, Katz KW, Konnerth A (2002) Neurotrophin-evoked depo-
and receptor function. Pharmacol Rev 50:291313. larization requires the sodium channel Na(V)1.9. Nature
Bartos M, Vida I, Frotscher M, Geiger JR, Jonas P (2001) Rapid sig- 419:687693.
naling at inhibitory synapses in a dentate gyrus interneuron net- Bollmann JH, Sakmann B, Borst JG (2000) Calcium sensitivity of
work. J Neurosci 21:26872698. glutamate release in a calyx-type terminal. Science 289:953957.
Baude A, Nusser Z, Roberts JD, Mulvihill E, McIlhinney RA, Somogyi Bolshakov VY, Golan H, Kandel ER, Siegelbaum SA (1997)
 Synaptic Function 233

Recruitment of new sites of synaptic transmission during the AMPA receptor subunits in NOS-positive neurons of cortex,
cAMP- dependent late phase of LTP at CA3-CA1 synapses in striatum, and hippocampus. J Neurosci 15:70467061.
the hippocampus. Neuron 19:635651. Chang Y, Wang R, Barot S, Weiss DS (1996) Stoichiometry of a
Bolton MM, Blanpied TA, Ehlers MD (2000) Localization and stabi- recombinant GABAA receptor. J Neurosci 16:54155424.
lization of ionotropic glutamate receptors at synapses. Cell Mol Chaudhry FA, Reimer RJ, Bellocchio EE, Danbolt NC, Osen KK,
Life Sci 57:15171525. Edwards RH, Storm-Mathisen J (1998) The vesicular GABA
Bormann J (2000) The ABC of GABA receptors. Trends Pharmacol transporter, VGAT, localizes to synaptic vesicles in sets of
Sci 21:1619. glycinergic as well as GABAergic neurons. J Neurosci 18:
Borst JG, Sakmann B (1996) Calcium inux and transmitter release 97339750.
in a fast CNS synapse. Nature 383:431434. Chen YA, Scheller RH (2001) SNARE-mediated membrane fusion.
Bortolotto ZA, Clarke VR, Delany CM, Parry MC, Smolders I, Vignes Nat Rev Mol Cell Biol 2:98106.
M, Ho KH, Miu P, Brinton BT, Fantaske R, Ogden A, Gates M, Chittajallu R, Vignes M, Dev KK, Barnes JM, Collingridge GL, Henley
Ornstein PL, Lodge D, Bleakman D, Collingridge GL (1999) JM (1996) Regulation of glutamate release by presynaptic
Kainate receptors are involved in synaptic plasticity. Nature kainate receptors in the hippocampus. Nature 379:7881.
402:297301. Chittajallu R, Braithwaite SP, Clarke VR, Henley JM (1999) Kainate
Bowie D, Mayer ML (1995) Inward rectication of both AMPA and receptors: subunits, synaptic localization and function. Trends
kainate subtype glutamate receptors generated by polyamine- Pharmacol Sci 20:2635.
mediated ion channel block. Neuron 15:453462. Choi S, Klingauf J, Tsien RW (2000) Postfusional regulation of cleft
Bracci E, Vreugdenhil M, Hack SP, Jefferys JG (2001) Dynamic mod- glutamate concentration during LTP at silent synapses. Nat
ulation of excitation and inhibition during stimulation at Neurosci 3:330336.
gamma and beta frequencies in the CA1 hippocampal region. Chow RH, von Ruden L, Neher E (1992) Delay in vesicle fusion
J Neurophysiol 85:24122422. revealed by electrochemical monitoring of single secretory
Brakeman PR, Lanahan AA, OBrien R, Roche K, Barnes CA, Huganir events in adrenal chromaffin cells. Nature 356:6063.
RL, Worley PF (1997) Homer: a protein that selectively binds Clements JD, Lester RA, Tong G, Jahr CE, Westbrook GL (1992) The
metabotropic glutamate receptors. Nature 386:284288. time course of glutamate in the synaptic cleft. Science
Breckenridge LJ, Almers W (1987) Final steps in exocytosis observed 258:14981501.
in a cell with giant secretory granules. Proc Natl Acad Sci USA Colquhoun D, Jonas P, Sakmann B (1992) Action of brief pulses
84:19451949. of glutamate on AMPA/kainate receptors in patches from differ-
Brenowitz S, Trussell LO (2001) Minimizing synaptic depression by ent neurones of rat hippocampal slices. J Physiol (Lond)
control of release probability. J Neurosci 21:18571867. 458:261287.
Brickley SG, Cull-Candy SG, Farrant M (1996) Development of a Contractor A, Swanson GT, Sailer A, OGorman S, Heinemann SF
tonic form of synaptic inhibition in rat cerebellar granule (2000) Identication of the kainate receptor subunits underly-
cells resulting from persistent activation of GABAA receptors. ing modulation of excitatory synaptic transmission in the CA3
J Physiol (Lond) 497:753759. region of the hippocampus. J Neurosci 20:82698278.
Brown DA, Adams PR (1980) Muscarinic suppression of a novel Contractor A, Swanson G, Heinemann SF (2001) Kainate receptors
voltage-sensitive K current in a vertebrate neurone. Nature are involved in short- and long-term plasticity at mossy ber
283:673676. synapses in the hippocampus. Neuron 29:209216.
Bruns D, Jahn R (1995) Real-time measurement of transmitter Cossart R, Esclapez M, Hirsch JC, Bernard C, Ben-Ari Y (1998)
release from single synaptic vesicles. Nature 377:6265. GluR5 kainate receptor activation in interneurons increases
Caillard O, Moreno H, Schwaller B, Llano I, Celio MR, Marty A tonic inhibition of pyramidal cells. Nat Neurosci 1:470478.
(2000) Role of the calcium-binding protein parvalbumin in Cossart R, Tyzio R, Dinocourt C, Esclapez M, Hirsch JC, Ben-Ari Y,
short-term synaptic plasticity. Proc Natl Acad Sci USA Bernard C (2001) Presynaptic kainate receptors that enhance
97:1337213377. the release of GABA on CA1 hippocampal interneurons. Neuron
Calabresi P, Pisani A, Centonze D, Bernardi G (1997) Synaptic plas- 29:497508.
ticity and physiological interactions between dopamine and Cossart R, Epsztein J, Tyzio R, Becq H, Hirsch J, Ben-Ari Y, Crepel V
glutamate in the striatum. Neurosci Biobehav Rev 21:519523. (2002) Quantal release of glutamate generates pure kainate and
Cammack JN, Rakhilin SV, Schwartz EA (1994) A GABA transporter mixed AMPA/kainate EPSCs in hippocampal neurons. Neuron
operates asymmetrically and with variable stoichiometry. 35:147159.
Neuron 13:949960. Cowan WM, Sdhof TC, Stevens CF (eds) (2003) Synapses.
Capogna M, Gahwiler BH, Thompson SM (1996) Presynaptic inhibi- Baltimore: Johns Hopkins University Press.
tion of calcium-dependent and -independent release elicited Cremona O, De Camilli P (1997) Synaptic vesicle endocytosis. Curr
with ionomycin, gadolinium, and alpha-latrotoxin in the hip- Opin Neurobiol 7:323330.
pocampus. J Neurophysiol 75:20172028. Dahl D, Sarvey JM (1989) Norepinephrine induces pathway-specic
Castillo PE, Janz R, Sudhof TC, Tzounopoulos T, Malenka RC, Nicoll long-lasting potentiation and depression in the hippocampal
RA (1997a) Rab3A is essential for mossy bre long-term poten- dentate gyrus. Proc Natl Acad Sci USA 86:47764780.
tiation in the hippocampus. Nature 388:590593. Danbolt NC (2001) Glutamate uptake. Prog Neurobiol 65:1105.
Castillo PE, Malenka RC, Nicoll RA (1997b) Kainate receptors medi- Dani JA (2001) Overview of nicotinic receptors and their roles in the
ate a slow postsynaptic current in hippocampal CA3 neurons. central nervous system. Biol Psychiatry 49:166174.
Nature 388:182186. Davies SN, Pertwee RG, Riedel G (2002) Functions of cannabinoid
Catania MV, Tolle TR, Monyer H (1995) Differential expression of receptors in the hippocampus. Neuropharmacology 42:9931007.
234 The Hippocampus Book

Derkach V, Barria A, Soderling TR (1999) Ca2/calmodulin-kinase II Figenschou A, Hu GY, Storm JF (1996) Cholinergic modulation of
enhances channel conductance of alpha-amino-3- hydroxy-5- the action potential in rat hippocampal neurons. Eur J Neurosci
methyl-4-isoxazolepropionate type glutamate receptors. Proc 8:211219.
Natl Acad Sci USA 96:32693274. Fischer Y, Gahwiler BH, Thompson SM (1999) Activation of intrinsic
Destexhe A, Sejnowski TJ (1995) G protein activation kinetics and hippocampal theta oscillations by acetylcholine in rat septo-
spillover of gamma-aminobutyric acid may account for differ- hippocampal cocultures. J Physiol 519(Pt 2):405413.
ences between inhibitory responses in the hippocampus and Forsythe ID, Westbrook GL (1988) Slow excitatory postsynaptic cur-
thalamus. Proc Natl Acad Sci USA 92:95159519. rents mediated by N-methyl-D-aspartate receptors on cultured
Diamond JS (2001) Neuronal glutamate transporters limit activation mouse central neurones. J Physiol (Lond) 396:515533.
of NMDA receptors by neurotransmitter spillover on CA1 Frederickson CJ, Suh SW, Silva D, Thompson RB (2000) Importance
pyramidal cells. J Neurosci 21:83288338. of zinc in the central nervous system: the zinc-containing neu-
Diamond JS, Jahr CE (1997) Transporters buffer synaptically released ron. J Nutr 130:1471S(1483S.
glutamate on a submillisecond time scale. J Neurosci Frerking M, Nicoll RA (2000) Synaptic kainate receptors. Curr Opin
17:46724687. Neurobiol 10:342351.
Dodge FA Jr, Rahamimoff R (1967) Co-operative action a calcium Frerking M, Malenka RC, Nicoll RA (1998) Synaptic activation of
ions in transmitter release at the neuromuscular junction. kainate receptors on hippocampal interneurons. Nat Neurosci
J Physiol 193:419432. 1:479486.
Dolphin AC (1998) Mechanisms of modulation of voltage-depend- Fritschy JM, Meskenaite V, Weinmann O, Honer M, Benke D, Mohler
ent calcium channels by G proteins. J Physiol 506:311. H (1999) GABAB-receptor splice variants GB1a and GB1b in rat
Dolphin AC, Errington ML, Bliss TV (1982) Long-term potentiation brain: developmental regulation, cellular distribution and
of the perforant path in vivo is associated with increased gluta- extrasynaptic localization. Eur J Neurosci 11:761768.
mate release. Nature 297:496498. Fukuda T, Kosaka T (2000) Gap junctions linking the dendritic
Donevan SD, Rogawski MA (1995) Intracellular polyamines mediate network of GABAergic interneurons in the hippocampus.
inward rectication of Ca(2)- permeable alpha-amino-3- J Neurosci 20:15191528.
hydroxy-5-methyl-4-isoxazolepropionic acid receptors. Proc Fykse EM, Fonnum F (1996) Amino acid neurotransmission: dynam-
Natl Acad Sci USA 92:92989302. ics of vesicular uptake. Neurochem Res 21:10531060.
Draguhn A, Verdorn TA, Ewert M, Seeburg PH, Sakmann B (1990) Galarreta M, Hestrin S (1999) A network of fast-spiking cells in
Functional and molecular distinction between recombinant rat the neocortex connected by electrical synapses. Nature
GABAA receptor subtypes by Zn2. Neuron 5:781788. 402:7275.
Dunlap K, Luebke JI, Turner TJ (1995) Exocytotic Ca2 channels in Garcia EP, Mehta S, Blair LA, Wells DG, Shang J, Fukushima T, Fallon
mammalian central neurons. Trends Neurosci 18:8998. JR, Garner CC, Marshall J (1998) SAP90 binds and clusters
Durand GM, Kovalchuk Y, Konnerth A (1996) Long-term potentia- kainate receptors causing incomplete desensitization. Neuron
tion and functional synapse induction in developing hippocam- 21:727739.
pus. Nature 381:7175. Gasparini S, Saviane C, Voronin LL, Cherubini E (2000) Silent
Edwards FA, Konnerth A, Sakmann B (1990) Quantal analysis of synapses in the developing hippocampus: lack of functional
inhibitory synaptic transmission in the dentate gyrus of rat hip- AMPA receptors or low probability of glutamate release? Proc
pocampal slices: a patch-clamp study. J Physiol 430:213249. Natl Acad Sci USA 97:97419746.
Eghbali M, Curmi JP, Birnir B, Gage PW (1997) Hippocampal Geiger JR, Melcher T, Koh DS, Sakmann B, Seeburg PH, Jonas P,
GABA(A) channel conductance increased by diazepam. Nature Monyer H (1995) Relative abundance of subunit mRNAs deter-
388:7175. mines gating and Ca2 permeability of AMPA receptors in prin-
Emptage N, Bliss TV, Fine A (1999) Single synaptic events evoke cipal neurons and interneurons in rat CNS. Neuron 15:193204.
NMDA receptor-mediated release of calcium from internal Geiger JR, Lubke J, Roth A, Frotscher M, Jonas P (1997)
stores in hippocampal dendritic spines. Neuron 22:115124. Submillisecond AMPA receptor-mediated signaling at a princi-
Essrich C, Lorez M, Benson JA, Fritschy JM, Luscher B (1998) pal neuroninterneuron synapse. Neuron 18:10091023.
Postsynaptic clustering of major GABAA receptor subtypes Geppert M, Goda Y, Hammer RE, Li C, Rosahl TW, Stevens CF,
requires the gamma 2 subunit and gephyrin. Nat Neurosci Sudhof TC (1994) Synaptotagmin I: a major Ca2 sensor for
1:563571. transmitter release at a central synapse. Cell 79:717727.
Fabian-Fine R, Skehel P, Errington ML, Davies HA, Sher E, Stewart Gibb AJ, Colquhoun D (1992) Activation of N-methyl-D-aspartate
MG, Fine A (2001) Ultrastructural distribution of the alpha7 receptors by L-glutamate in cells dissociated from adult rat hip-
nicotinic acetylcholine receptor subunit in rat hippocampus. pocampus. J Physiol (Lond) 456:143179.
J Neurosci 21:79938003. Gibson JR, Beierlein M, Connors BW (1999) Two networks of elec-
Fagni L, Chavis P, Ango F, Bockaert J (2000) Complex interactions trically coupled inhibitory neurons in neocortex. Nature
between mGluRs, intracellular Ca2 stores and ion channels in 402:7579.
neurons. Trends Neurosci 23:8088. Gray EG (1959) Electron microscopy of synaptic contacts on den-
Farrar SJ, Whiting PJ, Bonnert TP, McKernan RM (1999) drite spines of the cerebral cortex. Nature 183:15921593.
Stoichiometry of a ligand-gated ion channel determined by u- Gulyas AI, Miles R, Sik A, Toth K, Tamamaki N, Freund TF (1993)
orescence energy transfer. J Biol Chem 274:1010010104. Hippocampal pyramidal cells excite inhibitory neurons through
Fernandez-Chacon R, Konigstorfer A, Gerber SH, Garcia J, Matos a single release site. Nature 366:683687.
MF, Stevens CF, Brose N, Rizo J, Rosenmund C, Sudhof TC Gundersen V, Chaudhry FA, Bjaalie JG, Fonnum F, Ottersen OP,
(2001) Synaptotagmin I functions as a calcium regulator of Storm-Mathisen J (1998) Synaptic vesicular localization and
release probability. Nature 410:4149. exocytosis of L-aspartate in excitatory nerve terminals: a quanti-
 Synaptic Function 235

tative immunogold analysis in rat hippocampus. J Neurosci Husi H, Ward MA, Choudhary JS, Blackstock WP, Grant SG (2000)
18:60596070. Proteomic analysis of NMDA receptor(adhesion protein signal-
Gutierrez R (2000) Seizures induce simultaneous GABAergic and ing complexes. Nat Neurosci 3:661669.
glutamatergic transmission in the dentate gyrusCA3 system. Isaac JT, Nicoll RA, Malenka RC (1995) Evidence for silent
J Neurophysiol 84:30883090. synapses: implications for the expression of LTP. Neuron
Haas HL, Konnerth A (1983) Histamine and noradrenaline decrease 15:427434.
calcium-activated potassium conductance in hippocampal Isaacson JS, Solis JM, Nicoll RA (1993) Local and diffuse synaptic
pyramidal cells. Nature 302:432434. actions of GABA in the hippocampus. Neuron 10:165175.
Hajos N, Nusser Z, Rancz EA, Freund TF, Mody I (2000) Cell type- Jahn R, Lang T, Sudhof TC (2003) Membrane fusion. Cell
and synapse-specic variability in synaptic GABAA receptor 112:519533.
occupancy. Eur J Neurosci 12:810818. Jahr CE (1992) High probability opening of NMDA receptor chan-
Hardingham NR, Larkman AU (1998) Rapid report: the reliability of nels by L-glutamate. Science 255:470472.
excitatory synaptic transmission in slices of rat visual cortex in Jahr CE, Stevens CF (1987) Glutamate activates multiple single
vitro is temperature dependent. J Physiol 507:249256. channel conductances in hippocampal neurons. Nature
Harris KM, Landis DM (1986) Membrane structure at synaptic junc- 325:522525.
tions in area CA1 of the rat hippocampus. Neuroscience Jefferys JG (1995) Nonsynaptic modulation of neuronal activity in
19:857872. the brain: electric currents and extracellular ions. Physiol Rev
Harris KM, Sultan P (1995) Variation in the number, location and 75:689723.
size of synaptic vesicles provides an anatomical basis for the Jiang L, Xu J, Nedergaard M, Kang J (2001) A kainate receptor
nonuniform probability of release at hippocampal CA1 increases the efficacy of GABAergic synapses. Neuron
synapses. Neuropharmacology 34:13871395. 30:503513.
Hausser M, Major G, Stuart GJ (2001) Differential shunting of EPSPs Johnson JW, Ascher P (1987) Glycine potentiates the NMDA
by action potentials. Science 291:138141. response in cultured mouse brain neurons. Nature 325:529531.
Hefft S, Kraushaar U, Geiger JR, Jonas P (2002) Presynaptic short- Johnston D, Magee JC, Colbert CM, Cristie BR (1996) Active proper-
term depression is maintained during regulation of transmitter ties of neuronal dendrites. Annu Rev Neurosci 19:165186.
release at a GABAergic synapse in rat hippocampus. J Physiol Jonas P, Major G, Sakmann B (1993) Quantal components of unitary
539:201208. EPSCs at the mossy bre synapse on CA3 pyramidal cells of rat
Henze DA, Urban NN, Barrionuevo G (2000) The multifarious hip- hippocampus. J Physiol (Lond) 472:615663.
pocampal mossy ber pathway: a review. Neuroscience 98: Jonas P, Racca C, Sakmann B, Seeburg PH, Monyer H (1994)
407427. Differences in Ca2 permeability of AMPA-type glutamate
Henze DA, Wittner L, Buzsaki G (2002) Single granule cells reliably receptor channels in neocortical neurons caused by differential
discharge targets in the hippocampal CA3 network in vivo. Nat GluR-B subunit expression. Neuron 12:12811289.
Neurosci 5:790795. Jones KA, Borowsky B, Tamm JA, Craig DA, Durkin MM, Dai M, Yao
Hessler NA, Shirke AM, Malinow R (1993) The probability of trans- WJ, Johnson M, Gunwaldsen C, Huang LY, Tang C, Shen Q,
mitter release at a mammalian central synapse. Nature Salon JA, Morse K, Laz T, Smith KE, Nagarathnam D, Noble SA,
366:569572. Branchek TA, Gerald C (1998) GABA(B) receptors function as a
Hestrin S, Nicoll RA, Perkel DJ, Sah P (1990) Analysis of excitatory heteromeric assembly of the subunits GABA(B)R1 and
synaptic action in pyramidal cells using whole-cell recording GABA(B)R2. Nature 396:674679.
from rat hippocampal slices. J Physiol (Lond) 422:203225. Jordan BA, Devi LA (1999) G-protein-coupled receptor het-
Heuss C, Scanziani M, Gahwiler BH, Gerber U (1999) G-protein- erodimerization modulates receptor function. Nature 399:
independent signaling mediated by metabotropic glutamate 697700.
receptors. Nat Neurosci 2:10701077. Kahle JS, Cotman CW (1989) Carbachol depresses synaptic responses
Hille B (2001) Ion channels of excitable membranes, 3rd ed. in the medial but not the lateral perforant path. Brain Res
Sunderland, MA: Sinauer. 482:159163.
Hjelmstad GO, Isaac JT, Nicoll RA, Malenka RC (1999) Lack of Kaila K (1994) Ionic basis of GABAA receptor channel function in the
AMPA receptor desensitization during basal synaptic transmis- nervous system. Prog Neurobiol 42:489537.
sion in the hippocampal slice. J Neurophysiol 81:30963099. Kaila K, Lamsa K, Smirnov S, Taira T, Voipio J (1997) Long-lasting
Hoffman AF, Lupica CR (2000) Mechanisms of cannabinoid inhibi- GABA-mediated depolarization evoked by high-frequency
tion of GABA(A) synaptic transmission in the hippocampus. stimulation in pyramidal neurons of rat hippocampal slice is
J Neurosci 20:24702479. attributable to a network-driven, bicarbonate-dependent K
Hkfelt T, Johansson O, Ljungdahl A, Lundberg JM, Schultzberg M transient. J Neurosci 17:76627672.
(1980) Peptidergic neurones. Nature 284:515521. Kamboj SK, Swanson GT, Cull-Candy SG (1995) Intracellular sper-
Hollmann M, Heinemann S (1994) Cloned glutamate receptors. mine confers rectication on rat calcium-permeable AMPA and
Annu Rev Neurosci 17:31108. kainate receptors. J Physiol 486:297303.
Hollmann M, OShea-Greeneld A, Rogers SW, Heinemann S (1989) Kamiya H, Ozawa S (2000) Kainate receptor-mediated presynaptic
Cloning by functional expression of a member of the glutamate inhibition at the mouse hippocampal mossy bre synapse.
receptor family. Nature 342:643648. J Physiol (Lond) 523(Pt 3):653665.
Hormuzdi SG, Pais I, LeBeau FE, Towers SK, Rozov A, Buhl EH, Kamiya H, Zucker RS (1994) Residual Ca2 and short-term synaptic
Whittington MA, Monyer H (2001) Impaired electrical signal- plasticity. Nature 371:603606.
ing disrupts gamma frequency oscillations in connexin 36- Kamiya H, Shinozaki H, Yamamoto C (1996) Activation of
decient mice. Neuron 31:487495. metabotropic glutamate receptor type 2/3 suppresses transmis-
236 The Hippocampus Book

sion at rat hippocampal mossy bre synapses. J Physiol (Lond) Kullmann DM, Asztely F (1998) Extrasynaptic glutamate spillover in
493:447455. the hippocampus: evidence and implications. Trends Neurosci
Kandel ER, Spencer WA (1961) Electrophysiology of hippocampal 21:814.
neurons. II. Afterpotentials and repetitive ring. J Neurophysiol Kullmann DM, Nicoll RA (1992) Long-term potentiation is associ-
24:305322. ated with increases in quantal content and quantal amplitude.
Katona I, Sperlagh B, Sik A, Kafalvi A, Vizi ES, Mackie K, Freund TF Nature 357:240244.
(1999) Presynaptically located CB1 cannabinoid receptors regu- Kuner R, Kohr G, Grunewald S, Eisenhardt G, Bach A, Kornau HC
late GABA release from axon terminals of specic hippocampal (1999) Role of heteromer formation in GABAB receptor func-
interneurons. J Neurosci 19:45444558. tion. Science 283:7477.
Katz B (1969) The release of neural transmitter substances. Livepool, Kunishima N, Shimada Y, Tsuji Y, Sato T, Yamamoto M, Kumasaka T,
UK: Liverpool University Press. Nakanishi S, Jingami H, Morikawa K (2000) Structural basis of
Kavanaugh MP, Arriza JL, North RA, Amara SG (1992) Electro- glutamate recognition by a dimeric metabotropic glutamate
genic uptake of gamma-aminobutyric acid by a cloned receptor. Nature 407:971977.
transporter expressed in Xenopus oocytes. J Biol Chem Laezza F, Doherty JJ, Dingledine R (1999) Long-term depression in
267:2200722009. hippocampal interneurons: joint requirement for pre- and post-
Keinanen K, Wisden W, Sommer B, Werner P, Herb A, Verdoorn TA, synaptic events. Science 285:14111414.
Sakmann B, Seeburg PH (1990) A family of AMPA-selective glu- Lambert NA, Wilson WA (1993) Heterogeneity in presynaptic regu-
tamate receptors. Science 249:556560. lation of GABA release from hippocampal inhibitory neurons.
Kennedy MB (1997) The postsynaptic density at glutamatergic Neuron 11:10571067.
synapses. Trends Neurosci 20:264268. Lambolez B, Ropert N, Perrais D, Rossier J, Hestrin S (1996)
Kirov SA, Sorra KE, Harris KM (1999) Slices have more synapses Correlation between kinetics and RNA splicing of alpha-amino-
than perfusion-xed hippocampus from both young and 3-hydroxy- 5-methylisoxazole-4-propionic acid receptors in
mature rats. J Neurosci 19:28762886. neocortical neurons. Proc Natl Acad Sci USA 93:17971802.
Klausberger T, Roberts JD, Somogyi P (2002) Cell type- and input- Lanthorn TH, Ganong AH, Cotman CW (1984) 2-Amino-4-phos-
specic differences in the number and subtypes of synaptic phonobutyrate selectively blocks mossy ber-CA3 responses in
GABA(A) receptors in the hippocampus. J Neurosci 22: guinea pig but not rat hippocampus. Brain Res 290:174178.
25132521. Larkum ME, Zhu JJ, Sakmann B (1999) A new cellular mechanism
Kleckner NW, Dingledine R (1988) Requirement for glycine in acti- for coupling inputs arriving at different cortical layers. Nature
vation of NMDA-receptors expressed in Xenopus oocytes. 398:338341.
Science 241:835837. Laube B, Hirai H, Sturgess M, Betz H, Kuhse J (1997) Molecular
Klingauf J, Neher E (1997) Modeling buffered Ca2 diffusion near the determinants of agonist discrimination by NMDA receptor
membrane: implications for secretion in neuroendocrine cells. subunits: analysis of the glutamate binding site on the NR2B
Biophys J 72:674690. subunit. Neuron 18:493503.
Koester HJ, Sakmann B (2000) Calcium dynamics associated with Lauri SE, Bortolotto ZA, Bleakman D, Ornstein PL, Lodge D, Isaac JT,
action potentials in single nerve terminals of pyramidal cells in Collingridge GL (2001) A critical role of a facilitatory presynap-
layer 2/3 of the young rat neocortex. J Physiol 529(Pt 3): tic kainate receptor in mossy ber LTP. Neuron 32:697709.
625646. Lehre KP, Levy LM, Ottersen OP, Storm-Mathisen J, Danbolt NC
Koh DS, Burnashev N, Jonas P (1995) Block of native Ca(2)-per- (1995) Differential expression of two glial glutamate trans-
meable AMPA receptors in rat brain by intracellular polyamines porters in the rat brain: quantitative and immunocytochemical
generates double rectication. J Physiol 486:305312. observations. J Neurosci 15:18351853.
Kotter R (1994) Postsynaptic integration of glutamatergic and Lei S, McBain CJ (2002) Distinct NMDA receptors provide differen-
dopaminergic signals in the striatum. Prog Neurobiol tial modes of transmission at mossy berinterneuron synapses.
44:163196. Neuron 33:921933.
Kovalchuk Y, Eilers J, Lisman J, Konnerth A (2000) NMDA receptor- Lerma J, Morales M, Vicente MA, Herreras O (1997) Glutamate
mediated subthreshold Ca(2) signals in spines of hippocam- receptors of the kainate type and synaptic transmission. Trends
pal neurons. J Neurosci 20:17911799. Neurosci 20:912.
Kraushaar U, Jonas P (2000) Efficacy and stability of quantal GABA Lerma J, Paternain AV, Rodriguez-Moreno A, Lopez-Garcia JC (2001)
release at a hippocampal interneuronprincipal neuron Molecular physiology of kainate receptors. Physiol Rev
synapse. J Neurosci 20:55945607. 81:971998.
Kullmann DM (1994) Amplitude uctuations of dual-component Lester RA, Clements JD, Westbrook GL, Jahr CE (1990) Channel
EPSCs in hippocampal pyramidal cells: implications for long- kinetics determine the time course of NMDA receptor-
term potentiation. Neuron 12:11111120. mediated synaptic currents. Nature 346:565567.
Kullmann DM (1999) Excitatory synapses: neither too loud nor too Liao D, Hessler NA, Malinow R (1995) Activation of postsynaptically
quiet. Nature 399:111112. silent synapses during pairing-induced LTP in CA1 region of
Kullmann DM (2001) Presynaptic kainate receptors in the hip- hippocampal slice. Nature 375:400404.
pocampus: slowly emerging from obscurity. Neuron 32: Lin RC, Scheller RH (2000) Mechanisms of synaptic vesicle exocyto-
561564. sis. Annu Rev Cell Dev Biol 16:1949.
Kullmann DM (2003) Silent synapses: what are they telling us about Lisman JE (1997) Bursts as a unit of neural information: making
long-term potentiation? Philos Trans R Soc Lond B Biol Sci unreliable synapses reliable. Trends Neurosci 20:3843.
358:727733. Lozovaya NA, Kopanitsa MV, Boychuk YA, Krishtal OA (1999)
 Synaptic Function 237

Enhancement of glutamate release uncovers spillover-mediated McNaughton BL (1980) Evidence for two physiologically distinct
transmission by N-methyl-D-aspartate receptors in the rat hip- perforant pathways to the fascia dentata. Brain Res 199:119.
pocampus. Neuroscience 91:13211330. Mehta AK, Ticku MK (1999) An update on GABAA receptors. Brain
Lu YM, Mansuy IM, Kandel ER, Roder J (2000) Calcineurin-medi- Res Brain Res Rev 29:196217.
ated LTD of GABAergic inhibition underlies the increased Min MY, Asztely F, Kokaia M, Kullmann DM (1998a) Long-term
excitability of CA1 neurons associated with LTP. Neuron potentiation and dual-component quantal signaling in the den-
26:197205. tate gyrus. Proc Natl Acad Sci USA 95:47024707.
Luscher C, Jan LY, Stoffel M, Malenka RC, Nicoll RA (1997) G pro- Min MY, Rusakov DA, Kullmann DM (1998b) Activation of AMPA,
tein-coupled inwardly rectifying K channels (GIRKs) mediate kainate, and metabotropic receptors at hippocampal mossy
postsynaptic but not presynaptic transmitter actions in hip- ber synapses: role of glutamate diffusion. Neuron 21:561570.
pocampal neurons. Neuron 19:687695. Erratum: Neuron Mintz IM, Bean BP (1993) GABAB receptor inhibition of P-type
1997;19;following 945. Ca2 channels in central neurons. Neuron 10:889898.
Maccaferri G, Toth K, McBain CJ (1998) Target-specic expression of Misgeld U, Bijak M, Jarolimek W (1995) A physiological role for
presynaptic mossy ber plasticity. Science 279:13681370. GABAB receptors and the effects of baclofen in the mammalian
Maccaferri G, Roberts JD, Szucs P, Cottingham CA, Somogyi P central nervous system. Prog Neurobiol 46:423462.
(2000) Cell surface domain specic postsynaptic currents Monaghan DT, Holets VR, Toy DW, Cotman CW (1983) Anatomical
evoked by identied GABAergic neurones in rat hippocampus distributions of four pharmacologically distinct 3H-L- gluta-
in vitro. J Physiol 524(Pt 1):91116. mate binding sites. Nature 306:176179.
MacDermott AB, Role LW, Siegelbaum SA (1999) Presynaptic Mori M, Heuss C, Gahwiler BH, Gerber U (2001) Fast synaptic trans-
ionotropic receptors and the control of transmitter release. mission mediated by P2X receptors in CA3 pyramidal cells of
Annu Rev Neurosci 22:443485. rat hippocampal slice cultures. J Physiol 535:115123.
Macek TA, Winder DG, Gereau RWt, Ladd CO, Conn PJ (1996) Mori M, Gahwiler BH, Gerber U (2002) Beta-alanine and taurine as
Differential involvement of group II and group III mGluRs endogenous agonists at glycine receptors in rat hippocampus in
as autoreceptors at lateral and medial perforant path synapses. vitro. J Physiol 539:191200.
J Neurophysiol 76:37983806. Mosbacher J, Schoepfer R, Monyer H, Burnashev N, Seeburg PH,
Madison DV, Nicoll RA (1982) Noradrenaline blocks accommoda- Ruppersberg JP (1994) A molecular determinant for submil-
tion of pyramidal cell discharge in the hippocampus. Nature lisecond desensitization in glutamate receptors. Science
299:636638. 266:10591062.
Madison DV, Nicoll RA (1984) Control of the repetitive discharge of Mulle C, Sailer A, Perez-Otano I, Dickinson-Anson H, Castillo PE,
rat CA 1 pyramidal neurones in vitro. J Physiol 354:319331. Bureau I, Maron C, Gage FH, Mann JR, Bettler B, Heinemann
Magee JC, Cook EP (2000) Somatic EPSP amplitude is independent SF (1998) Altered synaptic physiology and reduced susceptibil-
of synapse location in hippocampal pyramidal neurons. Nat ity to kainate- induced seizures in GluR6-decient mice. Nature
Neurosci 3:895903. 392:601605.
Mainen ZF, Malinow R, Svoboda K (1999) Synaptic calcium tran- Mulle C, Sailer A, Swanson GT, Brana C, OGorman S, Bettler B,
sients in single spines indicate that NMDA receptors are not sat- Heinemann SF (2000) Subunit composition of kainate recep-
urated. Nature 399:151155. tors in hippocampal interneurons. Neuron 28:475484.
Manabe T, Renner P, Nicoll RA (1992) Postsynaptic contribution to Murthy VN, Stevens CF (1998) Synaptic vesicles retain their identity
long-term potentiation revealed by the analysis of miniature through the endocytic cycle. Nature 392:497501.
synaptic currents. Nature 355:5055. Murthy VN, Stevens CF (1999) Reversal of synaptic vesicle docking at
Markram H, Tsodyks M (1996) Redistribution of synaptic efficacy central synapses. Nat Neurosci 2:503507.
between neocortical pyramidal neurons. Nature 382:807810. Nadler JV, Martin D, Bustos GA, Burke SP, Bowe MA (1990)
Marsicano G, Lutz B (1999) Expression of the cannabinoid receptor Regulation of glutamate and aspartate release from the Schaffer
CB1 in distinct neuronal subpopulations in the adult mouse collaterals and other projections of CA3 hippocampal pyrami-
forebrain. Eur J Neurosci 11:42134225. dal cells. Prog Brain Res 83:115130.
Mathews GC, Diamond JS (2003) Neuronal glutamate uptake con- Nakanishi S, Nakajima Y, Masu M, Ueda Y, Nakahara K, Watanabe D,
tributes to GABA synthesis and inhibitory synaptic strength. Yamaguchi S, Kawabata S, Okada M (1998) Glutamate recep-
J Neurosci 23:20402048. tors: brain function and signal transduction. Brain Res Brain Res
Mayer ML, Westbrook GL, Guthrie PB (1984) Voltage-dependent Rev 26:230235.
block by Mg2 of NMDA responses in spinal cord neurones. Nicoll RA, Malenka RC, Kauer JA (1990) Functional comparison of
Nature 309:261263. neurotransmitter receptor subtypes in mammalian central
McBain CJ, Mayer ML (1994) N-Methyl-D-aspartic acid receptor nervous system. Physiol Rev 70:513565.
structure and function. Physiol Rev 74:723760. Nowak L, Bregestovski P, Ascher P, Herbet A, Prochiantz A (1984)
McGehee DS, Heath MJ, Gelber S, Devay P, Role LW (1995) Nicotine Magnesium gates glutamate-activated channels in mouse cen-
enhancement of fast excitatory synaptic transmission in CNS by tral neurones. Nature 307:462465.
presynaptic receptors. Science 269:16921696. Nusser Z, Mody I (2002) Selective modulation of tonic and phasic
McIntire SL, Reimer RJ, Schuske K, Edwards RH, Jorgensen EM inhibitions in dentate gyrus granule cells. J Neurophysiol
(1997) Identication and characterization of the vesicular 87:26242628.
GABA transporter. Nature 389:870876. Nusser Z, Sieghart W, Benke D, Fritschy JM, Somogyi P (1996)
McMahon LL, Kauer JA (1997) Hippocampal interneurons express a Differential synaptic localization of two major gamma-
novel form of synaptic plasticity. Neuron 18:295305. aminobutyric acid type A receptor alpha subunits on hip-
238 The Hippocampus Book

pocampal pyramidal cells. Proc Natl Acad Sci USA 93: Pin JP, Duvoisin R (1995) The metabotropic glutamate receptors:
1193911944. structure and functions. Neuropharmacology 34:126.
Nusser Z, Lujan R, Laube G, Roberts JD, Molnar E, Somogyi P (1998) Piomelli D (1994) Eicosanoids in synaptic transmission. Crit Rev
Cell type and pathway dependence of synaptic AMPA receptor Neurobiol 8:6583.
number and variability in the hippocampus. Neuron Pitler TA, Alger BE (1992) Cholinergic excitation of GABAergic
21:545559. interneurons in the rat hippocampal slice. J Physiol
Oertner TG, Sabatini BL, Nimchinsky EA, Svoboda K (2002) 450:127142.
Facilitation at single synapses probed with optical quantal Poncer JC, McKinney RA, Gahwiler BH, Thompson SM (1997)
analysis. Nat Neurosci 5:657664. Either N- or P-type calcium channels mediate GABA release
Ohno-Shosaku T, Maejima T, Kano M (2001) Endogenous cannabi- at distinct hippocampal inhibitory synapses. Neuron
noids mediate retrograde signals from depolarized postsynaptic 18:463472.
neurons to presynaptic terminals. Neuron 29:729738. Porter JT, Cauli B, Tsuzuki K, Lambolez B, Rossier J, Audinat E (1999)
Otmakhova NA, Lisman JE (1999) Dopamine selectively inhibits Selective excitation of subtypes of neocortical interneurons by
the direct cortical pathway to the CA1 hippocampal region. nicotinic receptors. J Neurosci 19:52285235.
J Neurosci 19:14371445. Pouille F, Scanziani M (2001) Enforcement of temporal delity in
Otmakhov N, Shirke AM, Malinow R (1993) Measuring the impact pyramidal cells by somatic feed-forward inhibition. Science
of probabilistic transmission on neuronal output. Neuron 293:11591163.
10:11011111. Raastad M (1995) Extracellular activation of unitary excitatory
Otmakhova NA, Otmakhov N, Lisman JE (2002) Pathway-specic synapses between hippocampal CA3 and CA1 pyramidal cells.
properties of AMPA and NMDA-mediated transmission in CA1 Eur J Neurosci 7:18821888.
hippocampal pyramidal cells. J Neurosci 22:11991207. Racca C, Stephenson FA, Streit P, Roberts JD, Somogyi P (2000)
Ottersen OP, Storm-Mathisen J, Bramham C, Torp R, Laake J, NMDA receptor content of synapses in stratum radiatum of the
Gundersen V (1990) A quantitative electron microscopic hippocampal CA1 area. J Neurosci 20:25122522.
immunocytochemical study of the distribution and synaptic Regehr WG, Delaney KR, Tank DW (1994) The role of presynaptic
handling of glutamate in rat hippocampus. Prog Brain Res calcium in short-term enhancement at the hippocampal mossy
83:99114. ber synapse. J Neurosci 14:523537.
Overstreet LS, Westbrook GL (2001) Paradoxical reduction of synap- Renger JJ, Egles C, Liu G (2001) A developmental switch in neuro-
tic inhibition by vigabatrin. J Neurophysiol 86:596603. transmitter ux enhances synaptic efficacy by affecting AMPA
Ozawa S, Kamiya H, Tsuzuki K (1998) Glutamate receptors in the receptor activation. Neuron 29:469484.
mammalian central nervous system. Prog Neurobiol 54:581618. Reyes A, Lujan R, Rozov A, Burnashev N, Somogyi P, Sakmann B
Pankratov Y, Castro E, Miras-Portugal MT, Krishtal O (1998) A (1998) Target-cell-specic facilitation and depression in neo-
purinergic component of the excitatory postsynaptic current cortical circuits. Nat Neurosci 1:279285.
mediated by P2X receptors in the CA1 neurons of the rat hip- Richerson GB, Wu Y (2003) Dynamic equilibrium of neurotransmit-
pocampus. Eur J Neurosci 10:38983902. ter transporters: not just for reuptake anymore. J Neurophysiol
Patneau DK, Mayer ML (1990) Structure-activity relationships for 90:13631374.
amino acid transmitter candidates acting at N-methyl-D-aspar- Rivera C, Voipio J, Payne JA, Ruusuvuori E, Lahtinen H, Lamsa K,
tate and quisqualate receptors. J Neurosci 10:23852399. Pirvola U, Saarma M, Kaila K (1999) The K/Cl- co-transporter
Pawelzik H, Bannister AP, Deuchars J, Ilia M, Thomson AM (1999) KCC2 renders GABA hyperpolarizing during neuronal matura-
Modulation of bistratied cell IPSPs and basket cell IPSPs by tion. Nature 397:251255.
pentobarbitone sodium, diazepam and Zn2: dual recordings in Rodriguez-Moreno A, Lerma J (1998) Kainate receptor modulation
slices of adult rat hippocampus. Eur J Neurosci 11:35523564. of GABA release involves a metabotropic function. Neuron
Pearce RA (1993) Physiological evidence for two distinct GABAA 20:12111218.
responses in rat hippocampus. Neuron 10:189200. Rodriguez-Moreno A, Herreras O, Lerma J (1997) Kainate receptors
Pedarzani P, Storm JF (1993) PKA mediates the effects of monoamine presynaptically downregulate GABAergic inhibition in the rat
transmitters on the K current underlying the slow spike fre- hippocampus. Neuron 19:893901.
quency adaptation in hippocampal neurons. Neuron Rosahl TW, Spillane D, Missler M, Herz J, Selig DK, Wolff JR,
11:10231035. Hammer RE, Malenka RC, Sudhof TC (1995) Essential func-
Pedarzani P, Storm JF (1996) Evidence that Ca/calmodulin-depend- tions of synapsins I and II in synaptic vesicle regulation. Nature
ent protein kinase mediates the modulation of the Ca2- 375:488493.
dependent K current, IAHP, by acetylcholine, but not by Rosenmund C, Stevens CF (1996) Denition of the readily releasable
glutamate, in hippocampal neurons. Pugers Arch 431:723728. pool of vesicles at hippocampal synapses. Neuron 16:11971207.
Perrais D, Ropert N (1999) Effect of zolpidem on miniature IPSCs Rosenmund C, Clements JD, Westbrook GL (1993) Nonuniform
and occupancy of postsynaptic GABAA receptors in central probability of glutamate release at a hippocampal synapse.
synapses. J Neurosci 19:578588. Science 262:754757.
Perroy J, El Far O, Bertaso F, Pin JP, Betz H, Bockaert J, Fagni L (2002) Rosenmund C, Feltz A, Westbrook GL (1995) Synaptic NMDA recep-
PICK1 is required for the control of synaptic transmission by tor channels have a low open probability. J Neurosci
the metabotropic glutamate receptor 7. EMBO J 21:29902999. 15:27882795.
Petralia RS, Esteban JA, Wang YX, Partridge JG, Zhao HM, Wenthold Rosenmund C, Stern-Bach Y, Stevens CF (1998) The tetrameric
RJ, Malinow R (1999) Selective acquisition of AMPA receptors structure of a glutamate receptor channel. Science 280:
over postnatal development suggests a molecular basis for silent 15961599.
synapses. Nat Neurosci 2:3136. Rossi DJ, Oshima T, Attwell D (2000) Glutamate release in severe
 Synaptic Function 239

brain ischaemia is mainly by reversed uptake. Nature Schwartzkroin PA (1975) Characteristics of CA1 neurons recorded
403:316321. intracellularly in the hippocampal in vitro slice preparation.
Rozov A, Burnashev N (1999) Polyamine-dependent facilitation of Brain Res 85:423436.
postsynaptic AMPA receptors counteracts paired-pulse depres- Sealfon SC, Olanow CW (2000) Dopamine receptors: from structure
sion. Nature 401:594598. to behavior. Trends Neurosci 23:S34S40.
Rozov A, Zilberter Y, Wollmuth LP, Burnashev N (1998) Facilitation Seeger T, Alzheimer C (2001) Muscarinic activation of inwardly rec-
of currents through rat Ca2-permeable AMPA receptor chan- tifying K() conductance reduces EPSPs in rat hippocampal
nels by activity-dependent relief from polyamine block. J Physiol CA1 pyramidal cells. J Physiol 535:383396.
511:361377. Selyanko AA, Hadley JK, Wood IC, Abogadie FC, Delmas P, Buckley
Rudolph U, Crestani F, Benke D, Brunig I, Benson JA, Fritschy JM, NJ, London B, Brown DA (1999) Two types of K() channel
Martin JR, Bluethmann H, Mohler H (1999) Benzodiazepine subunit, Erg1 and KCNQ2/3, contribute to the M-like current in
actions mediated by specic gamma-aminobutyric acid(A) a mammalian neuronal cell. J Neurosci 19:77427756.
receptor subtypes. Nature 401:796800. Semyanov A, Kullmann DM (2000) Modulation of GABAergic sig-
Rusakov DA, Kullmann DM (1998) Extrasynaptic glutamate diffu- naling among interneurons by metabotropic glutamate recep-
sion in the hippocampus: ultrastructural constraints, uptake, tors. Neuron 25:663672.
and receptor activation. J Neurosci 18:31583170. Semyanov A, Kullmann DM (2001) Kainate receptor-dependent
Rusakov DA, Harrison E, Stewart MG (1998) Synapses in hippocam- axonal depolarization and action potential initiation in
pus occupy only 12% of cell membranes and are spaced interneurons. Nat Neurosci 4:718723.
less than half-micron apart: a quantitative ultrastructural Semyanov A, Walker MC, Kullmann DM (2003) GABA uptake regu-
analysis with discussion of physiological implications. lates cortical excitability via cell type-specic tonic inhibition.
Neuropharmacology 37:513521. Nat Neurosci 6:484490.
Sandler R, Smith AD (1991) Coexistence of GABA and glutamate in Sepkuty JP, Cohen AS, Eccles C, Raq A, Behar K, Ganel R, Coulter
mossy ber terminals of the primate hippocampus: an ultra- DA, Rothstein JD (2002) A neuronal glutamate transporter
structural study. J Comp Neurol 303:177192. contributes to neurotransmitter GABA synthesis and epilepsy.
Scanziani M (2000) GABA spillover activates postsynaptic GABA(B) J Neurosci 22:63726379.
receptors to control rhythmic hippocampal activity. Neuron Shapiro MS, Roche JP, Kaftan EJ, Cruzblanca H, Mackie K, Hille B
25:673681. (2000) Reconstitution of muscarinic modulation of the
Scanziani M, Gahwiler BH, Thompson SM (1995) Presynaptic inhi- KCNQ2/KCNQ3 K() channels that underlie the neuronal M
bition of excitatory synaptic transmission by muscarinic and current. J Neurosci 20:17101721.
metabotropic glutamate receptor activation in the hippocam- Sheng M, Sala C (2001) PDZ domains and the organization of
pus: are Ca2 channels involved? Neuropharmacology supramolecular complexes. Annu Rev Neurosci 24:129.
34:15491557. Shepherd GM, Harris KM (1998) Three-dimensional structure and
Scanziani M, Salin PA, Vogt KE, Malenka RC, Nicoll RA (1997) Use- composition of CA3CA1 axons in rat hippocampal slices:
dependent increases in glutamate concentration activate presy- implications for presynaptic connectivity and compartmental-
naptic metabotropic glutamate receptors. Nature 385:630634. ization. J Neurosci 18:83008310.
Scanziani M, Gahwiler BH, Charpak S (1998) Target cell-specic Shepherd GM, Raastad M, Andersen P (2002) General and variable
modulation of transmitter release at terminals from a single features of varicosity spacing along unmyelinated axons in the
axon. Proc Natl Acad Sci USA 95:1200412009. hippocampus and cerebellum. Proc Natl Acad Sci USA
Schell MJ, Molliver ME, Snyder SH (1995) D-Serine, an endogenous 99:63406345.
synaptic modulator: localization to astrocytes and glutamate- Shi S, Hayashi Y, Esteban JA, Malinow R (2001) Subunit-specic rules
stimulated release. Proc Natl Acad Sci USA 92:39483952. governing AMPA receptor trafficking to synapses in hippocam-
Schikorski T, Stevens CF (1997) Quantitative ultrastructural analysis pal pyramidal neurons. Cell 105:331343.
of hippocampal excitatory synapses. J Neurosci 17:58585867. Shigemoto R, Kulik A, Roberts JD, Ohishi H, Nusser Z, Kaneko T,
Schmitz D, Mellor J, Nicoll RA (2001c) Presynaptic kainate receptor Somogyi P (1996) Target-cell-specific concentration of a
mediation of frequency facilitation at hippocampal mossy ber metabotropic glutamate receptor in the presynaptic active zone.
synapses. Science 291:19721976. Nature 381:523525.
Schmitz D, Mellor J, Frerking M, Nicoll RA (2001a) Presynaptic Shigemoto R, Kinoshita A, Wada E, Nomura S, Ohishi H, Takada M,
kainate receptors at hippocampal mossy ber synapses. Proc Flor PJ, Neki A, Abe T, Nakanishi S, Mizuno N (1997)
Natl Acad Sci USA 98:1100311008. Differential presynaptic localization of metabotropic glutamate
Schmitz D, Schuchmann S, Fisahn A, Draguhn A, Buhl EH, Petrasch- receptor subtypes in the rat hippocampus. J Neurosci
Parwez E, Dermietzel R, Heinemann U, Traub RD (2001b) Axo- 17:75037522.
axonal coupling. a novel mechanism for ultrafast neuronal Simmons ML, Chavkin C (1996) Endogenous opioid regulation of
communication. Neuron 31:831840. hippocampal function. Int Rev Neurobiol 39:145196.
Schneggenburger R, Neher E (2000) Intracellular calcium depend- Sloviter RS, Dichter MA, Rachinsky TL, Dean E, Goodman JH, Sollas
ence of transmitter release rates at a fast central synapse. Nature AL, Martin DL (1996) Basal expression and induction of gluta-
406:889893. mate decarboxylase and GABA in excitatory granule cells of the
Schoepp DD (2002) Metabotropic glutamate receptors. Pharmacol rat and monkey hippocampal dentate gyrus. J Comp Neurol
Biochem Behav 73:285286. 373:593618.
Schousboe A (2000) Pharmacological and functional characteriza- Smart TG, Xie X, Krishek BJ (1994) Modulation of inhibitory and
tion of astrocytic GABA transport: a short review. Neurochem excitatory amino acid receptor ion channels by zinc. Prog
Res 25:12411244. Neurobiol 42:393341.
240 The Hippocampus Book

Soghomonian JJ, Martin DL (1998) Two isoforms of glutamate glutamate released and retained in hippocampal slices during
decarboxylase: why? Trends Pharmacol Sci 19:500505. stimulation. J Neurochem 50:219224.
Sommer B, Kohler M, Sprengel R, Seeburg PH (1991) RNA editing in Takamori S, Rhee JS, Rosenmund C, Jahn R (2000) Identication of
brain controls a determinant of ion ow in glutamate-gated a vesicular glutamate transporter that denes a glutamatergic
channels. Cell 67:1119. phenotype in neurons. Nature 407:189194.
Sorra KE, Harris KM (1993) Occurrence and three-dimensional Takumi Y, Bergersen L, Landsend AS, Rinvik E, Ottersen OP (1998)
structure of multiple synapses between individual radiatum Synaptic arrangement of glutamate receptors. Prog Brain Res
axons and their target pyramidal cells in hippocampal area CA1. 116:105121.
J Neurosci 13:37363748. Takumi Y, Ramirez-Leon V, Laake P, Rinvik E, Ottersen OP (1999)
Spacek J (1985) Three-dimensional analysis of dendritic spines. III. Different modes of expression of AMPA and NMDA receptors
Glial sheath. Anat Embryol 171:245252. in hippocampal synapses. Nat Neurosci 2:618624.
Spencer WA, Kandel ER (1961) Electrophysiology of hippocampal Tang Y, Zucker RS (1997) Mitochondrial involvement in post-tetanic
neurons. IV. Fast pre-potentials. J Neurophysiol 24:272285. potentiation of synaptic transmission. Neuron 18:483491.
Sperk G, Schwarzer C, Tsunashima K, Fuchs K, Sieghart W (1997) Thomson AM (2000) Facilitation, augmentation and potentiation at
GABA(A) receptor subunits in the rat hippocampus. I. central synapses. Trends Neurosci 23:305312.
Immunocytochemical distribution of 13 subunits. Neuroscience Toth K, McBain CJ (1998) Afferent-specic innervation of two dis-
80:9871000. tinct AMPA receptor subtypes on single hippocampal interneu-
Spiridon M, Kamm D, Billups B, Mobbs P, Attwell D (1998) rons. Nat Neurosci 1:572578.
Modulation by zinc of the glutamate transporters in glial cells Toth K, Suares G, Lawrence JJ, Philips-Tansey E, McBain CJ (2000)
and cones isolated from the tiger salamander retina. J Physiol Differential mechanisms of transmission at three types of mossy
506:363376. ber synapse. J Neurosci 20:82798289.
Staley KJ, Mody I (1992) Shunting of excitatory input to dentate Trommald M, Hulleberg G (1997) Dimensions and density of den-
gyrus granule cells by a depolarizing GABAA receptor-mediated dritic spines from rat dentate granule cells based on reconstruc-
postsynaptic conductance. J Neurophysiol 68:197212. tions from serial electron micrographs. J Comp Neurol
Staley KJ, Soldo BL, Proctor WR (1995) Ionic mechanisms of neu- 377:1528.
ronal excitation by inhibitory GABAA receptors. Science Tsou K, Mackie K, Sanudo-Pena MC, Walker JM (1999) Cannabinoid
269:977981. CB1 receptors are localized primarily on cholecystokinin-con-
Stell BM, Mody I (2002) Receptors with different affinities mediate taining GABAergic interneurons in the rat hippocampal forma-
phasic and tonic GABA(A) conductances in hippocampal neu- tion. Neuroscience 93:969975.
rons. J Neurosci 22:RC223. Tzounopoulos T, Janz R, Sudhof TC, Nicoll RA, Malenka RC (1998)
Stevens CF, Tsujimoto T (1995) Estimates for the pool size of A role for cAMP in long-term depression at hippocampal mossy
releasable quanta at a single central synapse and for the time ber synapses. Neuron 21:837845.
required to rell the pool. Proc Natl Acad Sci USA 92:846849. Ventura R, Harris KM (1999) Three-dimensional relationships
Steyer JA, Horstmann H, Almers W (1997) Transport, docking and between hippocampal synapses and astrocytes. J Neurosci
exocytosis of single secretory granules in live chromaffin cells. 19:68976906.
Nature 388:474478. Vignes M, Collingridge GL (1997) The synaptic activation of kainate
Storm-Mathisen J, Iversen LL (1979) Uptake of [3H]glutamic acid in receptors. Nature 388:179182.
excitatory nerve endings: light and electron microscopic obser- Vignes M, Clarke VR, Parry MJ, Bleakman D, Lodge D, Ornstein
vations in the hippocampal formation of the rat. Neuroscience PL, Collingridge GL (1998) The GluR5 subtype of kainate
4:12371253. receptor regulates excitatory synaptic transmission in areas CA1
Stuart GJ, Hausser M (2001) Dendritic coincidence detection of and CA3 of the rat hippocampus. Neuropharmacology 37:
EPSPs and action potentials. Nat Neurosci 4:6371. 12691277.
Study RE, Barker JL (1981) Diazepam and ()-pentobarbital: uctu- Vogt KE, Nicoll RA (1999) Glutamate and gamma-aminobutyric acid
ation analysis reveals different mechanisms for potentiation of mediate a heterosynaptic depression at mossy ber synapses in
gamma-aminobutyric acid responses in cultured central neu- the hippocampus. Proc Natl Acad Sci USA 96:11181122.
rons. Proc Natl Acad Sci USA 78:71807184. Vogt K, Mellor J, Tong G, Nicoll R (2000) The actions of synaptically
Sudhof TC (2004) The synaptic vesicle cycle. Annu Rev Neurosci released zinc at hippocampal mossy ber synapses. Neuron
27:509547. 26:187196.
Suh B, Hille B (2002) Recovery from muscarinic modulation of m Wafford KA, Thompson SA, Thomas D, Sikela J, Wilcox AS, Whiting
current channels requires phosphatidylinositol 4,5-bisphos- PJ (1996) Functional characterization of human gamma-
phate synthesis. Neuron 35:507. aminobutyric acid A receptors containing the alpha 4 subunit.
Sundstrom-Poromaa I, Smith DH, Gong QH, Sabado TN, Li X, Light Mol Pharmacol 50:670678.
A, Wiedmann M, Williams K, Smith SS (2002) Hormonally reg- Wahl LM, Pouzat C, Stratford KJ (1996) Monte Carlo simulation of
ulated alpha(4)beta(2)delta GABA(A) receptors are a target for fast excitatory synaptic transmission at a hippocampal synapse.
alcohol. Nat Neurosci 5:721722. J Neurophysiol 75:597608.
Swanson GT, Kamboj SK, Cull-Candy SG (1997) Single-channel Walker MC, Galley PT, Errington ML, Shorvon SD, Jefferys JG (1995)
properties of recombinant AMPA receptors depend on RNA Ascorbate and glutamate release in the rat hippocampus
editing, splice variation, and subunit composition. J Neurosci after perforant path stimulation: a dialysis electrode study.
17:5869. J Neurochem 65:725731.
Szerb JC (1988) Changes in the relative amounts of aspartate and Walker MC, Ruiz A, Kullmann DM (2001) Monosynaptic GABAergic
 Synaptic Function 241

signaling from dentate to CA3 with a pharmacological and Wu LG, Saggau P (1994) Adenosine inhibits evoked synaptic trans-
physiological prole typical of mossy ber synapses. Neuron mission primarily by reducing presynaptic calcium inux in
29:703715. area CA1 of hippocampus. Neuron 12:11391148.
Wall MJ, Usowicz MM (1997) Development of action potential- Wu S, Wright RA, Rockey PK, Burgett SG, Arnold JS, Rosteck PR, Jr.,
dependent and independent spontaneous GABAA receptor- Johnson BG, Schoepp DD, Belagaje RM (1998) Group III
mediated currents in granule cells of postnatal rat cerebellum. human metabotropic glutamate receptors 4, 7 and 8: molecular
Eur J Neurosci 9:533548. cloning, functional expression, and comparison of pharmaco-
Wan Q, Xiong ZG, Man HY, Ackerley CA, Braunton J, Lu WY, Becker logical properties in RGT cells. Brain Res Mol Brain Res
LE, MacDonald JF, Wang YT (1997) Recruitment of functional 53:8897.
GABA(A) receptors to postsynaptic domains by insulin. Nature Yamamoto C, Sawada S, Takada S (1983) Suppressing action of 2-
388:686690. amino-4-phosphonobutyric acid on mossy ber-induced exci-
Wang HS, Pan Z, Shi W, Brown BS, Wymore RS, Cohen IS, Dixon JE, tation in the guinea pig hippocampus. Exp Brain Res
McKinnon D (1998) KCNQ2 and KCNQ3 potassium channel 51:128134.
subunits: molecular correlates of the M-channel. Science 282: Yeckel MF, Kapur A, Johnston D (1999) Multiple forms of LTP in hip-
18901893. pocampal CA3 neurons use a common postsynaptic mecha-
Watkins JC, Evans RH (1981) Excitatory amino acid transmitters. nism. Nat Neurosci 2:625633.
Annu Rev Pharmacol Toxicol 21:165204. Yokoi M, Kobayashi K, Manabe T, Takahashi T, Sakaguchi I, Katsuura
Weisskopf MG, Zalutsky RA, Nicoll RA (1993) The opioid peptide G, Shigemoto R, Ohishi H, Nomura S, Nakamura K, Nakao K,
dynorphin mediates heterosynaptic depression of hippocampal Katsuki M, Nakanishi S (1996) Impairment of hippocampal
mossy bre synapses and modulates long-term potentiation. mossy fiber LTD in mice lacking mGluR2. Science 273:
Nature 362:423427. 645647.
Wenthold RJ, Petralia RS, Blahos J, II, Niedzielski AS (1996) Evidence Yoon KW, Rothman SM (1991) Adenosine inhibits excitatory but not
for multiple AMPA receptor complexes in hippocampal CA1/ inhibitory synaptic transmission in the hippocampus. J Neurosci
CA2 neurons. J Neurosci 16:19821989. 11:13751380.
Westrum LE, Blackstad TW (1962) An electron microscopic study of Yuste R, Majewska A, Cash SS, Denk W (1999) Mechanisms of cal-
the stratum radiatum of the rat hippocampus (regio superior cium inux into hippocampal spines: heterogeneity among
CA1) with particular emphasis on synaptology. J Comp Neurol spines, coincidence detection by NMDA receptors, and optical
119:281309. quantal analysis. J Neurosci 19:19761987.
Whiting PJ, Bonnert TP, McKernan RM, Farrar S, Le Bourdelles B, Yuste R, Majewska A, Holthoff K (2000) From form to function: cal-
Heavens RP, Smith DW, Hewson L, Rigby MR, Sirinathsinghji cium compartmentalization in dendritic spines. Nat Neurosci
DJ, Thompson SA, Wafford KA (1999) Molecular and func- 3:653659.
tional diversity of the expanding GABA-A receptor gene family. Zerangue N, Kavanaugh MP (1996) Flux coupling in a neuronal glu-
Ann NY Acad Sci 868:645653. tamate transporter. Nature 383:634637.
Wigstrom H, Gustafsson B (1986) Postsynaptic control of hip- Zhang JF, Ellinor PT, Aldrich RW, Tsien RW (1996) Multiple struc-
pocampal long-term potentiation. J Physiol 81:228236. tural elements in voltage-dependent Ca2 channels support
Wilson RI, Nicoll RA (2001) Endogenous cannabinoids mediate ret- their inhibition by G proteins. Neuron 17:9911003.
rograde signalling at hippocampal synapses. Nature 410: Zhou Q, Petersen CC, Nicoll RA (2000) Effects of reduced vesicular
588592. lling on synaptic transmission in rat hippocampal neurones.
Wilson RI, Kunos G, Nicoll RA (2001) Presynaptic specicity of J Physiol 525(Pt 1):195206.
endocannabinoid signaling in the hippocampus. Neuron Zucker RS, Regehr WG (2002) Short-term synaptic plasticity. Annu
31:453462. Rev Physiol 64:355405.
Wroblewska B, Wroblewski JT, Saab OH, Neale JH (1993) N- Zucker RS, Kullmann DM, Bennett M (1998) Release of neurotrans-
Acetylaspartylglutamate inhibits forskolin-stimulated cyclic mitters. In: Fundamental neuroscience (Zigmond MJ, Bloom FE,
AMP levels via a metabotropic glutamate receptor in cultured Landis SC, Roberts JL, Squire LR, eds), pp 155192. San Diego:
cerebellar granule cells. J Neurochem 61:943948. Academic Press.
This page intentionally left blank
7 Pavel Osten, William Wisden, and Rolf Sprengel

Molecular Mechanisms of Synaptic Function

in the Hippocampus: Neurotransmitter
Exocytosis and Glutamatergic, GABAergic,
and Cholinergic Transmission

modulators such as serotonin, steroids, and endocannabinoids

7.1 Overview play important roles in the hippocampus, as does electrical
transmission through gap junctions.
In the hippocampus, excitatory (also termed principal or In this chapter, we primarily restrict ourselves to a descrip-
pyramidal) neurons, the basic working units of the mam- tion of the molecular biology of excitatory and inhibitory
malian central nervous system (CNS), form a highly organ- transmission. We start by describing the basis of all signal
ized three-layer circuit. Excitatory input from the entorhinal transmission processes, the release of the signaling transmitter
cortex enters the hippocampus largely via dentate gyrus gran- from the presynaptic terminal. Next we give details of the key
ular cells, which in turn connect through the mossy ber glutamate and GABA receptors responsible for the nuts and
synapses to the CA3 pyramidal neurons. CA3 pyramids bolts of hippocampal chemical transmission. We end by dis-
process the input via a dense network of recurrent connec- cussingas an example of modulatory transmissionhow
tions and send their output through the Schaffer collaterals acetylcholine muscarinic and nicotinic receptors inuence
synapsing onto the CA1 pyramidal cells. Finally, CA1 pyra- information ow in the hippocampal cell types. To fully inte-
mids send the hippocampus-processed information via the grate our chapter into the overall theme of the book, we focus
subiculum back to the cortex. Within this circuit, fast excita- in detail on hippocampal-specic features of synaptic trans-
tory synaptic transmission is mediated by inward excitatory mission, such as the hippocampal distribution of the individ-
postsynaptic currents (EPSCs) owing through the ionotropic ual receptor types. In addition, we discuss the growing
-amino-3-hydroxy-5-methyl-4-isoxazolepropionate number of mouse models aimed at genetic analysis of presy-
(AMPA) and, to a lesser extent, kainate types of glutamate naptic or postsynaptic functions, which are typically assayed
receptors. Slower hippocampal EPSCs are the result of activa- for behavioral phenotype in a battery of hippocampus-
tion of N-methyl-D-aspartate (NMDA) receptors and certain dependent tasks. As these mouse models are becoming
subtypes of metabotropic glutamate receptors. increasingly more rened, particularly with improved spa-
The activity of hippocampal principal neurons is tightly tiotemporal control over the onset of the genetic manipula-
controlled and synchronized by an inhibitory system com- tion, this line of work is beginning to uncover previously
prised of -aminobutyric acid (GABA)ergic interneurons unsuspected differences in the molecular and cellular mecha-
interspersed throughout the hippocampus. Fast inhibitory nisms of various hippocampus-based functions, such as the
postsynaptic currents (IPSCs) are the result of ionotropic differences in synaptic transmission and plasticity underlying
GABAA receptor activation, and slow IPSCs are due to the acti- hippocampal working and spatial memory.
vation of metabotropic GABAB receptors. In addition, activa-
tion of some metabotropic glutamate receptors may also have
inhibitory rather than excitatory effects. Cholinergic projec-
tions in the hippocampus comprise a prominent modulatory 7.2 Neurotransmitter Exocytosis
system, where acetylcholine binds to muscarinic and nicotinic
receptors and produces either excitatory or inhibitory effects 7.2.1 Introduction: Proteins Involved
depending on the receptor type, the cell type that expresses the in Synaptic Release
receptor, and the synaptic location of the receptors in a given
cell. Finally, a set of neuropeptide receptors [e.g., receptors for Neurotransmitter release occurs at specialized sites of the
somatostatin, neuropeptide Y (NPY)] and receptors that bind presynaptic axon terminals, termed the active zones, and is

243
244 The Hippocampus Book

tightly linked to intraterminal inux of extracellular Ca2. which indeed have a decreased number of reserve vesicles at
The event itself is a culmination of numerous trafficking steps hippocampal synapses (Rosahl et al., 1993, 1995; Li et al.,
of the synaptic vesicle cycle that was introduced in general 1995; Takei et al., 1995). In addition, the mice show interest-
terms in Chapter 6. The temporal and spatial complexity of ing phenotypes with respect to synapsin I and II functions in
these processes is accomplished by the highly synchronized the hippocampus. First, synapsin I, II, or I/II knockout mice
collaboration of a large number of presynaptic proteins. On all have normal long-term synaptic plasticity as assessed by
the basis of their function during the synaptic vesicle cycle, measurements of long-term potentiation (LTP) at either
they can be grouped as proteins regulating (1) the reserve pool mossy ber to CA3 or CA3 to CA1 synapses. Thus maintain-
of vesicles, (2) mobilization of vesicles toward the active zone, ing a reserve pool of synaptic vesicles is not necessary for LTP
(3) docking and priming of vesicles at the active zone, (4) in the hippocampus. However, the mice show changes in
Ca2-triggered neurotransmitter release, and (5) vesicle endo- short-term plasticity. Synapsin I knockout mice show an
cytosis, recycling, and relling. increase in paired pulse facilitation (PPF) but no change in
posttetanic potentiation (PTP) at CA3 to CA1 synapses,
7.2.2 Reserve Pool of Synaptic Vesicles whereas mice lacking synapsin II and I/II show a loss of PTP
but no change of PPF. Even more interestingly, synapsin II
Synaptic vesicles are kept in two spatially and functionally dis- and I/II but not synapsin I knockout mice show impaired
tinct pools: the reserve pool and the readily releasable pool. hippocampus-dependent contextual learning (Silva et al.,
The reserve pool, approximately 200 to 500 nm away from the 1996). This suggests that a decit in short-term plasticity (i.e.,
active zone plasma membrane, provides a supply of vesicles in PTP) in the presence of normal LTP can cause an impair-
after depletion of the readily releasable pool. The equilibrium ment of hippocampus-dependent learning; this nding,
between the reserve and the releasable pool is maintained although perhaps surprising, is supported by a similar result
largely by abundant synaptic vesicle proteins termed from animals lacking the presynaptic protein RIM1 (see
synapsins, which tether the vesicles to the presynaptic Section 7.2.5).
cytoskeletal network made up of lamentous actin (F-actin),
microtubules and brain spectrin (for review see Hilker et al., 7.2.3 Synaptic Vesicle Docking and Priming at
1999; Doussau and Augustine, 2000). Synapsins comprise a the Active Zone: Role of the SNARE Complex,
family of three phosphoproteins, termed synapsin I, II, and Munc18, and Munc13
III, each of which undergoes alternative splicing in the vari-
able C-terminal part of the molecule, generating isoforms Synaptic vesicles, after their mobilization toward the active
termed a and b. Whereas the distribution of synapsin I and zone, undergo a number of trafficking steps before achieving
II is highly overlapping throughout the CNS (Sdhof et al., a state at which they are ready to fuse with the presynaptic
1989), differences in function between the two isoforms in the plasma membrane, the so called readily releasable state.
hippocampus were recently revealed in synapsin I and II Although the functions of numerous proteins involved in
knockout mice (see below). In contrast, synapsin III is largely these processes are now understood in some detail, a complete
restricted to mossy ber terminals of the granule cells in the unied model that would explain how synaptic vesicles
adult hippocampus (Pieribone et al., 2002). dock and become primed at the active zone is yet to be estab-
The function of synapsins is tightly regulated by protein lished. Perhaps the least-understood step during the vesicle
kinases: During repetitive synaptic release, activated protein mobilization concerns the initial docking (tethering) of
kinases phosphorylate synapsins on several residues, inducing synaptic vesicles at the active zone membrane. It seems likely,
their unbinding from synaptic vesicles and subsequent mobi- based on data from membrane fusion in many secretory path-
lization of the vesicles toward the active zone. Two kinases ways, that vesicle tethering is mediated by small guanosine
implicated in the phosphorylation of synapsin I are triphosphate (GTP)-binding proteins of the Rab family: A
Ca2/calmodulin-dependent protein kinase (CaM kinase) and vesicle-associated, GTP-bound Rab links the vesicle to a Rab
mitogen-activated protein kinase (MAPK). Interestingly, the effector protein (tether) that is present selectively at the target
two kinases regulate synapsin I function in response to differ- membrane. Several Rab proteins are associated with synaptic
ent rates of activity: A lower rate of presynaptic stimulation vesicles, the most abundant of which is Rab3A (other Rab3
activates phosphorylation of synapsin I by CaM kinase, isoforms are Rab3B, -C, and -D). Analysis of knockout mice,
whereas a higher rate is more selective for MAPK phosphory- however, revealed that Rab3A functions during neurotrans-
lation (Chi et al., 2003). Even though the functional signi- mitter release, after docking and priming (Geppert et al.,
cance of the two kinase pathways is presently not clear, these 1994a, 1997). (See Section 7.2.5 on the role of Rab3A in the
observations suggest that synapsins may be ne-tuned to dif- later stages of the synaptic vesicle cycle.) Furthermore,
ferent ring frequencies in vivo and may participate in regu- quadruple Rab3A/B/C/D knockout mice have normal gross
lating the mobilization of synaptic vesicles during different brain and synaptic morphology (the mice die at birth owing
patterns of activity. to respiratory failure), and cultured hippocampal neurons
Direct experimental support for the above proposed role of prepared from these mice have normal spontaneous and only
synapsins comes from mice lacking synapsin I and/or II, partially reduced evoked release (Schlter et al., 2004). Thus,
 Molecular Mechanisms of Synaptic Function 245

other proteins, perhaps yet another Rab isoform or a protein Syntaxin1 comprises a large cytoplasmic domain, one trans-
with a different function altogether, are sufficient for docking membrane domain, and a short extracellular C-terminus.
of synaptic vesicles at hippocampal synapses in the absence of Functionally, Syntaxin1 exists in three conformational states
Rab3. (Fig. 71) (reviewed in Brose et al., 2000; Rizo and Sdhof,
Past the initial docking step, proteins principally involved 2002). In the closed state, syntaxin1 binds to Munc18-1 (also
in the regulation of priming of synaptic vesicles at the active known as nSec1, a neuronal isoform of the sec1/Munc18 pro-
zone include the SNARE (SNAp REceptor) complex, tein family). Because the binding of syntaxin1 to Munc18-1
Sec1/Munc18 homologues (SM proteins), Munc13-1, and and to SNAP25 is mutually exclusive, it is postulated that
complexins (reviewed in Rizo and Sdhof, 2002). The role of Munc18-1 controls the pool of free syntaxin1 and inhibits for-
synaptotagmin 1 is discussed in the section relating directly to mation of the SNARE complex. However, surprisingly, genetic
Ca2-triggered release (see Section 7.2.4). The SNARE com- deletion of Munc18-1 in mice caused complete loss of neuro-
plex is a critical component of most if not all intracellular transmitter release despite an apparently normal number of
membrane fusion events. Assembly of the SNARE complex docked vesicles at the presynaptic membrane (Verhage et al.,
from proteins of the vesicular and target membranes (v- and 2000). This indicates that in addition to its restrictive function
t-SNAREs) is believed to bring the two membranes into close in binding to syntaxin1 Munc18-1 plays a more positive and
contact; this primed state is a prerequisite for initiation of yet to be identied role during late steps of the synaptic vesi-
membrane fusion (Fig. 71) (see Chapter 6). At the presynap- cle cycle prior to synaptic release. Strikingly, these mice are
tic terminals, the SNAREs involved in neurotransmitter born with essentially normal gross brain anatomy and mor-
release are a v-SNARE termed synaptobrevin (also VAMP) phologically dened synapses, demonstrating that SNARE-
and two t-SNAREs present at the active zone termed SNAP25 dependent neurotransmitter release is not required for brain
and syntaxin1; synaptobrevin and syntaxin are single trans- assembly or formation of synaptic contacts; the mice, how-
membrane proteins, and SNAP25 is a membrane-attached ever, die immediately after birth owing to complete paralysis
protein. The critical importance of the SNAREs for synaptic and exhibit widespread apoptosis in the brain.
functions was first demonstrated in experiments with In the next step, before formation of the SNARE complex,
clostridial neurotoxins, which cause site-specic cleavage of syntaxin1 undergoes conformational transition, from a
the individual SNARE proteins: A loss of SNARE function closed to an open state, which disinhibits the binding
abolished neurotransmitter release (reviewed in Rizo and of Munc18-1. Based on the following evidence, the main can-
Sdhof, 2002). These experiments also showed that in the didate for initiating this transition is Munc13 (reviewed in
absence of the SNARE assembly synaptic vesicles can accumu- Brose et al., 2000; Rizo and Sdhof, 2002). First, the carboxy-
late at the active zone in a similar way as at wild-type synapses, terminal region of Munc13-1 binds to syntaxin1 and displaces
indicating that SNAREs are not involved in synaptic vesicle the binding of Munc18-1 in vitro. Second, double knockout
docking. mice lacking two Munc13 isoforms (Munc13-1 and Munc13-
The molecule key to understanding the function of the 2) exhibit complete loss of neurotransmitter release at hip-
SNARE complex during synaptic release is the t-SNARE syn- pocampal synapses but no change in the total or docked
taxin1, a neuronal isoform of the syntaxin protein family. number of vesicles at synapses; this demonstrates that

Figure 71. Three conformations of syntaxin during assembly of from syntaxin by Munc13. In the open state, syntaxin can interact
the ternary SNARE complex. A. Syntaxin in a close conformation with SNAP25 and synaptobrevin from a docked vesicle. C. The
is bound with Munc18 via its N-terminal Habc (also termed resulting assembled state corresponds to syntaxin binding within
HAHBHC) domain (checkered region of the syntaxin molecule) and the ternary SNARE complex. The vesicle is now primed for synaptic
a part of its juxtamembrane H3 domain (dashed region). B. During release triggered by Ca2 entry into the terminal (see Fig. 72).
a switch from a closed to an open state, Munc18 is displaced (Source: Brose et al., 2000, with permission.)
246 The Hippocampus Book

Munc13-1 and 13-2 function after synaptic vesicle docking, During or closely following formation of the SNARE com-
during priming, and prior to synaptic release (Varoqueaux et plex, low-molecular-weight proteins termed complexins bind
al., 2002). Importantly, a constitutively open mutant of syn- to the core complex at 1:1 stoichiometry in a groove between
taxin1, which does not bind Munc18-1, rescues the loss of the synaptobrevin and syntaxin coil-coil domains. It has been
Unc-13 (homologue of Munc13) in Caenorhabditis elegans proposed that complexins stabilize the SNARE complex prior
(Richmond et al., 2001); these data provide further evidence to evoked neurotransmitter release. Mice lacking both com-
that the Munc-13/Unc-13 proteins function during a confor- plexin 1 and 2 isoforms show normal brain anatomy but die a
mational switch of syntaxin1 into the open state (see Section few hours after birth (Reim et al., 2001). Analysis of synaptic
7.2.5 for another role of Munc13-1 in binding with RIM1 and functions in hippocampal cultured neurons revealed normal
Rab3A). Other potential mechanisms that may modulate syn- spontaneous release but dramatically reduced probability of
taxin1 transition to the open state include syntaxin1 binding evoked synaptic release. Complexins thus appear to ne-tune
to tomosyn, which also displaces Munc18-1, and/or a direct the competence of the docked vesicles to respond to a cyto-
modulation of Munc18-1 by phosphorylation or by an inter- plasmic inux of Ca2.
action with other presynaptic proteins (reviewed in Rizo and
Sdhof, 2002). 7.2.4 Ca2-triggered Synaptic Release:
The third conformational state of syntaxin corresponds to Role of Synaptotagmin
its binding in the SNARE complex, initially forming a binary
complex with SNAP25, which is followed by assembly of the The rst step in the sequence of events underlying Ca2-trig-
ternary SNARE complex with SNAP25 and synaptobrevin gered exocytosis is the activation of voltage-gated Ca2 chan-
(Fig. 71). The complete zipped SNARE complex is formed nels of the N and P/Q types by action potential-driven
by a tight four-helix bundle formed among coil-coil domains depolarization of the nerve terminal membrane (for review
of all three SNAREs (one coil domain of syntaxin and synap- see Meir et al., 1999). For transmitter release to occur, the
tobrevin, and two of SNAP25). The orientation of the coil-coil intracellular Ca2 concentration must reach a threshold,
domains in the complex is in parallel, which brings the docked which is set by the sensitivity of a Ca2 sensor responsible for
vesicle in close contact with the active zone membrane, result- triggering the release. The bulk Ca2 concentration in the
ing in a primed readily releasable state. The unstable nature of presynaptic terminal reaches about 500 nM following an
the primed vesicles is reected by occasional spontaneous action potential. However, the Ca2 concentration required
miniature release events, electrophysiologically measured as for synaptic release is estimated to be as high as 10 M (from
miniature excitatory and inhibitory postsynaptic currents: measurements made at the large calyx of the Held synapse)
mEPSCs and mIPSCs, respectively; these events can thus be (Bollmann et al., 2000; Schneggenburger and Neher, 2000).
seen as failures of the mechanism(s) that normally prevent Because such a high Ca2 concentration is likely to be reached
Ca2-independent release at the synaptic terminals (see only in close proximity to the openings of the Ca2 channels,
Section 7.1.5 for the role of synaptotagmin in the initiation of it is important to maintain a close link between the channels
Ca2-triggered synaptic exocytosis). and proteins of the presynaptic release machinery. This is
Mice engineered to lack either synaptobrevin or SNAP-25 mediated, at least in part, by direct proteinprotein coupling
were generated to explore the SNARE complex function in of Ca2 channels to the SNARE complex as well as to synapto-
vivo (Schoch et al., 2001; Washbourne et al., 2002). Both geno- tagmin 1 and 2, vesicular proteins proposed to function as
types were born with normal brain anatomy but died imme- presynaptic Ca2 sensors (fora review on Ca2 channel inter-
diately after birth, likely owing to respiratory failure. actions with the active zone proteins, see Seagar et al., 1999).
Strikingly, although they showed drastically reduced (in the In addition, because neurotransmitter release at the CNS
case of synaptobrevin knockout mice) or completely abol- synapses is comprised of two kinetically distinct compo-
ished (SNAP-25 knockout) action potential-evoked neuro- nentsa major fast component followed by a smaller slower
transmitter release in hippocampal cultured neurons, componentit is likely that at least two distinct Ca2 sensors
spontaneous synaptic release was either only 10-fold are employed to regulate the event.
decreased (synaptobrevin knockout) or occurred at wild-type Synaptotagmins constitute an 11-gene family of Ca2-
frequency and amplitude (SNAP-25 knockout). These mice binding proteins, the most abundant of which in the brain is
thus conrm that the SNARE complex is not required for synaptotagmin 1. The molecular structure of synaptotagmin
synaptic vesicle docking, as previously indicated by the exper- consists of a small intravesicular domain, a single transmem-
iments with clostridial neurotoxins (see above). In addition, brane domain, and a large cytoplasmic C-terminal domain,
the persistence of spontaneous release indicates that the which contains two Ca2-binding C2-domains (C2A and C2B).
SNARE complex is not essential for fusion of docked vesicles The Ca2 affinity of the C2 domains of synaptotagmin is
at the active zone membrane. It remains to be determined if highly increased by binding to phospholipid membranes. In
the release in the absence of synaptobrevin or SNAP-25 is due the primed state, synaptotagmin C2 domains bind to the
to compensation by other cellular SNAREs or is mediated via SNARE complex and to phospholipids of the active zone
a truly SNARE-independent mechanism. plasma membrane; upon Ca2 entry, the C2 domains insert
 Molecular Mechanisms of Synaptic Function 247

Figure 72. Synaptotagmin role as a Ca2 sensor for synaptic (marked as white dots; three per C2A and two per C2B), which may
release. A. In the primed state, synaptotagmin C2 domains (C2A and further insert the C2 domains into the plasma membrane phospho-
C2B) bind with phospholipids of the active zone plasma membrane lipids. Such conformational change was proposed to act as the nal
and the SNARE complex (the SNARE components are shown as in trigger for synaptic exocytosis (Source: Sdhof, 2004, with permis-
Fig. 71). B. Upon Ca2 entry, synaptotagmin binds Ca2 ions sion.)

themselves farther into the plasma membrane, a step that may GDP/GTP exchange factor, and Rab3A-GTP rebinds with
act as a nal trigger for fusion between the synaptic vesicle synaptic vesicles.
and active zone membranes during synaptic exocytosis Rab3A interacts, in a GTP-dependent manner, with two
(reviewed in Sdhof, 2004) (Fig. 72). proteins proposed to function as its effectors, rabphilin and
Perhaps the best evidence for the role of synaptotagmin in RIM1 (reviewed in Sdhof, 2004). Both proteins have an N-
Ca2-induced synaptic release comes from synaptotagmin 1 terminal zinc nger domain (which mediates the binding with
knockout mice, which exhibit a severely depressed fast phase Rab3A) and two C-terminal Ca2-binding C2 domains.
and a normal slow (asynchronous) phase of Ca2-dependent Rabphilin, a cytoplasmic protein, is recruited to synaptic vesi-
synaptic release in cultured hippocampal neurons. Note that cles by its interaction with Rab3A-GTP. RIM1, which is stably
the mice die after birth, indicating that the fast Ca2-depend- associated with the active zone membrane, binds via its zinc
ent release is essential for life (Geppert et al., 1994b). As nger to Rab3A and Munc13-1, via its C2B domain to synap-
expected, presynaptic vesicle docking appears normal in these totagmin 1 and 2, and to liprins.
mice, with the readily releasable pool of synaptic vesicles, fre- Rab3A does not play an essential role in docking or prim-
quency of mEPSCs, and Ca2-independent neurotransmitter ing of synaptic vesicles as initially believed, as Rab3A knock-
release all comparable to that in wild-type terminals. In addi- out mice have a normal gross brain as well as synaptic
tion, Fernandez-Chacon et al. (2001) used a sophisticated morphology and normal spontaneous transmitter release
knock-in approach to introduce a single point mutation in the (Geppert et al., 1994a). Rather, the function of Rab3A seems
synaptotagmin 1 C2A domain, which decreased its Ca2 affin- to be to modulate synaptic release, with distinct roles at vari-
ity by 50%. This genetic manipulation resulted in a propor- ous hippocampal synapses. At the CA3 to CA1 synapses, loss
tional decrease of Ca2 sensitivity of neurotransmitter release of Rab3A results in altered short-term plasticity with
in cultured hippocampal neurons prepared from the mutant enhanced paired-pulse facilitation (Geppert et al., 1994a,
mice (Fernandez-Chacon et al., 2001). (See Section 7.2.5 for a 1997). In contrast, the mossy ber to CA3 synapse shows nor-
discussion of behavioral phenotype of synaptotagmin mutant mal short-term plasticity, but mossy ber LTP (mfLTP) is
mice.) Taken together, these ndings strongly indicate that abolished (Castillo et al., 1997a). Somewhat surprisingly,
synaptotagmin 1 functions after docking and priming, during knockout of rabphilin showed no phenotype, indicating that
the fast phase of Ca2-triggered release. rabphilin is not an essential Rab3A effector in the synaptic
vesicle cycle (Schlter et al., 1999). However, knockout of the
7.2.5 Ca2-triggered Synaptic Release: major brain RIM isoform, RIM1, showed a broader pheno-
Role of Rab3A and RIM1 type than that of Rab3A, indicating additional roles for
RIM1 in synaptic release. First, mfLTP is abolished, as in the
The abundant synaptic vesicle-associated protein Rab3A, a Rab3A knockout mice, suggesting that RIM1 is the main
member of the Rab family of low-molecular-weight GTP- effector of Rab3A at the mossy ber to CA3 synapse (Castillo
binding proteins, cycles between two functional states: a et al., 2002). Second, both the probability of evoked synaptic
synaptic vesicle-associated GTP-bound form and a cytoplas- release and short-term plasticity are altered, but LTP can still
mic GDP-bound form. Rab3A GTP hydrolysis, which occurs be induced at CA3 to CA1 synapses (Schoch et al., 2002;
during or after neurotransmitter exocytosis, releases Rab3A Calakos et al., 2004).
from its association with the vesicular membrane. Although the exact molecular mechanisms underlying the
Cytoplasmic Rab3A-GDP is then recognized by the Rab3- RIM1-based decits remain to be explained, recent behav-
248 The Hippocampus Book

ioral experiments showed that the synaptic functions that are 7.2.7 Synaptic Vesicle Endocytosis,
disturbed in the RIM1 knockout mice are crucial for hip- Recycling, and Relling
pocampus-based behavior. The mice showed normal coordi-
nation and anxiety-like behavior but were decient in two Currently, the model of recycling of the synaptic vesicular
forms of hippocampus-dependent associative learning and membrane and its integral proteins at the CNS synapses
memory: context-dependent fear conditioning and spatial includes two pathways: kiss-and-run and clathrin-mediated
learning in the Morris watermaze (Powell et al., 2004). In par- endocytosis.
allel experiments, the same authors found that mice lacking The kiss-and-run hypothesis proposes that the lumen of
Rab3A, which are selectively decient in mfLTP (see above), the synaptic vesicle briey opens and closes again, without the
showed normal behavior. Thus, mfLTP is not essential for vesicle membrane collapsing to the active zone plasma mem-
these hippocampus-dependent forms of learning. In addition, brane; in the next step, the vesicle is recaptured mostly intact.
mice with a synaptotagmin C2A point mutation, which Compelling evidence for kiss-and-run recycling comes from
exhibit a decrease in the probability of release at hippocampal uorescence imaging of single-vesicle release in hippocampal
excitatory synapses similar to that seen in RIM1 knockout primary cultures, documenting partial loss of the vesicular
mice (see Section 7.2.4), also showed normal behavior. The content and recovery of the same vesicle afterward (Aravanis
RIM1 knockout behavioral phenotype thus cannot be sim- et al., 2003; Gandhi and Stevens, 2003). However, many ques-
ply correlated with the decreased release probability at excita- tions about the molecular mechanisms that may regulate this
tory synapses. The observed behavioral decits in the RIM1 form of release remain unanswered; for example, it is not
knockout may result from the combined alterations in mossy known what prevents full collapse of the vesicle into the presy-
ber to CA3 transmission and CA3 to CA1 transmission. naptic membrane or how the vesicle is pinched off before it
Alternatively, short-term plasticity, even in the presence of can run again.
normal LTP, may play an important role in learning, as was The mechanisms of clathrin-mediated endocytosis of
already suggested with experiments using the synapsin II and synaptic vesicles, in contrast, are understood in much more
I/II knockout mice (Silva et al., 1996) (see Section 7.2.2). detail (reviewed in Slepnev and De Camilli, 2000). The general
outline of the distinct molecular steps include formation of
7.2.6 NSF-mediated Disassembly clathrin-coated invaginations (pits) followed by ssion of the
of the SNARE Complex coated pits to generate free clathrin-coated vesicles, rapid loss
of the clathrin coat, and reentry into the synaptic vesicle cycle.
The SNARE complex, assembled by the -helical domains of The formation of the clathrin-coated pits is initiated by bind-
t- and v-SNAREs during the steps of docking and priming, is ing of adaptor proteins AP-2 and AP-180 to synaptic vesicle
thermodynamically stable. To initiate recycling of SNARE proteins, such as synaptotagmin, that are undergoing sorting
proteins for the next round of synaptic vesicle cycle after at the plasma membrane after neurotransmitter release and to
synaptic exocytosis, the SNARE complex must be rst disas- phospholipids. The adapters then recruit the three-legged
sembled by a chaperon protein, the hexameric ATPase N- clathrin triskelia and synergistically induce clathrin polymer-
ethylmaleimide-sensitive fusion protein (NSF) (for review see ization, resulting in membrane invagination.
May et al., 2001). The NSF molecule consists of three In addition to adaptor proteins, a number of accessory
domains: an N-terminal domain responsible for binding with proteins are involved in the mechanisms of invagination of
the -SNAPSNARE complex (below) and two C-terminal the clathrin-coated pit and formation of a uniform-size vesi-
ATP-binding domains, termed D1 and D2. The D1 domain cle. Amphiphysin 1 and 2 are multifunctional adapters that
is essential for NSF ATPase activity, and the D2 domain is bind to clathrin, AP-2, and synaptojanin and dynamin, which
required for NSF hexameric assembly. The association are involved in uncoating and ssion of the vesicles (see
between NSF and the SNARE complex is not direct but medi- below). In addition to their adaptor role, amphiphysins may
ated by an adaptor protein termed -SNAP (soluble NSF- also participate in the formation of membrane curvature dur-
attachment protein). NSF binds to the -SNAPSNARE ing endocytosis via its N-terminal Bin/Amphiphysin/Rvs
complex only in the presence of ATP, forming the so-called 20 (BAR) domain; amphiphysins directly bind the membrane
S particle. The quick-freeze/deep-etch electron microscopic lipid bilayer and force membrane curving via BAR domains
image of the 20 S particle shows -SNAP and NSF wrapping dimerized into a banana-like shape. As with many other
as a sleeve along the rod-like -helical domains of the synaptic proteins, the role of amphiphysins in the CNS has
SNARE complex (Hanson et al., 1997). Binding of NSF to the been examined by mouse genetics. Amphiphysin 1 knockout
-SNAPSNARE complex stimulates NSF ATPase activity, mice showed drastically diminished levels of amphiphysin 2,
resulting in subsequent disassembly of the SNARE complex indicating that the proteins exist as functional heterodimers;
driven by NSF-mediated ATP hydrolysis. The monomeric and in vitro assays using lysates and cultured hippocampal
SNARE proteins syntaxin1, synaptobrevinm and SNAP25 are cells from the knockout animals revealed defects in the assem-
then sorted for the next round of vesicular cycling during bly of the synaptic endocytotic protein machinery as well as
endocytotic uptake of the components of the vesicular mem- decreased rates of activity-dependent endocytosis and recy-
brane (see below). cling of the vesicles (Di Paolo et al., 2002). Interestingly, the
 Molecular Mechanisms of Synaptic Function 249

mice have morphologically normal presynaptic terminals and a specic cell type determines the neurotransmitter that is
unaffected evoked glutamate release, but they are prone to loaded into vesicles and subsequently released at axon termi-
seizures and show cognitive decits in hippocampus-based nals.
spatial learning (Di Paolo et al., 2002). Thus the exact role of The vesicular monoamine transporter 2 (VMAT2) is
amphiphysins in clathrin-mediated endocytosis remains to be responsible for synaptic vesicle uptake of catecholamines,
established. serotonin, and histamine; and the vesicular acetylcholine
Another gene family of clathrin accessory proteins com- transporter (VAChT) mediates uptake of acetylcholine. The
prises the endophilins (endophilin 1, 2, and 3, of which two transporters are closely related, with the same protein
endophilin 1 is the major isoform in the brain). Endophilins topology consisting of 12 transmembrane domains and cyto-
share the BAR and SH3-domain structure with amphiphysins plasmic N- and C-terminal domains. The vesicular GABA
and, in addition, exhibit enzymatic lysophosphatidic acid acyl transporter (VGAT), also termed the vesicular inhibitory
transferase (LPAAT) activity. Endophilin 1 is proposed to pro- amino acid transporter (VIAAT), is responsible for loading of
mote negative membrane curvature directly during endocyto- neurotransmitters GABA and glycine. VGAT has a topology of
sis by LPAAT activity-induced change in the composition of 10 predicted transmembrane domains. Finally, vesicular
the plasma membrane and to act as an adaptor in recruiting uptake of glutamate is mediated by a protein family of the
synaptojanin and dynamin. vesicular glutamate transporters (VGLUT1-3). Interestingly,
The key molecule during the ssion of clathrin-coated the rst identied vesicular transporter for glutamate (termed
vesicles is the GTPase dynamin (for review see Schmid et al., VGLUT1) was originally cloned as a gene upregulated in
1998). GTP-bound dynamin, which is recruited to the cerebellar granule cells after treatment with NMDA; the gene
clathrin-coated membrane by its interaction with SH3 was termed Brain, Na-dependent Pi transporter1 (BNP1)
domains of amphiphysin, endophilin, and other accessory because of its 32% amino acid identity with the Na-depend-
proteins, forms rings around the stalk of the clathrin pit. It has ent Pi transporter from the kidney. However, 5 years later,
been proposed that dynamin-driven GTP hydrolysis induces BNP1 was localized selectively to glutamatergic synaptic vesi-
conformational change in the dynamin ring and pinches off cles and was shown to mediate vesicular glutamate uptake.
the coated vesicle by either restricting or elongating the stalk The fact that VGLUT1 expression can be regulated by activa-
membrane. tion of NMDA-type glutamate receptors suggests that the
It has been proposed that clathrin uncoating, a nal step levels of VGLUT1 may have a functional relevance for regula-
in clathrin-dependent endocytosis, is mediated by synapto- tion of synaptic transmission. At the molecular level, VGLUT1
janins (synaptojanin 1 and 2synaptojanin 1 is the main iso- is a protein with predicted six to eight transmembrane
form in the brain). Synaptojanins have polyphosphoinositide domains. VGLUT1 and VGLUT2 show largely complementary
phosphatase activity and degrade phoshatidylinositol-4,5- expression in the CNS, with hippocampal glutamatergic neu-
biphosphate, PI(4,5)P2, to phosphatidylinositol (for review rons expressing mainly the VGLUT1 transporter (Fremeau
see Slepnev and De Camilli, 2000). because PI(4,5)P2 is a pos- et al., 2004).
itive regulator of clathrin assembly, its degradation is believed
to be the mechanism for vesicle uncoating after dynamin-
mediated ssion. Mice lacking synaptojanin 1 die shortly after
birth (indicating a crucial role of the gene for life) and show 7.3 Glutamate Receptors: Structure, Function,
increased levels of PI(4,5)P2 in cortical neurons, accumulation and Hippocampal Distribution
of clathrin-coated vesicles in nerve terminals, and depression
of hippocampal Schaffer collateral CA1 transmission, rather 7.3.1 Introduction: Ionotropic
than LTP, after high-frequency stimulation (Cremona et al., and Metabotropic Receptors
1999).
Finally, before the vesicles are re-sorted back to the active Glutamatergic synaptic transmission and, in particular,
zone membrane or to the reserve pool, they must be relled synaptic plasticity play a central role in hippocampus-
with a corresponding neurotransmitter. The energy necessary dependent learning and memory. Activation of the ionotropic
for uptake of a specic neurotransmitter is provided by an glutamate receptors provides an essential mechanism required
electrochemical gradient with positively charged, acidic vesi- for the induction of at least some forms of Hebbian synaptic
cle lumen; this gradient is generated by an ATP-driven proton plasticity; the metabotropic receptors play more modulatory
pump termed vacuolar H-ATPase (V-ATPase) (reviewed in roles in these processes. After a brief introduction to the
Nishi and Forgac, 2002). The uptake is mediated by vesicular nomenclature of the receptors (see Fig. 73 for an overview of
neurotransmitter transporters (VNTs), which pump neuro- ligand-gated receptor types), we discuss the molecular princi-
transmitter in exchange for protons. There are three types of ples that enable the receptors to mediate their functions, the
VNT in the brain: vesicular amine transporters, vesicular distribution of receptors in hippocampal neurons, and the
inhibitory amino acid transporters, and vesicular excitatory mechanisms of the transport of receptors to excitatory
amino acid transporters (reviewed in Eiden, 2000; Torres et synapses, with a particular focus on how glutamate receptor
al., 2003; Fremeau et al., 2004). The type of VNT expressed in trafficking may regulate synaptic plasticity. Finally, we con-
250 The Hippocampus Book

Figure 73. Molecular organization of neurotransmitter receptors. served regions between subunits of a receptor family have a gray
From top to down: Receptors are grouped into NMDA, AMPA, and background. (4) Membrane topology of the receptor subunits.
kainate receptor ion channels (iGluRs), GABAA receptor ion chan- Membrane spanning alpha helixes are numbered and depicted as
nels (GABAARs), and GABAB, mGlu, and mACh receptors barrels; the cell membrane areas are in gray. (5) The subunit com-
(GABABRs, mGluRs, mAChRs). From left to right: (1) Receptor position of the receptors is given as a formula. For example,
nomenclature. (2) Individual subunit nomenclature. (3) Primary NMDAR  2  [NR1/NR2(3)] NMDA receptors are tetramers
structure of the subunits with indicated amino-termini (N) and consisting of two NR1 and two NR2 or NR3 subunits. (6) The
carboxy-termini (C). The N-terminal signal peptides and the posi- oligomeric receptor structure shows the formation of the receptors
tions of membrane domains are indicated in dark gray and black from the individual subunits; in the cell membrane they are given as
boxes, respectively; the positions of alternative spliced exons are gray background.
indicated in striped boxes and RNA-edited sites in open circles; con-

sider how genetically modied mice can be models of gluta- intracellular and carries binding sites for several proteins reg-
mate receptor function in vivo. ulating trafficking and signaling of the receptors at synapses.
Ionotropic glutamate receptors (iGluRs) are ligand-gated The expression of most iGluR subunits in the hippocam-
ion channels that mediate most of the excitatory neurotrans- pus increases with development and reaches peak levels
mission in the hippocampus (and elsewhere in the brain). approximately 2 weeks after birth (Fig. 74). This develop-
Traditionally, iGluRs are classied into NMDA and non- mental prole closely resembles that of the  subunit of
NMDA (AMPA and kainate) receptors. The receptors are het- CaMKII, a kinase that critically contributes to the conversion
eromeric assemblies of four subunits, with each distinct of glutamatergic synaptic transmission into intracellular sig-
subunit encoded by its own gene. The molecular naling during synaptic plasticity and learning. The iGluRs and
structurefunction complexity is further extended by alterna- -CaMKII can be viewed as the molecular core of hippocam-
tive splicing of the primary transcripts and by recoding pal synaptic plasticity and learning.
through posttranscriptional RNA editing. The receptor sub- In addition to the direct feedforward excitatory role medi-
units are glycosylated membrane-spanning polypeptides with ated by activation of the iGluRs, glutamate plays modulatory
an average length of 900 amino acids. According to our cur- roles in regulating its own release as well as GABA release and
rent understanding, the subunit segments M1, M3, and M4 neuronal excitability in the hippocampus. This is achieved by
are membrane-spanning; the M2 segment, which does not activation of the kainate and metabotropic glutamate recep-
traverse the membrane bilayer, forms a pore-loop region, tors (mGluRs). mGluRs are G protein-coupled receptors,
which determines the ion selectivity of the channel. The com- which exhibit excitatory or inhibitory effects depending on
plete channel pore is built by amino acid residues of M1 and the receptor subtype and the cell context: EPSPs are often
M3 together with the M2 loop; it is permeable to monovalent reduced by presynaptic mGluR activation and enhanced by
and, in some instances, divalent cations. The extracellular lig- activation of postsynaptic mGluRs. Localization of mGluRs is
and-binding site is formed by the S1 region of the N-terminal often outside of pre- and postsynaptic sites, which suggests
domain and the S2 region of the extracellular loop between that optimal activation of mGluRs takes place during intense
M3 and M4. The carboxy-terminal domain of each subunit is synaptic activity resulting in neurotransmitter spillover.
 Molecular Mechanisms of Synaptic Function 251

Figure 74. AMPA receptor subunit expression in mouse hip- sion, higher magnication inserts demonstrate strong expression
pocampus. A. Immunoblots of mouse hippocampal extracts at in interneurons in the pyramidal cell layer. (Source: Modied from
postnatal day 2 (P2), P7, P14, P28, and P42 probed with antibodies Jensen et al., 2003b, with permission). C. GluR3 expression was
against ionotropic glutamate receptor subunits and the -subunit visualized by an anti-GluR2/3 antibody in the hippocampus of
of CaMKII (indicated on the right side of the panel). Anti- -actin adult GluR2 knockout mice (GluR2 KO, left panel); the right
staining was used for normalization of the amount of protein panel shows staining with the same antibody in wt hippo-
loaded (bottom lane). B. Expression of the GluR1, 2, and 4 subunits campus; the GluR2/3 signal represents mainly the GluR2
in the mouse hippocampus at P14 and P42 visualized by immuno- subunit (compare to GluR2-specic staining in the GluR2
histochemistry with subunit-specic antibodies. For GluR4 expres- panel, second from the top).

7.3.2 AMPA Receptors channels have low Ca2 permeability and a linear I/V function
relating evoked current to membrane potential. This electro-
The non-NMDA ionotropic receptors consist of two sub- physiological prole is determined by a single amino acid
groups: AMPA and kainate receptors. In the hippocampus, residue of the GluR2 M2 pore-loop segment. Whereas GluR1,
AMPA receptors are responsible for general purpose trans- 3, and 4 subunits contain glutamine (Q) in this position, the
mission at all excitatory synapses. The properties of AMPA GluR2 primary transcript undergoes selective RNA editing
receptors allow high temporal precision and short latency of that changes the glutamine into an arginine (R)so-called
action potential initiation. Q/R site editing (for review see Seeburg et al., 1998). In the
hippocampus, the mature GluR2 transcript is 100% edited,
Posttranscriptional Modications and the Q-to-R coding change is essential for normal brain
of AMPA Receptor Subunits function (see Section 7.4.3 for consequences of disrupted Q/R
editing in mice). In addition, GluR2, 3, and 4 transcripts are
AMPA receptors are heterotetramers of GluR1 to GluR4 (or edited at an R/G site (the glycine codon replacing genomically
GluR-A to -D) subunits, all of which are expressed in the hip- encoded arginine), which is located just before the ip/op
pocampus. Of the four subunits, GluR2 plays the most critical exons (see below). R/G editing is somewhat less efficient,
role in determining ion channel properties: GluR2-containing resulting in the editing of approximately 80% to 90% of
252 The Hippocampus Book

GluR2, 3, and 4, except for GluR4 ip, which is only 50% GluR3). The prevalent splice of the GluR2 transcript codes for
edited. Editing at the R/G site was shown to reduce the ampli- a short C-terminal tail (50 amino acids); the 68 amino acid
tude of and accelerate recovery from desensitization for long splice form, termed GluR2long, is expressed at signi-
recombinant GluR3 and GluR4; the physiological role for this cantly lower levels but may be functionally important in the
AMPA receptor modication, however, is still unclear. young postnatal hippocampus, where it mediates a juvenile
In addition to RNA editing, a second type of posttran- GluR1-independent form of LTP (see Section 7.3.1). The
scriptional modicationalternative splicingmodulates GluR4 subunit is expressed in the hippocampus only in its
the properties of AMPA receptor channels in the hippocam- long form, with a 68-amino-acid residue C-terminal domain.
pus at distinct synapses (for review, see Dingledine et al., The role of the AMPA receptor subunit C-terminal peptides in
1999) (see below for hippocampal distribution of different mediating subunit-specic interactions with proteins regulat-
subunit splice forms). All AMPA receptor primary transcripts ing intracellular trafficking of the receptors is discussed fur-
undergo alternative splicing of two exons encoding a 38- ther in Section 7.4.
amino-acid segment of the extracellular loop between M3 and
M4, just before the start of the M4 domain. Each subunit thus Hippocampal Distribution
exists as either a ip or a op variant, depending on which of AMPA Receptor Subunits
exon is retained. The ip/op domain inuences receptor
desensitization: Flip forms desensitize with slower kinetics AMPA receptor subunit distribution varies significantly
compared to op forms. The GluR2 and GluR4 primary tran- among hippocampal cell types (Fig. 74), resulting in differ-
scripts undergo, in addition to ip/op splicing, alternative ent receptor properties between, for example, excitatory and
splicing of their C-terminal domain-coding sequences. The inhibitory neurons. Hippocampal principal cells express
AMPA receptor subunit C-terminal domains are grouped into mainly GluR1 and GluR2, whereas GluR3 is present at lower
long (~70 to 80 amino acids; e.g., 81 aa in the rat GluR1 levels and GluR4 only in embryonic and early postnatal prin-
sequence) and short (~50 amino acids; e.g., 50 aa in rat cipal cells (see mRNA in situ detection in Keinanen et al.,

Figure 75. Expression of AMPA receptor-subunit splice variants in the mouse hippocampus, visualized by in situ hybridization with
in rodent hippocampus. A. In situ hybridization with specic GluR2long-specic oligonucleotides at P7, P15, and mice older than
oligonucleotides detects the mRNAs of the ip and op P42 (P>42). For an approximate comparison of mRNA expression
versions of each AMPA receptor subunit in rat hippocampus at of both GluR2 isoforms, GluR2short (marked as GluR2) in situ is
P15. (Source: Modied from Monyer et al., 1991, with permission). shown on the right.
B. Developmentally regulated expression of the GluR2long subunit
 Molecular Mechanisms of Synaptic Function 253

1990; protein detection in Petralia and Wenthold, 1992). The are selectively distributed in the proximal versus the distal
receptor assemblies are predominantly GluR1/2 channels, dendritic segments of hippocampal CA1 pyramidal neurons,
with some GluR2/3 and GluR1 homomeric channels. The fact with an increasing density toward more distal localization
that most AMPA receptor assemblies in the principal cells (Magee and Cook, 2000). This distance-dependent scaling of
contain the edited GluR2 subunit guarantees that they con- AMPA receptor distribution provides a mechanism that may
duct AMPA currents with low Ca2 permeability. (Note that help normalize the amplitudes of distal versus proximal
GluR1 does not appear to form channel assemblies with synaptic inputs onto the same cell.
GluR3 in the principal cells, even though the GluR1/GluR3
channels are efficiently formed from recombinant subunits in 7.3.3 NMDA Receptors
heterologous expression systems; the mechanism of this selec-
tive AMPA receptor assembly is not understood.) NMDA receptors differ from AMPA receptors in several
Both ip and op splice versions of GluR1, 2, and 3 are important properties, including voltage-dependent Mg2
expressed in CA1 pyramidal cells (Fig. 75). In contrast, CA3 block of the channel pore, high permeability to Ca2 ions,
pyramidal neurons synthesize only the ip versions of GluR1, slower activation kinetics, longer open time, and higher affin-
2, and 3. Dentate granule cells express GluR1, 2, and 3 ip and ity for glutamate (reviewed in Dingledine et al., 1999). At rest-
op forms, with op > ip forms (Sommer et al., 1990; ing and slightly depolarized membrane potentials, the ion
Monyer et al., 1991) (Fig. 75). In addition, the ip and op channel is blocked by the binding of extracellular Mg2 in the
splice variants are also differently regulated during the devel- ion pore. However, this Mg2 block is released when the post-
opment. The op versions are expressed at low levels prior to synaptic membrane is sufficiently depolarized (for example,
postnatal day 8 (P8), and their characteristic high expression due to strong activation of AMPA receptors). This unique fea-
in CA1 pyramidal cells becomes apparent only during the sec- ture gives NMDA receptors the ability to act as coincidence
ond postnatal week, in parallel with synaptic maturation. In detectors for presynaptic activity (glutamate release) and
contrast, ip mRNA levels are already substantial in pyrami- postsynaptic activity (sufficient depolarization of the postsy-
dal cells at birth (Monyer et al., 1991). Hippocampal pyrami- naptic membrane). The subsequent postsynaptic inux of
dal cells also express both C-terminal splice forms GluR2 and Ca2 transduces the activation of NMDA receptors into Ca2-
GluR2long. GluR2long is expressed as early as the embryonic dependent intracellular signaling. These integrative and ion
stage and reaches peak expression between P7 and P15. channel properties of NMDA receptor function underlie the
However, GluR2long is considerably less abundant than GluR2 induction of NMDA receptor-dependent synaptic plasticity, a
(at the protein level, the GluR2long/GluR2 ratio is 1:5 in young process involved in many types of learning and memory for-
hippocampus and 1:20 in adult hippocampus) (Kolleker et al., mation (see Section 7.5.2 for genetic dissection of NMDA
2003). receptor-dependent plasticity in hippocampal learning in
Hippocampal interneurons express mainly GluR1 and mice and Chapter 10 for a full account of synaptic plasticity in
GluR4 subunits. The fact that interneurons lack signicant the hippocampus).
levels of GluR2 means that their synaptic AMPA currents Seven genes encode the family of NMDA receptor subunits
show voltage dependence with high permeability to Ca2 ions termed NR1; NR2A, 2B, 2C, 2D; and NR3A and 3B. The NR1
(Geiger et al., 1995). Both GluR1 and GluR4 are expressed and NR3 subunits contain a binding site for glycine; and NR2
mainly as the op splice form in hippocampal interneurons subunits bind glutamate. As demonstrated in mice with
(Monyer et al., 1991). genetic deletion of the NR1 gene, NR1 is the essential subunit
In addition to the cell type-specic expression patterns, of all NMDA receptor channels; and in its absence, no func-
there is now evidence that receptors with different subunit tional NMDA receptors are formed in the brain (Forrest et al.,
compositions are selectively targeted to distinct subcellular 1994). Native tetrameric NMDA receptor assemblies are prob-
locations in individual neurons; this could potentially allow ably built of two NR1 and two NR2 or NR3 subunits.
selective tuning of postsynaptic responses to glutamate release Depending on the presence of the NR2 and NR3 subunits, the
at different anatomical inputs projecting onto a particular cell. NMDA receptors can be subdivided into the NR2A subtype
In the hippocampus, stratum lucidum interneurons express (tetramer of two NR1 plus two NR2A), NR2B subtype (two
both Ca2-permeable (GluR2-lacking) and Ca2-imperme- NR1 plus two NR2B), and so on. The NR2 subunits determine
able (GluR2-containing) AMPA receptors at synapses receiv- the electrophysiological and pharmacological prole of the
ing mossy ber input, whereas synapses receiving input from channels, affecting its open time, conductance, and Mg2 sen-
CA3 pyramidal neurons express selectively Ca2-imperme- sitivity. The NR2B receptors stay open longer, providing a
able AMPA receptors (Toth and McBain, 1998). Functionally, larger time window for coincidence detection than the NR2A
whereas induction of synaptic plasticity (LTD) requires an ele- receptor type. Compared to the NR2A or NR2B subtypes, the
vation of postsynaptic Ca2 at both synapse types, the require- NR2D-type receptors show slow activation and deactivation
ment for NMDA receptor activation is restricted to the kinetics (in the range of seconds rather than hundreds of mil-
Ca2-impermeable synapses (Laezza et al., 1999; Lei and liseconds) as well as lower sensitivity to voltage-dependent
McBain, 2002). Mg2 block. Similarly, NR3 subunits determine that NR1/NR3
In addition to synapse-specic targeting, AMPA receptors assemblies function as glycine-gated ion channels (Chatterton
254 The Hippocampus Book

et al., 2002). With the exception of the NR3B subtype, all at the mbrialCA3 synapses (Ito et al., 1997). The NR2B-
NMDA receptor subtypes are expressed in the hippocampus; type receptors are also reported to be selectively targeted to
the NR2A and NR2B subtypes are the major hippocampal the apical versus the basal dendrites of CA1 neurons of the
forms. mouse hippocampus at synapses receiving either ipsilateral or
contralateral CA3 inputs (Kawakami et al., 2003). Finally, the
Hippocampal Distribution NR2A- and NR2B-type receptors are distinctly localized in the
of NR1 and NR1 Splice Variants same synapse, with the NR2B type placed more extrasynapti-
cally and the NR2A type restricted to the center of gluta-
The NR1 subunit is expressed in dentate granule cells, in all matergic synapses in hippocampal principal neurons (Tovar
pyramidal neurons, and in many interneurons (Laurie and and Westbrook, 1999). The distinct expression proles of the
Seeburg, 1994; Monyer et al., 1994). Alternative splicing gen- two subunits thus provide a mechanism for expression of dif-
erates eight splice forms of the NR1 subunit. First, there are ferent types of NMDA receptor-dependent synaptic plasticity
NR1 splice forms that either lack (NR1a) or contain (NR1b) at different synapses and developmental stages in the hip-
exon 5 of the NR1 gene, which encodes the so-called N1 cas- pocampus (see Section 7.5.2 for NR2 mutant mice).
sette; this cassette inuences gating kinetics and is a part of the The NR2C receptor type is not signicantly expressed in
extracellular N-terminal domain of the receptor. A second the hippocampus. NR2D expression is restricted to hip-
group of NR1 forms is generated by alternative splicing of pocampal GAD67-, parvalbumin-, and somatostatin-positive
the last two carboxyl-terminal exons, 21 and 22. Exon 21 interneurons in the stratum radiatum of the CA1 and CA3
encodes the C-terminal domain C1 and exon 22 the C-termi- regions (Standaert et al., 1996). Functionally, as noted above,
nal domains C2 and C2. In the rst set of splice forms, exon the NR1/NR2D receptors show slower activation and deacti-
21 is excluded and C2 or C2 are used for coding the NR1 C- vation kinetics (in the range of seconds rather than hundreds
terminal domain (NR1-2 and NR1-4). In the second set, exon of milliseconds) and lower sensitivity to voltage-dependent
21 is present, which leads to C1-C2 and C1-C2 carboxy-ter- Mg2 block than do the NR1/NR2A or NR1/NR2B receptors
mini (NR1-1 and NR1-3). The eight NR1 isoforms are then (Monyer et al., 1994). Cells expressing the NR1/NR2D recep-
termed NR1a-1, NR1a-2, NR1a-3, NR1a-4, NR1b-1, NR1b-2, tors would thus be expected to conduct NMDA currents
NR1b-3, and NR1b-4. without the need for strong coincident depolarization.
The NR1 splice forms are differentially expressed in the However, hippocampal interneurons do not show the
hippocampus (Laurie and Seeburg, 1994) (Fig. 76). The expected NR2D-type channel characteristics, possibly because
NR1a, NR1-1, and NR1-2 splice forms are prominently NR2B has dominance over channel properties in
expressed in all pyramidal cell layers of the hippocampus NR1/NR2B/NR2D assemblies (cf. the NMDA receptor
throughout development. In general, NR1-4 and NR1b are responses of NR1-, NR2B-, and NR2D-expressing cholinergic
less expressed but show more prominent levels in CA3 pyram- interneurons in the caudate-putamen) (Gtz et al., 1997).
idal cells. NR1-3 expression is not detectable in the embryonic As with NR2D, the NR3A subunit is expressed in the hip-
hippocampus and remains low in adult rats. The proposed pocampus during embryogenesis and early postnatally,
roles of the various NR1 C-terminal splice forms in NMDA mainly in area CA1 and the subiculum (Ciabarra et al., 1995;
receptor trafficking and function, through mediating splice Sucher et al., 1995). In the adult hippocampus, NR3A mRNA
form-specic proteinprotein interactions, are discussed in remains expressed in CA1 pyramidal cells, possibly forming
Section 7.4.2. NR1/NR3A glycine-gated excitatory cation channels. NR3B
subtype receptors are not expressed in the hippocampus.
Hippocampal Distribution of NR2 Subunits
7.3.4 Kainate Receptors
Functionally, NR2A and NR2B receptor subtypes differ in
their channel kinetics. As already discussed, NR2B receptors Kainate receptors are tetrameric assemblies composed of
stay open longer, providing a larger time window for coinci- GluR5, -6, -7 and KA1 and KA2 subunits, depending on which
dence detection compared to the NR2A receptor type. Both genes are expressed in a given cell type (Wisden et al., 2000).
NR2A and NR2B are prominently expressed in all cell layers Recombinant GluR5, -6, and -7 subunits form functional
of the hippocampus (Fig. 76). The NR2A/NR2B expression homomeric as well as heteromeric channels; but most native
ratio increases strongly during postnatal development, receptors probably coassemble from both GluR and KA sub-
indicating that the NR2B-type NMDA receptors are more units.
important at the early postnatal stage, whereas the NR2A Characterization of native kainate currents in the hip-
receptors play a major role at mature hippocampal synapses pocampus has been made possible by using an antagonist
(Monyer et al., 1994). In addition to the developmental differ- (GYKI 53655), which selectively inhibits AMPA but not
ences, NR2A and NR2B receptors are also selectively targeted kainate receptors. With this pharmacological approach,
to different synapses in the CA1 and CA3 principal cells. In kainate receptor-mediated currents were identied in the hip-
CA3 cells, the NR2A receptors are preferentially located at pocampus at the mossy ber synapses onto CA3 principal
commissural/associational synapses and the NR2B receptors cells and the Schaffer collateral projections onto CA1
 Figure 76. Expression of NMDA receptor subunits in rat hippocampus. C. Expression of kainate receptor subunit mRNAs in the adult rat hippocam-
A. In situ hybridization with specic oligonucleotides detects the mRNAs of pus. (Sources: A: Modied from Monyer et al., 1994; B: modied from Laurie
NR1 and NR2A-D subunits. B. Expression of NR1 splice variant mRNAs and Seeburg, 1994; C: modied from Wisden and Seeburg, 1993, all with
visualized by in situ hybridization with splice form-specic oligonucleotides. permission.)

255
256 The Hippocampus Book

interneurons (Castillo et al., 1997b; Vignes and Collingridge, GluR5 and 80% of GluR6 are edited at the Q/R site. In addi-
1997; Cossart et al., 1998; Frerking et al., 1998). Further char- tion, GluR6 undergoes editing at the I/V and Y/C sites in the
acterization of kainate receptor function was made possible by M1 segment. Editing efficiency at these site is variable, giving
the development of knockout mice for the receptor subunit a rise to a high number of functional variants (reviewed in
genes. At the mossy berCA3 pyramidal cell synapse, kainate Seeburg et al., 1998).
receptors participate in multiple aspects of transmission both
pre- and postsynaptically. GluR6/KA2 receptors provide the Hippocampal Distribution
postsynaptic response; GluR6/KA2/KA1 receptors contribute of Kainate Receptor Subunits
to presynaptic modulation (Contractor et al., 2003) (see more
below). For the presynaptic component, applying low concen- In hippocampal principal cells, the KA1 gene is mainly
trations of kainate facilitates glutamate release; high concen- expressed in CA3 cells and dentate granule cells, with
trations depress release because of a depolarizing block of markedly weaker expression in CA1 pyramidal cells (Werner
axonal conduction. et al., 1991; Wisden and Seeburg, 1993) (Fig. 76). At the elec-
Kainate receptor currents constitute only about 10% of the tron microscopic level, KA1 immunoreactivity was found on
total peak current generated by AMPA receptors on hip- presynaptic mossy ber boutons and, to a lesser degree, in
pocampal pyramidal cells. However, the slow deactivation postsynaptic spines (Darstein et al., 2003). KA2 mRNA is
kinetics of kainate receptor currents means that during a abundant in both CA1 and CA3 pyramidal cells and in the
mossy ber composite EPSC a signicant proportion of the dentate granule cells (Bahn et al., 1994). The GluR6 subunit is
total charge transfers through kainate receptors. Kainate moderately expressed in all CA pyramidal cells and in the den-
receptors thus contribute to glutamate transmission by tate granule cells, with expression in CA3 higher than in CA1
enhancing as well as extending the postsynaptic membrane (Egebjerg et al., 1991; Wisden and Seeburg, 1993). Using a
depolarization. In addition, repetitive stimulation of the GluR7 knockout mouse and a GluR6/7-selective antibody,
mossy bers markedly increases kainate receptor currents; GluR6 immunoreactivity was strong in the stratum lucidum
there is evidence that kainate receptors on mossy ber (suggesting mossy bers) and the CA3 pyramidal cell layer
terminals positively modulate synaptic release at higher ring (Darstein et al., 2003). GluR7 mRNA is present in dentate
frequencies of dentate granule cells (reviewed in Lerma, granule cells but absent from CA pyramidal cells (Bettler et
2003). al., 1992). Using a GluR6 knockout mouse and the GluR6/7-
selective antibody showed that GluR7 immunoreactivity was
Posttranscriptional Modications conned to the somatodendritic regions of dentate gyrus
of Kainate Receptor Subunits granule cells (Darstein et al., 2003); thus, GluR7 may not con-
tribute to the KA1/KA2/GluR6 receptors on the mossy ber
As with AMPA receptors, the structural/functional complexity terminals (alternatively, GluR7 axonal transport may require
of kainate receptors is increased by posttranscriptional modi- subunit coassembly with GluR6). Taken together, based on
cations (reviewed in Dingledine et al., 1999). The subunits the subunit distributions, immunoprecipitation assays, and
GluR6 and GluR7 each come in two splice forms, with differ- electrophysiological analyses from knockout mice, kainate
ent C-terminal sequences (GluR6-1, 2; GluR7a, b). The GluR5 receptors in hippocampal principal cells are assemblies of
subunit displays four alternatively splice C-tails; additionally, KA2/GluR6 in CA1 pyramidal cells; KA2/GluR6 and
an exon encoding 15 amino acids in the N-terminal domain KA1/GluR6 or KA1/KA2/GluR6 in CA3 pyramidal cells; and
occurs in some transcripts. Functionally, recombinant homo- KA1-, KA2-, GluR6-, and GluR7-derived receptors in dentate
meric GluR7a channels have 5- to 10-fold larger currents in granule cells. A signicant expression of KA1/KA2/GluR6
HEK 293 cells compared to GluR7b, suggesting that the C-tail receptors is presynaptically assembled on mossy ber termi-
splicing may regulate kainate receptor function. In addition, nals. Note that KA1 and KA2 subunits selectively assemble
the various C-terminal domains are likely to play distinct roles with GluR6 but not GluR7 subunits in the mouse hippocam-
in receptor trafficking by binding with splice form-specic pus (Darstein et al., 2003).
regulatory proteins. In hippocampal interneurons, the most prominent kainate
The GluR5 and GluR6 subunits are also regulated by site- receptor subunit is GluR5 (Bettler et al., 1990; Wisden and
selective RNA editing. Like the AMPA receptor GluR2 subunit Seeburg, 1993; Bahn et al., 1994); approximately half of the
(see Section 7.2.2), RNA editing at the Q/R site (a glutamine interneurons in the stratum oriens of adult CA1 express
to arginine coding change) in the M2 domain of the GluR5 GluR5 (Paternain et al., 2000). Some of these GluR5-positive
and GluR6 subunits reduces the single-channel conductance cells are probably oriens/alveus/lacunosum-moleculare
and Ca2 permeability and changes the IV relation. In con- (OLM) interneurons. In addition, there are a few GluR5-
trast to GluR2, however, the glutamine residue is edited only positive cells in the stratum radiatum (approximately 14% of
in some GluR5 and GluR6 transcripts, with a gradual increase all GABAergic cells), and in the pyramidal cell layer itself
in the expression of the edited forms of both subunits during (approximately 30% of all GABAergic cells located in the
postnatal development. In the adult, approximately 50% of pyramidal cell layer). However, disruption of the GluR5 gene
 Molecular Mechanisms of Synaptic Function 257

does not cause a loss or reduction of functional kainate recep- (considered later in the chapter). Unlike the heteromeric
tors in CA1 stratum radiatum interneurons, indicating that GABAB receptors (see Section 7.5.3), all mGluR receptors are
other kainate receptor subunits are also expressed in hip- homodimers. The mGluR family subdivides into three groups
pocampal interneurons (Mulle et al., 2000). Indeed, most of based on their pharmacological and functional properties:
the GABAergic cells in the pyramidal cell layer also express group I (mGluR1, mGluR5), II (mGluR2, mGluR3), and III
GluR6 (Paternain et al., 2000), and combined deletion of (mGluR4, mGluR6, mGluR7, mGluR8) (Shigemoto et al.,
GluR5 and GluR6 abolishes their kainate receptor-mediated 1997). Group I mGluRs are selectively activated by 3,5-dihy-
currents. In addition, presynaptic GluR6- but not GluR5- droxyphenylglycine (DHPG) and positively couple via Gq to
containing receptors modulate synaptic transmission between phospholipase C, thus stimulating protein kinase C activation
inhibitory interneurons, indicating synapse-specic distribu- and inositol triphosate (IP3) production. Group I mGluRs
tion of kainate receptor subunits in interneurons (Mulle et al., have a somatodendritic distribution (see below); and their
2000). activation leads to enhanced excitability of hippocampal neu-
Considering the GluR-6 expressing interneurons in more rons via modulation of Ca2, K, and nonselective cation
detail, there are a few GluR6-positive cells in theCA1 stratum channels and to an increase in postsynaptic intracellular Ca2
oriens (6% of total GABAergic cells) and the stratum radia- via IP3-mediated release from internal Ca2 stores. Group II
tum (approximately 3%); however, there are many GluR6- and III mGluRs are selectively activated by 2-(2,3-dicarboxy-
positive cells in the CA3 stratum lucidum; 85% of these cyclopropyl)glycine (DCG-IV) and 2-amino-4-phospho-
GluR6-expressing stratum lucidum cells are also GAD65- nobutyrate (L-AP4), respectively, and negatively couple via Gi
positive. There are a few GluR6 mRNA-positive cells in the to adenylate cyclase, thus inhibiting cAMP production. The
mouse stratum oriens and radiatum layers of both CA1 and mGluRs 2, 3, 4, 7, and 8 are located in axons near and within
CA3 (Bureau et al., 1999). Kainate receptor-mediated modu- presynaptic terminals (see below). They suppress neurotrans-
lation of synaptic transmission between inhibitory interneu- mitter release by inhibiting voltage-dependent Ca2 channels
rons in the stratum radiatum is impaired in GluR6 knockout and/or by interfering directly with the release machinery. In
mice (Mulle et al., 1998). However, excitation of the interneu- the hippocampus, group III mGluRs mediate inhibition of
rons by activation of kainate receptors was still possible, sug- excitatory transmission in the Schaffer collateral(CA1 cell
gesting the existence of two populations of kainate receptors synapses, whereas both group II and group III mGluRs con-
in hippocampal interneurons (Mulle et al., 2000). GluR7 is tribute to presynaptic inhibition in the perforant pathgran-
expressed in occasional cells in the pyramidal cell layer; it is ule cell synapses.
unknown if they are pyramidal cells or interneurons. There
are a few GluR7 mRNA-positive cells in the mouse stratum mGluR Distribution in the Hippocampus
oriens and stratum radiatum (Bureau et al., 1999).
All mGluR genes, except mGluR6, are expressed in the hip-
7.3.5 Metabotropic Glutamate Receptors pocampus and show a diverse cellular distribution (Fig. 77).
Their expression has been studied in detail by in situ
Metabotropic glutamate receptors (mGluRs) are G protein- hybridization and with subtype-specic antibodies.
coupled and modulate glutamate release, GABA release, and From group I mGluRs, mGluR1 mRNA is in all principal
neuronal excitability (for a review, see De Blasi et al., 2001). cells, with the order of expression level DG  CA3  CA1
Metabotropic glutamate receptor activation inhibits or (Shigemoto et al., 1992). Many interneurons in the CA1 stra-
excites, depending on the receptor subtype and cell context. tum oriens and the stratum oriens and radiatum in CA3
EPSPs are often reduced by presynaptic mGluR activation and are strong mGluR1 expressors (Shigemoto et al., 1992).
enhanced by activation of postsynaptic mGluRs. An interest- Within the detection limits of the method, double-labeling in
ing aspect of the function of mGluRs in synaptic transmission situ hybridization shows that mGluR1 mRNA is in somato-
comes from their subcellular localization: The mGluRs are statin-positive but not parvalbumin-positive interneurons
typically outside of the pre- and postsynaptic membranes, (Kerner et al., 1997). mGluR-5 mRNA is abundant in the hip-
indicating that they are selectively activated during intense pocampus, expressed strongly in CA pyramidal cells, dentate
synaptic activity, such as repetitive high-frequency ring, granule cells, and many types of GABAergic interneurons
leading to a spillover of neurotransmitter into the perisynap- (Fotuhi et al., 1994). mGluR-5 mRNA was also detected by
tic space (Baude et al., 1993; Lujan et al., 1996; Shigemoto et double-labeling in situ hybridization in somatostatin-positive
al., 1997). and parvalbumin-positive interneurons (Kerner et al., 1997).
The eight mGluR receptor subtypes, each consisting of a Group I mGluRs have a selective somatodendritic perisynap-
large extracellular N-terminal domain with a binding site for tic distribution in principal neurons and are typically located
glutamate, seven transmembrane domains, and an intracellu- at the outer edge of postsynaptic densities of dendritic spines
lar C-terminal peptide, are closely related to the family of (Lujan et al., 1996; Shigemoto et al., 1997).
GABAB receptors but have low sequence identity with classic From group II and III mGluRs, mGluR3 mRNA is abun-
G protein-coupled receptors such as the muscarinic receptors dant in dentate granule cells but absent from pyramidal cells;
258 The Hippocampus Book

Figure 77. Expression of mGluRs in the rat hippocampus. Immunohistochemistry with sub-
unit-specic antibodies was used to visualize expression patterns of individual mGluRs in adult
hippocampus. (Source: Modied from Shigemoto et al., 1997, with permission.)

mGluR3 is also expressed in hippocampal white matter tracts, whereas group II receptors are found in the preterminal por-
the mbria, and the fornix (Ohishi et al., 1993). In contrast to tions of axons (Shigemoto et al., 1997).
the largely presynaptic localization of most group II and III
mGluRs, mGluR3 protein is expressed perisynaptically in
dendrites of the dentate granule cells in a pattern reminiscent
of mGluR1 and mGluR5 expression (Tamaru et al., 2001). 7.4 Trafficking of Glutamate Receptors
mGluR4 and mGluR7 mRNAs are both expressed in CA1 and and Hippocampal Synaptic Plasticity
CA3 pyramidal cells, dentate granule cells, and scattered
interneurons (Ohishi et al., 1995). mGluR4 is more marked in In mammalian neurons, subcellular compartments, often
CA2 pyramidal cells. The mGluR2 and mGluR7a proteins are located in near proximity to each other, comprise distinct pro-
on axons and terminals of the medial perforant path, whereas tein complexes. For example, in principal neurons, excitatory
the lateral perforant path is prominently immunoreactive for synapses are positioned on dendritic spines, which contain
mGluR8 in the dentate gyrus and the CA3 area. Mossy bers glutamatergic receptors and cytoskeletal and signaling pro-
contain mGluR2, mGluR7a, and mGluR7b, whereas Schaffer teins relevant to their functionsall components of the so-
collateral axons display only mGluR7a. Group II (mGluR2) called postsynaptic density (PSD). GABA receptors and their
and group III (mGluR4, 7, and 8) mGluRs are in different specialized cytoskeletal proteins, on the other hand, are posi-
parts of the presynaptic elements: Group III receptors pre- tioned at GABAergic synapses scattered between the dendritic
dominate in presynaptic active zones of both asymmetrical spines; whereas dendritic voltage-gated ion channels, regulat-
(glutamatergic) and symmetrical (GABAergic) synapses, ing integration of the excitatory and inhibitory inputs along
 Molecular Mechanisms of Synaptic Function 259

the dendritic branches, are mostly excluded from the synaptic tors, either homomeric GluR1 or heteromeric GluR1/2 chan-
sites and show either uniform or gradient-like distributions in nels, was shown to be triggered during the induction of LTP
the dendrites. To maintain such function-specic subcellular and to depend on the activation of NMDA receptors as well as
distribution of proteins, neurons have a variety of trafficking -CaMKII. Direct evidence of the role of GluR1 in LTP and in
proteins that mediate intracellular targeting, retention, and specic forms of learning came from mice that either lack the
removal of particular receptors, channels, kinases, phos- GluR1 gene altogether or contain the gene with knock-in
phatases, and other molecules at their corresponding destina- mutations in its C-terminal phosphorylation sites (Zamanillo
tion sites. et al., 1999; Reisel et al., 2002; Lee et al., 2003) (these mouse
The most prevalent type of proteinprotein interaction models are discussed in detail in Section 7.5.3).
underlying intracellular trafficking is between a short amino The search for trafficking proteins binding specically to
acid motif typically present at the C-terminal end of the traf- the GluR1 C-terminal domain has yet to yield a clear candi-
cked protein and a PDZ domain of its cytoskeletal interactor. date for mediating synaptic delivery of GluR1-containing
The PDZ abbreviation is derived from three proteins origi- AMPA receptors. To date, the list of GluR1 C-tail interactors
nally identied to contain this approximately 90-amino-acid includes 4.1N protein, which binds at juxtamembrane C-tail
structural motif: PSD-95 (postsynaptic density protein of 95 region; PDZ domain-containing protein SAP-97, which binds
kD molecular weight), DlgA (Drosophila discs-large protein; at the class I PDZ binding motif at the C-tail carboxyl end;
present in septate junctions) and ZO-1 (protein of epithelial and PDZ and LIM domain-containing protein RIL (reverse-
tight junctions). There are three PDZ domains that bind dif- induced LIM domain protein), which binds at the last 10
ferent amino acid sequences. The class 1 PDZ domain recog- amino residues of GluR1 C-tail via its LIM domain (reviewed
nizes the X-S/T-X-V/L motif (where X is any amino acid), in Malinow and Malenka, 2002; Song and Huganir, 2002).
class 2 the X-f-X-f motif (f  hydrophobic aa), and class 3 the Note that the PDZ domain-binding motif of the GluR1 C-tail
X-D-X-V motif (reviewed in Sheng and Sala, 2001). Next to is not required for synaptic delivery/accumulation of the
the carboxyl sequence, PDZ domains can also bind to an receptors or for synaptic plasticity in the CA1 hippocampus
internal motif that forms a -hairpin structure, although this (Lee et al., 2003). Functionally, 4.1N protein was proposed to
form of PDZ interaction is much less frequent. anchor GluR1-containing AMPA receptors at the synaptic
surface by a linkage to spectrin/actin cytoskeleton; SAP-97
7.4.1 Synaptic Transport of AMPA may regulate the transport of newly synthesized receptors
Receptors in LTP and LTD through Golgi and ER compartments; and RIL was proposed
to link the intracellular transport of the receptors in the den-
Intracellular trafficking of ionotropic glutamate receptors dritic spine compartment to the actin/-actinin cytoskeleton
includes selective somatodendritic transport to the postsynap- (Song and Huganir, 2002; Schulz et al., 2004).
tic membrane of glutamatergic synapses and anchoring of the
receptors by cytoskeletal proteins. In addition, AMPA recep- GluR2-dependent Trafficking and GluR2-binding
tors undergo activity-dependent cycling between two pools: Proteins: Relevance to Hippocampal LTD
the functional pool of AMPA receptors inserted at the synap-
tic surface and the reserve pool of AMPA receptors present It has been proposed that the prevalent C-terminal splice form
either intracellularly (in which case the movement is regulated of GluR2, with its 50-amino-acid residue short C-terminal
by local endo- and exocytosis) or at the extrasynaptic surface tail, regulates activity-independent (constitutive) synaptic
membrane (in which case the receptors are transported by lat- delivery of GluR2/3 receptors in the CA1 pyramidal neurons
eral diffusion). Such activity-dependent local synaptic recy- (this model of transport is again primarily based on electro-
cling of AMPA receptors is believed to be one mechanism by physiology experiments with recombinant channels)
which hippocampal pyramidal neuronsparticularly in the (reviewed in Malinow and Malenka, 2002). Interestingly,
CA1 regioncan regulate the strength of their synaptic trans- the GluR2/3 receptor insertion does not change the overall
mission. strength of synaptic transmission but, rather, replaces the
GluR1/2-containing receptors. The GluR2-containing recep-
GluR1-dependent Trafficking and GluR1-binding tors, however, undergo activity-dependent endocytosis from
Proteins: Relevance to Hippocampal LTP the synaptic surface during protocols resulting in the induc-
tion of LTD. Based on these experiments, it was suggested
Trafficking mechanisms mediated by the GluR1 C-terminal that the synaptic transport of AMPA receptors is regulated
domain were proposed to regulate synaptic insertion of by subunit-specic interactions where the GluR1 subunit
GluR1-containing AMPA receptors at the Schaffer collateral to plays a critical role in upregulating the strength of AMPA
CA1 synapses. The initial evidence for this came from a series receptor-mediated transmission during LTP, and the GluR2
of electrophysiological experiments studying synaptic delivery subunit mediates downregulation of the glutamatergic trans-
of recombinant receptors in organotypic slice cultures mission seen in LTD.
(reviewed in Malinow and Malenka, 2002). In this work, Trafficking proteins that bind to the GluR2 C-terminal
synaptic insertion of recombinant GFP-tagged GluR1 recep- domain include N-ethylmaleimide-sensitive fusion protein
260 The Hippocampus Book

(NSF), glutamate receptor interacting protein (GRIP) and a AMPA receptor-lacking synapses in the young hippocampus
closely related AMPA receptor-binding protein (ABP, also (Nusser et al., 1998b; Petralia et al., 1999).
termed GRIP2), protein interacting with C-kinase (PICK1), The initial delivery of AMPA receptors to silent synapses
and clathrin adaptor protein complex AP2 (reviewed in Song the conversion of silent to functional AMPA receptor-express-
and Huganir, 2002). In general, the function of individual ing synapseshas been proposed to be mediated primarily by
GluR2 interactors is not clearly resolved; the overall conclu- two subunits, GluR4 and GluR2long, with closely related C-
sions, however, tend to agree with the proposed role of GluR2- terminal domains (Zhu et al., 2000; Kolleker et al., 2003);
interacting proteins in endocytosis and recycling of the GluR2long is a C-terminal splice form of GluR2 (see section
receptors and support the electrophysiology ndings that 7.2.2). Both subunits are synaptically inserted during the
GluR2-dependent transport plays a role in LTD. NSF is an induction of LTP as well as by synchronized spontaneous
ATPase that is essential for SNARE-dependent intracellular synaptic activity resembling the giant depolarizing potentials
membrane fusion (see Section 7.2.6 for the role of NSF in (GDPs) that frequently occur in the juvenile hippocampus.
neurotransmitter release). It was proposed that NSF binding Interacting proteins specic for the GluR4 and/or GluR2long
to the GluR2 C-tail regulates the recycling of GluR2-contain- subunits, which could potentially regulate the trafficking of
ing AMPA receptors after their endocytosis from synaptic sites AMPA receptors to the silent synapses, are yet to be identied.
toward new synaptic insertion. One model of NSF function
suggests that the ATPase regulates disassembly of the internal- AMPA Receptor Pan-subunit-interacting
ized receptors from their binding with PICK1 (Hanley et al., Proteins: Narp and Stargazin
2002). The binding by GRIP/ABP (proteins containing 6/7
PDZ domains) and PICK1 (single PDZ domain protein) at Two proteins, neuronal activity-regulated pentraxin (Narp)
the same class II PDZ-binding motif of the GluR2 C-tail is and stargazin (also termed 2), were shown to bind to AMPA
regulated by phosphorylation: PICK1, but not GRIP/ABP, receptors without selectivity for individual subunits (reviewed
binds to the phosphorylated TEV motif. Functionally, in Bredt and Nicoll, 2003). Narp is a secreted immediate early
GRIP/ABP binding was proposed to mediate anchoring of the gene product that binds to the extracellular portions of the
GluR2-containing receptors at the hippocampal CA1 synapses GluR1/2/3 subunits. Narp was shown to induce AMPA recep-
(Osten et al., 2000), and PICK1 binding may regulate inter- tor clustering and may function during glutamatergic synap-
nalization of GluR2-containing receptors during LTD (Kim et togenesis (OBrien et al., 1999).
al., 2001). Interestingly, PICK1-based GluR2 trafficking may The stargazin gene was originally identied because of its
also enhance the proportion of synaptic Ca2-permeable spontaneous loss-of-function mutation in a mouse line with
(GluR2-lacking) AMPA receptors, suggesting a further, more head-tossing movements (hence the name stargazer for the
complex role for PICK1 in hippocampal synaptic functions mouse and stargazin for the gene), ataxic gait, impaired eye-
(Terashima et al., 2004). The role of the AP2 adaptor, in agree- blink conditioning, and spike-wave seizures characteristic of
ment with its well established role in clathrin-dependent absence epilepsy (reviewed in Letts, 2005). Stargazin is a four-
endocytosis (see also Section 7.1.7), appears to be necessary transmembrane protein, with three closely related brain-
for activity-dependent internalization of the GluR2-contain- specic isoforms: 3, 4, and 8. The trafficking function of
ing receptors during LTD (Lee et al., 2002). stargazin with respect to AMPA receptors was revealed rst in
stargazer cerebellar granule cells, which show a striking failure
Glutamatergic Silent Synapses: of AMPA receptor synaptic delivery (Chen et al., 2000). Note
GluR2long-dependent and GluR4-dependent Trafficking that these cells appear to express only stargazin, whereas most
other CNS neurons express multiple stargazin isoforms
Mature glutamatergic synapses contain both AMPA- and (Tomita et al., 2003). Thus, the cerebellar granule cells can be
NMDA-type receptors. In contrast, electrophysiological considered a cellular model of a complete 2, 3, 4, and 8
recordings of synaptic currents in hippocampal slices pre- knockout. In the hippocampus, 8 is the most prominent iso-
pared from rats less than 2-weeks old demonstrated that a form but 2, 3, and 4 are also expressed; cellularly, stargazin
large portion of young synapses contain only NMDA recep- and its isoforms are expressed in pyramidal neurons and
tors and lack functional AMPA receptors. Because NMDA interneurons (Tomita et al., 2003). In agreement with the
receptors are inhibited by their voltage-dependent Mg2 block overlapping expression in the hippocampus, genetic deletion
at a resting membrane potential (see Section 7.3.3), glutamate of 8 did not cause complete breakdown of AMPA receptor
release at these sites fails to evoke postsynaptic currents; and trafficking: Synaptic AMPA receptor-mediated currents in the
hence such synapses were named silent synapses (reviewed CA1 pyramidal neurons were only modestly decreased in the
in Malinow and Malenka, 2002). In addition to the electro- 8 knockout; however, the total protein levels of GluR1 and
physiological demonstration of silent synapses, anatomical GluR2/3 receptors were strongly reduced, a portion of the
support for their existence was provided by an electron remaining receptors appeared retained somatically, and
microscopy-based analysis of AMPA and NMDA receptor dis- extrasynaptic receptor pools were depleted; furthermore, LTP
tribution, which conrmed the predicted large portion of in the CA1 neurons was also impaired (Rouach et al., 2005).
 Molecular Mechanisms of Synaptic Function 261

How does this phenotype, which strikingly resembles genetic with the entire 627-amino-acid residue NR2A C-terminal
deletion of GluR1 in the hippocampus (see Section 7.4.3), domain (Kornau et al., 1995). The yeast two-hybrid method
compare to what is known about stargazin functions? is used to identify protein interactions by screening a cDNA
Stargazin appears to be stably associated (as an auxiliary sub- library, such as a brain or hippocampal library, with a tran-
unit) with a large portion of native AMPA receptors scription-reporter system in yeast; it has been used for isola-
(Nakagawa et al., 2005; Vandenberghe et al., 2005) and was tion of most of the glutamate receptor-interacting proteins
proposed to chaperone the transport of the receptors through described to date. NR2A, as well as the other NR2 subunits,
early endoplasmatic reticulum and Golgi compartments binds to the second and third PDZ domains of PSD-95 via a
(Vandenberghe et al., 2005). In the hippocampal CA1 pyram- conserved class I PDZ-binding motif; in addition, NR1 splice
idal neurons, most mature AMPA receptor complexes thus forms containing the C2 segment, NR1-3, and NR-4, may
appear to require the presence of GluR1 (see Section 7.4.3) also bind to PSD-95 via a different PDZ-binding sequence
and 8. In addition, stargazin and its  isoforms are also likely (reviewed in Sheng and Sala, 2001).
to play direct roles in synaptic functions of AMPA receptors in PSD-95 is an abundant postsynaptic protein that contains
the hippocampus: rst, by anchoring the receptors to the three PDZ domains, an SH3 domain, and a nonfunctional
postsynaptic cytoskeletal protein PSD-95 (Chen et al., 2000) guanylate kinase (GK) domain. The PSD-95 protein family
(see Section 7.3.2 for PSD-95 role in synaptic function of consists of three additional closely related proteins: SAP102,
NMDA receptors) and, second, by reducing the receptor SAP97, and PSD-93 (note that SAP97 binds to GluR1) (see
desensitization in response to synaptic glutamate (Priel et al., Section 7.3.1). Functionally, PSD-95 acts as a scaffold linking
2005; Tomita et al., 2005; Turetsky et al., 2005). These func- a number of other postsynaptic proteins to the NMDA recep-
tions may be sufficiently compensated by 2, 3, and/or 4 in tor complex (reviewed in Sheng and Pak, 2000). For example,
the 8-knockout hippocampus. PSD-95 binds, via its GK domain, to a protein termed GKAP.
GKAP binds to Shank, and Shank in turn binds to Homers
7.4.2 NMDA Receptor-associated and cortactin. Dimers of Homer may link the NMDA recep-
Cytoskeletal and Signaling Proteins tors to internal Ca2 stores or to the metabotropic receptors,
as Homer also binds to the IP3 and/or ryanodine receptors of
Until recently, NMDA receptors were believed to be less the Ca2 stores and to the C-terminal domains of
synaptically mobile than AMPA receptors, and the traffick- mGluR1a/mGluR5 (see more on Homer in Section 7.3.3).
ing of NMDA receptors was originally proposed to be largely Cortactin, on the other hand, links the NMDA receptors to the
limited to their synaptic delivery and subsequently transport actin cytoskeleton. Other proteins are associated with PSD-95
for degradation. However, it is becoming evident that NMDA via binding to one of its three PDZ domains; these proteins
receptors undergo activity-dependent trafficking, which may include the neuronal nitric oxide synthase (nNOS); neuronal
be quite similar (for example, with regard to the subunit- cell adhesion proteins neuroligins, microtubule-binding pro-
specic transport mechanisms) to that of AMPA receptors, tein CRIPT; synaptic GTPase-activating protein for Ras
though perhaps happening on a slower time scale. At the termed SynGAP; Rho effector citron; nonreceptor tyrosine
same time, synaptic functions of NMDA receptors appear to kinase Fyn; and possibly Src. Taken together, the scaffold func-
depend critically on interactions with cytoskeletal and signal- tion of PSD-95 is twofold: First, PSD-95 anchors the receptors
ing proteins of the postsynaptic density, which directly or to postsynaptic cytoskeletal proteins; and second, it links sig-
indirectly bind to the C-terminal domains of the receptor sub- naling molecules to the near proximity of the receptors. In
units. As described in Section 7.2.3, the major NMDA recep- addition, PSD-95 directly clusters NMDA receptors at the
tor types in the hippocampus are the NR1/NR2A, NR1/NR2B, postsynaptic membrane, a function that depends on its ability
and to a lesser extent NR1/NR2A/NR2B subunit assemblies. to multimerize and to associate with membrane lipids via
The NMDA receptor C-terminal domains are quite large, palmitoylation of cysteine residues in the N-terminal portion
varying from approximately 100 amino acid residues for the of the PSD-95 molecule (reviewed in el-Husseini Ael and
longer NR1 splice forms to more than 600 amino acid residues Bredt, 2002).
for the NR2A/2B subunits. These receptor domains are thus The importance of PSD-95 for NMDA receptor function
well suited to accommodate a large number of interacting was directly tested in mice engineered to lack the PSD-95 gene
proteins and their associated protein complexes. (Migaud et al., 1998). Somewhat surprisingly, the subcellular
distribution of the receptors in hippocampal CA1 neurons
Regulation of NMDA Receptor Function was normal, indicating that PSD-95 is not essential for NMDA
via PSD-95 and Its Interacting Proteins receptor targeting to and accumulation at excitatory synapses.
In contrast, synaptic plasticity at the Schaffer collateral to CA1
The identication of PSD-95 as an NMDA receptor-binding synapses was altered: The frequency threshold for the induc-
protein (which incidentally was also the rst identication of tion of NMDA receptor-dependent LTD and LTP was shifted
a proteinprotein interaction at a C-terminal tail of a gluta- in favor of LTP, resulting in strikingly enhanced LTP at stimu-
mate receptor subunit) came from a yeast two-hybrid screen lation frequencies ranging from 1 to 100 Hz. In agreement
262 The Hippocampus Book

with the belief that LTP and LTD are both required for normal 7.5.1 on decits in LTP and learning in mice lacking the -iso-
learning and memory, the PSD-95 knockout mice showed a form of the kinase). The current model of CaMKII in LTP
profound decit in hippocampus-based spatial learning in the suggests that the kinase, after its activation by postsynaptic
Morris watermaze (Migaud et al., 1998). One possible expla- Ca2 inux through the NMDA receptors, undergoes a molec-
nation of the PSD-95 null phenotype is that other members of ular switch by autophosphorylation of a single residue in the
the PSD-95 family are able to compensate for the trafficking regulatory domain (Thr286 in -CaMKII), which allows the
and clustering functions but are not sufficient to bind one or dodecameric holoenzyme to become persistently active even
more of the signaling molecules that are normally linked by in the absence of Ca2/calmodulin; as a next step, the kinase
PSD-95 in the proximity of NMDA receptors (see above). translocates to the NMDA receptor complex and binds to the
Presently, it is not clear which PSD-95-dependent signaling C-tails of NR2A/B and/or NR1 (reviewed in Lisman et al.,
molecules/cascades may be impaired in the knockout mice. 2002). The switch back to a normal state requires dephospho-
The cellular and behavioral phenotypes do, however, provide rylation of the kinase; similar dephosphorylation by PP1 may
evidence regarding the importance of the NMDA receptor occur during LTD (note that PP1 is also localized to the
subunit C-terminal domains for normal NMDA receptor NMDA receptor complex by its interactions with yotiao, see
function. This conclusion was even more dramatically estab- above). In addition, binding of activated CaMKII selectively to
lished in mice engineered to lack the complete C-terminal the NR2B C-tail locks the kinase in an autoactive state that
peptides of the NRA, NR2B, or NR2C (Sprengel et al., 1998) does not require autophosphorylation (Bayer et al., 2001).
(see Section 7.5.2). Once at the PSD site, CaMKII can phosphorylate the GluR1
C-terminal Ser831, which directly increases the channel con-
NMDA Receptor-interacting Proteins ductance of GluR1-containing AMPA receptors; furthermore,
(Other than PSD-95) CaMKII-mediated phosphorylation of a yet to be identied
protein(s) was proposed to regulate the synaptic insertion of
In addition to PSD-95, a number of other postsynaptic pro- AMPA receptors during the induction of LTP (see Section
teins bind directly to the NMDA receptor complexprima- 7.4.1) (reviewed in Malinow and Malenka, 2002).
rily to the NR1 C-terminal domain (see above for NR1
C-terminal splice forms). The juxtamembrane region of the Synaptic Trafficking of NMDA Receptors:
NR1 C-terminal peptide, termed C0, serves as a binding site Relevance to Homeostatic Plasticity and LTP
for -actinin, an actin-binding protein. This interaction may
contribute to anchoring of the receptors at the postsynapse, as Intracellular transport of NMDA receptors appears to be
actin is the major cytoskeleton of dendritic spines. (Note that tightly regulated at several levels, from the assembly and traf-
there are at least three other possible ways for linking NMDA cking in the endoplasmatic reticulum to agonist induced
receptors to actin cytoskeleton: via NR2B binding to - internalization at synapses. As discussed in Section 7.3.3, the
actinin, via NR1/2A/2B binding to spectrin, and via the PSD- NR2B receptors are more prevalent at young and NR2A at
95-GKAP-Shank-spectrin complexas reviewed by Sheng mature hippocampal synapses; in the adult, NR2B receptors
and Pak, 2000.) Other interactions give rise to functional are expressed at lower levels and perhaps mainly extrasynapti-
changes. Calmodulin binds to the C0 and C1 segment of NR1; cally. Some of the rst evidence for regulated synaptic traf-
this interaction is Ca2-dependent and causes an approxi- cking of NMDA receptors came from studies showing that
mately fourfold reduction of open NMDA channel probabil- long-term changes in overall synaptic activity regulate the
ity. It has been proposed that competitive binding between synaptic expression of both AMPA and NMDA receptors, a
-actinin and Ca2/calmodulin at the C0 segment mediates phenomenon termed homeostatic plasticity. Chronic (sev-
activity/Ca2-dependent inactivation of NMDA receptors eral hours or even days) suppression of excitatory network
(Zhang et al., 1998; Krupp et al., 1999). In addition to calmod- activity upregulates AMPA and NMDA receptor-mediated
ulin, two other proteins bind to the NR1 C1 segment: yotiao responses, whereas chronic elevation of excitatory network
and neurolament L. Yotiao may serve to link the type I pro- activity has the opposite effect (reviewed in Turrigiano and
tein phosphatase (PP1) and the adenosine 3,5-monophos- Nelson, 2004). This form of plasticity was proposed to protect
phate (cAMP)-dependent protein kinase (PKA) to the excitatory circuits from both runaway excitation and too little
proximity of the receptor (Sheng and Pak, 2000). This close excitation (in other words, to maintain an overall balance
physical association of kinases and phosphatases with the between synaptic input and spiking output in excitatory net-
NMDA receptors is an elegant mechanism that enhances the works).
coupling of the Ca2 inux to downstream signaling mole- Synaptic delivery of the NR2A and NR2B receptors at hip-
cules. Perhaps the best example of such symbiosis is the pocampal synapses is also differently regulated by activity:
binding of the most abundant postsynaptic kinase, CaMKII, NR2B insertion at synapses occurs constitutively, whereas the
to the NR2A/NR2B subunits. exchange of NR2B for NR2A receptors is driven by synaptic
CaMKII plays a critical role in plasticity at Schaffer collat- activity (Barria and Malinow, 2002). This transport mecha-
eral to CA1 synapses, where it has been shown to be both nec- nism may guarantee that NR2B receptors are freely inserted
essary and sufficient for the induction of LTP (see Section during development and prior to and/or during the formation
 Molecular Mechanisms of Synaptic Function 263

of synaptic contacts, whereas the NR2A receptors are deliv- inhibit the constitutive activity of this mGluR species. Under
ered only to preexisting synapses. Recently, synaptic recruit- this last scenario, the induction of Homer1a would not only
ment of NMDA receptors was also observed during LTP, increase the surface expression of mGluR1a/5, as described
analogous to the delivery of AMPA receptors but with a delay above, but also enhance mGluR activation even in the absence
of approximately 2 hours (Watt et al., 2004); this nding pro- of glutamate.
vides a mechanism for maintaining a constant ratio of AMPA Axonal transport of the presynaptic mGluR7 in hippocam-
to NMDA receptors at synapses, as observed on a longer time pal neurons is directed by a sequence in the mid-region of its
scale during the homeostatic form of plasticity. C-terminal peptide (Stowell and Craig, 1999). The distal
region of the mGluR7 C-terminal tail interacts with the
7.4.3 Proteins Regulating Transport PDZ domain-containing protein PICK1 (Boudin et al., 2000),
and Function of mGluRs a protein that also binds to the C-terminal domains of
the AMPA receptor subunits GluR2 and GluR3. The
As with ionotropic glutamate receptors, the subcellular traf- mGluR7PICK1 interaction may mediate the clustering of
cking and function of mGluRs are also regulated via proteins metabotropic receptors at the presynaptic terminal mem-
binding at their intracellular C-terminal domains. However, brane.
fewer mGluR-binding partners have been identied so far (see
below). In view of the numerous mGluR C-terminal splice
variants, further mGluR trafficking proteins may be expected
to be found in the coming years (mGluR C-terminal varia- 7.5 Glutamate Receptor Mutant Mice:
tions due to alternative RNA splicing include four versions of Genetic Analysis of Hippocampal Function
mGluR1, termed mGluR1a to d; and two versions each of
mGluR4, 7, and 8, termed mGluR4a and b, mGluR5a and b, 7.5.1 Introduction: Building
mGluR7a and b, and mGluR8a and b). of Hippocampus-specic Genetic Models
Both members of the mGluR I group (mGluR1 and
mGluR5) bind to proteins of the Homer gene family, which, As discussed in detail in Chapter 10, modulation of excitatory
in turn, interact with the NMDA receptorPSD-95 complex synaptic transmission between principal neurons, typically via
(see above). Homers exist in two forms: constitutively Hebbian synaptic plasticity, is believed to constitute the prin-
expressed Homers (Homer 1b and c, 2, and 3), which contain cipal neural mechanism of hippocampus-dependent learning
an N-terminal EVH domain and a C-terminal dimerizing and memory. Mouse models engineered to lack a specic gene
coil-coil domain; and an immediate early gene (IEG) homer related to functions underlying synaptic plasticity or to
(Homer 1a), which lacks the coil-coil domain (reviewed in express a modied version of such a gene with an altered
Xiao et al., 2000; Fagni et al., 2002). Due to the lack of the coil- function, are becoming increasingly powerful in addressing
coil domain, Homer1a cannot dimerize; it can, however, act as molecular and cellular mechanisms underlying the various
a competitive decoy and disrupt the interactions between full- types of hippocampal learning and memory that are studied
length Homer dimers and their associated proteins. Homers in rodents (see Chapter 10 for hippocampus-based learning in
bind via their EVH domain to the ProProXXPhe motif rats and mice). A pioneering work in this direction was the
present in the distal part of the mGluR1a/5 C-terminal generation of mice lacking the a subunit of the Ca2/calmod-
domains. The mGluRHomer interaction may serve several ulin-dependent protein kinase (CaMKII), an enzyme postu-
functions. First, Homer proteins may direct the targeting of lated to be a critically important player in the process of
mGluR1a/5 to dendrites rather than axons, and anchor memory formation in the hippocampus (see Section 7.4.2
mGluR1a/5 within intracellular membrane compartments. and Chapter 10 for activation of CaMKII in LTP). The -
In this scenario, the induction of the IEG Homer1a would CaMKII knockout mice showed impairment of LTP at CA3 to
act to release the intracellular anchor and lead to the delivery CA1 synapses, as well as decits in hippocampus-dependent
of the receptors at the perisynaptic membrane. Second, as learning (Silva et al., 1992a,b). This was a critical piece of evi-
with NMDA receptors, the Homer dimer complex may link dence highlighting the importance of -CaMKII in LTP; even
group I mGluRs to the cytoskeleton of dendritic spines more importantly, together with the contemporaneous study
via a mGluR1a/5-Homer-Homer-Shank complex. However, of the fyn mutant (Grant et al., 1992), it provided the rst
because NMDARs and mGluR1a/5 are differentially localized genetic evidence that LTP is involved in hippocampal memory
in spines (the former in the postsynaptic density, the latter formation.
perisynaptically), it seems likely that protein(s) other than In many other genetic studies, however, the correlation
Homers mediate the precise localization of mGluR1a/5 between the introduced manipulation and hippocampus-
at the cellular membrane. Third, Homers may couple mGlu- based memory formation is not as easily interpretable.
R1a/5 directly to intracellular Ca2 stores via the mGluR1a/ Typically, this is due to lack of a spatiotemporal regulation
5-Homer-Homer-IP3/ryanodine receptor complex. Fourth, over the onset of the manipulation, resulting in alteration of a
and nally, Homers may directly regulate the function of specic gene function that is present in many different brain
mGluR1a/5, as full-length Homer forms have been shown to regions and throughout development. The straightforward
264 The Hippocampus Book

explanation of the results of Silva et al. was possible because of hours after birth). Mice with reduced expression of the NR1
the prominent expression of -CaMKII in the hippocampus gene were generated by insertion of the neo gene cassette in
as well as the fact that the onset of -CaMKII expression is an intronic region of the NR1 gene, which resulted in silenc-
limited to a later stage of postnatal development. For other ing NR1 gene expression and subsequent reduction of NR1
genes one wishes to study with respect to their hippocampal protein expression down to 5% of the wild-type level (Mohn
functions, this problem has been overcome, at least to some et al., 1999). Note that the neo gene cassette serves as a selec-
extent, by the development of genetic techniques that are spa- tion marker during preparation of knockout targeting con-
tially restricted to the frontal cortex including the hippocam- structs (see Box 71); the presence of the cassette sometimes
pus or to hippocampal pyramidal cell types (e.g., CA1 interferes with normal mRNA processing of the transcribed
neurons or dentate gyrus granule cells) (see Box 71). So far, gene product, causing indirect silencing of gene expression.
such hippocampus-specic mutations have been mainly used The NR1 neo mice survived to adulthood, showed normal
to examine the functions of AMPA and NMDA receptors. As AMPA and kainate receptor functions, but displayed decits
described below, these studies not only support the conclu- in social and sexual interactions, increased motor activity, and
sions of Silva et al. and Grant et al. and those of the earlier stereotypical behavior that was similar to the behavior
pharmacological studies (Morris et al., 1986) regarding the observed in pharmacologically induced animal models of
role of LTP in learning, they also threw light on the impor- schizophrenia.
tance of glutamate receptors for both the induction and
expression of synaptic plasticity. The experiments showed that Hippocampal Region-specic NR1 Knockout Mice
different mechanisms are used to establish and maintain dif-
ferent forms of LTP used in different forms of learning (for Hippocampal CA1-specic NR1 knockout mice were gener-
example, spatial working versus reference memory). ated using the Cre/lox-recombination system, with the Cre
recombinase expressed under the control of the -CaMKII
7.5.2 NMDA Receptor Mutant Mice promoter (see Box 71) (Tsien et al., 1996a,b). Although this
promoter is typically active in most postnatal excitatory fore-
The rst mouse model aimed at studying NMDA receptor brain neurons, a transgenic line with Cre expression restricted
function was a knockout of the NR1 subunit (Forrest et al., almost exclusively to the CA1 region (due to a chromosomal
1994); these mice showed that NR1 is essential for formation integration effect) was selected. The CA1 cell-specic NR1
of functional NMDA receptor channels (the mice died 815 gene deletion had no apparent effect on mRNA levels of
(text continued on page 265)

Box 71
Tools to Generate Conditional Mouse Mutants

Conditional mouse mutants are used for temporally and spatially regulated transgene expression, gene
deletion, or gene activation. The two most commonly used systems are the tTA system and the Cre
system. In both systems, two genetic modules are introduced separately into the mouse genome: the
activator gene, which regulates expression of the second module, and the responsive target gene, the
function of which one wishes to study. After rst obtaining independent, activator and responder
mouse lines, the two are bred together to create the conditional mouse line (see Box Table 71).
Activator mice function in two ways (see Box Fig. 71). First, the expression of the activator gene is
controlled spatially and, to some extent, temporally by the selection of a specic promoter. The pro-
moters that have been used to drive activator genes include the following: the -CaMKII promoter,
which restricts expression of the activator to pyramidal neurons in postnatal brain (spatially ranging
from most pyramidal cells of the forebrain to only selective hippocampal CAI expression, depending
on the integration of the transgenic construct, and the kainate receptor subunit KA-I promoter, which
restricts the expression to CA1 and CA3 pyramidal neurons. Second, the activator itself can be con-
trolled pharmacologically, typically by systemic drug application (for example, in drinking water). The
transactivators tTA and rtTA are regulated by tetracycline (or its analogue doxycycline); tTA is inacti-
vated, which means that it does not bind to the tet operators in the promoter of the target gene in the
presence of doxycycline (often referred to as a let off protocol); rtTA is regulated in the opposite way
(tet on). Fusion protein of Cre recombinase and the hormone-binding portion of the estrogen
receptor, named CreERt2, was developed for temporal activation of the Cre recombinase: In the cell,
CreERt2 is present in the cytoplasm, thus restricting the nuclear localization of Cre; but it translocates
into the nucleus in the presence of drugs such as tamoxifen. Finally, expression of Cre recombinase
alone from various cell-specic promoters, such as the -CaMKII promoter, has been used for condi-
tional knockout of target genes in selected neuronal populations.
 Molecular Mechanisms of Synaptic Function 265

Box Table 71.

Activator, Responder, and Conditional Mice

Activator Mice Responder Mice

Activator genes Responder Genes

tTA or rtTA Any transgene (wt or mutant) under the control
of a promoter with tet operon regulation
Cre or CreERt2 Floxed genes for conditional gene knockout, oxed
hypomorphic alleles for gene activation
Activator genes determine where and Responder genes determine the type of the genetic
when genetic manuipulation occurs manipulation in the responder gene
No phenotype No phenotype

Conditional Mice (obtained by breeding activator and responder mice)

Type of Genetic Manipulation Responder Involved Activator Involved

Knockout rescue wt agged transgene tTA or rtTA in knockout mice

Expression of mutant genes Mutant agged transgenes tTA or rtTA
Conditional gene knockout Floxed gene Cre or CreERt2
Conditional gene activation Floxed hypomorphic alleles Cre or CreERt2
Conditional phenotype is mediated by activation of the genetic manipulation of the responder gene in
brain regions where activator gene is expressed.

There are two types of responder mice (see Box Fig. 71), depending on the activator used. Mice
regulated by tTA or rtTA expression are typically generated by transgenic insertion of the responder
gene under control of a pol II transcription initiation site containing several upstream tet operators.
After binding of tTA/rtTA at the tet operators, transcription of the target gene is initiated by the
tTA/rtTa VP16 activation domain. In contrast, the target genes in mice regulated by the expression
of Cre recombinase are endogenous genes with genetically introduced Cre-recognition (lox P) sites
in two intronic regions anking one or several functionally important exons. The Cre recombinase
binds to the loxP elements and eliminates the in-between genomic sequence by DNA recombina-
tion, creating a knockout form of the target gene. As an alternative use of the Cre-based approach,
a loxP-anked neo gene is sometimes inserted into a small intron of a target gene, which in most
cases inhibits the target gene expression by disturbing its transcription or RNA processing (such
target genes are called hypomorphic). Expression of the hypomorphic target allele can be activated
by Cre-mediated removal of the oxed neo gene. Because this activation can be performed in mice
heterozygous or homozygous for the hypomorphic alleles, it can be also used for studying gene-
dose effects.
tTA = tetracycline-regulated transactivator; rtTA-reverse tetracycline-regulated transactivator;
Floxed = exon(s) of a gene anked by loxP site; agged = proteins tagged by GFP or a short epitope
recognized by selective antibody, such as Myc or Flag -epitopes.
Conditional mutagenesis: an approach to disease models. In: Handbook of Experimental
Pharmacology (Starke K, Editor-in-chief Feil R, Series editors Metzger). D, Heidelberg: Springer Verlag.
(Continued)
 266
Box 71
Tools to Generate Conditional Mouse Mutants (Continued)

The Hippocampus Book

Box Figure 71. Principles of conditional expression systems. Left. tTA System. tTA nor POI are expressed. Right, Cre system. Upper box: -CaMKII promoter
Upper box: -CaMKII promoter regulates the expression of tTA, which can regulates the expression of Cre recombinase, which can bind to loxP elements
bind to the tet operator region (TetO7) in front of a transcription unit for a inserted by gene targeting into a mouse gene. In cells where -CaMKII pro-
protein of interest (POI) labeled with a ag (GEF) or epitopes; agged POI). In moter is not active, Cre is not expressed, which leaves expression of the oxed
neurons with active a-CaMKII promoter, the binding of tTA at TetO7 initiates gene (POI) intact. Middle box: In cells with active -CaMKII promoter, Cre is
transcription of agged POI. Middle box: In the presence of doxycycline expressed and binds to the loxP elements in both alleles of a gene. Lower box:
(DOX), the binding affinity of tTA to its operators is drastically reduced, and In Cre-expressing cells, the oxed gene segment is removed in both alleles, and
POI is not expressed. Lower box: In cells where -CaMKII is not active, neither the POI is knocked-out.
 Molecular Mechanisms of Synaptic Function 267

NR2A or NR2B but caused severe reduction of NR2A and vated and NR1-GFP expression is switched off. Using this
NR2B protein; in addition, both subunits were retained and onoff mouse model, the study showed that when NR1-
aggregated in the somata of CA1 cells, indicating that the NR1 GFP expression was switched off in CA1 at day 1 to 6 after
subunit is necessary for the release of NR2 subunits from the learning a spatial memory task retention of the acquired
endoplasmic reticulum in hippocampal pyramidal cells, and memory was signicantly impaired compared to control mice
that in the absence of NR1 no functional NMDA receptor or mice with the NR1-GFP expression switched off at post-
channels can be formed in CA1 cells (Fukaya et al., 2003). training day 9 to 14 (Shimizu et al., 2000). Contextual fear
The loss of NMDA receptor-mediated synaptic currents in memory was affected in a similar way. These experiments
CA1-specic knockout mice resulted in impaired LTP at showed that memory retrieval relies on NMDA receptor-
Schaffer collateral to CA1 synapses. Conrming the specicity mediated synaptic transmission/plasticity for the rst 6 days
of the genetic manipulation, no changes in synaptic transmis- after learning; thereafter, retrieval becomes NR1-independent.
sion or plasticity were found at other synaptic connections in In a follow-up study, -CaMKII promoter-Cre transgenic line
the hippocampus. Behaviorally, the mice showed regular with Cre expression in forebrain principal neurons was used
development and normal nonspatial learning in the water- to switch off NR1-GFP expression during memory storage, 6
maze. However, adult mice had impaired learning during the months after the initial training, and at least 2 months prior to
acquisition phase of the hidden version of the Morris water- memory retrieval. The retention of 9-month contextual and
maze and in the postacquisition probe (Tsien et al., 1996b). cued fear memories was severely disrupted by a long (30-day)
These mice thus provided the rst direct genetic evidence that but not short (7-day) doxycycline treatment suppressing
NMDA receptor-mediated synaptic plasticity at the CA1 NR1-GFP expression, indicating an unexpected role for
synapses is involved in the formation of spatial memory in NMDA receptors in long-term memory storage (Cui et al.,
mice. In addition, further analysis showed a number of other 2004). The proposed role of NMDA receptors in memory
cellular and behavioral decits. The mice failed to learn an consolidation and storage, however, contradicts earlier phar-
association between a conditional and an unconditional stim- macological experiments (reviewed in Riedel et al., 2003).
ulus (separated by 30 seconds) in trace fear conditioning, Further experiments are necessary to settle the question of
indicating that in addition to spatial information temporal NMDA receptor function in memory consolidation, storage,
information is processed in an NMDA receptor-dependent and retrieval.
manner in the mouse hippocampus (Huerta et al., 2000). Hippocampal CA3-specic NR1 knockout mice were gen-
Mice also had impaired nonspatial memory formation, erated by Cre recombinase driven from a promoter of the
including object recognition, olfactory discrimination, and kainate receptor subunit KA-1 (Nakazawa et al., 2002). In this
contextual fear memory; interestingly, the last three decits case, electrophysiological analysis of the mice revealed
were rescued by keeping the mice in an enriched environment impaired LTP in the recurrent commissural/associational
(Rampon et al., 2000). The same study showed that the pathway but not in mossy ber synapses of CA3 cells. The
enriched environment led to an increase in synapse density in mice showed regular performance in the Morris watermaze
the CA1 region in the knockouts (as well as in the control lit- but had clear decits when some of the spatial guiding cues
termates), suggesting that NMDA receptor activity is not were removed or when rapid memory formation was tested in
essential for experience-induced synaptic structural changes, a delayed matching-to-place version of the watermaze. This
and that such structural plasticity may compensate for normal led to a model proposing a critical role for the CA3 region in
synaptic plasticity at least to some extent. Further decits were lling in the blanks (pattern completion) after partial cue
found in place cell properties in area CA1: McHugh and col- removal. Similarly, CA1 pyramidal cells in these mice dis-
leagues found a striking decit in the coordinated ring of played normal place-related activity in a full-cue environment
pairs of neurons tuned to similar spatial locations. However but reduced activity upon partial cue removal and increased
CA1 cells retained place-related activity, indicating that the eld size in a novel environment (Nakazawa et al., 2002,
formation of place cells as such is not NMDA receptor- 2003).
dependent (McHugh et al., 1996). In summary, the NR1 cell-type specific knockouts
CA1-specic NR1 knockout mice were also used to study unequivocally demonstrated the critical role of NMDA recep-
the consolidation phase of learning. Shimizu et al. described a tors in synaptic plasticity and learning in both CA3 and CA1
mouse model termed iCA1-KO in which gene expression in regions of the hippocampus. This work initiated a novel strat-
CA1 neurons could be temporally controlled (Shimizu et al., egy to determine the role of specic synapses in a complex
2000). As with the previous model for NMDA receptor deple- memory system.
tion, the -CaMKII promoter-driven Cre transgene is used to
excise the oxed NR1 gene, but now it leads to simultaneous Mice with Point Mutations in the NR1 Subunit
induction of a doxycycline-sensitive transcriptional activator
(tTA) in the affected CA1 cells (see Box 71). In the absence of Two knock-in mouse models were designed to study the role
doxycycline, tTA is active and induces expression of the tTA- of glycine modulation in NMDA receptor function: NR1
controlled transgene, which in the case of the iCA1-KO of K483Q mutation causes severe reduction of glycine affinity of
NR1 was the GFP-tagged wild-type NR1 subunit (NR1-GFP). the NMDA receptors, whereas NR1 D481N mutation has a
However, when the mice are fed doxycycline, tTA is inacti- milder effect (Kew et al., 2000). Mice homozygous for the
268 The Hippocampus Book

K483Q mutation do not feed and die early after birth, essen- mediated synaptic currents in CA1 neurons. Nonetheless, LTP
tially exhibiting a phenotype similar to that of the complete at the hippocampal Schaffer collateral to CA1 synapses and
NR1 knockout. This indicates that the high-affinity glycine spatial learning were only slightly impaired. This indicates
binding site is crucial for normal NMDA receptor function. that in the absence of the NR2A subunit the NR1/NR2B
The mild reduction of glycine affinity in NR1 D481N mice heteromeric receptors are generally sufficient for adult hip-
was not lethal, and the mutants exhibited reduced theta burst- pocampal LTP and spatial learning (Sakimura et al., 1995;
induced LTP and slightly impaired acquisition in the Morris Kiyama et al., 1998). The same phenotype was also observed
watermaze. when only the NR2A C-terminal domain was genetically
The postulated function of the NMDA receptor with deleted (Sprengel et al., 1998). The truncated NMDA receptor
respect to synaptic plasticity in the hippocampus, and else- subunits NR2AC participated in synaptic hippocampal
where in the brain, is to act as a Hebbian-type coincidence NMDA receptor functions, as indicated by immunoisolation
detection mechanism for post- and presynaptic activity owing of synaptic NR2AC-containing receptor complexes,
to its dual ligand and voltage-dependent ion channel activa- NR2AC immunogold labeling at hippocampal synapses, and
tion (see Section 7.2.3); once the receptors open, Ca2 inux the existence of spinous Ca2 transients in the presence of
initiates postsynaptic signaling cascades leading to changes in NR2B blockers. LTP induced by a single tetanic stimulation in
synaptic efficacy. To test this model of NMDA receptor func- CA1 neurons was strongly reduced but could be restored to
tion directly, mice with altered NMDA receptor ion perme- wild-type levels by repeated tetanic stimulation; this paradigm
ability were generated. A single asparagine residue N598 in the thus may preferentially activate NMDA receptors containing
M2 segment of the NR1 subunit, which determines Ca2 per- the NR2B subtype. Because the NR2AC-containing recep-
meability, was changed to glutamine (Q) or arginine (R) tors allowed synaptically evoked Ca2 inux in dendritic
(Single et al., 2000). Like NR1 knockout mice, mice exclusively spines, these results indicate that NR2A C-terminal domain-
expressing the mutated NR1 allelesNR1(Q/Q) and NR1 based function (e.g., a scaffold role for various cytoskeletal
(-/R) micedied perinatally. Electrophysiological recordings and signaling proteins) is critical for normal LTP induction. In
from nucleated patches of CA1 cells in slices from NR1 juvenile (P14) mice, when NR2B-mediated LTP induction
(Q/Q) mice revealed the expected reduction of Ca2 perme- pathway was more pronounced, the NR2AC mice showed
ability and voltage-dependent Mg2 block. Furthermore, even no decits in synaptic plasticity (Khr et al., 2003). Notably,
heterozygous NR1(/Q) mice exhibited reduced life NR2A-dependent LTP as well as hippocampus-dependent
expectancy, and NR1(/R) mice displayed signs of underde- spatial and associative learning are impaired in mice with
velopment and died before weaning. This shows that Ca2 forebrain-specic knockout of the transcription factor c-fos,
inux via NMDA receptors remains critically important indicating that NR2A-dependent plasticity involves activation
throughout life. of the c-fos pathways (note that NR2B-dependent LTP
To address the importance of NMDA receptor-controlled induced by repetitive tetanic stimulation was normal)
Ca2 inux for adult hippocampal functions, the N598R (Fleischmann et al., 2003). The results from genetic mouse
mutation was expressed conditionally in the forebrain using models, however, remain to be reconciled with pharmacolog-
the Cre/lox system (see Box 71). First, expression of the dom- ical studies indicating a selective role of NR2A in LTP and
inant negative NR1(R) allele was inhibited by a oxed neo NR2B in LTD (Liu et al., 2004; but see also Berberich et al.,
gene cassette in an intron of the NR1(R) allele. In these mice, 2005). Thus, it seems most likely that the two NMDA receptor
the intronic neo gene can be deleted by Cre recombinase subunits play distinct roles in different synaptic plasticity par-
expressed, for example, from the -CaMKII promoter, result- adigms by activating selective signaling pathways; however,
ing in activation of the NR1(R) allele. Analysis of the NR1(R) further experimental work is necessary to determine the exact
forebrain-specic mice revealed a role for NMDA receptor- mechanisms and pathways.
mediated Ca2 influx in hippocampal homeostatic and NR2B knockout mice die shortly after birth owing to lack
Hebbian synaptic plasticity. The developmental synaptic of a suckling response. In mice kept alive for several days by
scaling was lacking in CA1 cells, and LTP was induced by hand-feeding, synaptic NMDA receptor currents and LTD
high-frequency but not low-frequency stimulation (Pawlak were abolished at CA1 synapses; this conrms that the NR2B
et al., 2005a,b). The mortality of these mutants, unfortu- subunit is critical for NMDA receptor function during early
nately, was still elevated; to assess the role of NMDA receptor postnatal development (Kutsuwada et al., 1996).
Ca2 signaling for learning and memory, it is thus necessary Selective truncation of the NR2B C-terminal domain in
to develop more restricted NR1(R) expression, limited, mice caused the same phenotype as complete deletion of the
for example, to expression only in CA1 or CA3 pyramidal subunit (Mori et al., 1998; Sprengel et al., 1998); homozygous
neurons. NR2BC mice died a few days after birth, NMDA receptor-
mediated EPSCs were reduced, with an accompanying reduc-
NR2 Mutant Mice tion in synaptic NR2B immunosignal in the dendritic regions
of the hippocampus. This suggests that the NR2B C-terminal
In agreement with the high expression of NR2A in mature domain is important for efficient synaptic transport of the
compared to juvenile hippocampus, adult NR2A knockout NR2B-type receptors.
mice showed a signicant reduction of NMDA receptor- A striking nding regarding the function of the NR2B sub-
 Molecular Mechanisms of Synaptic Function 269

unit and of NMDA receptors in general came from mice with currents were only slightly reduced but showed a loss of a den-
forebrain-specic overexpression of the NR2B gene, driven sity scaling toward the apical dendrites (Zamanillo et al., 1999;
from the -CaMKII promoter; these mice showed not only Andrasfalvy et al., 2003). These ndings demonstrated that
increased LTP at the CA3 to CA1 synapses but also enhanced GluR1 is necessary to establish and/or maintain an extrasy-
associative learning and memory (Tang et al., 1999). The naptic AMPA receptor pool as well as distance-dependent
authors proposed that the enhanced synaptic and behavioral synaptic scaling. The GluR2 subunit, in the absence of GluR1,
performance was primarily the result of the increased opening showed a strong reduction in overall dendritic distribution
time of the NR2B receptors, leading to overall enhanced coin- and increased somatic accumulation in CA1 principal neu-
cidence detection by NMDA receptors. rons when examined by immunostaining at the light
The phenotype of NR2C knockout mice relates to the microscopy level. However, consistent with the preservation of
prominent expression of the NR2C subunit in cerebellar synaptic currents, synaptic GluR2 expression, as detected by
granule cells: The mice show decits in ne-tuning motor immunogold labeling, was comparable to that in wild-type
coordination. The same impairment was also observed in cells. Taken together, these data indicate that in wild-type ani-
mice with the C-terminally truncated NR2C subunit, provid- mals a large proportion of AMPA receptors is maintained as a
ing a third conrmation of the importance of the C-terminal reserve pool at extrasynaptic sites, not immediately participat-
intracellular domain for normal synaptic function of each NR ing in synaptic signal transduction.
subunit (Ebralidze et al., 1996; Kadotani et al., 1996; Sprengel The function of the extrasynaptic pool of AMPA receptors
et al., 1998). No decits were observed in hippocampus-based became evident from analysis of synaptic plasticity in GluR1
functions. knockout mice; the NMDA receptor-dependent LTP at adult
The NR2D subunit shows its highest expression in the Schaffer collateral to CA1 synapses was largely abolished
embryonic brain and declines by P2. NR2D knockout mice (Zamanillo et al., 1999). Interestingly, whereas the loss of
are viable, indicating that the subunit function is not essential LTP in CA1 was observed after various stimulation proto-
for development; in addition, no synaptic or memory de- colse.g., tetanic stimulation (1100 Hz or 4100 Hz
ciencies have been described in NR2D knockouts (Ikeda et al., eld LTP) or 3-minute pairing of postsynaptic depolariza-
1995). tion and low-frequency Schaffer collateral stimulation (cellu-
In contrast to the NR2B-overexpressing mice, -CaMKII lar LTP)a protocol using theta-burst frequency of pairing
promoter-driven forebrain-specic overexpression of NR2D pre- and postsynaptic activity induced a slow-onset LTP that
did not improve CA3 to CA1 LTP or memory performance in reached normal levels within ~20 minutes after induction.
mice (Okabe et al., 1998). In NR2D-overexpressing mice, This was the rst evidence that two forms of LTP coexist at
NMDA receptor-mediated currents in CA1 pyramidal cells the Schaffer collateral to CA1 synapses: one dependent on
had smaller amplitudes and slower kinetics, with selective and the other independent of the extracellular pool of GluR1-
impairment of LTD at juvenile and of LTP at mature Schaffer containing AMPA receptors (Hoffman et al., 2002). Similarly,
collateral to CA1 synapses. Thus, expression of NMDA recep- at the perforant path to dentate gyrus granular cell synapses,
tors with NR2D-specic channel characteristicsleading to GluR1-independent LTP was detected in adult GluR1 knock-
high affinity for glutamate and glycine, slow gating, and low out mice (Zamanillo et al., 1999); this indicates that in wild-
sensitivity to Mg2signicantly altered NMDA receptor- type animals LTP at these synapses is at least partly
dependent plasticity; surprisingly, these changes did not affect GluR1-independent.
learning in the Morris watermaze task (Ikeda et al., 1995). Molecular mechanisms underlying synaptic plasticity in
the hippocampus also change during development. In con-
7.5.3 AMPA Receptor Mutant Mice trast to adult GluR1-decient mice, juvenile (2-week-old)
knockout mice have normal eld and cellular LTP in CA1
As discussed in the section about AMPA receptor trafficking (Jensen et al., 2003b). This juvenile LTP may rely on expres-
(see Section 7.4.1), the two major subunits GluR1 and GluR2 sion of other AMPA receptor subunits, such as GluR2long, a
appear to have distinct functions in hippocampal synaptic GluR2 form with an alternatively spliced C-terminal domain
plasticity. Using mouse genetics, the GluR1 subunit was (Kolleker et al., 2003) (see Section 7.3.2). The notion that
unequivocally identied as a key molecule in LTP at the hip- GluR1-dependent function is not critical in juvenile mice is
pocampal CA3 to CA1 connections (Zamanillo et al., 1999). consistent with the observation that hippocampal LTP in
The importance of the GluR2 subunit at the behavioral level adult GluR1 knockout mice can be rescued by transgenic
is at the moment less clear, and the GluR3 subunit appears not forebrain-specic expression of GFP-tagged GluR1 with an
to play a critical role in hippocampal functions. onset from the second to the third postnatal week (Mack et al.,
2001). This indicates that a GluR1-independent form(s) of
GluR1 Mutant Mice synaptic plasticity is (are) sufficient for development of nor-
mal hippocampal synaptic circuits. Later in development, as
Genetic depletion of the GluR1 subunit has major effects on the extrasynaptic AMPA receptor pool increases, GluR1-
hippocampal functions (Fig. 78). CA1 pyramidal neurons dependent plasticity becomes the dominant form of LTP
showed a dramatic loss of AMPA receptor-mediated extrasy- (Jensen et al., 2003b).
naptic currents in both the soma and the dendrites; synaptic Does hippocampal CA3 to CA1 LTP in the adult rely on
270 The Hippocampus Book

Figure 78. Various forms of LTP at CA3-to-CA1 synapses revealed stimulation and pairing LTP protocols (Jensen et al., 2003b).
by analysis of GluR1-decient mice. A. GluR1 knockout was gener- These two forms of LTP are absent at P42 in the GluR1-decient
ated by gene targeting of the M1- and M2-encoding exon 11 mice. However, theta-burst pairing can still induce LTP at P42 in
(Zamanillo et al., 1999). Right panel: GluR1 immunoreactivity was these mice (Hoffman et al., 2002). C. GluR1-decient mice show
detected in the hippocampi of wild-type (WT) but not GluR1- regular learning in the hidden platform task of the Morris water-
decient mice (-/-) at P14 and P42. B. Residual CA3-to-CA1 LTP maze (left) but cannot learn the rewarded alternation in the
is present in the GluR1-decient mice at P14, as revealed by tetanic T-maze task (Reisel et al., 2002).

any extracellular reserve pool of AMPA receptors, or does it receptors after LTP induction (Esteban et al., 2003); and sec-
specically require a GluR1-containing reserve pool? A spe- ond, gene-targeted mice carrying a mutation of GluR1 serine
cic role for the GluR1 subunit was suggested from studies 845 showed reduced LTP despite normal levels of extrasynap-
examining the function of GluR1 C-terminal phosphoryla- tic AMPA receptors (Lee et al., 2003). These results provide
tion. First, experiments in hippocampal slice cultures sug- evidence for a specic role of the GluR1 subunit in LTP at CA1
gested that PKA phosphorylation of GluR1 C-tail serine 845 is synapses.
necessary for synaptic insertion of GluR1-containing AMPA Behavioral analysis of the GluR1-decient mice revealed
 Molecular Mechanisms of Synaptic Function 271

two striking ndings. First, it came as a surprise that these hypothesis directly and to examine the overall role of GluR2
mice showed normal spatial reference memory in the Morris in synaptic transmission, mice were generated that either
watermaze task, the paddling-pool escape task, and the appet- completely lack GluR2 or contain targeted mutations elimi-
itive elevated plus-maze and Y-maze tasks (Zamanillo et al., nating the editing of the Q codon in the M2 membrane
1999; Reisel et al., 2002; Bannerman et al., 2003; Schmitt et al., segment.
2004b). Importantly, the spatial learning in the watermaze In contrast to the GluR1-decient mice, GluR2 knockout
and Y-maze was still dependent on hippocampal function, as mice are compromised in development and overall behavior.
bilateral hippocampal lesions in the GluR1-decient mice Juvenile GluR2-decient mice are about half the size of their
profoundly impaired their performance in both tasks. These wild-type littermates, and some succumb to preweaning mor-
ndings thus established that GluR1-dependent LTP in the tality. Even though the surviving mice catch up in weight by
CA1 region of the hippocampus is not necessary for the 8 weeks of age, they remain impaired in several aspects of their
formation of spatial reference memory. Second, the GluR1- behavior, ranging from reduced sensory stimulus processing
decient mice did show a drastic reduction in spatial working to reduced motivation and learning; decits are associated
memory during tests of nonmatching to place in a Y-maze and with hippocampus-dependent behaviors as well as behaviors
elevated T-maze. This impairment was partially rescued by associated with other brain regions, including the neocortex
transgenic expression of GFP-GluR1 that was largely and cerebellum. The overall lower activity of the GluR2-
restricted to the dorsal hippocampus of GluR1 knockout mice decient mice and the reduction in synaptic AMPA receptor-
(Schmitt et al., 2005). These experiments thus provide a close mediated currents (Jia et al., 1996; Meng et al., 2003) suggest
link between hippocampal CA1 LTP and a specic form of that GluR2 is necessary for maintaining a steady-state level of
memory formation, as a genetic LTP rescue in adult mice was AMPA receptor-mediated transmission, in contrast to GluR1,
paralleled by a gain of function in the behavior of the mice. which is more critical for activity-dependent synaptic plastic-
GluR1-dependent synaptic plasticity appears to be critical ity (see also discussion of subunit-specic AMPA receptor
for a form of learning that involves processing rapid interac- trafficking in Section 7.4.1).
tions between spatial cues and response requirements At the cellular level, GluR1 distribution in the absence of
(Schmitt et al., 2004b). When mice were allowed to navigate GluR2 was unchanged in the hippocampus, indicating that
through a T-maze by using textural conditional cues (all oor GluR2 is not necessary for the extracellular AMPA receptor
areas were covered with either white Perspex or wire mesh, pool in dendrites; the aberrant receptor assemblies of
depending on the position of the food reward) the knockout GluR1/GluR3 and GluR1 or GluR3 homomeric channels are
mice learned as well as wild-type mice; this is consistent with less efficiently expressed at synapses, again supporting the idea
their having normal spatial reference memory. When the oor that GluR2 is necessary for synaptic stabilization of AMPA
cues were present only in the start arm (not in the goal arms) receptors (Sans et al., 2003). However, the GluR1/3-based
of the maze, the GluR1-decient mice again failed to learn the receptors, which are permeable to Ca2, are sufficient even
task. This shows that the absence of the conditional cue at the at their low synaptic levels to mediate CA3 to CA1 LTP;
time when the place/reward association was processed criti- furthermore, this LTP is partially independent of NMDA
cally determines whether the mice can learn. Hippocampal receptor activationabout twofold enhanced compared to
GluR1-dependent synaptic plasticity at the CA1 region thus that of their wild-type littermates and nonsaturating (Jia et
contributes to a memory system that in rodents encodes both al., 1996). Thus, it appears that activity-dependent GluR1-
the spatial and temporal contexts associated with a particular based synaptic delivery of AMPA receptors is still operative,
event, a paradigm that may be analogous to human episodic but now changes in synaptic transmission can be achieved in
memory (Schmitt et al., 2004a). an atypical manner, without relying on NMDA receptor-based
The lack of GluR1 in the brain, despite the prominent, coincidence detection. It is unfortunate that the overall low
widespread expression of the subunit, does not have major behavioral performance of the mice makes it difficult to ana-
effects on general sensorimotor behavior. The GluR1 knock- lyze learning under these altered synaptic conditions. One
out mice are slightly more hyperactive, anxious, and aggres- approach to getting around this problem was to analyze hip-
sive, with only a subtle lack of motor coordination pocampal place cell ring, as there is a strong correlation
(Bannerman et al., 2004; Vekovischeva et al., 2004). between Hebbian NMDA receptor-dependent LTP, place-cell
function, and learning and memory. In the GluR2 knockout
GluR2 and GluR2/3 Mutant Mice mice, CA1 place cells were impaired, with only 25% (down
from 46%) of CA1 neurons showing detectable place elds;
According to the model for NMDA receptor-dependent LTP moreover, the remaining place cells were less precise and
at Schaffer collateral to CA1 synapses, it is important that unstable (Yan et al., 2002). This suggests that CA1 depletion of
AMPA receptors have low permeability for Ca2, which is GluR2 leads to impaired hippocampal learning and memory
determined by the presence of the Q to R edited GluR2 sub- functions.
unit (see Section 7.3.2); NMDA receptor-mediated Ca2 Mice lacking GluR2 selectively in postnatal forebrain were
inux then acts as a selective trigger for Hebbian coincidence generated to directly circumvent the global behavioral impair-
detection of pre- and postsynaptic activity. To test this ments in the complete knockout animals (a oxed GluR2 gene
272 The Hippocampus Book

was again excised by CRE recombinase under control of the was increased but was still fully dependent on activation of
-CaMKII promoter (Shimshek et al., 2005, 2006). The mice NMDA receptors (Meng et al., 2003). In addition,
showed less overall behavioral decits but had impaired spa- GluR2/GluR3 double knockout mice showed a phenotype
tial reference and working memory; interestingly, LTP at similar to that of the GluR2-decient mice, with general retar-
the CA3 to CA1 synapses was induced to a wild-type level dation of postnatal development and increased mortality,
of potentiation and was entirely dependent on activation without detectable abnormalities in the gross anatomy of the
of NMDA receptors.. Detailed analysis of hippocampal central nervous system, including the hippocampus (Meng et
functions in these mice revealed the following morphologi- al., 2003). Although synaptic transmission at CA1 synapses
cal and synaptic changes: loss of CA3 pyramidal cells and was reduced several-fold, the GluR2/3-decient mice were
parvalbumin-positive interneurons in the dentate gyrus; competent in establishing LTD and LTP as well as depotentia-
diminished neurogenesis in the subgranular zone; reduced tion and dedepression at the Schaffer collateral to CA1
presynaptic ber excitability at CA3 to CA1 synapses and synapses. This indicates that GluR1 receptors are sufficient, in
reduced threshold for evoking population spikes at the CA3 to the absence of GluR2/3, to mediate the major types of synap-
CA1 synapses. tic plasticity in CA1 principal neurons. The overall behavioral
At least some of the changes observed in the forebrain- retardation of the animals, however, prevented tests of learn-
specic GluR2 knockout mice are likely due to aberrant Ca2 ing and memory in the double knockout mice.
inux through the remaining GluR1/3 receptors; this is also
supported by observations in mice with Ca2-permeable (but 7.5.4 Kainate Receptor Mutant Mice
GluR2-containing) AMPA receptors due to genetically
induced deciency in the GluR2 Q to R editing (Brusa et al., At present, there are no reports on mice lacking the KA1 sub-
1995; Feldmeyer et al., 1999; Krestel et al., 2004; Shimshek unit; however, the other kainate receptor subunit mutants
et al., 2005). In these mice, AMPA receptors are stabilized have been extremely useful in mapping kainate receptor func-
by GluR2 at excitatory synapses, and the overall Ca2 load tion in different hippocampal cell types and at distinct subcel-
during neurotransmission is larger. When the unedited GluR2 lular localizations (see also Section 7.3.4).
is expressed in the whole brain, the mice develop severe
seizures and typically die by 3 weeks of age (Brusa et al., 1995). KA2, GluR5, and GluR6 Knockout Mice
The severity of the phenotype correlates with the level of
GluR2(Q586) expression, as mice with GluR2(Q586) Genetic depletion of KA2 identied this subunit as an impor-
expressed in the brain at only the 5% level of the endogenous tant functional component of both pre- and postsynaptic
GluR2 survive substantially longer and show a low tendency kainate receptors at the mossy fiber to CA3 synapses
to epileptic attacks (Feldmeyer et al., 1999). In young mice, the (Contractor et al., 2003). First, whereas the function of the
deciency of GluR2 Q to R editing induces an NMDA recep- presynaptic facilitatory kainate receptor, which is responsible
tor-independent LTP at Schaffer collateral to CA1 synapses, for the marked frequency facilitation at the mossy ber termi-
the magnitude of which correlates with the expression level of nals, was normal, heterosynaptic facilitation resulting from a
the GluR2(Q586) subunit (Feldmeyer et al., 1999). spillover of glutamate from CA3 collateral synapses was
Interestingly, lethality and seizures do not occur in mice absent. This nding agrees with the idea that KA2 plays a role
with GluR2(Q586) expression conditionally restricted to as a high-affinity glutamate-binding subunit, specically
postnatal hippocampus, suggesting that additional expression required for sensing low concentrations of glutamate. Second,
of GluR2(Q586) in forebrain, striatum, and/or amygdala is postsynaptic kainate receptor-mediated EPSCs in CA3
necessary for global epileptic attacks (Krestel et al., 2004). pyramidal cells showed faster decay. Mossy ber LTP in these
In mice with forebrain-restricted GluR2(Q586) expression, mice, however, was normal (Contractor et al., 2003).
the presynaptic ber excitability at CA3 to CA1 synapses is Whereas the GluR5 subunit is prominently expressed in
unchanged, but excitatory postsynaptic responses are smaller, hippocampal interneurons, disruption of the GluR5 gene did
suggesting compensatory downscaling of the AMPA receptor- not cause a loss or reduction of functional kainate receptors in
mediated Ca2-permeable currents; nonetheless, CA1 neu- these cells; this suggests that other kainate receptor subunits
rons are more excitable, as revealed by a lower threshold for an can compensate for the lack of GluR5 in hippocampal
evoked population spike and epileptic activity in these mice; interneurons (Mulle et al., 2000). In contrast, kainate recep-
increased gliosis and mossy ber sprouting are clear signs of tor-mediated enhancement of the amplitude of synaptic cur-
recurrent seizures. Thus, in both postnatally induced fore- rents at the perforant path to CA3 synapses was abolished in
brain-specic GluR2 knockout and editing-decient animals, the GluR5 and the GluR6 knockout animals (Contractor et al.,
an AMPA receptor-mediated increase of a postsynaptic Ca2 2000).
load induces long-term changes in hippocampal circuits Gene-targeted mice lacking GluR6 revealed that the sub-
affecting normal hippocampal functions and spatial learning. unit plays a critical role at the mossy ber to CA3 synapses,
GluR3 is not as highly expressed as GluR1 and GluR2 in where postsynaptic kainate receptors mediate a small EPSC
the hippocampus. GluR3 knockout mice showed normal basal component that is completely abolished in the GluR6-de-
synaptic transmission in CA1; LTP at CA3 to CA1 synapses cient mice (Mulle et al., 1998). Importantly, synaptic plasticity
 Molecular Mechanisms of Synaptic Function 273

of mossy ber connections is also affected; mossy ber LTP mGluR1 null mice showed impaired context-specic associa-
the expression of which is known to depend on presynaptic tive learning (Aiba et al., 1994; Conquet et al., 1994); note that
mechanismsis reduced, and short-term synaptic facilitation the mice showed much stronger decits in cerebellar plasticity
is impaired in the absence GluR6. This suggests that kainate and related behavior. Mice decient in the mGluR5 subunit
receptors act as presynaptic autoreceptors that mediate fre- had a phenotype similar to that of the mGluR1 null mice, with
quency-dependent synaptic facilitation at the mossy ber ter- reduced LTP in CA1 neurons, intact LTP at the mossy ber to
minals, a form of short-term plasticity in which the strength CA3 neurons, and somewhat impaired spatial learning and
of transmission increases with repetitive stimulation contextual fear conditioning. Thus, these animal models
(Contractor et al., 2001). In addition, several other roles for reveal that group I mGluRs play at least modulatory roles in
GluR6 in the hippocampus were revealed during analysis of hippocampal learning and memory.
the GluR6 knockout mice. Kainate receptor-mediated modu- Mice with targeted deletions of the group II/III mGluRs
lation of synaptic transmission between inhibitory interneu- were also generated in a number of laboratories during the
rons in the stratum radiatum was impaired; however, a loss of late 1990s. Among them, mice lacking the mGluR2 subtype, in
postsynaptic kainate currents was observed only in double agreement with its prominent expression in mossy bers,
GluR5/GluR6 knockouts, suggesting the existence of two pop- show a loss of mossy fiber LTD (Yokoi et al., 1996).
ulations of kainate receptors in hippocampal interneurons Interestingly, in the same study, no decits were found in
(Mulle et al., 1998, 2000). GluR6-decient mice also provided mossy ber LTP or in watermaze tasks, suggesting that neither
evidence of a functional role of GluR6-containing kainate mGluR2 nor mossy ber LTD are required for hippocampal
receptors in CA1 cells, which showed reduced kainate- and spatial learning. Mice decient in mGluR7 were found to be
domoate-induced CA1 inward currents in the mutant mice. more prone to developing seizures, with a concomitant
increase in population excitability of principal neurons in
GluR5(Q636R) and GluR6(Q636R) Mutant Mice hippocampal slices (Sansig et al., 2001). Thus, mGluR7 may
play an important role in presynaptic control of hippocampal
The Ca2 permeability of GluR5- and GluR6-containing excitability.
receptors is controlled by site-selective RNA editing in the
coding region for the channel pore-forming segment M2 (see 7.5.6 Synopsis of the Section
Section 7.3.4). Mice with decient editing of the GluR6 Q/R
site show several phenotypic changes at the levels of synaptic Studies on genetic mouse mutants with altered excitatory
transmission and behavior. First, LTP at medial perforant path neurotransmission in the hippocampus have greatly enhanced
to dentate gyrus synapses had an NMDA receptor-independ- our understanding of the principles of glutamatergic synaptic
ent component, suggesting that, at this synapse too, kainate transmission and synaptic plasticity, as well as of complex
receptor-mediated Ca2-signaling can lead to an atypical hippocampus-dependent behavior, learning, and memory.
NMDAR coincidence-decient form of synaptic plasticity First, work with NMDA receptor mutant mice has provided
(Vissel et al., 2001). The GluR6(Q) mice were also more vul- important evidence for a causal link between NMDA recep-
nerable to kainate-induced seizures than were their wild-type tor-mediated LTP at CA3 to CA1 synapses and spatial mem-
littermates. Together with the seizure-prone phenotypes of ory formation, as well as between this form of LTP and the
the GluR2(Q856) mice, these results provide further support formation of stable hippocampal place elds. In addition,
for the hypothesis that increased postsynaptic Ca2 inux and NMDA receptor mutant mice have also revealed that different
NMDAR coincidence-decient synaptic plasticity lead to NMDA receptor subtypes contribute to different signaling
changes in hippocampal neuronal circuits, resulting in epilep- pathways for LTP induction, such as the NR2A- and NR2B-
tic activity. In addition, the GluR6(Q) mice showed that specic pathways. During juvenile development the NR2B
GluR6-containing receptors are located on dendrites of den- pathway is dominant, whereas in adult mice the NR2A path-
tate gyrus granule cells as well as on mossy ber boutons. No way is more prominent.
phenotypic changes as yet have been found in the GluR5(Q) The importance of hippocampal LTP in learning and
mice carrying the same mutation as the GluR6(Q) mice memory, as well as the existence of subunit-specic forms of
(Sailer et al., 1999). LTP in CA1 neurons were further substantiated by experi-
ments using mice with altered AMPA receptors. Adult GluR1-
7.5.5 mGluR Mutant Mice decient mice showed complete loss of eld CA1 LTP, which
correlated with spatial working memory impairment. The fact
The role of group I metabotropic receptors in synaptic plas- that spatial reference memory in these mice is intact suggests
ticity is well documented at both mossy ber to CA3 and that different forms of NMDA receptor-dependent plasticity
Schaffer collateral to CA1 synapses (see Chapter 10). may underlie different forms of hippocampal memory. In
Accordingly, mice lacking the mGluR1 subtype show reduced juvenile mice, GluR1 is not essential for normal synaptic plas-
LTP at the Schaffer collateral CA1 synapse (Aiba et al., 1994) ticity; juvenile LTP may be mediated by other AMPA receptor
and possibly also at the mossy ber to CA3 synapse (Conquet subunits (e.g., GluR4 and/or GluR2long). Genetic conversion of
et al., 1994; but see also Hsia et al., 1995). Behaviorally, the AMPA receptors to Ca2-permeable ion channels in pyrami-
274 The Hippocampus Book

dal neurons by depletion of GluR2 or by the inhibition of vates GABAB receptors more effectively than single stimuli:
GluR2 Q/R site editing leads to long-term changes in hip- Enough GABA has to build up and diffuse away from the cleft.
pocampal activity with increased excitability of CA1 cells and GABAB receptors are perhaps optimally activated by short
subsequent epileptic activity. This points to the critical impor- bursts of inhibitory afferent activation coming from groups of
tance of the edited GluR2 subunit, which determines the interneurons. However, even the GABA released by a single
monovalent ion selectivity of the AMPA receptors in estab- interneuronprovided it res at least three action potentials
lishing and/or maintaining normal hippocampal functions at high frequencyis sufficient to activate postsynaptic
and the survival of the animal. Finally, genetically modied GABAB receptors. On the other hand, extrasynaptic activation
mice showed unequivocally that kainate receptors act as of GABAA receptors is likely to be continually present, consti-
important modulators of synaptic transmission in the hip- tuting background conductance.
pocampus; their function is critical for maintaining normal The functions of background inhibitory conductances in
synaptic transmission, especially for the dentate gyrus and neuronal circuits are unknown. Small, persistent tonic chlo-
CA3 neurons. ride conductances alter input resistance and membrane time
In summary, it seems clear that mouse genetics will con- constants; these in turn inuence synaptic efficacy and inte-
tinue to be an important source for unraveling complex gration. Manipulating background inhibitory currents, for
molecular processes that underlie hippocampal functions. example, may inuence epileptic thresholds. Additionally,
Currently, the global gene knockout strategy is being replaced allosteric modulators active on GABAA receptors, such as ben-
by an approach aimed at generating mice decient for recep- zodiazepines, steroids, and general anesthetics, may be more
tor subtypes at specic hippocampal connections and at potent on extrasynaptic than synaptic receptors because of the
dened developmental stages. This will enable geneticists to much lower GABA concentration outside the synapse
study the role of glutamate receptors at distinct hippocampal (allosteric modulators produce effects at the receptor only at
connections without the interference of developmental dis- submaximal GABA concentrations).
turbances. These new generations of mouse mutants promise GABA transporter activity helps set the extracellular GABA
to enhance our understanding of the relations between vari- level and so governs the extent of extrasynaptic tonic and
ous forms of hippocampal synaptic plasticity and hippocam- perisynaptic GABAA receptor activation and presynaptic
pus-dependent learning and memory. GABAB activation. For example, in the hippocampus of GABA
transporter-1 (GAT-1)-decient mice, complex layers of feed-
back emerge: A large increase in tonic postsynaptic GABAA
conductance occurs in these mice, but the frequency of spon-
7.6 GABAergic Receptors: Structure, taneous quantal GABA release is one-third of normal (Jensen
Function, and Hippocampal Distribution et al., 2003a).
In summary, the GABAergic system in the hippocampus
7.6.1 Introduction: Synaptic and Extrasynaptic has various modes. First, there is point-to-point GABAA
GABAergic Receptors Mediate Tonic and receptor-mediated synaptic transmission and a continually
Phasic Inhibition in the Hippocampus present background of extrasynaptic tonic GABAA-mediated
inhibition. In this case, interneurons act as individual units to
The ow of excitation through the hippocampus (dentate release GABA. This mode does not provide enough GABA to
granule cells to CA3 pyramidal cells to CA1 pyramidal cells) is activate GABAB receptors (because of their low affinity/effi-
paced by GABA pulses from interneurons. A typical synaptic cacy). Second, during oscillatory activity, with synaptic activ-
pulse of GABA (1 mM, 1 ms) initiates fast (milliseconds) ity from many inhibitory neurons, substantially more GABA
point-to-point inhibition via ionotropic type A (GABAA) is released, allowing the concentration of GABA to build up
receptors forming Cl-permeable channels. This fast GABAA extrasynaptically. When this diffuse mode comes into play, a
receptor activation controls action potential frequency and new population of receptors, extrasynaptic GABAB receptors,
inhibits dendritic Ca2 spikes. Slower responses (e.g., late is recruited as well as more extrasynaptic GABAA receptors.
inhibitory postsynaptic potentials caused by opening of Specic behavioral states might use different levels of diffuse
K channels), lasting 500 to 2000 ms, originate from inhibitory transmission.
metabotropic GABAB receptors and extrasynaptic or perisy-
naptic GABAA receptors. Both receptor types can be pre- or 7.6.2 GABAA Receptors
postsynaptic, as well as extrasynaptic or perisynaptic
(reviewed in Kullmann et al., 2005). Indeed, there is likely to GABAA receptors are anion-permeable channels, with an
be substantial extrasynaptic activation of both GABAA and HCO3-/Cl- permeability ratio of approximately 0.2 to 0.4. In
GABAB receptors via GABA spillover. Extrasynaptic GABA adult cells, Cl- ions usually move into the cell to produce
does not provide a clean time-dependent signal, unlike the strong inhibitory hyperpolarization, as the reversal potential
phasic GABA pulses at the synapse. Hence extrasynaptic inhi- for Cl- is 15 to 20 mV more negative than the resting mem-
bition is termed tonic inhibition. Repetitive stimulation acti- brane potential. The Cl- gradient is maintained by K/Cl co-
 Molecular Mechanisms of Synaptic Function 275

transporters. HCO3- ions move out of the cells through the Some receptors specialize as extra or perisynaptic sensors of
GABAA receptor channel; the HCO3- efflux is mildly depolar- GABA; others enrich in particular synaptic locations. For
izing (HCO3- has a reversal potential of 12 mV), but this is some types of extrasynaptic or perisynaptic receptors that
normally offset by the Cl- hyperpolarization. The HCO3- contain the  subunit, the key properties are high affinity for
efflux is encouraged by the actions of extracellular carbonic neurotransmitter and limited desensitization, enabling them
anhydrase; in fact, these enzymes are essential for maintaining to contribute to tonic background conductance. Another
the HCO3- gradient. With an increased HCO3-/Cl- permeabil- important consequence of differences in subunit expression
ity ratio, the outward HCO3- ux would depolarize the mem- between cells or on different parts of the same cell is that
brane at the resting membrane potential. Depending on the GABAA receptor kinetics differ at different synapses (from,
local internal Cl- concentration, Cl- ions can also move out for example, distinct interneurons synapsing onto the pyram-
after GABAA receptor activation and depolarize the cell; this idal cell). This may inuence the oscillation frequency of
happens especially during embryonic and postnatal hip- principal cells, as different interneuronal subtypes innervate
pocampal development. Even in the adult hippocampus, different domains of pyramidal or dentate granule cells or
when interneurons are strongly stimulated in the presence of each other (Bartos et al., 2002). In recombinant n 2 com-
ionotropic glutamate receptor antagonists, a biphasic postsy- binations, the  subunits determine receptor kinetics
naptic GABAergic response is produced in pyramidal cells. An (Goldstein et al., 2002). The decay time constants of 1 2
initial hyperpolarizing IPSP is followed by slow depolarization subunits are fast, whereas those of 2- and 3-containing
of the pyramidal cell. The slow depolarization causes action receptors are slower; for example, currents from recombinant
potential ring, and so is excitatoryit is termed a GABA- 3 22 receptors decay more slowly than 1 22 receptors
mediated depolarizing postsynaptic potential (GDSP) (Kaila in response to a synaptic pulse of GABA. The 1 subunit
et al., 1997). knockout mice provide direct support for this. Synaptic
GABAA receptors form as pentameric assemblies of sub- GABAA receptor-mediated currents in hippocampal neurons
units (reviewed in Ernst et al., 2003). A gene family encodes (pyramidal and interneurons) from wild-type mice decay
the subunits (16, 1 3, 13, , , , , and 1-3). The more rapidly than those in 1 knockout mice; loss of the
GABAA receptor subunits belong to the acetylcholine receptor 1 subunit in synaptically localized receptors leads to a popu-
subunit gene superfamily; consequently, our working models lation of synaptic receptors, probably containing the 2
of the GABAA receptor structures are based on the extracellu- subunit, with slower deactivation kinetics (Goldstein et al.,
lar domain of the nicotinic acetylcholine receptor structure. 2002). Differences in GABAA receptor subunit expression
As for other brain regions, the GABAA receptor subunit genes between hippocampal cells and synapses likely directly inu-
are differentially transcribed in the hippocampus (Fig. 79) ence circuit properties. Parvalbumin interneurons, for exam-
(Persohn et al., 1992; Wisden et al., 1992; Sperk et al., 1997). ple, express mainly 1 2-containing GABAA receptors
Genes expressed at signicant levels in the hippocampus are: (Klausberger et al., 2002). The fast GABAA-conveyed inhibi-
1 2 4 5, 1- 3, 2 and . The subunits assemble into dif- tion between these cell types helps pace the oscillation
ferent combinations, depending on the cell-type. The large frequency in the pyramidal cell network (Bartos et al., 2002).
family of subunit genes produces considerable receptor diver- Similarly, the mean decay time constant of the IPSC at the
sity. These GABAA receptors differ in their affinity for neuro- basket cell to basket cell synapse in the dentate gyrus is
transmitter and allosteric modulators, activation rate, twofold faster than at the basket cell to granule cell synapse
desensitization rate, channel conductance and location on the (Bartos et al., 2001). Again, this almost certainly reects dif-
cell (reviewed in Korpi et al., 2002). ferences in receptor subunit expression.
Principal GABAA receptor subtypes relevant for hip-
pocampal function contain: 1 2, 2 2, 3 2, 4 2, Distribution of GABAA Receptor Subunits
5 2, and   (mainly 4 ). Most hippocampal GABAA in Hippocampal Dentate Granule Cells
receptors are probably  2 combinations, and a small num-
ber are  : The subunit ratio in the receptor pentamer is Inhibitory terminals originate from several types of interneu-
probably 2/2 /1 or 22 1 (Ernst et al., 2003). Note that ron. Of the total number of inhibitory GABAergic terminals
some receptors may contain a mixture of  and subunits synapsing onto hippocampal granule cells, 25% occur on the
(e.g., 12 22, and perhaps x4  where x is 1, 2, 3, or 5). cell body or axon initial segments, with the remaining 75%
The single channel conductance of  2 or   receptors is synapsing onto dendrites in the molecular layer (Halasy and
25 to 30 pS. As discussed in detail below, receptors with the 2 Somogyi, 1993). At GABAA receptors on the somatic synapses
subunit can be both synaptic and extrasynaptic; receptors of rat dentate granule cells, there is a high probability of
with the  subunit are probably principally perisynaptic, receptor opening (Po  0.8), and all receptors are occupied by
localized around the edge of synapses (Wei et al., 2003). the released transmitter (reviewed in Nusser, 1999); these
The question of the function(s) of GABAA receptor diver- synapses have, on average, 26 GABAA receptors, with an aver-
sity has been partially addressed by pharmacological and age conductance of 21 pS. Rat dentate granule cells have been
genetic studies (reviewed in Rudolph and Mohler, 2004). shown to express an extensive set of subunit mRNAs: 15,
276 The Hippocampus Book

Figure 79. Expression of GABAA receptor subunits in the rat hippocampus. Expression pat-
tern of the 12 GABAA subunits in the adult dorsal rat hippocampus was visualized by immuno-
histochemistry with subunit-specic antibodies. Note: the 2 subunit staining is articially
reduced because of the low antibody concentration used. (Source: Modied from Sperk et al.,
1997, with permission.)

1 3, and 13, although the 3, 1, and 3 mRNAs are dentate granule cells: Double labeling immunocytochemistry
rare (Persohn et al., 1992; Wisden et al., 1992; Brooks-Kayal et shows that most of the 2-positive synapses also contains 1.
al., 2001). The 2 and 4 subunit mRNAs are strongly Autoradiography with an 5-selective ligand, [3H]L-655,708,
expressed; the 1 and 5 mRNAs are at moderate levels. Of highlights the molecular layer of the dentate gyrus, conrm-
the s, the 3 gene has the highest expression level, but 1 ing 5 protein enrichment on granule cell dendrites (Sur et
and 2 are also signicantly expressed. al., 1999). The 4 and/or 1 and  subunits probably assem-
Immunocytochemical results generally agree with the in ble as perisynaptic 4 ,, 1 -, or 14 -type receptors
situ hybridization data. The 2, 4, 5, 1, 3, and  protein localized around the edge of the synapse (Peng et al., 2002;
staining is high in the dentate molecular layer, indicating a Wei et al., 2003). These receptors, based on analysis of their
dendritic location on granule cells (Pirker et al., 2000; Peng et cloned counterparts, are predicted to be high-affinity and
al., 2002). There is weaker staining of the granule cell soma. nondesensitizing, designed for their role in sensing extrasy-
Many of the  subunits co-localize in the same synapses on naptic GABA and contributing to tonic inhibition.
 Molecular Mechanisms of Synaptic Function 277

Kullmanns group has identied presynaptic GABAA recep- 3400 times higher on  receptors than on  2 receptors
tors on mossy bers of dentate granule cells that inhibit (reviewed in Hosie et al., 2003). Thus the 2 subunit lowers
axonal excitability; these receptors are tonically active in the the sensitivity of the GABAA receptor complex to zinc. Hosie
absence of evoked GABA release or exogenous agonist appli- et al. suggested that the concentrations of extracellular zinc
cation (Ruiz et al., 2003). In the presence of glutamate recep- occurring in the hippocampus inhibit   receptors but do
tor blockers, stimulation of mossy fibers provokes not affect  2 receptors (presumably 1-, 2-, 3-contain-
monosynaptic GABA-mediated responses in their target cells; ing), so zinc would be a specic modulator of tonic inhibition.
this mossy ber-released GABA may also contribute to nega- They hypothesized that in addition to its role in promoting
tive feedback via the presynaptic GABAA receptors. synaptic targeting and single channel conductance the 2 sub-
Alternatively, or in addition, these receptors may sense ambi- unit evolved to retain the delity of GABAergic inhibition in
ent GABA released from interneurons, allowing axons from the presence of zinc. Nevertheless, zincs potency is also  sub-
granule cells to rapidly integrate the activity of surrounding unit- and  subunit-dependent. The IC50 for zinc inhibition of
neurons. 4 3 and 4 32 receptors is, in fact, similar (2 M).
What is the identity of this presynaptic receptor? It is acti- Blood alcohol levels of 1 to 3 mM can result from drinking
vated by muscimol, is antagonized by SR 95531, and has half a glass of wine or less. Ethanol inuences various chan-
GABA responses potentiated by zolpidem (0.2 M) (Ruiz et nels, including the NMDA glutamate receptor; but among
al., 2003), a selective agonist for 1 22 receptors if used at GABAA receptor subunit combinations, low concentrations of
the appropriate concentration, although at higher concentra- ethanol (about 3 mM, a concentration six times lower than
tions zolpidem also potentiates GABAs actions at 2 2 and the legal blood alcohol limit for driving in many states in the
3 2 receptors. The zolpidem sensitivity of these presynap- United States) specically potentiate GABA responses of
tic receptors suggests an 1 2 or 2 2 combination. Ruiz cloned 4 2 receptors expressed in Xenopus oocytes; this
et al. found punctate 2 subunit immunoreactivity in rat stra- effect is subunit-dependent, as 3 subunits are needed for
tum lucidum, the termination zone of mossy bers (Ruiz et maximal sensitivity to ethanol (Sundstrom-Poromaa et al.,
al., 2003); however, immunoreactivity for the 3 subunit was 2002; Wallner et al., 2003).
also reported at the rat mossy bers (Pirker et al., 2000).
Several modulators of hippocampal GABAA receptors were Distribution of GABAA Receptors
shown to have special relevance to dentate granule cells. in Hippocampal Pyramidal Cells
Neuroactive steroids modulate GABAA receptor function in
the hippocampus and other brain regions. Naturally occur- A pyramidal cell is covered with GABAergic terminals; a typi-
ring steroid metabolites form locally in the brain, independ- cal rat CA1 pyramidal cell receives around 1700 GABAergic
ently of their peripheral concentrations. 5-Reductase synapses, with the highest density on the perisomatic region
transforms progesterone to 5-DPH, which in turn is reduced (Megias et al., 2001). Inhibition on the various domains
by 3-hydroxysteroid oxidoreductase to allopregnanalone. affects different aspects of pyramidal cell function. GABAA
Allopregnanalone potently activates GABAA receptors. No receptor-mediated IPSCs can be generated along the whole
absolute specicity of neurosteroids for particular GABAA somatodendritic domain and on the axon initial segment of
receptor subunit combinations has been found. Many GABAA CA1 pyramidal cells (Maccaferri et al., 2000). The apical den-
receptors are sensitive to the steroid tetrahydrodeoxycorticos- dritic trunk has a high density of GABAergic terminals relative
terone (THDOC), but receptors with the  subunit are partic- to the rest of the dendrite (Papp et al., 2001); strong
ularly sensitive: 30 nM THDOC enhances the peak currents of GABAergic stimulation on this apical trunk region isolates the
1 3 GABAA receptors (with 1 M GABA) by up to 800%; dendritic compartment from the cell body. Inhibition on the
currents of other receptor isoforms (e.g., 1 32) are more distal dendrites controls Ca2 spike propagation.
enhanced by less (1550%) (Mihalek et al., 1999; Wohlfarth et Inhibiting the axon initial segment powerfully clamps down
al., 2002; Stell et al., 2003). Thus, endogenous allopreg- the pyramidal cells activity, as this is where action potentials
nanolone may act on perisynaptic   GABAA receptor iso- initiate: A typical rat CA1 pyramidal cell has more than 25
forms on dentate granule cells to increase basal levels of GABAergic terminals per 50 m of the axon initial segment
inhibition. Mice without functional  subunits have decreased (AIS). Thus, GABAA receptors in the AIS deliver inhibition
sensitivity to the sedative/hypnotic, anxiolytic, and pro- that controls the overall level of output activity of pyramidal
absence effects of neuroactive steroids (Mihalek et al., 1999). cells.
Zinc inhibits GABAA receptors. Glutamatergic neurons Hippocampal pyramidal cells assemble many GABAA
provide most of the extracellular zinc in the hippocampus. receptor subtypes (Fritschy and Brunig, 2003). By in situ
Zinc is stored in synaptic vesicles, notably in mossy ber ter- hybridization and immunocytochemistry, rat pyramidal cells
minals; strong stimulation of mossy bers induces co-release are seen to express the 1, 2, 4, 5, 1, 2, 3, 1, and 2
of glutamate and zinc into the synaptic cleft. In hippocampal genes; expression of 3 and 1 genes is present but weak
neurons, zinc can reduce the amplitude, slow the rise time, (Wisden et al., 1992). For the  subunits, 2 mRNA is the
and accelerate the decay of mIPSCs (e.g., Wei et al., 2003). On most abundant; of the genes, 1 and 3 are more abundant
recombinant GABAA receptors, zinc has an inhibitory potency than 2; 2 mRNA is at about the same level as 1 and 2
278 The Hippocampus Book

(Wisden et al., 1992). The in situ hybridization and immuno- diazepam. Thus, by pharmacological inference, 1 2/32
cytochemical data are in agreement. 1 immunoreactivity is receptors dominate postsynaptic responses in synapses
enriched at the CA2 pyramidal cell boundary (Pirker et al., formed on pyramidal cells from fast-spiking cells, whereas the
2000), as is 1 mRNA (Wisden et al., 1992). 2 (and 3) subunits also contribute to receptors, together
As seen by immunocytochemistry with the light micro- with 1, in the regular spiking basket cell to pyramidal cell
scope, 1, 2, 5, 1, 2, 3, and 2 proteins distribute synapses. This is directly supported by anatomical work (see
evenly on pyramidal cell dendrites and are sparser on the below).
cell body (Sperk et al., 1997). Autoradiography with the 5- Indirect evidence for 4 2-type GABAA receptors partic-
selective ligand [3H]L-655708 highlights the stratum radia- ipating in synaptic transmission on the soma of CA1 pyrami-
tum, conrming 5 protein enrichment on pyramidal cell dal cells comes from studies on chronic intermittent ethanol
dendrites (Sur et al., 1999). Subcellularly, the 5 subunit (CIE)-treated rats. In this model, rats are exposed to intermit-
appears predominantly extrasynaptic, with about 25% synap- tent episodes of ethanol and then ethanol withdrawal, pro-
tic component (Brunig et al., 2002; Crestani et al., 2002). Note ducing a kindled-like state of excitability; this is accompanied
that the reduction in spontaneous IPSC amplitude on CA1 by changes in GABAA receptor expression in, among other
pyramidal cells in hippocampal slices of 5 knockout mice areas, the hippocampus (Cagetti et al., 2003). In CA1 pyrami-
supports the conclusion that at least some 5 subunits are dal cells of CIE-treated rats, mIPSCs become insensitive to
synaptically located (Collinson et al., 2002). diazepam but can still be modulated by the benzodiazepine
As already mentioned, the cell bodies of pyramidal cells do bretazenil (diazepam is active on 1 2 receptors, but not
not heavily stain with GABAA receptor subunit-specic anti- 4 2 receptors, whereas bretazenil is active on both). In CA1
bodies. However, gold-labeled antibodies combined with pyramidal cells of control rats, mIPSCs are enhanced only
immunocytochemistry at the electron microscopic level have slightly by the benzodiazepine partial inverse agonist Ro 15-
revealed GABAA subunits at synapses on the cell body; these 4513 but are substantially enhanced by this drug in CIE-
somatic GABAergic synapses do not all have the same GABAA treated rats.
receptor subunit composition (Nusser et al., 1996; Somogyi et GABAA receptors also show differential subunit distribu-
al., 1996; Nyiri et al., 2001) (see below). tion at different synapses of the same pyramidal cell. On hip-
GABAA receptors are also differentially sorted to specic pocampal pyramidal cells, the 2 subunit is enriched in the
pyramidal cell synapses. This is indicated by three levels of axon initial segment but is present at only a small number of
evidence: kinetically distinct IPSPs depending on subcellular cell body synapses and synapses on dendrites (Nusser et al.,
location; GABAA receptors with different pharmacology at 1996; Fritschy et al., 1998). The axon initial segment synapse
different synaptic inputs; and direct visualization of different also contains the 1 subunit and possibly the 4 and 5 sub-
receptor subunits at different synapses and extrasynaptic sites. units. GABAA receptors at the axon initial segment are posi-
Different synaptic locations on pyramidal neurons have tioned to exert strong inhibitory influence over action
kinetically distinct IPSCs (Banks et al., 1998). At least two dif- potential generation.
ferent IPSCs are found on CA1 pyramidal cells. A fast compo- Synapses formed by two types of basket cell on CA1
nent (decay time constant of 9 ms) on the soma is mediated pyramidal cells contain distinct GABAA receptor subtypes:
by basket cells and other interneurons; the slow component Only those synapses coming from parvalbumin-
is dendritic, with a decay time constant of 50 ms, and is acti- negative/CCK and VIP-positive basket cell interneurons con-
vated by interneurons in the stratum lacunosum-moleculare. tain the 2 subunit; synapses from parvalbumin-positive
These two decay types may be due to different receptors, basket cells are often 2-immunonegative (Nyiri et al., 2001).
although there are no subunits on the dendrite that are not on This is preferential targeting or enrichment of the 2 in a par-
the cell body (Banks et al., 1998). In fact, a better explanation ticular type of synapse, not absolute selectivity. Both synapse
for location-dependent IPSCs is electrotonic ltering and/or types contain the same level of immunoreactivity for the 2/3
the lack of voltage-clamp at the more distal recording loca- subunits; both types of synapse made by basket cells onto
tions (Maccaferri et al., 2000). pyramidal cell somata also contain the 1 and the 2 subunits
GABAA receptors at different GABAergic inputs onto (Somogyi et al., 1996). However, the 1 subunit is less abun-
pyramidal cells have differing sensitivities to zinc and benzo- dant in the parvalbumin-negative pyramidal cell synapses
diazepines, suggestive of different subunit compositions. than in the parvalbumin-positive ones. This work ts well
Zolpidem sensitivity (a selective agonist for 1 2/32 recep- with the pharmacological experiments (zolpidem sensitivity)
tors if used at the appropriate concentration) was used to described above. Thus, (fast-spiking) PV-basket cells probably
probe the subunit composition of synapses on the pyramidal act through 1 2/32, 2 2/32, and/or 12 2/32 sub-
cell body (Thomson et al., 2000). IPSPs in pyramidal cells unit-containing receptors. The subcellular location of other 
elicited by fast-spiking basket cells (possibly corresponding to subunits (4 and 5) on pyramidal cells has not yet been
the parvalbumin-positive basket cells) were enhanced more by established by electron microscopy. Despite these differences
zolpidem than IPSCs elicited by regular spiking basket cells. in GABAA receptor subunit composition among synapses,
Both types of synapse have IPSCs equally enhanced by paired intracellular recordings between pyramidal cells and
 Molecular Mechanisms of Synaptic Function 279

basket cells did not reveal signicant differences in the kinetic GABAA receptor. For example, as seen by the single-cell
parameters of postsynaptic responses evoked by PV- or CCK- reverse transcription-polymerase chain reaction (RT-PCR),
containing basket cells (Maccaferri et al., 2000; Pawelzik et al., the most prominently expressed GABAA receptor subunit
2002). Different GABAA receptor subunit compositions at the genes in rat dentate basket-like cells are 2, 3, and 2 (Berger
synapse do not necessarily produce different kinetics. et al., 1998), although by immunocytochemistry there is only
GABAA receptors are also expressed at the dentate granule weak (Fritschy and Mohler, 1995) or no (Sperk et al., 1997) 2
cell mossy ber to CA3 pyramidal cell synapse, a synapse tra- subunit immunoreactivity detectable in this cell type. Some
ditionally considered glutamatergic (Walker et al., 2001). other interneurons (unidentied) in the polymorph cell layer
Dentate granule cells can induce a partial GABAergic pheno- of the dentate gyrus may assemble multiple GABAA receptor
type after seizure by expressing glutamic acid decarboxylase types: Such cells are immunoreactive for the 3, 1, and the 
(reviewed in Gutierrez, 2003). The 1 and 2/3 subunits were subunits as well as for the 1, 2, and 2 subunits (Fritschy
found by immunogold cytochemistry in some rat mossy ber and Mohler, 1995; Pirker et al., 2000). The 3-positive cells in
to CA3 synapses and co-localized with GluR2/3 AMPA recep- the hilus may be mossy cells. Some 3-positive cells (small
tor subunits (95% of AMPA receptor-positive mossy ber to cells, three to ve parallel dendrites) are present in the stratum
CA3 synapses were also GABAA receptor-positive) (Bergersen oriens of the CA1 region (Brunig et al., 2002). Peng et al
et al., 2003). What is the function of GABA release and GABAA (2002) and Wei et al (2003) reported  subunit immunoreac-
receptor activation in mossy ber synapses? It may be a brake tivity in scattered, unidentied interneurons throughout the
on the AMPA system, although the GABAergic signal is small hippocampus. Double-labeling studies are needed to conrm
compared with the AMPA receptor EPSP. Note also that mossy subunit co-expression and interneuron type.
bers have a particularly high concentration of zinc, which
when released may inhibit extrasynaptic  subunit-containing GABAA Receptor Subtypes and Network Oscillation
receptors; on the other hand, the 2 subunit may serve to Frequency in the Hippocampus
reduce zinc sensitivity of the synaptic GABAA receptor com-
plex (Hosie et al., 2003). Modeling predicts that IPSC kinetics between basket cellbas-
ket cell groupings and between basket cells and principal
Distribution of GABAA Receptors cells strongly inuence network oscillations in the gamma
in Hippocampal Interneurons frequency range, although models differ in their assumption
of the actual values of IPSC parameters (reviewed in Bartos
GABAergic interneuronsand innervating principal cells also et al., 2002). Empirically, synapses in the hippocampus
innervate each other and thus reciprocally inhibit each other. differ with respect to their IPSC kinetics; the kinetics of
The amount of GABAergic input onto hippocampal interneu- interneuroninterneuron IPSCs are always faster than
rons varies with cell type, as seen from a detailed study of interneuronprincipal cell IPSCs (Bartos et al., 2002). For
three types of CA1 interneuron: The ratio of GABAergic to example, the time course of the GABAergic postsynaptic con-
glutamatergic inputs was about 30% for calbindin-expressing ductance change at basket cell to basket cell synapses in the
D28K cells, 20% for calretinin-expressing cells, and only 6% for dentate gyrus is fast (decay time constant 1.8 ms), whereas
parvalbumin-expressing cells (Gulyas et al., 1999). Regardless basket cell to granule cell synapses are slower (decay time con-
of the exact interneuronal type, GABAergic terminals onto stant 5.2 ms). The differing subunit combinations between
interneurons are concentrated in the perisomatic region and cell types can explain the differing IPSC kinetics; the predom-
on proximal dendrites; GABAergic synapses are also on distal inant receptor subtype on many hippocampal interneurons is
dendrites, although at reduced density. Because of the high 1 2, and this receptor subtype has faster decay kinetics
diversity of the interneuronal GABAergic population in the than those of the 2 2 receptors enriched at some synapses
hippocampus, the available information on GABAA subunit on pyramidal and dentate cell bodies. Whether these IPSC
expression in hippocampal interneurons is patchy. Based on kinetic differences are really due to the distinct GABAA recep-
immunocytochemistry and zolpidem sensitivity, many rat tor subunit compositions at specic synapses or have some
hippocampal interneurons probably use the 1 22 receptor other explanation remains to be tested. Indeed, the decay time
subunit combination. The 1 immunoreactivity is found in constants between basket cellbasket cell pairs vary among
all parvalbumin-positive interneurons and about half of cal- hippocampal regions (DG faster than CA3, in turn faster than
retinin-positive cells but in no calbindin-containing cells (Gao CA1), although they are always faster than the corresponding
and Fritschy, 1994; Klausberger et al., 2002); much of the 1 interneuronprincipal cell IPSCs. Factors in addition to sub-
immunoreactivity on parvalbumin-positive cells is attributa- unit composition that might affect IPSC kinetics include the
ble to extrasynaptic receptors. The 1 immunoreactivity is phosphorylation status of the GABAA receptors; synaptic
found in most NPY- and some somatostatin-containing cells, geometry (differing distances of receptors from transmitter
but not in CCK- or VIP-positive cells (Gao and Fritschy, release sites); and differences in GABA reuptake kinetics via
1994). However, at least 40% of hippocampal interneurons do transporters. These factors could be synapse- and cell type-
not express the 1 subunit and so probably use other types of dependent.
280 The Hippocampus Book

7.6.3 GABAB Receptors neurons, immunocytochemistry shows that GABAB receptors

and ATF4 co-cluster in the soma and at the dendritic mem-
The GABAB receptors consist of seven transmembrane- brane surface. This suggests a model in which activation of the
domain proteins, GABABR1, and GABABR2 (reviewed in receptor releases the transcription factors, which then translo-
Bettler et al., 2004). Most GABAB receptors are heterodimers cate to the nucleus to activate/repress target gene expression.
of GABABR1 and GABABR2. Inactivation of the GABABR1 NGF and BDNF genes are candidate target genes (White et al.,
gene in mice removes all biochemical and electrophysiological 2000).
GABAB responses, demonstrating that GABABR1 is an essen-
tial component of all GABAB receptors. The lack of the Distribution of GABAB Receptors
GABAB response causes generalized epilepsy (Prosser et al., in Hippocampal Principal Cells
2001; Schuler et al., 2001). There are two R1 versions,
GABABR1a and GABABR1b, produced by different promoters Both GABABR1 and GABABR2 are expressed in hippocampal
of the GABABR1 gene; the proteins differ in their N-termini, pyramidal cells and dentate granule cells, as shown by in situ
the rst 147 amino acids in GABABR1a being replaced by hybridization and immunocytochemistry (Fig. 710A) (Kulik
an equivalent region in GABABR1b differing in 18 amino et al., 2003). Most GABAB receptors are on dendrites. There is
acids. In many neurons, such as pyramidal cells and dentate a differential subcellular distribution of GABAB receptors on
granule cells, GABAB receptors are formed as heteromers of hippocampal pyramidal cells: GABABR1b immunoreactivity
GABABR1 and GABABR2 in a stoichiometric 1:1 ratio. is enriched on distal dendrites (stratum lacunosum-molecu-
Most GABAB receptors are likely to be GABABR1a/GABABR2 lare) of rat pyramidal cells, whereas the GABABR1a form is
or GABABR1b/GABABR2 heteromers. In the GABABR com- found predominantly on the cell body. The somatic receptors
plex, the GABABR1 subunit binds GABA and all competitive are probably not functional; electron microscopy with gold
GABAB ligands, whereas the GABABR2 subunit is essential particle-tagged antibodies shows that the strong somatic
for the translocation of GABABR1 subunits to the cell surface labeling results from GABAB receptor immunoreactivity in
and for targeting of the receptor to distal dendrites (Pagano the endoplasmic reticulum (Kulik et al., 2003). On dentate
et al., 2001; Gassmann et al., 2004); the GABABR2 subunit granule cells, GABAB receptor immunoreactivity is enriched
contributes to activating the G protein but is not absolutely in the molecular layer, suggesting a dendritic location.
essential, as GABABR2 knockout mice still have (atypical) Throughout the hippocampus, GABABR2 immunoreactivity
electrophysiological GABAB responses (Gassmann et al., corresponds to the sum of the R1a and R1b patterns. The sub-
2004). units co-localize in the same intracellular compartments.
GABAB receptors act through G proteins to, for example, The targeting of GABAB receptors to distal dendrites requires
open K channels (GIRK type), inhibit high-voltage-activated the GABABR2 subunit; in mouse hippocampal neurons
Ca2 channels (P/Q- and N-type) and inhibit/stimulate lacking GABABR2 the soma of pyramidal cells and dentate
adenyl cyclase; the opposite effects on cAMP production granule cells become prominently labeled by a GABABR1
might be due to different G proteins (Bettler et al., 2004). antibody, whereas in wild-type animals cell bodies have
GABAB receptor effects on K channels are mainly postsynap- only weak, diffuse staining, and GABABR1 immunoreactivity
tic and cause long-lasting membrane hyperpolarization, is distributed throughout the neuropil (Gassmann et al.,
which can take up to 200 ms to peak and 450 to 2000 msec 2004).
to decay. In contrast, Ca2 effects are mainly presynaptic.
GABAB receptors can act as inhibitory autoreceptors to block Distribution of GABAB Receptors
GABA release from interneurons (Poncer et al., 2000); the in Hippocampal Interneurons
resulting disinhibition plays an important role in the induc-
tion of LTP (see Chapter 10). In CA1 pyramidal cells of mice GABAB receptor responses are detected on some interneurons
lacking the GABABR2 subunit, GABAB receptors inhibit rather and are similar to those of principal cells (Mott et al., 1999).
than activate K channels, but whether this is physiologically Both GABABR1a- and GABABR1b-like immunoreactivity is
relevant is unclearit may be a side effect of GABABR1 hav- found on subsets of hippocampal GABAergic interneurons
ing no proper partner (Gassmann et al., 2004). (Kulik et al., 2003; Gassmann et al., 2004); however, these
In keeping with the long-term effects produced by GABAB interneurons do not stain with an antibody recognizing the
receptor activation, various signaling proteins, including tran- GABAA receptor 2/ 3 subunits. This implies that GABAA
scription factors, may link to the GABAB receptor complex, and GABAB receptors may be on distinct interneuronal cell
forming a signaling scaffold. Examples include the direct populations, or that these GABAB-expressing interneurons
interactions between the C-terminal tails of GABABR1 and use GABAA receptors with the 1 subunit (Fritschy et al.,
GABABR2 with two related transcription factors, CREB2 1999). Not all hippocampal interneurons seem to have GABAB
(ATF4) and ATFx (Nehring et al., 2000; White et al., 2000); receptors; there are some in the dentate hilus border region in
both these transcription factors activate target genes with CRE which GABAB-activated K currents are not produced (Mott
sites in their regulatory regions. On cultured hippocampal et al., 1999). By double labeling for glutamic acid decarboxy-
 Molecular Mechanisms of Synaptic Function 281

Figure 710. Expression of GABAB and mAChR receptors in the hippocampus.

A. Immunohistochemistry with subunit-specic antibodies was used to visualize the expression
pattern of GABAB receptor subunits. (Source: Modied from Kulik et al., 2003, with permis-
sion.) B. Immunohistochemistry with mAChR type-specic antibodies in the adult rat hip-
pocampus. (Source: Modied from Levey et al., 1995, with permission.)

lase 67 (GAD67) mRNA, Kulik et al, estimated that only 50% terminals autoinhibit GABA release. Presynaptic GABAB
of GABAergic interneurons contain GABAB receptor receptors on glutamatergic terminals likely function as het-
immunoreactivity (Kulik et al., 2003). eroreceptors regulating glutamate receptor release (Kulik et
al., 2003).
Subcellular Localization of GABAB Postsynaptically, the most intense labeling for GABAB
Receptors in Hippocampal Neurons receptor subunits was found in CA1 and CA3 pyramidal cell
spines associated with glutamatergic terminals (Kulik et al.,
Presynaptic GABAB receptor subunits are localized to 2003). The functional relevance of this is unclear. In dendritic
GABAergic and putative glutamatergic axon terminals. shafts, the GABAB receptor subunits localize to the extrasy-
Functional GABAB receptors can be found, for example, on naptic membrane, with no association with inhibitory
mossy ber terminals, where they provide heterosynaptic synapses. Thus, activation of these receptors requires GABA
depression (Vogt and Nicoll, 1999). The subunits are found at spillover. Activation of GABAB receptors depends on tran-
the extrasynaptic membrane but also at the presynaptic mem- sients in the ambient GABA concentration and may serve to
brane specialization. GABAB receptors on inhibitory synaptic detect enhanced and/or simultaneous activity of GABAergic
282 The Hippocampus Book

interneurons, as occurs in population oscillations when many ectopically expressed in hippocampal pyramidal cells, these
interneurons re simultaneously. receptors remain extrasynaptic (Wisden et al., 1992). The
determinants of synaptic/extrasynaptic GABAA receptor loca-
tion may be cell type-specic and subunit-specic. In hip-
pocampal fireball interneurons, the 3 subunit is
7.7 Trafficking of GABA Receptors extrasynaptic (not co-localized with gephyrin); whereas in
and Hippocampal Synaptic Function interneurons described as small multipolar, 3 immunore-
activity clusters with that of gephyrin (Brunig et al., 2002).
How GABAA receptor subunits/subtypes target GABAergic Thus, in addition to gephyrin, other clustering proteins must
synapses, and differentially target between synapses on the contribute to the synaptic localization of selected GABAA
same hippocampal cell (e.g., the 2 subunit enriched in the receptor subtypes.
axon initial segment of pyramidal cells) is not well under-
stood. In vivo most GABAA receptors on hippocampal neu- 7.7.2 Role of Dystrophin-associated Protein
rons are located opposite GABAergic terminals, not opposite Complex in GABAA Receptor Function
glutamatergic terminals. How does this selective targeting
occur? There is evidence that targeting of GABAA receptors to The dystrophin-associated protein complex (DAPC) connects
specic domains of hippocampal neurons is partially cell- the cytoskeleton to the extracellular matrix via a transmem-
intrinsic. In cultured hippocampal neurons, 40% of pyrami- brane link. The DAPCs integrative role has been best studied
dal axon initial segments have enrichment of the 2 subunit, in muscle and Duchenne muscular dystrophy. However,
as seen by co-staining with the axon-specic type II voltage- DAPC components are also expressed in both neurons and
gated Na channel (Brunig et al., 2002), and a similar propor- glia, and some that are not found at excitatory synapses seem
tion is seen in vivo (Nusser et al., 1996). Thus, specic to be specialized for GABAergic synaptic function (Kneussel
innervation patterns on the hippocampal pyramidal cells et al., 1999; Brunig et al., 2002; Levi et al., 2002; Fritschy and
(e.g., from the axo-axonic interneurons in vivo) are not Brunig, 2003). Because the DAPC binds intracellular and
required to establish 2 enrichment at the axon initial seg- extracellular components, it could function as an organizer of
ment during development. GABAergic synaptic structure in the hippocampus. The
DAPC comprises dystroglycan (extracellular  subunits and
7.7.1 Role of Gephyrin in GABAA transmembrane subunits derived from a common precur-
Receptor Localization sor protein by cleavage). The -dystroglycan binds laminin
and agrin and the transmembrane -dystroglycan. The -
Placing GABAA receptors at synapses requires specic proteins dystroglycan binds the intracellularly located dystrophin
that interact directly or indirectly with the 2 subunit. For and/or related utrophin. Dystrophin (Dp) is a cytoskeletal
example, targeting some GABAA receptor subtypes to protein of the -actinin/ -spectrin family. The main brain
GABAergic terminals involves the microtubule-binding pro- isoform of dystrophin is Dp71. Dp71 binds to -dystobrevin.
tein gephyrin. In hippocampal neurons in vitro and in vivo, Dystrophin and -dystobrevin each bind up to two syn-
gephyrin either helps convey some GABAA receptor subtypes trophins (pleckstrin homology domain and PDZ-containing
to the synapse or anchors them there; this requires the 2 sub- proteins). The PDZ domain of the syntrophins serves as an
unit (Kneussel et al., 1999; Brunig et al., 2002). Without the 2 adaptor to bind membrane channels, receptors, kinases, and
subunit, no GABAA receptors are found in synapses in the other signaling proteins. Dystrophin and dystroglycan selec-
developing or adult hippocampus (Gunther et al., 1995; tively associate with subsets of GABAergic synapses in the hip-
Essrich et al., 1998; Schweizer et al., 2003); and without pocampus (Kneussel et al., 1999; Levi et al., 2002). Dystrophin
gephyrin, much reduced numbers of some synaptic GABAA immunoreactivity is present in the CA1 and CA3 regions but
receptor subtypes, especially 2-containing, are found. Some not in the dentate gyrus; extensive co-localization between the
receptor clusters, especially those containing the 1 subunit, 2, 2, and dystrophin immunoreactivities was observed in
persist in hippocampal gephyrin knockout neurons (Levi the soma and dendrites of hippocampal pyramidal cells; but
et al., 2004). Gephyrin, however, does not bind the 2 GABAA the axon initial segment, which was prominently stained for
receptor subunit intracellular loop directly, as it does for the GABAA receptor subunits, was negative for dystrophin stain-
glycine receptor subunit. The identity of the missing link(s) ing (Kneussel et al., 1999). Similarly, -dystroglycan is present
between gephyrin and GABAA receptor subunits is unknown. at only a subset of GABAergic synapses in cultured hip-
As mentioned above, the targeting of 2-containing recep- pocampal neurons, including synapses positive for 1, 2, 2,
tors to hippocampal synapses must depend on both the  sub- and 2/3 subunits (Levi et al., 2002). However, many clusters
unit and the 2 subunit. According to some investigators of glutamic acid decarboxylase, gephyrin, and 2 subunits
5 2 receptors seem largely extrasynaptic and not co-local- were not co-localized with -dystroglycan, which, like dys-
ized with gephyrin (Crestani et al., 2002), and when 6 2 trophin, is present only in a subset of GABAergic synapses.
receptors (normally found only in cerebellar granule cells) are There is no interaction between the gephyrin and DAPC sys-
 Molecular Mechanisms of Synaptic Function 283

tems. The number and size of gephyrin-immunoreactive clus- subunit mRNA levels, as well as increases in and  subunit
ters were unaffected by the absence of dystrophin in mdx transcript levels (Brooks-Kayal et al., 1998).
mice; hippocampal neurons from gephyrin knockout mice Other forms of stimulation eliciting GABAA receptor plas-
have normal dystroglycan and dystrophin immunoreactive ticity in dentate granule cells include ethanol, steroids, and
clusters. prenatal handling. Expression of the 4 subunit gene seems
particularly plastic with regard to ethanol and steroid treat-
7.7.3 Plasticity of GABAA Receptor ments. Chronic in vivo administration and withdrawal of
Expression at Hippocampal Synapses progesterone, which mimics hormonal changes of the men-
strual cycle, increases the expression of hippocampal 4 
It is important to keep in mind that, as for ionotropic gluta- receptors (Smith et al., 1998a,b); long-term treatment and
mate receptors, GABAA receptor expression on the surface of withdrawal from ethanol (in CIE-treated rats) elevates hip-
hippocampal cells is dynamic; receptors rapidly recycle and pocampal 4 mRNA and protein in CA1 pyramidal cells
leave from or insert into the synapse (reviewed in Kittler and (Mahmoudi et al., 1997). Brief traumatic stress during infancy
Moss, 2003). For some inhibitory hippocampal synapses, a can permanently alter expression of GABAA receptors in den-
direct relation exists between the number of synaptic GABAA tate granule cells of the adult hippocampus (Hsu et al., 2003).
receptors and the strength of the synapse (reviewed in Nusser, In adult rats that have experienced two episodes of separation
1999). As for glutamate receptors at excitatory synapses, neu- from their mothers and human handling, dentate granule cells
rons probably recycle GABAA receptors as an important strat- recorded in vitro signicantly reduced responses to zolpidem
egy for setting their degree of excitability. GABAA receptors application; by single-cell RNA amplication, 1 subunit
constitutively internalize by clathrin-dependent endocytosis; mRNA levels decrease in handled individuals, whereas those
this requires interactions between the and 2 subunits and of the 2 subunit increase.
the AP2 adaptin complex. Again, as for glutamatergic
synapses, the properties of inhibitory GABAergic synapses
alter with activity through changes in GABAA receptor prop-
erties or receptor number. 7.8 Genetic Analysis of GABA Receptor
Function in the Hippocampus
Plasticity in GABA Receptor Expression
in Dentate Granule Cells 7.8.1 GABAA Receptor Mutant Mice

Dentate granule cells exhibit pronounced plasticity in their The widespread expression of most GABAA receptor subunit
patterns of gene expression; although their phenotype is pre- genes makes it difficult to attribute behavioral changes in
dominantly glutamatergic, after prolonged seizure activity knockout animals specically to changes in the hippocampus
granule cells can display a partial GABAergic phenotype by (reviewed in Rudolph and Mohler, 2004). Nevertheless, vari-
expressing glutamic acid decarboxylase (reviewed in ous studies on genetically engineered mice have provided
Gutierrez, 2003). Moreover, seizure activity in the hippocam- insight into the role of GABAA receptors in hippocampal
pusfor example, pilocarpine-induced epilepsy or kin- function.
dlingcauses changes in GABAA receptor subunit expression Expression of the 5 gene is highly enriched in pyramidal
in dentate granule cells (Brooks-Kayal et al., 1998). There are cells of the hippocampus compared with other brain regions,
also long-term changes in receptor number. After kindling, making the knockout phenotype relatively easy to interpret as
the amplitudes of elementary synaptic GABAA currents on hippocampal (Collinson et al., 2002). Homozygous 5 sub-
dentate granule cells increased 66% owing to a corresponding unit knockout mice show no overt phenotypic abnormalities,
75% increase in GABAA receptor number in GABAergic breed normally, and have no spontaneous seizures. However,
synapses on granule cell somata and axon initial segments, as in a matching-to-place watermaze task, in which the hidden
assessed with immunogold labeling with an antibody to the platform is moved between trials, 5 knockout mice had
2 and 3 subunits (Nusser et al., 1998a); immunostaining of enhanced performance compared with their wild-type litter-
granule cells with 1, 2, and 2 subunit antibodies was also mates, whereas performance in non-hippocampal-dependent
increased. This demonstrates directly that under these condi- learning and anxiety tasks was unaltered. In the CA1 region of
tions the efficacy of GABAergic synapses is enhanced by the hippocampal slices from 5 knockout mice, the IPSC ampli-
insertion of extra GABAA receptors postsynaptically. Using tude was decreased and paired-pulse facilitation of eld EPSP
single-cell mRNA amplication, Brooks-Kayal et al. (1998) amplitudes was enhanced. In complementary data, mice engi-
studied the relative changes in GABAA receptor subunit neered to express 5 subunits with a histidine to arginine
expression in isolated dentate granule cells from normal rats mutation at position 105 of 5 have a fortuitous CA1-specic
and chronically epileptic rats in which temporal lobe epilepsy knockdown of the 5 subunit expression (Crestani et al.,
was produced by pilocarpine injection. The largest changes 2002). In these 5 CA1 knockdown mice, trace fear condi-
found were a marked decease in 1 and a strong increase in 4 tioning (a hippocampal-dependent process) is selectively
284 The Hippocampus Book

enhanced. These combined data suggest that 5-containing was observed in hippocampal slices from mice that had repet-
GABAA receptors contribute to cognitive processes by con- itive seizures (Prosser et al., 2001).
trolling a component of synaptic transmission in CA1 A knockout mouse line has also been made for the other
(Collinson et al., 2002). GABAB receptor subunit gene, GABABR2 (Gassmann et al.,
In terms of network properties,  oscillations (2080 Hz) 2004). Adult GABABR2 knockout mice (BALB/c) display sev-
are modied in 5 knockout hippocampal slices (Towers et eral episodes of spontaneous seizures per day. The recorded
al., 2004). The peak power of oscillations evoked by kainic seizures are exclusively clonic. This is in contrast to GABABR1
acid was signicantly greater in 5 knockout than in wild- knockout mice (BALB/c), in which additionally absence-type
type slices; however, the frequency change of  oscillations and spontaneous tonic-clonic seizures occur with low fre-
that normally occurs with increasing network drive was quency. GABABR2 knockout mice exhibit impaired passive
absent in 5 knockout slices. The enriched expression of the avoidance learning, similar to GABABR1 knockout mice.
5 subunit in the hippocampus has therapeutic potential. Thus, the GABABR2 subunit is needed for functional GABAB
The drugs L-655,708 (partial inverse agonist at the BZ site) receptors in vivo.
and 43 (a thiophene analogue and a full inverse agonist at
the BZ site) are 5 2-selective modulatory ligands
(Chambers et al., 2003); 43 can increase, relatively selec-
tively, neuronal activity in the hippocampus; 43 enhances 7.9 Cholinergic Receptors
cognitive performance in rats in the delayed matching-
to-place Morris watermaze. In principle, it and similar 5 7.9.1 Introduction: Muscarinic
subunit-selective drugs are potential cognition enhancers and Nicotinic Receptors
in humans.
Further useful information on hippocampal signaling has Cholinergic neurotransmission modulates glutamatergic and
also been gleaned from some of the other GABAA receptor GABAergic transmission in the hippocampus, and cholinergic
subunit knockouts. The  subunit knockout mice were instru- projections into the hippocampus regulate the hippocampal
mental in showing the selective contribution of the 4- theta rhythm in vivo (reviewed in Kimura, 2000). The neuro-
containing GABAA receptors to extrasynaptic tonic inhibition transmitter acetylcholine activates both metabotropic G
on dentate granule cells (Stell et al., 2003). In the hippocam- protein-coupled muscarinic receptors and neuronal nicotinic
pus of 1 subunit knockout mice, synaptic GABAA receptor- receptors constituting ligand-gated cation channels.
mediated currents (pyramidal and interneuronal) decay more Acetylcholine is delivered to the hippocampus primarily by
slowly than those from wild-type mice (Goldstein et al., 2002). cholinergic neurons projecting from the medial septum/diag-
Loss of the 1 subunit in synaptically localized receptors leads onal band complex; in addition, infrequent choline acetyl
to a population of synaptic receptors with slower deactivation transferase (ChAT)-positive interneurons in the hippocampus
kinetics, in which the 1 subunit is probably replaced by an produce local acetylcholine (ACh).
2 or 3 subunit, thus conrming the role of  subunits in Activation of muscarinic receptors on both principal cells
determining GABAA receptor kinetics. and interneurons causes a variety of cellular effects, including
inhibition of K currents, membrane depolarization, and IP3
7.8.2 GABAB Receptor Mutant Mice production (reviewed in Wess, 2004). A prominent effect of
muscarinic agonists on pyramidal cells is the potentiation of
Two independent GABABR1 knockout lines were established NMDA-induced currents (Marino et al., 1998). Muscarinic
(Prosser et al., 2001; Schuler et al., 2001). Although both lines effects on hippocampal function can be modulated by activa-
show a loss of GABAB responses in hippocampal pyramidal tion of autoreceptors on cholinergic terminals of the afferents
cells, demonstrating the absolute requirement for the from the septo-hippocampal pathway, with a resultant reduc-
GABABR1 subunit in forming GABAB receptors, the genetic tion in hippocampal ACh release.
background of the two lines dramatically affects their mortal- Nicotinic receptors are used for fast transmission at the
ity. The responsible strain-dependent modier genes have not nervemuscle synapse and at autonomic nervous system gan-
been identied. glia; however, in the brain, nicotinic receptors play modula-
Balb/c GABABR1 knockout mice (BALB/c ( BALB/c, F1) tory rolesan interesting case of a fast-ligand-gated
generated from BALB/c embryonic stem cells have sponta- neurotransmitter receptor working as a modulator. The tim-
neous epileptiform activity but are otherwise viable. In addi- ing of nicotinic receptor activation depends on activity in the
tion, the mice are hyperalgesic, show hyperlocomotor activity, medial septumdiagonal band complex (Ji et al., 2001).
and are severely impaired in passive avoidance learning Presynaptic nicotinic acetylcholine receptors (nAChRs) can
(Schuler et al., 2001). The second linegenetic background increase the probability of neurotransmitter release, increas-
129SVJ  C57Bl6/J,GABABR1 knockout mice have sponta- ing the delity of synaptic transmission. Postsynaptic nAChRs
neous generalized seizures and die after several seizure events; can increase the depolarization and Ca2 signal associated
an increased ictal discharge pattern of epileptiform activity with successful transmission, helping initiate intracellular cas-
 Molecular Mechanisms of Synaptic Function 285

cades. On the other hand, nicotinic activity can strongly excite In the hippocampus, M3 mRNA and protein have distri-
GABAergic interneurons, thus regulating the excitability of butions similar to those of M4 mRNA (see above) (Buckley
circuits. To quote Ji et al., the location of nAChR activity and et al., 1988; Levey et al., 1995). M5 mRNA is restricted to
the moment-by-moment change in that activity can tip the CA1 pyramidal cells, but the expression is weak (Vilaro et al.,
balance in favor or against the induction of synaptic plastic- 1990); M5 protein has not been detected in the hippocampus.
ity (Ji et al., 2001).
Nicotinic acetylcholine receptors in the hippocampus are Muscarinic Receptor Mutant Mice
important for learning and memory. In addition to ndings
from mice lacking distinct nicotinic acetylcholine receptor M1 knockout mice exhibit a mild reduction in hippocampal
genes (see Section 7.8.3), evidence for this role comes also LTP (Schaffer-CA1 synapse) in response to theta burst stimu-
from behavioral pharmacology showing that infusion of the lation (Anagnostaras et al., 2003). M1 knockouts have normal
nicotinic receptor antagonist mecamylamine into the ventral or enhanced memory for tasks involving contextual fear con-
hippocampus impairs memory performance in rats; con- ditioning and the Morris watermaze, but they are impaired in
versely, injection of nicotine and nicotinic receptor agonists non-matching-to-sample working memory and consolidation
improves memory performance (reviewed in Levin et al., (win-shift radial arm and social discrimination learning). The
2002). In the ventral hippocampus, cholinergic transmission conclusion drawn from these mice was that M1 receptors are
thus boosts memory-related functions. In contrast, the not essential for memory formation or initial stability of
cholinergic tone in the dorsal hippocampus is associated with memory in the hippocampus but are more likely to contribute
anxiolysis (reviewed in File et al., 2000). to processes requiring interactions between the neocortex and
hippocampus. A possible substrate for binding distant popu-
7.9.2 Hippocampal Muscarinic Receptors lations of cells is  oscillations. Muscarine-induced  oscilla-
tions are absent in hippocampal slices (CA3 area) from M1
Five subtypes (M1M5) of muscarinic receptors exist: M1, knockout mice but are still present in the M2 to M5 knockouts
M3, and M5 couple to Gq/G11 G proteins; M2 and M4 prefer (Fisahn et al., 2002). These muscarine-induced  oscillations
Gi/Go (reviewed in Wess, 2004). Muscarinic receptors are depend on M1-mediated depolarization of CA3 pyramidal
classic G protein-coupled receptors that are distantly cells by activation of the mixed Na/K current (Ih) and the
related, if at all, to metabotropic glutamate and GABAB recep- Ca2-dependent nonspecic cation current (Icat).
tors. As seen by in situ hybridization and immunocytochem- M2 knockout mice have signicant performance decits in
istry, M1 is the most prominent muscarinic receptor in the passive avoidance tests. The increase in hippocampal ACh
hippocampus. The M1 gene is abundantly expressed in CA1 induced by local administration of scopolamine was markedly
and CA3 pyramidal cells, dentate granule cells, and scattered reduced in M2 knockout mice and completely abolished in
interneurons; M1 protein is on somata and dendrites (Buckley M2/M4 double knockouts (as measured by in vivo microdial-
et al., 1988; Levey et al., 1995). In CA1 pyramidal cells, M1 and ysis), probably due to a lack of M2 and M4 autoreceptors on
the NR1a NMDA receptor subunits co-localize and may septal-hippocampal afferents (Tzavara et al., 2003). In M2 and
directly interact via an intracellular signaling scaffold (Marino M4 knockouts, and to a much greater extent in M2/M4 dou-
et al., 1998). This may explain how muscarinic agonists poten- ble knockouts, the increase in hippocampal ACh triggered by
tiate NMDA-induced currents (see above). Mice lacking the exposure to a novel environment was more pronounced than
M1 gene show various cellular and behavioral phenotypes in wild types (both amplitude and duration).
related to hippocampal functions (see below), conrming the
importance of M1 in the hippocampus. 7.9.3 Hippocampal Nicotinic Receptors
M2 and M4 couple to the Gi family and share similar lig-
and-binding properties, making it difficult to distinguish Nicotinic acetylcholine receptors are ligand-gated cation
pharmacologically between the two subtypes. However, stud- channels of the superfamily that also includes the GABAA
ies with M2 and M4 knockouts revealed that various physio- receptors; and like the GABAA receptors, they are pentameric
logical functions are mediated by a mixture of M2 and M4 assemblies of subunits. Neuronal nicotinic receptor subunit
activation (Wess, 2004) (see below). M2 is expressed mostly in genes comprise 2 to 10 and 2 to 4 subtypes. In the hip-
interneurons in the oriens/alveus border; the M2 protein is on pocampus, prominent neuronal nicotinic receptors are
the GABAergic terminals of the interneurons and might (4)2( 2)3, (3)2( 4)3, and (7)5 (Zoli et al., 1998; Fabian-
account for the observation that presynaptic muscarinic Fine et al., 2001). The 7 receptors are fast-inactivating nons-
receptors depress inhibitory responses in the hippocampus elective cation channels with high Ca2 permeability. The
(Levey et al., 1995). M2 is not found in pyramidal cells or den- (4)2( 2)3 receptors are mainly permeable to Na and K
tate granule cells. In contrast, M4 mRNA and protein are and are the main targets for nicotine. Both the 7 and
mainly in CA pyramidal cells; they have lower expression in (4)2( 2)3 receptors can be pre- and postsynaptic. Although
the dentate granule cells, with CA3 pyramidal cells containing the (4)2( 2)3 and (3)2( 4)3 receptors are expressed on
slightly more M4 than CA1 cells. pyramidal cells, their function is not clear.
286 The Hippocampus Book

The 7 gene is expressed in many hippocampal cell types, tors (Fonck et al., 2003). In these mice, the 4 subunits have a
including interneurons, dentate granule cells, and pyramidal leucine to serine mutation in the channel pore-lining region
cells (Fabian-Fine et al., 2001). Many GABAergic and gluta- of TM2, a 9 position that renders 4 receptors hypersensitive
matergic synapses in the hippocampus contain 7 receptors; to agonists. Homozygous animals die soon after birth (the
postsynaptically, 7 receptors are on dendritic spines in a same lethal phenotype was also observed for knock-in mice
perisynaptic annulus. The large Ca2 entry through 7 nico- with an L250T mutation in the 7 subunit gene that increases
tinic receptors may play a key role in synaptic plasticity, evok- the 7 currents) (Orr-Urtreger et al., 2000). However, mice
ing transmitter release (e.g., from mossy ber terminals) heterozygous for the 4 TM2 mutation survive but show
and/or triggering Ca2-induced Ca2 release from internal increased anxiety. In these mice, the hippocampus is hyper-
stores (Khiroug et al., 2003). In hippocampal principal neu- sensitive to nicotine; much lower doses of nicotine are needed
rons, nicotinic 7 receptors mediate activation of to elicit increases in the amplitude of the electroencephalog-
Ca2/calmodulin-dependent protein kinase and the raphy trace; and compared with wild-type mice, a large
ERK/MAPK cascade, resulting in sustained phosphorylation (threefold) nicotine-induced increase in power density at
of the transcription factor CREB. Presynaptic 7 receptors  frequencies (7 Hz) occurs (Fonck et al., 2003). Thus 4-
enhance the probability of hippocampal LTP, presumably via containing nicotinic receptors could contribute to oscillatory
Ca2 entry through these receptors (Dajas-Bailador and network behavior in the wild-type hippocampus.
Wonnacott, 2004). Interestingly, -amyloid peptide (A 1-42)
antagonizes 7 receptors on hippocampal neurons (Liu et al.,
2001), suggesting that in patients with early Alzheimers dis- ACKNOWLEDGMENTS
ease this antagonist effect could be an additional factor that
contributes to cognitive defects. This also provides indirect We thank Peter H. Seeburg for long-term support. We also gratefully
support for the role of the 7 subunit in cognition under nor- acknowledge our collaborators Per Andersen, David Bannerman,
mal conditions. In hippocampal interneurons, 7 receptors vind Hvalby, Vidar Jensen, Nick Rawlins, and Bert Sakmann. W.W.
constitute 38 pS, inwardly rectifying channels that produce thanks Hannah Monyer for long-term support. The work in P.O.s
laboratory was funded in part by Deutsche Forschungsgemeinschaft,
strong excitatory effects, including generation of action
the German Israeli Foundation, and the Human Frontier Science
potentials. Program. The work in W.W.s laboratory was funded by the
Volkswagen Stiftung, the Fonds der Chemischen Industrie, the
Nicotinic Receptor Mutant Mice Deutsche Forschungsgemeinschaft, and the Schilling Foundation
(grant to Prof. H. Monyer). The work in R.S.s laboratory was funded
Strong pharmacological evidence suggests that nicotinic in part by the Deutsche Forschungsgemeinschaft and Volkswagen
transmission is essential for healthy cognition and learning; Stiftung.
but as is often the case with knockouts of modulatory systems,
studies on nicotinic receptor subunit knockout mice have
failed to provide convincing support for this conclusion (see REFERENCES
the informed commentary by Champtiaux and Changeux,
2002, for possible reasons). Thus, despite the proposed roles of Aiba A, Chen C, Herrup K, Rosenmund C, Stevens CF, Tonegawa S
7 receptors in hippocampal function alluded to above, mice (1994) Reduced hippocampal long-term potentiation and con-
text-specic decit in associative learning in mGluR1 mutant
lacking the 7 subunit have no decits in hippocampal learn-
mice. Cell 79:365375.
ing tasks (Paylor et al., 1998). Thus, either there is compensa-
Anagnostaras SG, Murphy GG, Hamilton SE, Mitchell SL, Rahnama
tion (always a possibility), or the 7 receptors regulate NP, Nathanson NM, Silva AJ (2003) Selective cognitive dysfunc-
behavioral effects too subtle to pick up with current tech- tion in acetylcholine M1 muscarinic receptor mutant mice. Nat
niques. Similarly, young adult 2 knockout animals per- Neurosci 6:5158.
formed normally in spatial learning (Morris watermaze) and Andrasfalvy BK, Smith MA, Borchardt T, Sprengel R, Magee JC
Pavlovian fear conditioning (reviewed in Champtiaux and (2003) Impaired regulation of synaptic strength in hippocampal
Changeux, 2002). However, the performance of aged 2 neurons from GluR1-decient mice. J Physiol 552:3545.
knockout mice was signicantly impaired in the watermaze Aravanis AM, Pyle JL, Tsien RW (2003) Single synaptic vesicles fus-
compared with age-matched wild-type mice. These decits ing transiently and successively without loss of identity. Nature
were mainly attributed to increased neurodegeneration in 423:643647.
Bahn S, Volk B, Wisden W (1994) Kainate receptor gene expression in
hippocampal and neocortical areas of 2 knockout brains.
the developing rat brain. J Neurosci 14:55255547.
Instead of studying loss of function, dominant-negative
Banks MI, Li TB, Pearce RA (1998) The synaptic basis of
mutations can sometimes provide more insight into receptor GABA(A,slow). J Neurosci 18:13051317.
function. Non-7 nACh responses (e.g., from 4 2 receptors) Bannerman DM, Deacon RM, Seeburg PH, Rawlins JN (2003) GluR-
are more difficult to characterize on central neurons, so A-decient mice display normal acquisition of a hippocampus-
knock-in mice with hypersensitive 4 subunit-containing dependent spatial reference memory task but are impaired
receptors have been used to amplify the effect of these recep- during spatial reversal. Behav Neurosci 117:866870.
 Molecular Mechanisms of Synaptic Function 287

Bannerman DM, Deacon RM, Brady S, Bruce A, Sprengel R, Seeburg Brose N, Rosenmund C, Rettig J (2000) Regulation of transmitter
PH, Rawlins JN (2004) A comparison of GluR-A-decient and release by Unc-13 and its homologues. Curr Opin Neurobiol
wild-type mice on a test battery assessing sensorimotor, affec- 10:303311.
tive, and cognitive behaviors. Behav Neurosci 118:643647. Brunig I, Scotti E, Sidler C, Fritschy JM (2002) Intact sorting, target-
Barria A, Malinow R (2002) Subunit-specic NMDA receptor traf- ing, and clustering of gamma-aminobutyric acid A receptor
cking to synapses. Neuron 35:345353. subtypes in hippocampal neurons in vitro. J Comp Neurol
Bartos M, Vida I, Frotscher M, Geiger JR, Jonas P (2001) Rapid sig- 443:4355.
naling at inhibitory synapses in a dentate gyrus interneuron net- Brusa R, Zimmermann F, Koh D-S, Feldmeyer D, Gass P, Seeburg PH,
work. J Neurosci 21:26872698. Sprengel R (1995) Early-onset epilepsy and postnatal lethality
Bartos M, Vida I, Frotscher M, Meyer A, Monyer H, Geiger JR, Jonas associated with an editing-decient GluR-B allele in mice.
P (2002) Fast synaptic inhibition promotes synchronized Science 270:16771680.
gamma oscillations in hippocampal interneuron networks. Proc Buckley NJ, Bonner TI, Brann MR (1988) Localization of a family of
Natl Acad Sci USA 99:1322213227. muscarinic receptor mRNAs in rat brain. J Neurosci
Baude A, Nusser Z, Roberts JD, Mulvihill E, McIlhinney RA, Somogyi 8:46464652.
P (1993) The metabotropic glutamate receptor (mGluR1 alpha) Bureau I, Bischoff S, Heinemann SF, Mulle C (1999) Kainate recep-
is concentrated at perisynaptic membrane of neuronal subpop- tor-mediated responses in the CA1 eld of wild-type and
ulations as detected by immunogold reaction. Neuron GluR6-decient mice. J Neurosci 19:653663.
11:771787. Cagetti E, Liang J, Spigelman I, Olsen RW (2003) Withdrawal from
Bayer KU, De Koninck P, Leonard AS, Hell JW, Schulman H (2001) chronic intermittent ethanol treatment changes subunit com-
Interaction with the NMDA receptor locks CaMKII in an active position, reduces synaptic function, and decreases behavioral
conformation. Nature 411:801805. responses to positive allosteric modulators of GABAA receptors.
Berberich S, Punnakkal P, Jensen V, Pawlak V, Seeburg PH, Hvalby O, Mol Pharmacol 63:5364.
Khr G (2005) Lack of NMDAR subtype selectivity for hip- Calakos N, Schoch S, Sdhof TC, Malenka RC (2004) Multiple roles
pocampal LTP. J Neurosci 25:69076910. for the active zone protein RIM1alpha in late stages of neuro-
Berger T, Schwarz C, Kraushaar U, Monyer H (1998) Dentate gyrus transmitter release. Neuron 42:889896.
basket cell GABAA receptors are blocked by Zn2 via changes of Castillo PE, Janz R, Sdhof TC, Tzounopoulos T, Malenka RC, Nicoll
their desensitization kinetics: an in situ patch-clamp and single- RA (1997b) Rab3A is essential for mossy bre long-term poten-
cell PCR study. J Neurosci 18:24372448. tiation in the hippocampus. Nature 388:590593.
Bergersen L, Ruiz A, Bjaalie JG, Kullmann DM, Gundersen V (2003) Castillo PE, Malenka RC, Nicoll RA (1997a) Kainate receptors medi-
GABA and GABAA receptors at hippocampal mossy bre ate a slow postsynaptic current in hippocampal CA3 neurons.
synapses. Eur J Neurosci 18:931941. Nature 388:182186.
Bettler B, Boulter J, Hermans-Borgmeyer I, OShea-Greeneld A, Castillo PE, Schoch S, Schmitz F, Sdhof TC, Malenka RC (2002)
Deneris ES, Moll C, Borgmeyer U, Hollmann M, Heinemann S RIM1alpha is required for presynaptic long-term potentiation.
(1990) Cloning of a novel glutamate receptor subunit, GluR5: Nature 415:327330.
expression in the nervous system during development. Neuron Chambers MS, Atack JR, Broughton HB, Collinson N, Cook S,
5:583595. Dawson GR, Hobbs SC, Marshall G, Maubach KA, Pillai GV,
Bettler B, Egebjerg J, Sharma G, Pecht G, Hermans-Borgmeyer I, Moll Reeve AJ, MacLeod AM (2003) Identication of a novel, selec-
C, Stevens CF, Heinemann S (1992) Cloning of a putative gluta- tive GABA(A) alpha5 receptor inverse agonist which enhances
mate receptor: a low affinity kainate-binding subunit. Neuron cognition. J Med Chem 46:22272240.
8:257265. Champtiaux N, Changeux JP (2002) Knock-out and knock-in mice
Bettler B, Kaupmann K, Mosbacher J, Gassmann M (2004) Molecular to investigate the role of nicotinic receptors in the central nerv-
structure and physiological functions of GABAB receptors. ous system. Curr Drug Targets CNS Neurol Disord 1:319330.
Physiol Rev 84:835867. Chatterton JE, Awobuluyi M, Premkumar LS, Takahashi H, Talantova
Bollmann JH, Sakmann B, Borst JG (2000) Calcium sensitivity M, Shin Y, Cui J, Tu S, Sevarino KA, Nakanishi N, Tong G,
of glutamate release in a calyx-type terminal. Science Lipton SA, Zhang D (2002) Excitatory glycine receptors con-
289:953957. taining the NR3 family of NMDA receptor subunits. Nature
Boudin H, Doan A, Xia J, Shigemoto R, Huganir RL, Worley P, Craig 415:793798.
AM (2000) Presynaptic clustering of mGluR7a requires the Chen L, Chetkovich DM, Petralia RS, Sweeney NT, Kawasaki Y,
PICK1 PDZ domain binding site. Neuron 28:485497. Wenthold RJ, Bredt DS, Nicoll RA (2000) Stargazin regulates
Bredt DS, Nicoll RA (2003) AMPA receptor trafficking at excitatory synaptic targeting of AMPA receptors by two distinct mecha-
synapses. Neuron 40:361379. nisms. Nature 408:936943.
Brooks-Kayal AR, Shumate MD, Jin H, Rikhter TY, Coulter DA Chi P, Greengard P, Ryan TA (2003) Synaptic vesicle mobilization is
(1998) Selective changes in single cell GABA(A) receptor sub- regulated by distinct synapsin I phosphorylation pathways at
unit expression and function in temporal lobe epilepsy. Nat Med different frequencies. Neuron 38:6978.
4:11661172. Ciabarra AM, Sullivan JM, Gahn LG, Pecht G, Heinemann S,
Brooks-Kayal AR, Shumate MD, Jin H, Rikhter TY, Kelly ME, Coulter Sevarino KA (1995) Cloning and characterization of chi-1: a
DA (2001) Gamma-aminobutyric acid(A) receptor subunit developmentally regulated member of a novel class of the
expression predicts functional changes in hippocampal dentate ionotropic glutamate receptor family. J Neurosci 15:64986508.
granule cells during postnatal development. J Neurochem Collinson N, Kuenzi FM, Jarolimek W, Maubach KA, Cothliff R, Sur
77:12661278. C, Smith A, Otu FM, Howell O, Atack JR, McKernan RM,
288 The Hippocampus Book

Seabrook GR, Dawson GR, Whiting PJ, Rosahl TW (2002) El-Husseini Ael D, Bredt DS (2002) Protein palmitoylation: a regula-
Enhanced learning and memory and altered GABAergic synap- tor of neuronal development and function. Nat Rev Neurosci
tic transmission in mice lacking the alpha 5 subunit of the 3:791802.
GABAA receptor. J Neurosci 22:55725580. Ernst M, Brauchart D, Boresch S, Sieghart W (2003) Comparative
Conquet F, Bashir ZI, Davies CH, Daniel H, Ferraguti F, Bordi F, modeling of GABA(A) receptors: limits, insights, future devel-
Franz-Bacon K, Reggiani A, Matarese V, Conde F, et al. (1994) opments. Neuroscience 119:933943.
Motor decit and impairment of synaptic plasticity in mice Essrich C, Lorez M, Benson JA, Fritschy JM, Luscher B (1998)
lacking mGluR1. Nature 372:237243. Postsynaptic clustering of major GABAA receptor subtypes
Contractor A, Swanson GT, Sailer A, OGorman S, Heinemann SF requires the gamma 2 subunit and gephyrin. Nat Neurosci
(2000) Identication of the kainate receptor subunits underly- 1:563571.
ing modulation of excitatory synaptic transmission in the CA3 Esteban JA, Shi SH, Wilson C, Nuriya M, Huganir RL, Malinow R
region of the hippocampus. J Neurosci 20:82698278. (2003) PKA phosphorylation of AMPA receptor subunits con-
Contractor A, Swanson G, Heinemann SF (2001) Kainate receptors trols synaptic trafficking underlying plasticity. Nat Neurosci
are involved in short- and long-term plasticity at mossy ber 6:136143.
synapses in the hippocampus. Neuron 29:209216. Fabian-Fine R, Skehel P, Errington ML, Davies HA, Sher E, Stewart
Contractor A, Sailer AW, Darstein M, Maron C, Xu J, Swanson GT, MG, Fine A (2001) Ultrastructural distribution of the alpha7
Heinemann SF (2003) Loss of kainate receptor-mediated het- nicotinic acetylcholine receptor subunit in rat hippocampus.
erosynaptic facilitation of mossy-ber synapses in KA2-/- mice. J Neurosci 21:79938003.
J Neurosci 23:422429. Fagni L, Worley PF, Ango F (2002) Homer as both a scaffold and
Cossart R, Esclapez M, Hirsch JC, Bernard C, Ben-Ari Y (1998) transduction molecule. Sci STKE 2002:RE8.
GluR5 kainate receptor activation in interneurons increases Feldmeyer D, Kask K, Brusa R, Kornau HC, Kolhekar R, Rozov A,
tonic inhibition of pyramidal cells. Nat Neurosci 1:470478. Burnashev N, Jensen V, Hvalby O, Sprengel R, Seeburg PH
Cremona O, Di Paolo G, Wenk MR, Luthi A, Kim WT, Takei K, (1999) Neurological dysfunctions in mice expressing different
Daniell L, Nemoto Y, Shears SB, Flavell RA, McCormick DA, De levels of the Q/R site-unedited AMPAR subunit GluR-B. Nat
Camilli P (1999) Essential role of phosphoinositide metabolism Neurosci 2:5764.
in synaptic vesicle recycling. Cell 99:179188. Fernandez-Chacon R, Konigstorfer A, Gerber SH, Garcia J, Matos
Crestani F, Keist R, Fritschy JM, Benke D, Vogt K, Prut L, Bluthmann MF, Stevens CF, Brose N, Rizo J, Rosenmund C, Sdhof TC
H, Mohler H, Rudolph U (2002) Trace fear conditioning (2001) Synaptotagmin I functions as a calcium regulator of
involves hippocampal alpha5 GABA(A) receptors. Proc Natl release probability. Nature 410:4149.
Acad Sci USA 99:89808985. File SE, Kenny PJ, Cheeta S (2000) The role of the dorsal hippocam-
Cui Z, Wang H, Tan Y, Zaia KA, Zhang S, Tsien JZ (2004) Inducible pal serotonergic and cholinergic systems in the modulation of
and reversible NR1 knockout reveals crucial role of the NMDA anxiety. Pharmacol Biochem Behav 66:6572.
receptor in preserving remote memories in the brain. Neuron Fisahn A, Yamada M, Duttaroy A, Gan JW, Deng CX, McBain CJ,
41:781793. Wess J (2002) Muscarinic induction of hippocampal gamma
Dajas-Bailador F, Wonnacott S (2004) Nicotinic acetylcholine recep- oscillations requires coupling of the M1 receptor to two mixed
tors and the regulation of neuronal signalling. Trends Pharmacol cation currents. Neuron 33:615624.
Sci 25:317324. Fleischmann A, Hvalby O, Jensen V, Strekalova T, Zacher C, Layer LE,
Darstein M, Petralia RS, Swanson GT, Wenthold RJ, Heinemann SF Kvello A, Reschke M, Spanagel R, Sprengel R, Wagner EF, Gass P
(2003) Distribution of kainate receptor subunits at hippocam- (2003) Impaired long-term memory and NR2A-type NMDA
pal mossy ber synapses. J Neurosci 23:80138019. receptor-dependent synaptic plasticity in mice lacking c-Fos in
De Blasi A, Conn PJ, Pin J, Nicoletti F (2001) Molecular determinants the CNS. J Neurosci 23:91169122.
of metabotropic glutamate receptor signaling. Trends Pharmacol Fonck C, Nashmi R, Deshpande P, Damaj MI, Marks MJ, Riedel A,
Sci 22:114120. Schwarz J, Collins AC, Labarca C, Lester HA (2003) Increased
Dingledine R, Borges K, Bowie D, Traynelis SF (1999) The glutamate sensitivity to agonist-induced seizures, straub tail, and hip-
receptor ion channels. Pharmacol Rev 51:761. pocampal theta rhythm in knock-in mice carrying hypersensi-
Di Paolo G, Sankaranarayanan S, Wenk MR, Daniell L, Perucco E, tive alpha 4 nicotinic receptors. J Neurosci 23:25822590.
Caldarone BJ, Flavell R, Picciotto MR, Ryan TA, Cremona O, De Forrest D, Yuzaki M, Soares HD, Ng L, Luk DC, Sheng M, Stewart CL,
Camilli P (2002) Decreased synaptic vesicle recycling efficiency Morgan JI, Connor JA, Curran T (1994) Targeted disruption of
and cognitive decits in amphiphysin 1 knockout mice. Neuron NMDA receptor 1 gene abolishes NMDA response and results in
33:789804. neonatal death. Neuron 13:325338.
Doussau F, Augustine GJ (2000) The actin cytoskeleton and neuro- Fotuhi M, Standaert DG, Testa CM, Penney JB Jr, Young AB (1994)
transmitter release: an overview. Biochimie 82:353363. Differential expression of metabotropic glutamate receptors in
Ebralidze AK, Rossi DJ, Tonegawa S, Slater NT (1996) Modication the hippocampus and entorhinal cortex of the rat. Brain Res Mol
of NMDA receptor channels and synaptic transmission by tar- Brain Res 21:283292.
geted disruption of the NR2C gene. J Neurosci 16:50145025. Fremeau RT Jr, Voglmaier S, Seal RP, Edwards RH (2004) VGLUTs
Egebjerg J, Bettler B, Hermans-Borgmeyer I, Heinemann S (1991) dene subsets of excitatory neurons and suggest novel roles for
Cloning of a cDNA for a glutamate receptor subunit activated glutamate. Trends Neurosci 27:98103.
by kainate but not AMPA. Nature 351:745748. Frerking M, Malenka RC, Nicoll RA (1998) Synaptic activation of
Eiden LE (2000) The vesicular neurotransmitter transporters: cur- kainate receptors on hippocampal interneurons. Nat Neurosci
rent perspectives and future prospects. FASEB J 14:23962400. 1:479486.
 Molecular Mechanisms of Synaptic Function 289

Fritschy JM, Brunig I (2003) Formation and plasticity of GABAergic gle interneurons of different types in the CA1 area of the rat
synapses: physiological mechanisms and pathophysiological hippocampus. J Neurosci 19:1008210097.
implications. Pharmacol Ther 98:299323. Gunther U, Benson J, Benke D, Fritschy JM, Reyes G, Knoach F,
Fritschy JM, Mohler H (1995) GABAA-receptor heterogeneity in the Crestani F, Aguzzi A, Arigoni M, Lang Y, et al. (1995)
adult rat brain: differential regional and cellular distribution of Benzodiazepine-insensitive mice generated by targeted disrup-
seven major subunits. J Comp Neurol 359:154194. tion of the gamma 2 subunit gene of gamma-aminobutyric acid
Fritschy JM, Weinmann O, Wenzel A, Benke D (1998) Synapse- type A receptors. Proc Natl Acad Sci USA 92:77497753.
specic localization of NMDA and GABA(A) receptor subunits Gutierrez R (2003) The GABAergic phenotype of the glutamatergic
revealed by antigen-retrieval immunohistochemistry. J Comp granule cells of the dentate gyrus. Prog Neurobiol 71:
Neurol 390:194210. 337358.
Fritschy JM, Meskenaite V, Weinmann O, Honer M, Benke D, Mohler Halasy K, Somogyi P (1993) Subdivisions in the multiple GABAergic
H (1999) GABAB-receptor splice variants GB1a and GB1b in rat innervation of granule cells in the dentate gyrus of the rat hip-
brain: developmental regulation, cellular distribution and pocampus. Eur J Neurosci 5:411429.
extrasynaptic localization. Eur J Neurosci 11:761768. Hanley JG, Khatri L, Hanson PI, Ziff EB (2002) NSF ATPase and
Fukaya M, Kato A, Lovett C, Tonegawa S, Watanabe M (2003) alpha-/beta-SNAPs disassemble the AMPA receptor-PICK1
Retention of NMDA receptor NR2 subunits in the lumen of complex. Neuron 34:5367.
endoplasmic reticulum in targeted NR1 knockout mice. Proc Hanson PI, Roth R, Morisaki H, Jahn R, Heuser JE (1997) Structure
Natl Acad Sci USA 100:48554860. and conformational changes in NSF and its membrane receptor
Gandhi SP, Stevens CF (2003) Three modes of synaptic vesicular complexes visualized by quick-freeze/deep-etch electron
recycling revealed by single-vesicle imaging. Nature microscopy. Cell 90:523535.
423:607613. Hilker S, Pieribone VA, Czernik AJ, Kao HT, Augustine GJ,
Gao B, Fritschy JM (1994) Selective allocation of GABAA receptors Greengard P (1999) Synapsins as regulators of neurotransmitter
containing the alpha 1 subunit to neurochemically distinct sub- release. Philos Trans R Soc Lond B Biol Sci 354:269279.
populations of rat hippocampal interneurons. Eur J Neurosci Hoffman DA, Sprengel R, Sakmann B (2002) Molecular dissection of
6:837853. hippocampal theta-burst pairing potentiation. Proc Natl Acad
Gassmann M, Shaban H, Vigot R, Sansig G, Haller C, Barbieri S, Sci USA 99:77407745.
Humeau Y, Schuler V, Muller M, Kinzel B, Klebs K, Schmutz M, Hosie AM, Dunne EL, Harvey RJ, Smart TG (2003) Zinc-mediated
Froestl W, Heid J, Kelly PH, Gentry C, Jaton AL, Van der Putten inhibition of GABA(A) receptors: discrete binding sites under-
H, Mombereau C, Lecourtier L, Mosbacher J, Cryan JF, Fritschy lie subtype specicity. Nat Neurosci 6:362369.
JM, Luthi A, Kaupmann K, Bettler B (2004) Redistribution of Hsia AY, Salin PA, Castillo PE, Aiba A, Abeliovich A, Tonegawa S,
GABAB(1) protein and atypical GABAB responses in GABAB(2)- Nicoll RA (1995) Evidence against a role for metabotropic
decient mice. J Neurosci 24:60866097. glutamate receptors in mossy ber LTP: the use of mutant mice
Geiger JR, Melcher T, Koh DS, Sakmann B, Seeburg PH, Jonas P, and pharmacological antagonists. Neuropharmacology 34:
Monyer H (1995) Relative abundance of subunit mRNAs 15671572.
determines gating and Ca2 permeability of AMPA receptors in Hsu FC, Zhang GJ, Raol YS, Valentino RJ, Coulter DA, Brooks-Kayal
principal neurons and interneurons in rat CNS. Neuron 15: AR (2003) Repeated neonatal handling with maternal separa-
193204. tion permanently alters hippocampal GABAA receptors and
Geppert M, Bolshakov VY, Siegelbaum SA, Takei K, De Camilli P, behavioral stress responses. Proc Natl Acad Sci USA
Hammer RE, Sdhof TC (1994a) The role of Rab3A in neuro- 100:1221312218.
transmitter release. Nature 369:493497. Huerta PT, Sun LD, Wilson MA, Tonegawa S (2000) Formation of
Geppert M, Goda Y, Hammer RE, Li C, Rosahl TW, Stevens CF, temporal memory requires NMDA receptors within CA1
Sdhof TC (1994b) Synaptotagmin I: a major Ca2 sensor for pyramidal neurons. Neuron 25:473480.
transmitter release at a central synapse. Cell 79:717727. Ikeda K, Araki K, Takayama C, Inoue Y, Yagi T, Aizawa S, Mishina M
Geppert M, Goda Y, Stevens CF, Sdhof TC (1997) The small GTP- (1995) Reduced spontaneous activity of mice defective in the
binding protein Rab3A regulates a late step in synaptic vesicle epsilon 4 subunit of the NMDA receptor channel. Brain Res Mol
fusion. Nature 387:810814. Brain Res 33:6171.
Goldstein PA, Elsen FP, Ying SW, Ferguson C, Homanics GE, Ito I, Futai K, Katagiri H, Watanabe M, Sakimura K, Mishina M,
Harrison NL (2002) Prolongation of hippocampal miniature Sugiyama H (1997) Synapse-selective impairment of NMDA
inhibitory postsynaptic currents in mice lacking the GABA(A) receptor functions in mice lacking NMDA receptor epsilon 1 or
receptor alpha1 subunit. J Neurophysiol 88:32083217. epsilon 2 subunit. J Physiol 500:401408.
Gtz T, Kraushaar U, Geiger J, Lubke J, Berger T, Jonas P (1997) Jensen K, Chiu CS, Sokolova I, Lester HA, Mody I (2003a) GABA
Functional properties of AMPA and NMDA receptors expressed transporter-1 (GAT1)-decient mice: differential tonic activa-
in identified types of basal ganglia neurons. J Neurosci tion of GABAA versus GABAB receptors in the hippocampus. J
17:204215. Neurophysiol 90:26902701.
Grant SG, ODell TJ, Karl KA, Stein PL, Soriano P, Kandel ER (1992) Jensen V, Kaiser KM, Borchardt T, Adelmann G, Rozov A, Burnashev
Impaired long-term potentiation, spatial learning, and hip- N, Brix C, Frotscher M, Andersen P, Hvalby , Sakmann B,
pocampal development in fyn mutant mice. Science Seeburg PH, Sprengel R (2003b) A juvenile form of postsynap-
258:19031910. tic hippocampal long-term potentiation in mice decient for
Gulyas AI, Megias M, Emri Z, Freund TF (1999) Total number and the AMPA receptor subunit GluR-A. J Physiol 553:843856.
ratio of excitatory and inhibitory synapses converging onto sin- Ji D, Lape R, Dani JA (2001) Timing and location of nicotinic activ-
290 The Hippocampus Book

ity enhances or depresses hippocampal synaptic plasticity. Intracellular domains of NMDA receptor subtypes are determi-
Neuron 31:131141. nants for long-term potentiation induction. J Neurosci
Jia Z, Agopyan N, Miu P, Xiong Z, Henderson J, Gerlai R, Taverna FA, 23:1079110799.
Velumian A, MacDonald J, Carlen P, Abramow-Newerly W, Kolleker A, Zhu JJ, Schupp BJ, Qin Y, Mack V, Borchardt T, Khr G,
Roder J (1996) Enhanced LTP in mice decient in the AMPA Malinow R, Seeburg PH, Osten P (2003) Glutamatergic plastic-
receptor GluR2. Neuron 17:945956. ity by synaptic delivery of GluR-B(long)-containing AMPA
Kadotani H, Hirano T, Masugi M, Nakamura K, Nakao K, Katsuki M, receptors. Neuron 40:11991212.
Nakanishi S (1996) Motor discoordination results from com- Kornau HC, Schenker LT, Kennedy MB, Seeburg PH (1995) Domain
bined gene disruption of the NMDA receptor NR2A and NR2C interaction between NMDA receptor subunits and the postsy-
subunits, but not from single disruption of the NR2A or NR2C naptic density protein PSD-95. Science 269:17371740.
subunit. J Neurosci 16:78597867. Korpi ER, Grunder G, Luddens H (2002) Drug interactions at
Kaila K, Lamsa K, Smirnov S, Taira T, Voipio J (1997) Long-lasting GABA(A) receptors. Prog Neurobiol 67:113159.
GABA-mediated depolarization evoked by high-frequency Krestel HE, Shimshek DR, Jensen V, Nevian T, Kim J, Geng Y, Bast T,
stimulation in pyramidal neurons of rat hippocampal slice is Depaulis A, Schonig K, Schwenk F, Bujard H, Hvalby O,
attributable to a network-driven, bicarbonate-dependent K Sprengel R, Seeburg PH (2004) A genetic switch for epilepsy in
transient. J Neurosci 17:76627672. adult mice. J Neurosci 24:1056810578.
Kawakami R, Shinohara Y, Kato Y, Sugiyama H, Shigemoto R, Ito I Krupp JJ, Vissel B, Thomas CG, Heinemann SF, Westbrook GL (1999)
(2003) Asymmetrical allocation of NMDA receptor epsilon2 Interactions of calmodulin and alpha-actinin with the NR1 sub-
subunits in hippocampal circuitry. Science 300:990994. unit modulate Ca2-dependent inactivation of NMDA recep-
Keinanen K, Wisden W, Sommer B, Werner P, Herb A, Verdoorn TA, tors. J Neurosci 19:11651178.
Sakmann B, Seeburg PH (1990) A family of AMPA-selective glu- Kulik A, Vida I, Lujan R, Haas CA, Lopez-Bendito G, Shigemoto R,
tamate receptors. Science 249:556560. Frotscher M (2003) Subcellular localization of metabotropic
Kerner JA, Standaert DG, Penney JB Jr, Young AB, Landwehrmeyer GABA(B) receptor subunits GABA(B1a/b) and GABA(B2) in
GB (1997) Expression of group one metabotropic glutamate the rat hippocampus. J Neurosci 23:1102611035.
receptor subunit mRNAs in neurochemically identied neurons Kullmann DM, Ruiz A, Rusakov DM, Scott R, Semyanov A, Walker
in the rat neostriatum, neocortex, and hippocampus. Brain Res MC (2005) Presynaptic, extrasynaptic and axonal GABAA recep-
Mol Brain Res 48:259269. tors in the CNS: where and why? Prog Biophys Mol Biol
Kew JN, Koester A, Moreau JL, Jenck F, Ouagazzal AM, Mutel V, 87:3346.
Richards JG, Trube G, Fischer G, Montkowski A, Hundt W, Kutsuwada T, Sakimura K, Manabe T, Takayama C, Katakura N,
Reinscheid RK, Pauly-Evers M, Kemp JA, Bluethmann H (2000) Kushiya E, Natsume R, Watanabe M, Inoue Y, Yagi T, Aizawa S,
Functional consequences of reduction in NMDA receptor Arakawa M, Takahashi T, Nakamura Y, Mori H, Mishina M
glycine affinity in mice carrying targeted point mutations in the (1996) Impairment of suckling response, trigeminal neuronal
glycine binding site. J Neurosci 20:40374049. pattern formation, and hippocampal LTD in NMDA receptor
Khiroug L, Giniatullin R, Klein RC, Fayuk D, Yakel JL (2003) epsilon 2 subunit mutant mice. Neuron 16:333344.
Functional mapping and Ca2 regulation of nicotinic acetyl- Laezza F, Doherty JJ, Dingledine R (1999) Long-term depression in
choline receptor channels in rat hippocampal CA1 neurons. hippocampal interneurons: joint requirement for pre- and post-
J Neurosci 23:90249031. synaptic events. Science 285:14111414.
Kim CH, Chung HJ, Lee HK, Huganir RL (2001) Interaction of the Laurie DJ, Seeburg PH (1994) Regional and developmental hetero-
AMPA receptor subunit GluR2/3 with PDZ domains regulates geneity in splicing of the rat brain NMDAR1 mRNA. J Neurosci
hippocampal long-term depression. Proc Natl Acad Sci USA 14:31803194.
98:1172511730. Lee HK, Takamiya K, Han JS, Man H, Kim CH, Rumbaugh G, Yu S,
Kimura F (2000) Cholinergic modulation of cortical function: a Ding L, He C, Petralia RS, Wenthold RJ, Gallagher M, Huganir
hypothetical role in shifting the dynamics in cortical network. RL (2003) Phosphorylation of the AMPA receptor GluR1 sub-
Neurosci Res 38:1926. unit is required for synaptic plasticity and retention of spatial
Kittler JT, Moss SJ (2003) Modulation of GABAA receptor activity memory. Cell 112:631643.
by phosphorylation and receptor trafficking: implications for Lee SH, Liu L, Wang YT, Sheng M (2002) Clathrin adaptor AP2 and
the efficacy of synaptic inhibition. Curr Opin Neurobiol NSF interact with overlapping sites of GluR2 and play distinct
13:341347. roles in AMPA receptor trafficking and hippocampal LTD.
Kiyama Y, Manabe T, Sakimura K, Kawakami F, Mori H, Mishina M Neuron 36:661674.
(1998) Increased thresholds for long-term potentiation and Lei S, McBain CJ (2002) Distinct NMDA receptors provide differen-
contextual learning in mice lacking the NMDA-type glutamate tial modes of transmission at mossy berinterneuron synapses.
receptor epsilon1 subunit. J Neurosci 18:67046712. Neuron 33:921933.
Klausberger T, Roberts JD, Somogyi P (2002) Cell type- and input- Lerma J (2003) Roles and rules of kainate receptors in synaptic
specic differences in the number and subtypes of synaptic transmission. Nat Rev Neurosci 4:481495.
GABA(A) receptors in the hippocampus. J Neurosci Letts VA (2005) Stargazera mouse to seize! Epilepsy Curr
22:25132521. 5:161165.
Kneussel M, Brandstatter JH, Laube B, Stahl S, Muller U, Betz H Levey AI, Edmunds SM, Koliatsos V, Wiley RG, Heilman CJ (1995)
(1999) Loss of postsynaptic GABA(A) receptor clustering in Expression of m1-m4 muscarinic acetylcholine receptor pro-
gephyrin-decient mice. J Neurosci 19:92899297. teins in rat hippocampus and regulation by cholinergic inner-
Khr G, Jensen V, Koester HJ, Mihaljevic AL, Utvik JK, Kvello A, vation. J Neurosci 15:40774092.
Ottersen OP, Seeburg PH, Sprengel R, Hvalby O (2003) Levi S, Grady RM, Henry MD, Campbell KP, Sanes JR, Craig AM
 Molecular Mechanisms of Synaptic Function 291

(2002) Dystroglycan is selectively associated with inhibitory R, Demirgoren S, Rahamimoff R (1999) Ion channels in presy-
GABAergic synapses but is dispensable for their differentiation. naptic nerve terminals and control of transmitter release.
J Neurosci 22:42744285. Physiol Rev 79:10191088.
Levi S, Logan SM, Tovar KR, Craig AM (2004) Gephyrin is critical for Meng Y, Zhang Y, Jia Z (2003) Synaptic transmission and plasticity in
glycine receptor clustering but not for the formation of func- the absence of AMPA glutamate receptor GluR2 and GluR3.
tional GABAergic synapses in hippocampal neurons. J Neurosci Neuron 39:163176.
24:207217. Migaud M, Charlesworth P, Dempster M, Webster LC, Watabe AM,
Levin ED, Bradley A, Addy N, Sigurani N (2002) Hippocampal alpha Makhinson M, He Y, Ramsay MF, Morris RG, Morrison JH,
7 and alpha 4 beta 2 nicotinic receptors and working memory. ODell TJ, Grant SG (1998) Enhanced long-term potentiation
Neuroscience 109:757765. and impaired learning in mice with mutant postsynaptic
Li L, Chin LS, Shupliakov O, Brodin L, Sihra TS, Hvalby O, Jensen V, density-95 protein. Nature 396:433439.
Zheng D, McNamara JO, Greengard P, et al. (1995) Impairment Mihalek RM, Banerjee PK, Korpi ER, Quinlan JJ, Firestone LL,
of synaptic vesicle clustering and of synaptic transmission, and Mi ZP, Lagenaur C, Tretter V, Sieghart W, Anagnostaras SG,
increased seizure propensity, in synapsin I-decient mice. Proc Sage JR, Fanselow MS, Guidotti A, Spigelman I, Li Z, DeLorey
Natl Acad Sci USA 92:92359239. TM, Olsen RW, Homanics GE (1999) Attenuated sensitivity to
Lisman J, Schulman H, Cline H (2002) The molecular basis of neuroactive steroids in gamma-aminobutyrate type A receptor
CaMKII function in synaptic and behavioural memory. Nat Rev delta subunit knockout mice. Proc Natl Acad Sci USA
Neurosci 3:175190. 96:1290512910.
Liu L, Wong TP, Pozza MF, Lingenhoehl K, Wang Y, Sheng M, Mohn AR, Gainetdinov RR, Caron MG, Koller BH (1999) Mice with
Auberson YP, Wang YT (2004) Role of NMDA receptor subtypes reduced NMDA receptor expression display behaviors related to
in governing the direction of hippocampal synaptic plasticity. schizophrenia. Cell 98:427436.
Science 304:10211024. Monyer H, Seeburg PH, Wisden W (1991) Glutamate-operated chan-
Liu Q, Kawai H, Berg DK (2001) Beta-amyloid peptide blocks the nels: developmentally early and mature forms arise by alterna-
response of alpha 7-containing nicotinic receptors on hip- tive splicing. Neuron 6:799810.
pocampal neurons. Proc Natl Acad Sci USA 98:47344739. Monyer H, Burnashev N, Laurie DJ, Sakmann B, Seeburg PH (1994)
Lujan R, Nusser Z, Roberts JD, Shigemoto R, Somogyi P (1996) Developmental and regional expression in the rat brain and
Perisynaptic location of metabotropic glutamate receptors functional properties of four NMDA receptors. Neuron
mGluR1 and mGluR5 on dendrites and dendritic spines in the 12:529540.
rat hippocampus. Eur J Neurosci 8:14881500. Mori H, Manabe T, Watanabe M, Satoh Y, Suzuki N, Toki S,
Maccaferri G, Roberts JD, Szucs P, Cottingham CA, Somogyi P Nakamura K, Yagi T, Kushiya E, Takahashi T, Inoue Y, Sakimura
(2000) Cell surface domain specic postsynaptic currents K, Mishina M (1998) Role of the carboxy-terminal region of the
evoked by identied GABAergic neurones in rat hippocampus GluR epsilon2 subunit in synaptic localization of the NMDA
in vitro. J Physiol 524:91116. receptor channel. Neuron 21:571580.
Mack V, Burnashev N, Kaiser KM, Rozov A, Jensen V, Hvalby O, Morris RG, Anderson E, Lynch GS, Baudry M (1986) Selective
Seeburg PH, Sakmann B, Sprengel R (2001) Conditional impairment of learning and blockade of long-term potentiation
restoration of hippocampal synaptic potentiation in GluR-A by an N-methyl-D-aspartate receptor antagonist, AP5. Nature
decient mice. Science 292:25012504. 319:774776.
Magee JC, Cook EP (2000) Somatic EPSP amplitude is independent Mott DD, Li Q, Okazaki MM, Turner DA, Lewis DV (1999) GABAB-
of synapse location in hippocampal pyramidal neurons. Nat receptor-mediated currents in interneurons of the dentate-hilus
Neurosci 3:895903. border. J Neurophysiol 82:14381450.
Mahmoudi M, Kang MH, Tillakaratne N, Tobin AJ, Olsen RW (1997) Mulle C, Sailer A, Perez-Otano I, Dickinson-Anson H, Castillo PE,
Chronic intermittent ethanol treatment in rats increases Bureau I, Maron C, Gage FH, Mann JR, Bettler B, Heinemann
GABA(A) receptor alpha4-subunit expression: possible rele- SF (1998) Altered synaptic physiology and reduced susceptibil-
vance to alcohol dependence. J Neurochem 68:24852492. ity to kainate-induced seizures in GluR6-decient mice. Nature
Malinow R, Malenka RC (2002) AMPA receptor trafficking and 392:601605.
synaptic plasticity. Annu Rev Neurosci 25:103126. Mulle C, Sailer A, Swanson GT, Brana C, OGorman S, Bettler B,
Marino MJ, Rouse ST, Levey AI, Potter LT, Conn PJ (1998) Activation Heinemann SF (2000) Subunit composition of kainate recep-
of the genetically dened m1 muscarinic receptor potentiates tors in hippocampal interneurons. Neuron 28:475484.
N-methyl-D-aspartate (NMDA) receptor currents in hip- Nakagawa T, Cheng Y, Ramm E, Sheng M, Walz T (2005) Structure
pocampal pyramidal cells. Proc Natl Acad Sci USA 95: and different conformational states of native AMPA receptor
1146511470. complexes. Nature 433:545549.
May AP, Whiteheart SW, Weis WI (2001) Unraveling the mechanism Nakazawa K, Quirk MC, Chitwood RA, Watanabe M, Yeckel MF, Sun
of the vesicle transport ATPase NSF, the N-ethylmaleimide- LD, Kato A, Carr CA, Johnston D, Wilson MA, Tonegawa S
sensitive factor. J Biol Chem 276:2199121994. (2002) Requirement for hippocampal CA3 NMDA receptors in
McHugh TJ, Blum KI, Tsien JZ, Tonegawa S, Wilson MA (1996) associative memory recall. Science 297:211218.
Impaired hippocampal representation of space in CA1-specic Nakazawa K, Sun LD, Quirk MC, Rondi-Reig L, Wilson MA,
NMDAR1 knockout mice. Cell 87:13391349. Tonegawa S (2003) Hippocampal CA3 NMDA receptors are
Megias M, Emri Z, Freund TF, Gulyas AI (2001) Total number and crucial for memory acquisition of one-time experience. Neuron
distribution of inhibitory and excitatory synapses on hip- 38:305315.
pocampal CA1 pyramidal cells. Neuroscience 102:527540. Nehring RB, Horikawa HP, El Far O, Kneussel M, Brandstatter JH,
Meir A, Ginsburg S, Butkevich A, Kachalsky SG, Kaiserman I, Ahdut Stamm S, Wischmeyer E, Betz H, Karschin A (2000) The
292 The Hippocampus Book

metabotropic GABAB receptor directly interacts with the acti- bumin- and cholecystokinin-positive interneurones in CA1 of
vating transcription factor 4. J Biol Chem 275:3518535191. the adult rat hippocampus. J Comp Neurol 443:346367.
Nishi T, Forgac M (2002) The vacuolar (H)-ATPasesnatures Pawlak V, Jensen V, Schupp BJ, Kvello A, Hvalby O, Seeburg PH, Khr
most versatile proton pumps. Nat Rev Mol Cell Biol 3:94103. G (2005a) Frequency-dependent impairment of hippocampal
Nusser Z (1999) A new approach to estimate the number, density and LTP from NMDA receptors with reduced calcium permeability.
variability of receptors at central synapses. Eur J Neurosci Eur J Neurosci 22:476484.
11:745752. Pawlak V, Schupp BJ, Single FN, Seeburg PH, Khr G (2005b)
Nusser Z, Sieghart W, Benke D, Fritschy JM, Somogyi P (1996) Impaired synaptic scaling in mouse hippocampal neurones
Differential synaptic localization of two major gamma- expressing NMDA receptors with reduced calcium permeability.
aminobutyric acid type A receptor alpha subunits on hip- J Physiol 562:771783.
pocampal pyramidal cells. Proc Natl Acad Sci USA Paylor R, Nguyen M, Crawley JN, Patrick J, Beaudet A, Orr-Urtreger
93:1193911944. A (1998) Alpha7 nicotinic receptor subunits are not necessary
Nusser Z, Hajos N, Somogyi P, Mody I (1998a) Increased number of for hippocampal-dependent learning or sensorimotor gating: a
synaptic GABA(A) receptors underlies potentiation at hip- behavioral characterization of Acra7-decient mice. Learn Mem
pocampal inhibitory synapses. Nature 395:172177. 5:302316.
Nusser Z, Lujan R, Laube G, Roberts JD, Molnar E, Somogyi P Peng Z, Hauer B, Mihalek RM, Homanics GE, Sieghart W, Olsen RW,
(1998b) Cell type and pathway dependence of synaptic AMPA Houser CR (2002) GABA(A) receptor changes in delta subunit-
receptor number and variability in the hippocampus. Neuron decient mice: altered expression of alpha4 and gamma2 sub-
21:545559. units in the forebrain. J Comp Neurol 446:179197.
Nyiri G, Freund TF, Somogyi P (2001) Input-dependent synaptic tar- Persohn E, Malherbe P, Richards JG (1992) Comparative molecular
geting of alpha(2)-subunit-containing GABA(A) receptors in neuroanatomy of cloned GABAA receptor subunits in the rat
synapses of hippocampal pyramidal cells of the rat. Eur J CNS. J Comp Neurol 326:193216.
Neurosci 13:428442. Petralia RS, Wenthold RJ (1992) Light and electron immunocyto-
OBrien RJ, Xu D, Petralia RS, Steward O, Huganir RL, Worley P chemical localization of AMPA-selective glutamate receptors in
(1999) Synaptic clustering of AMPA receptors by the extracellu- the rat brain. J Comp Neurol 318:329354.
lar immediate-early gene product Narp. Neuron 23:309323. Petralia RS, Esteban JA, Wang YX, Partridge JG, Zhao HM, Wenthold
Ohishi H, Shigemoto R, Nakanishi S, Mizuno N (1993) Distribution RJ, Malinow R (1999) Selective acquisition of AMPA receptors
of the mRNA for a metabotropic glutamate receptor (mGluR3) over postnatal development suggests a molecular basis for silent
in the rat brain: an in situ hybridization study. J Comp Neurol synapses. Nat Neurosci 2:3136.
335:252266. Pieribone VA, Porton B, Rendon B, Feng J, Greengard P, Kao HT
Ohishi H, Akazawa C, Shigemoto R, Nakanishi S, Mizuno N (1995) (2002) Expression of synapsin III in nerve terminals and neuro-
Distributions of the mRNAs for L-2-amino-4-phosphonobu- genic regions of the adult brain. J Comp Neurol 454:105114.
tyrate-sensitive metabotropic glutamate receptors, mGluR4 and Pirker S, Schwarzer C, Wieselthaler A, Sieghart W, Sperk G (2000)
mGluR7, in the rat brain. J Comp Neurol 360:555570. GABA(A) receptors: immunocytochemical distribution of 13
Okabe S, Collin C, Auerbach JM, Meiri N, Bengzon J, Kennedy MB, subunits in the adult rat brain. Neuroscience 101:815850.
Segal M, McKay RD (1998) Hippocampal synaptic plasticity in Poncer JC, McKinney RA, Gahwiler BH, Thompson SM (2000)
mice overexpressing an embryonic subunit of the NMDA recep- Differential control of GABA release at synapses from
tor. J Neurosci 18:41774188. distinct interneurons in rat hippocampus. J Physiol 528
Orr-Urtreger A, Broide RS, Kasten MR, Dang H, Dani JA, Beaudet (Pt 1):123130.
AL, Patrick JW (2000) Mice homozygous for the L250T muta- Powell CM, Schoch S, Monteggia L, Barrot M, Matos MF, Feldmann
tion in the alpha7 nicotinic acetylcholine receptor show N, Sdhof TC, Nestler EJ (2004) The presynaptic active zone
increased neuronal apoptosis and die within 1 day of birth. protein RIM1alpha is critical for normal learning and memory.
J Neurochem 74:21542166. Neuron 42:143153.
Osten P, Khatri L, Perez JL, Khr G, Giese G, Daly C, Schulz TW, Priel A, Kolleker A, Ayalon G, Gillor M, Osten P, Stern-Bach Y (2005)
Wensky A, Lee LM, Ziff EB (2000) Mutagenesis reveals a role for Stargazin reduces desensitization and slows deactivation of the
ABP/GRIP binding to GluR2 in synaptic surface accumulation AMPA-type glutamate receptors. J Neurosci 25:26822686.
of the AMPA receptor. Neuron 27:313325. Prosser HM, Gill CH, Hirst WD, Grau E, Robbins M, Calver A, Soffin
Pagano A, Rovelli G, Mosbacher J, Lohmann T, Duthey B, Stauffer D, EM, Farmer CE, Lanneau C, Gray J, Schenck E, Warmerdam BS,
Ristig D, Schuler V, Meigel I, Lampert C, Stein T, Prezeau L, Clapham C, Reavill C, Rogers DC, Stean T, Upton N,
Blahos J, Pin J, Froestl W, Kuhn R, Heid J, Kaupmann K, Bettler Humphreys K, Randall A, Geppert M, Davies CH, Pangalos MN
B (2001) C-terminal interaction is essential for surface traffick- (2001) Epileptogenesis and enhanced prepulse inhibition in
ing but not for heteromeric assembly of GABA(b) receptors. J GABA(B1)-decient mice. Mol Cell Neurosci 17:10591070.
Neurosci 21:11891202. Rampon C, Jiang CH, Dong H, Tang YP, Lockhart DJ, Schultz PG,
Papp E, Leinekugel X, Henze DA, Lee J, Buzsaki G (2001) The apical Tsien JZ, Hu Y (2000) Effects of environmental enrichment on
shaft of CA1 pyramidal cells is under GABAergic interneuronal gene expression in the brain. Proc Natl Acad Sci USA
control. Neuroscience 102:715721. 97:1288012884.
Paternain AV, Herrera MT, Nieto MA, Lerma J (2000) GluR5 and Reim K, Mansour M, Varoqueaux F, McMahon HT, Sdhof TC, Brose
GluR6 kainate receptor subunits coexist in hippocampal neu- N, Rosenmund C (2001) Complexins regulate a late step in
rons and coassemble to form functional receptors. J Neurosci Ca2-dependent neurotransmitter release. Cell 104:7181.
20:196205. Reisel D, Bannerman DM, Schmitt WB, Deacon RM, Flint J,
Pawelzik H, Hughes DI, Thomson AM (2002) Physiological and mor- Borchardt T, Seeburg PH, Rawlins JN (2002) Spatial memory
phological diversity of immunocytochemically dened parval- dissociations in mice lacking GluR1. Nat Neurosci 5:868873.
 Molecular Mechanisms of Synaptic Function 293

Richmond JE, Weimer RM, Jorgensen EM (2001) An open form of A-decient mice using a novel hippocampal-dependent pad-
syntaxin bypasses the requirement for UNC-13 in vesicle prim- dling pool escape task. Hippocampus 14:216223.
ing. Nature 412:338341. Schmitt WB, Sprengel R, Mack V, Draft RW, Seeburg PH, Deacon
Riedel G, Platt B, Micheau J (2003) Glutamate receptor function in RM, Rawlins JN, Bannerman DM (2005) Restoration of spatial
learning and memory. Behav Brain Res 140:147. working memory by genetic rescue of GluR-A-decient mice.
Rizo J, Sdhof TC (2002) Snares and Munc18 in synaptic vesicle Nat Neurosci 8:270272.
fusion. Nat Rev Neurosci 3:641653. Schneggenburger R, Neher E (2000) Intracellular calcium depend-
Rosahl TW, Geppert M, Spillane D, Herz J, Hammer RE, Malenka RC, ence of transmitter release rates at a fast central synapse. Nature
Sdhof TC (1993) Short-term synaptic plasticity is altered in 406:889893.
mice lacking synapsin I. Cell 75:661670. Schoch S, Deak F, Konigstorfer A, Mozhayeva M, Sara Y, Sdhof
Rosahl TW, Spillane D, Missler M, Herz J, Selig DK, Wolff JR, TC, Kavalali ET (2001) SNARE function analyzed in synapto-
Hammer RE, Malenka RC, Sdhof TC (1995) Essential func- brevin/VAMP knockout mice. Science 294:11171122.
tions of synapsins I and II in synaptic vesicle regulation. Nature Schoch S, Castillo PE, Jo T, Mukherjee K, Geppert M, Wang Y,
375:488493. Schmitz F, Malenka RC, Sdhof TC (2002) RIM1alpha forms a
Rouach N, Byrd K, Petralia RS, Elias GM, Adesnik H, Tomita S, protein scaffold for regulating neurotransmitter release at the
Karimzadegan S, Kealey C, Bredt DS, Nicoll RA (2005) TARP active zone. Nature 415:321326.
gamma-8 controls hippocampal AMPA receptor number, distri- Schuler V, Luscher C, Blanchet C, Klix N, Sansig G, Klebs K, Schmutz
bution and synaptic plasticity. Nat Neurosci 8:15251533. M, Heid J, Gentry C, Urban L, Fox A, Spooren W, Jaton AL,
Rudolph U, Mohler H (2004) Analysis of GABAA receptor function Vigouret J, Pozza M, Kelly PH, Mosbacher J, Froestl W, Kaslin E,
and dissection of the pharmacology of benzodiazepines and Korn R, Bischoff S, Kaupmann K, van der Putten H, Bettler B
general anesthetics through mouse genetics. Annu Rev (2001) Epilepsy, hyperalgesia, impaired memory, and loss of
Pharmacol Toxicol 44:475498. pre- and postsynaptic GABA(B) responses in mice lacking
Ruiz A, Fabian-Fine R, Scott R, Walker MC, Rusakov DA, Kullmann GABA(B(1)). Neuron 31:4758.
DM (2003) GABAA receptors at hippocampal mossy bers. Schulz TW, Nakagawa T, Licznerski P, Pawlak V, Kolleker A, Rozov A,
Neuron 39:961973. Kim J, Dittgen T, Kohr G, Sheng M, Seeburg PH, Osten P (2004)
Sailer A, Swanson GT, Perez-Otano I, OLeary L, Malkmus SA, Dyck Actin/alpha-actinin-dependent transport of AMPA receptors in
RH, Dickinson-Anson H, Schiffer HH, Maron C, Yaksh TL, dendritic spines: role of the PDZ-LIM protein RIL. J Neurosci
Gage FH, OGorman S, Heinemann SF (1999) Generation and 24:85848594.
analysis of GluR5(Q636R) kainate receptor mutant mice. J Schweizer C, Balsiger S, Bluethmann H, Mansuy IM, Fritschy JM,
Neurosci 19:87578764. Mohler H, Luscher B (2003) The gamma 2 subunit of GABA(A)
Sakimura K, Kutsuwada T, Ito I, Manabe T, Takayama C, Kushiya E, receptors is required for maintenance of receptors at mature
Yagi T, Aizawa S, Inoue Y, Sugiyama H, et al. (1995) Reduced synapses. Mol Cell Neurosci 24:442450.
hippocampal LTP and spatial learning in mice lacking NMDA Seagar M, Leveque C, Charvin N, Marqueze B, Martin-Moutot N,
receptor epsilon 1 subunit. Nature 373:151155. Boudier JA, Boudier JL, Shoji-Kasai Y, Sato K, Takahashi M
Sans N, Vissel B, Petralia RS, Wang YX, Chang K, Royle GA, Wang CY, (1999) Interactions between proteins implicated in exocytosis
OGorman S, Heinemann SF, Wenthold RJ (2003) Aberrant for- and voltage-gated calcium channels. Philos Trans R Soc Lond B
mation of glutamate receptor complexes in hippocampal neu- Biol Sci 354:289297.
rons of mice lacking the GluR2 AMPA receptor subunit. J Seeburg PH, Higuchi M, Sprengel R (1998) RNA editing of brain glu-
Neurosci 23:93679373. tamate receptor channels: mechanism and physiology. Brain Res
Sansig G, Bushell TJ, Clarke VR, Rozov A, Burnashev N, Portet C, Brain Res Rev 26:217229.
Gasparini F, Schmutz M, Klebs K, Shigemoto R, Flor PJ, Kuhn R, Sheng M, Pak DT (2000) Ligand-gated ion channel interactions with
Knoepfel T, Schroeder M, Hampson DR, Collett VJ, Zhang C, cytoskeletal and signaling proteins. Annu Rev Physiol
Duvoisin RM, Collingridge GL, van Der Putten H (2001) 62:755778.
Increased seizure susceptibility in mice lacking metabotropic Sheng M, Sala C (2001) PDZ domains and the organization of
glutamate receptor 7. J Neurosci 21:87348745. supramolecular complexes. Annu Rev Neurosci 24:129.
Schlter OM, Schnell E, Verhage M, Tzonopoulos T, Nicoll RA, Janz Shigemoto R, Nakanishi S, Mizuno N (1992) Distribution of the
R, Malenka RC, Geppert M, Sdhof TC (1999) Rabphilin mRNA for a metabotropic glutamate receptor (mGluR1) in the
knock-out mice reveal that rabphilin is not required for rab3 central nervous system: an in situ hybridization study in adult
function in regulating neurotransmitter release. J Neurosci and developing rat. J Comp Neurol 322:121135.
19:58345846. Shigemoto R, Kinoshita A, Wada E, Nomura S, Ohishi H, Takada M,
Schlter OM, Schmitz F, Jahn R, Rosenmund C, Sdhof TC (2004) A Flor PJ, Neki A, Abe T, Nakanishi S, Mizuno N (1997)
complete genetic analysis of neuronal Rab3 function. J Neurosci Differential presynaptic localization of metabotropic glutamate
24:66296637. receptor subtypes in the rat hippocampus. J Neurosci
Schmid SL, McNiven MA, De Camilli P (1998) Dynamin and its 17:75037522.
partners: a progress report. Curr Opin Cell Biol 10:504512. Shimizu E, Tang YP, Rampon C, Tsien JZ (2000) NMDA receptor-
Schmitt WB, Arianpour R, Deacon RM, Seeburg PH, Sprengel R, dependent synaptic reinforcement as a crucial process for mem-
Rawlins JN, Bannerman DM (2004a) The role of hippocampal ory consolidation. Science 290:11701174.
glutamate receptor-A-dependent synaptic plasticity in condi- Shimshek DR, Bus T, Kim J, Mihaljevic A, Mack V, Seeburg PH,
tional learning: the importance of spatiotemporal discontiguity. Sprengel R, Schaefer AT (2005) Enhanced odor discrimination
J Neurosci 24:92779282. and impaired olfactory memory by spatially controlled switch
Schmitt WB, Deacon RM, Reisel D, Sprengel R, Seeburg PH, Rawlins of AMPA receptors. PLoS Biol 3:e354.
JN, Bannerman DM (2004b) Spatial reference memory in GluR- Shimshek DR, Jensen V, Celikel T, Geng Y, Schupp B, Bus T, Mack V,
294 The Hippocampus Book

Marx V, Hvalby , Seeburg PH, Sprengel R (2006) Forebrain- metabotropic glutamate receptors by their cytoplasmic carboxy-
specic glutamate receptor B deletion impairs spatial memory terminal domains. Neuron 22:525536.
but not hippocampal eld long-term potentiation. J Neurosci Sucher NJ, Akbarian S, Chi CL, Leclerc CL, Awobuluyi M, Deitcher
26:84288440. DL, Wu MK, Yuan JP, Jones EG, Lipton SA (1995)
Silva AJ, Paylor R, Wehner JM, Tonegawa S (1992a) Impaired spatial Developmental and regional expression pattern of a novel
learning in alpha-calcium-calmodulin kinase II mutant mice. NMDA receptor-like subunit (NMDAR-L) in the rodent brain.
Science 257:206211. J Neurosci 15:65096520.
Silva AJ, Stevens CF, Tonegawa S, Wang Y (1992b) Decient hip- Sdhof TC (2004) The synaptic vesicle cycle. Annu Rev Neurosci
pocampal long-term potentiation in alpha-calcium-calmodulin 27:509547.
kinase II mutant mice. Science 257:201206. Sdhof TC, Czernik AJ, Kao HT, Takei K, Johnston PA, Horiuchi A,
Silva AJ, Rosahl TW, Chapman PF, Marowitz Z, Friedman E, Kanazir SD, Wagner MA, Perin MS, De Camilli P, Greengard P
Frankland PW, Cestari V, Cioffi D, Sdhof TC, Bourtchuladze R (1989) Synapsins: mosaics of shared and individual domains in
(1996) Impaired learning in mice with abnormal short-lived a family of synaptic vesicle phosphoproteins. Science
plasticity. Curr Biol 6:15091518. 245:14741480.
Single FN, Rozov A, Burnashev N, Zimmermann F, Hanley DF, Forrest Sundstrom-Poromaa I, Smith DH, Gong QH, Sabado TN, Li X, Light
D, Curran T, Jensen V, Hvalby O, Sprengel R, Seeburg PH (2000) A, Wiedmann M, Williams K, Smith SS (2002) Hormonally reg-
Dysfunctions in mice by NMDA receptor point mutations ulated alpha(4)beta(2)delta GABA(A) receptors are a target for
NR1(N598Q) and NR1(N598R). J Neurosci 20:25582566. alcohol. Nat Neurosci 5:721722.
Slepnev VI, De Camilli P (2000) Accessory factors in clathrin-depend- Sur C, Fresu L, Howell O, McKernan RM, Atack JR (1999)
ent synaptic vesicle endocytosis. Nat Rev Neurosci 1:161172. Autoradiographic localization of alpha5 subunit-containing
Smith SS, Gong QH, Hsu FC, Markowitz RS, ffrench-Mullen JM, Li GABAA receptors in rat brain. Brain Res 822:265270.
X (1998a) GABA(A) receptor alpha4 subunit suppression pre- Takei Y, Harada A, Takeda S, Kobayashi K, Terada S, Noda T,
vents withdrawal properties of an endogenous steroid. Nature Takahashi T, Hirokawa N (1995) Synapsin I deciency results in
392:926930. the structural change in the presynaptic terminals in the murine
Smith SS, Gong QH, Li X, Moran MH, Bitran D, Frye CA, Hsu FC nervous system. J Cell Biol 131:17891800.
(1998b) Withdrawal from 3alpha-OH-5alpha-pregnan-20-one Tamaru Y, Nomura S, Mizuno N, Shigemoto R (2001) Distribution of
using a pseudopregnancy model alters the kinetics of hip- metabotropic glutamate receptor mGluR3 in the mouse CNS:
pocampal GABAA-gated current and increases the GABAA differential location relative to pre- and postsynaptic sites.
receptor alpha4 subunit in association with increased anxiety. Neuroscience 106:481503.
J Neurosci 18:52755284. Tang YP, Shimizu E, Dube GR, Rampon C, Kerchner GA, Zhuo M,
Sommer B, Keinanen K, Verdoorn TA, Wisden W, Burnashev N, Herb Liu G, Tsien JZ (1999) Genetic enhancement of learning and
A, Kohler M, Takagi T, Sakmann B, Seeburg PH (1990) Flip and memory in mice. Nature 401:6369.
op: a cell-specic functional switch in glutamate-operated Terashima A, Cotton L, Dev KK, Meyer G, Zaman S, Duprat F, Henley
channels of the CNS. Science 249:15801585. JM, Collingridge GL, Isaac JT (2004) Regulation of synaptic
Somogyi P, Fritschy JM, Benke D, Roberts JD, Sieghart W (1996) The strength and AMPA receptor subunit composition by PICK1. J
gamma 2 subunit of the GABAA receptor is concentrated in Neurosci 24:53815390.
synaptic junctions containing the alpha 1 and beta 2/3 subunits Thomson AM, Bannister AP, Hughes DI, Pawelzik H (2000)
in hippocampus, cerebellum and globus pallidus. Differential sensitivity to zolpidem of IPSPs activated by mor-
Neuropharmacology 35:14251444. phologically identied CA1 interneurons in slices of rat hip-
Song I, Huganir RL (2002) Regulation of AMPA receptors during pocampus. Eur J Neurosci 12:425436.
synaptic plasticity. Trends Neurosci 25:578588. Tomita S, Chen L, Kawasaki Y, Petralia RS, Wenthold RJ, Nicoll RA,
Sperk G, Schwarzer C, Tsunashima K, Fuchs K, Sieghart W (1997) Bredt DS (2003) Functional studies and distribution dene a
GABA(A) receptor subunits in the rat hippocampus. I. family of transmembrane AMPA receptor regulatory proteins. J
Immunocytochemical distribution of 13 subunits. Neuroscience Cell Biol 161:805816.
80:9871000. Tomita S, Adesnik H, Sekiguchi M, Zhang W, Wada K, Howe JR, Nicoll
Sprengel R, Suchanek B, Amico C, Brusa R, Burnashev N, Rozov A, RA, Bredt DS (2005) Stargazin modulates AMPA receptor gating
Hvalby O, Jensen V, Paulsen O, Andersen P, Kim JJ, Thompson and trafficking by distinct domains. Nature 435:10521058.
RF, Sun W, Webster LC, Grant SG, Eilers J, Konnerth A, Li J, Torres GE, Gainetdinov RR, Caron MG (2003) Plasma membrane
McNamara JO, Seeburg PH (1998) Importance of the intracel- monoamine transporters: structure, regulation and function.
lular domain of NR2 subunits for NMDA receptor function in Nat Rev Neurosci 4:1325.
vivo. Cell 92:279289. Toth K, McBain CJ (1998) Afferent-specic innervation of two dis-
Standaert DG, Landwehrmeyer GB, Kerner JA, Penney JB Jr, Young tinct AMPA receptor subtypes on single hippocampal interneu-
AB (1996) Expression of NMDAR2D glutamate receptor sub- rons. Nat Neurosci 1:572578.
unit mRNA in neurochemically identied interneurons in the Tovar KR, Westbrook GL (1999) The incorporation of NMDA recep-
rat neostriatum, neocortex and hippocampus. Brain Res Mol tors with a distinct subunit composition at nascent hippocam-
Brain Res 42:89102. pal synapses in vitro. J Neurosci 19:41804188.
Stell BM, Brickley SG, Tang CY, Farrant M, Mody I (2003) Towers SK, Gloveli T, Traub RD, Driver J, Engel D, Fradley R, Rosahl
Neuroactive steroids reduce neuronal excitability by selectively TW, Maubach K, Buhl EH, Whittington MA (2004) Alpha5 sub-
enhancing tonic inhibition mediated by delta subunit-contain- unit-containing GABAA receptors affect the dynamic range of
ing GABAA receptors. Proc Natl Acad Sci USA 100:1443914444. kainate-induced hippocampal gamma frequency oscillations in
Stowell JN, Craig AM (1999) Axon/dendrite targeting of vitro. J Physiol 559:721728.
 Molecular Mechanisms of Synaptic Function 295

Tsien JZ, Chen DF, Gerber D, Tom C, Mercer EH, Anderson DJ, SNARE SNAP-25 distinguishes mechanisms of neuroexocytosis.
Mayford M, Kandel ER, Tonegawa S (1996a) Subregion- and cell Nat Neurosci 5:1926.
type-restricted gene knockout in mouse brain. Cell Watt AJ, Sjostrom PJ, Hausser M, Nelson SB, Turrigiano GG (2004) A
87:13171326. proportional but slower NMDA potentiation follows AMPA
Tsien JZ, Huerta PT, Tonegawa S (1996b) The essential role of hip- potentiation in LTP. Nat Neurosci 7:518524.
pocampal CA1 NMDA receptor-dependent synaptic plasticity Wei W, Zhang N, Peng Z, Houser CR, Mody I (2003) Perisynaptic
in spatial memory. Cell 87:13271338. localization of delta subunit-containing GABA(A) receptors
Turetsky D, Garringer E, Patneau DK (2005) Stargazin modulates and their activation by GABA spillover in the mouse dentate
native AMPA receptor functional properties by two distinct gyrus. J Neurosci 23:1065010661.
mechanisms. J Neurosci 25:74387448. Werner P, Voigt M, Keinanen K, Wisden W, Seeburg PH (1991)
Turrigiano GG, Nelson SB (2004) Homeostatic plasticity in the devel- Cloning of a putative high-affinity kainate receptor expressed
oping nervous system. Nat Rev Neurosci 5:97107. predominantly in hippocampal CA3 cells. Nature 351:742744.
Tzavara ET, Bymaster FP, Felder CC, Wade M, Gomeza J, Wess J, Wess J (2004) Muscarinic acetylcholine receptor knockout mice:
McKinzie DL, Nomikos GG (2003) Dysregulated hippocampal novel phenotypes and clinical implications. Annu Rev
acetylcholine neurotransmission and impaired cognition in M2, Pharmacol Toxicol 44:423450.
M4 and M2/M4 muscarinic receptor knockout mice. Mol White JH, McIllhinney RA, Wise A, Ciruela F, Chan WY, Emson PC,
Psychiatry 8:673679. Billinton A, Marshall FH (2000) The GABAB receptor interacts
Vandenberghe W, Nicoll RA, Bredt DS (2005) Stargazin is an directly with the related transcription factors CREB2 and ATFx.
AMPA receptor auxiliary subunit. Proc Natl Acad Sci USA Proc Natl Acad Sci USA 97:1396713972.
102:485490. Wisden W, Seeburg PH (1993) A complex mosaic of high-affinity
Varoqueaux F, Sigler A, Rhee JS, Brose N, Enk C, Reim K, Rosenmund kainate receptors in rat brain. J Neurosci 13:35823598.
C (2002) Total arrest of spontaneous and evoked synaptic trans- Wisden W, Laurie DJ, Monyer H, Seeburg PH (1992) The distribu-
mission but normal synaptogenesis in the absence of Munc13- tion of 13 GABAA receptor subunit mRNAs in the rat brain. I.
mediated vesicle priming. Proc Natl Acad Sci USA Telencephalon, diencephalon, mesencephalon. J Neurosci
99:90379042. 12:10401062.
Vekovischeva OY, Aitta-Aho T, Echenko O, Kankaanpaa A, Seppala T, Wisden W, Seeburg PH, Monyer H (2000) AMPA, kainate and
Honkanen A, Sprengel R, Korpi ER (2004) Reduced aggression NMDA ionotropic glutamate receptor expressionan in situ
in AMPA-type glutamate receptor GluR-A subunit-decient hybridization atlas. In: Handbook of chemical neuroanatomy:
mice. Genes Brain Behav 3:253265. glutamate (Ottersen OPaS-M J, ed), pp 99143. Amsterdam:
Verhage M, Maia AS, Plomp JJ, Brussaard AB, Heeroma JH, Vermeer Elsevier.
H, Toonen RF, Hammer RE, van den Berg TK, Missler M, Geuze Wohlfarth KM, Bianchi MT, Macdonald RL (2002) Enhanced neu-
HJ, Sdhof TC (2000) Synaptic assembly of the brain in the rosteroid potentiation of ternary GABA(A) receptors contain-
absence of neurotransmitter secretion. Science 287:864869. ing the delta subunit. J Neurosci 22:15411549.
Vignes M, Collingridge GL (1997) The synaptic activation of kainate Xiao B, Tu JC, Worley PF (2000) Homer: a link between neural activ-
receptors. Nature 388:179182. ity and glutamate receptor function. Curr Opin Neurobiol
Vilaro MT, Palacios JM, Mengod G (1990) Localization of m5 mus- 10:370374.
carinic receptor mRNA in rat brain examined by in situ Yan J, Zhang Y, Jia Z, Taverna FA, McDonald RJ, Muller RU, Roder JC
hybridization histochemistry. Neurosci Lett 114:154159. (2002) Place-cell impairment in glutamate receptor 2 mutant
Vissel B, Royle GA, Christie BR, Schiffer HH, Ghetti A, Tritto T, mice. J Neurosci 22:RC204.
Perez-Otano I, Radcliffe RA, Seamans J, Sejnowski T, Wehner Yokoi M, Kobayashi K, Manabe T, Takahashi T, Sakaguchi I, Katsuura
JM, Collins AC, OGorman S, Heinemann SF (2001) The role of G, Shigemoto R, Ohishi H, Nomura S, Nakamura K, Nakao K,
RNA editing of kainate receptors in synaptic plasticity and Katsuki M, Nakanishi S (1996) Impairment of hippocampal
seizures. Neuron 29:217227. mossy ber LTD in mice lacking mGluR2. Science 273:645647.
Vogt KE, Nicoll RA (1999) Glutamate and gamma-aminobutyric acid Zamanillo D, Sprengel R, Hvalby O, Jensen V, Burnashev N, Rozov A,
mediate a heterosynaptic depression at mossy ber synapses in Kaiser KM, Koster HJ, Borchardt T, Worley P, Lubke J, Frotscher
the hippocampus. Proc Natl Acad Sci USA 96:11181122. M, Kelly PH, Sommer B, Andersen P, Seeburg PH, Sakmann B
Walker MC, Ruiz A, Kullmann DM (2001) Monosynaptic GABAergic (1999) Importance of AMPA receptors for hippocampal synap-
signaling from dentate to CA3 with a pharmacological and tic plasticity but not for spatial learning. Science 284:18051811.
physiological prole typical of mossy ber synapses. Neuron Zhang S, Ehlers MD, Bernhardt JP, Su CT, Huganir RL (1998)
29:703715. Calmodulin mediates calcium-dependent inactivation of N-
Wallner M, Hanchar HJ, Olsen RW (2003) Ethanol enhances alpha 4 methyl-D-aspartate receptors. Neuron 21:443453.
beta 3 delta and alpha 6 beta 3 delta gamma-aminobutyric acid Zhu JJ, Esteban JA, Hayashi Y, Malinow R (2000) Postnatal synaptic
type A receptors at low concentrations known to affect humans. potentiation: delivery of GluR4-containing AMPA receptors by
Proc Natl Acad Sci USA 100:1521815223. spontaneous activity. Nat Neurosci 3:10981106.
Washbourne P, Thompson PM, Carta M, Costa ET, Mathews JR, Zoli M, Lena C, Picciotto MR, Changeux JP (1998) Identication of
Lopez-Bendito G, Molnar Z, Becher MW, Valenzuela CF, four classes of brain nicotinic receptors using beta2 mutant
Partridge LD, Wilson MC (2002) Genetic ablation of the t- mice. J Neurosci 18:44614472.
This page intentionally left blank
8 Eberhard Buhl and Miles Whittington

Local Circuits

tion. Some interneurons, for example, show properties that

8.1 Overview are traditionally associated with principal, projection neu-
rons. Certain classes of hippocampal interneurons establish a
Local circuits play a critical role in determining the pattern of commissural axon collateral or even project to extrahip-
output from discrete brain regions receiving multiple inputs pocampal targets (Toth and Freund 1992), and others have
over time. Individual neurons have been proposed to act as been shown to have dendritic spines (Gulyas et al., 1992).
integrators of multiple afferent inputs and/or coincidence Thus, some degree of caution is warranted when using the
detectors for these inputs depending on the compartmental term local-circuit neuron and the associated morphological
organization of intrinsic membrane conductance and the type parameters as a synonym for interneuron. Conversely, virtu-
of synaptic input received. However, within a brain region, it ally all excitatory hippocampal projection neurons have a
is common to nd the pattern of inputs, and subsequent out- local axonal arbor and could therefore be also considered
puts, of a subpopulation of neurons affecting the behavior of bona de local-circuit neurons.
the population as a whole. This property of discrete brain Problems also arise when attempting to establish a more
regions is governed by the organization of local circuit con- functionally oriented operational denition of interneurons.
nections between homogeneous and heterogeneous cell types. Although traditionally interneurons have been viewed as sub-
The aim of this chapter is to demonstrate the basic principles serving an inhibitory function, recent anatomical studies indi-
involved in local circuits in general and to review evidence for cate that the interneurons that selectively innervate other
the organization of local circuits in the hippocampus. interneurons may have a net disinhibitory effect on principal
cells (for review see Freund and Buzsaki, 1996). Moreover, -
8.1.1 Neuronal Classication Issues aminobutyric acid (GABA)ergic inhibitory postsynaptic
potentials (IPSPs) may be depolarizing, for example during
Hippocampal neurons are divided into two major classes: early development (Cherubini et al. 1991). GABAergic
principal cells (which form the major output pathway from interneurons may also have pacemaker-like properties that are
hippocampal subregions) and interneurons (which are generally not associated with an inhibitory function.
thought to be mainly local-circuit neurons). Interneurons are Hippocampal interneurons can be dened operationally as
considerably less abundant ( 10%) than spiny principal GABAergic nonprincipal cells (Buckmaster and Soltesz 1996).
neurons, such as granule and mossy cells in the dentate gyrus Within this catch-all denition various attempts have been
and pyramidal neurons in the CA1 and CA3 areas. As yet (and made to subclassify hippocampal interneurons based on their
in contrast to the neocortex), there is relatively little evidence anatomical, electrophysiological, or neurochemical proper-
that hippocampal principal neurons fall into many distinct ties, including the following.
subclasses. However, despite their relatively small numbers,
recent technical and conceptual progress has indicated an 1. Dendritic morphology
astonishing variety of interneuron classes (for review see 2. Laminar position
Freund and Buzsaki, 1996). Apart from highlighting the diver- 3. Neuropeptide content
sity of interneuronal subtypes, these data have also revealed an 4. Type of calcium-binding proteins expressed
equally bewildering complexity of interneuronal organiza- 5. Efferent connectivity

297
298 The Hippocampus Book

6. Intrinsic conductance and ring properties operated calcium channels) is preferentially located on distal
7. Participation in specic circuits dendritic compartments (Magee, 1998; Wei et al., 2001). The
distribution of such conductance has profound effects on
A number of these classication schemes have their undis- the integrative and coincidence-detection abilities of target
puted merits, often facilitating the discrimination of interneu- neurons, suggesting that inputs with different laminar proles
rons into nonoverlapping subclasses; however, this is not are handled differently by neurons in the target regions.
always possible owing to common properties shared by other- In addition, the high degree of laminar specicity of gluta-
wise distinct interneurons. For example, with respect to useful matergic input to hippocampal subregions is also accompa-
classication schemes, several extensive quantitative electron nied by specity in interneuron targets. The axon domains
microscopic studies have provided compelling evidence that of interneurons targeted by a specic pathway often match
hippocampal interneurons show striking differences in their remarkably to the axon domains of that specic pathway
efferent connectivity properties (e.g., Gulyas et al., 1993a). (see Figs. 82 and 85, later in the chapter). This suggests
However, combining this property with others in the above that where compartment-specific excitatory inputs are
list leads to classication difficulties: Basket and axo-axonic present a specic feedforward inhibitory local circuit is also
(chandelier) cells both contain parvalbumin (Kosaka et al., co-activated to control those compartments (see also Section
1987) but show clear-cut differences with respect to their 8.1.3). Thus, the architecture exists to provide specic target
efferent target prole. In addition, attempts to classify hip- region responses to specic inputs at both the cellular com-
pocampal interneurons based on their membrane and ring partmental level and local circuit levels.
properties also cause problems (e.g., Lacaille and Williams, With respect to target cell selectivity, there are, as yet, no
1990; Mott et al, 1997). In short, it has, as yet, proved impos- data that suggest any clear biases toward particular target neu-
sible to nd a set of physiological parameters that allows a rons for glutamatergic pathways. In contrast, the GABAergic
meaningful classication scheme of hippocampal interneu- septal projection has no obvious laminar innervation pattern
rons to be established. For the future, the current inadequacy (Freund and Antal, 1988) but does have a precise target cell
of this approach could be addressed by either nding more preference. Synapses are formed exclusively with GABAergic
suitable criteria or, ideally, adopting a multiparametric, and interneurons, among them subpopulations co-localizing
therefore integrated, approach to rene the existing classica- parvalbumin, calbindin, calretinin, somatostatin, and chole-
tion schemes, similar to efforts being made in neocortical cystokinin (CCK) (Freund and Buzsaki, 1996). Interestingly,
areas (Gupta et al., 2000). the cholinergic afferents, also forming part of the septo-
hippocampal pathway, do not appear to have any particular
8.1.2 Input Specicity of Extrinsic Afferents target cell or laminar preference. Cholinergic axons branch
profusely in all hippocampal layers and establish synaptic
All hippocampal subelds receive an abundance of extrinsic contacts with both principal cells and interneurons. This
afferents, which can be grouped into three broad classes: (1) widespread, nonspecic inuence over target areas is also
glutamatergic inputs (e.g., originating in the entorhinal cortex seen to some extent for other neuromodulatory inputs.
and other ipsilateral and contralateral hippocampal subre- Noradrenergic afferents from the locus coeruleus do not
gions); (2) the septo-hippocampal GABAergic projection; and appear to have a particular target cell preference but are par-
(3) several pathways from brain stem and forebrain nuclei, ticularly dense in areas that receive mossy ber input (i.e., in
releasing neurotransmitters that are often referred to as neu- the hilus of the dentate gyrus and stratum lucidum of the CA3
romodulators, among them acetylcholine, dopamine, sero- area). One branch of the serotoninergic raphe-hippocampal
tonin, and noradrenaline (norepinephrine). projection also shows this lack of specicity. Most transmitter-
As far as excitatory, glutamatergic afferent pathways are containing boutons in this pathway (> 75%) do not appear to
concerned, it appears that most of them show a high degree of make synaptic specializations and show no obvious associa-
laminar selectivity. Entorhinal afferents target the distal apical tion with particular target cells or a domain-specic innerva-
dendrites of granule and pyramidal cells, whereas commis- tion pattern (Oleskevich et al., 1991). Less specic data are
sural/associational bers innervate the proximal dendrites of available for dopaminergic and noradrenergic projections to
granule cells and the apical and basal dendrites of pyramidal the hippocampus. In general, although most hippocampal
neurons in the strata radiatum and oriens of the CA3 and CA1 neurons express dopamine receptors, mesohippocampal pro-
subelds. In addition, submammillary and thalamic nucleus jections are predominantly to area CA1. Terminals are found
reuniens pathways also show a high degree of laminar speci- in all elds from the stratum oriens to the molecular layer
city. This laminar specicity of inputs can have powerful (Gasbarri et al., 1997), with projections from the ventral
effects on shaping any consequent output from the target tegmental area favoring the temporal hippocampus and nigral
population. projections favoring septal CA1 elds (Gasbarri et al., 1994).
Hippocampal principal cells show a marked distribution In contrast, noradrenergic bers from the locus coeruleus
pattern of intrinsic conductance. In general, conductance that project mainly to the dentate gyrus and stratum lacunosum-
favors slower postsynaptic responses (e.g., Ih- and voltage- moleculare (Swanson & Hartman, 1975; Samson et al., 1990).
 Local Circuits 299

Dense axonal varicosities are seen in the inner molecular layer, bistratied cells, and oriens and lacunosum-moleculare cells.
granule cell layer, and hilus (Ishida et al., 2000). The strength of excitatory synaptic innervation is such that
there is a high probability that the target interneuron will re
8.1.3 Subcellular Domain Specicity an action potential (see Section 8.3). In turn, interneurons
in Hippocampal Circuits target local principal cells, thus providing a feedback
inhibitory circuit (Fig. 81B). As with feedforward inhibition,
The pattern of co-lamination of specic extrinsic pathway these local feedback circuits can account for most of the local
inputs and their preferred interneuronal targets suggests a output from principal neurons. In area CA1, excitatory neu-
general pattern of organization whereby different interneu- rons have a strong preference for interneurons as targets (e.g.,
rons control the activity of different principal cell compart- Gulyas et al., 1998). Thus, generation of an output action
ments. Evidence for this division of labor by interneuron potential from a hippocampal subregion is followed rapidly
subtypes can also be found when considering other interneu- by a period of marked local inhibition. Many interneurons
ron subtypes. Axo-axonic cells show remarkably high-target can be involved in both feedforward and feedback inhibition
specicity for axon initial segments (Somogyi et al., 1985; (e.g., Buzsaki, 1983; Frotscher et al., 1984), providing a func-
Buhl et al., 1994b; Martinez et al., 1996). Conversely, basket tional link between afferent input patterns and any resulting
cells clearly favor somata and proximal dendritic shafts of output from the target area.
their postsynaptic targets (Gulyas et al., 1993; Halasy and In addition to circuits involving both principal excitatory
Somogyi, 1993; Han et al., 1993). Thus, it appears that despite neurons and inhibitory interneurons, there are many exam-
the spatial mixing of postsynaptic target domains at least ples of interneuroninterneuron synaptic interactions (see
some GABAergic cells maintain a remarkable degree of speci- Sections 8.2.6 and 8.3.5). Most interneurons that target prin-
city for particular membrane compartments. It is interest- cipal cells, such as basket and bistratied cells, also target
ing, however, to note that domain specicity and postsynaptic other interneurons. There is evidence of a broad heterogene-
target cell selectivity are not necessarily linked. Although ity of interneuroninterneuron connections, with single
hippocampal basket cells appear to be invariably domain- presynaptic interneurons innervating other interneurons
specic, an individual basket cell may provide divergent with different postsynaptic targets (e.g., the distal dendritic or
innervation to both principal cells as well as other interneu- perisomatic compartments of pyramidal cells) (Tamas et al.,
rons (Sik et al., 1995; Cobb et al., 1997). 1998). Unlike feedforward and feedback inhibition, this
mutual inhibition (Fig. 81D) can be relatively weak. For
8.1.4 Patterns of Local Circuit Connectivity example, CA1 basket cells are 25 times more likely to form
synapses with principal cells than are other basket cells.
The previous two sections dealt predominantly with the pat- However, there are also distinct subclasses of interneurons
tern of input to local hippocampal circuits. These patterns that appear to target other interneurons exclusively (Freund
involve two elements: the input prole directly onto local and Buzsaki, 1996). Networks of such interneurons, coupled
principal neurons and the concurrent activation of interneu- with interneurons that have principal cells as targets, mas-
rons generating a feedforward inhibitory signal (Fig. 81A). sively enrich the dynamics of local circuit responses to an
Feedforward inhibition serves to impose a temporal frame- afferent input. In particular, recurrent pathways provide
work on a target area on the basis of inputs received. The extra a mechanism by which rhythmic patterns of activity can
delay associated with an additional synapse in the circuit be j2generated in interneuron circuits and projected onto
ensures that the feedforward inhibitory synaptic event does principal cells, thus generating network oscillatory activity
not impinge on the initial, direct extrinsic synaptic event in (Whittington et al., 1995) (see Chapter 11).
principal neurons. However, further extrinsic inuence of The occurrence of recurrent connections is not limited
the target principal cells is reduced for a time equal to the to interneurons. Recurrent excitation is also seen in hip-
effective period of the feedforward inhibitory synaptic poten- pocampal local circuits (Fig. 81C). In the hippocampus
tial. Some interneurons receive input only from extrinsic there are probably few, if any, neurons without local axon col-
afferents and can therefore participate only in feedforward laterals synapsing on neighboring neurons. In general, prin-
inhibition. Examples are cells associated with the perforant cipal neurons recurrently excite other local (or neighboring)
path in the molecular layer of the dentate gyrus and neurogli- principal neurons as well as interneurons. Only dentate
aform cells in area CA1 (Han et al., 1993; Vida et al., 1998). granule cells in rodents appear to be wired differently (see
Feedforward inhibition can account for most of the input Section 8.2.1).
from extrinsic sources. For example, the mossy ber axons in
area CA3 make roughly 10 times as many contacts with CA3 8.1.5 Circuit Specic Receptor Distribution
interneurons as with the far more numerous CA3 principal
cells (Acsady et al., 1998). The basic principles of connectivity for interneurons
If principal neurons re, a number of interneurons receive described above also have a correlate in the types of GABA
synaptic excitation. Such interneurons include basket cells, receptor expressed at specic target sites. Thus, receptors that
300 The Hippocampus Book

Figure 81. Basic local circuit interactions. A. Feedforward inhibi- C. Recurrent excitation. Axon collaterals from local principal cells
tion. Axon collaterals from excitatory afferent bers contact local also contact other local principal cells, providing an excitatory
interneurons. The additional synaptic delay compared to direct mechanism for concerted, temporally coordinated population out-
afferent excitation onto principal cells provides a time-dependent put. D. Mutual inhibition. Some interneuron subtypes contact other
sequence of excitation and inhibition from single afferent inputs. interneurons as well as principal cells, and some interneurons con-
B. Feedback inhibition. Axon collaterals from local principal cells tact other interneurons exclusively. This pattern of connectivity can
contact local interneurons, providing a period of inhibition of prin- serve to impart spatiotemporally coordinated patterns of excitation
cipal cell activity following the generation of an output. Interneuron and inhibition in the local circuit leading to rhythm generation.
populations involved in A and B are not always mutually exclusive. Filled circles, excitatory synapses; open circles, inhibitory synapses.

contain 5 subunits are enriched in the dendritic regions tor distribution, as 1-containing receptors are matched with
of the hippocampal subelds and therefore are associated the terminals of parvalbumin-positive basket cells, whereas
with distinct presynaptic sources, such as the class of dendrite- those with an 2 subunit are associated with synapses estab-
targeting bistratied cells (Sperk et al., 1997; Thomson et al., lished by CCK-positive basket cells (Nyiri et al., 2001).
2000). Conversely, 1 and 2 subunits containing GABAA Apart from the domain-specic distribution of GABAA
receptors predominate in the perisomatic domain of pyrami- receptor  subunits, there is also evidence of remarkably selec-
dal neurons (Nusser et al., 1996; Nyiri et al., 2001). Those tive expression across cell classes. Immunocytochemical dou-
receptor complexes with an 2 subunit are primarily found ble-labeling studies have revealed that all parvalbumin- and
on the axon initial segment and are, by default, associated with most NPY-positive interneurons express GABAA receptor 1
presynaptic terminals of axo-axonic cells, which provide the subunits, whereas calbindin-, vasoactiveinhibitory peptide
near-exclusive innervation of this neuronal compartment. (VIP)-, and CCK-containing neurons are immunonegative for
The somata of pyramidal neurons, the postsynaptic target of 1 (Gao and Fritschy, 1994). In view of the different molecular
basket cells, are decorated with both 1 and 2 subunit-con- pharmacology of GABAA receptors containing, for example,
taining GABAA receptors (Nusser et al., 1996). Interestingly, 1, 2, or 5 subunits, it is thus likely that specic inhibitory
there is an even greater renement of the input-specic recep- microcircuits are differentially modulated by those allosteric
 Local Circuits 301

modulators of GABAA receptors that show a subunit-depend- specic antisera, a complete description may disclose clearer
ent pharmacological prole. Indeed, paired recordings from rules that govern the cell type-specic and/or circuit-specic
identied interneurons in the CA1 area have shown that IPSPs distribution of receptor proteins.
mediated by bistratied cells were not enhanced by zolpidem,
a benzodiazepine type 1 agonist requiring an 1 subunit. 8.1.6 Convergence and Divergence
However, they were powerfully enhanced by diazepam, which
is consistent with evidence that these synapses are associated In common with neocortical areas, the hippocampal forma-
with 5 subunitcontaining GABAA receptors (Thomson et tion is characterized by a high degree of neuronal intercon-
al., 2000). Conversely, and as predicted, zolpidem potentiated nectivity. This is due to a multitude of converging inputs and
the IPSPs mediated by fast-spiking basket cells. diverging outputs; that is, an individual neuron is generally
Some classes of glutamate receptors are also distributed targeted by multiple afferents (convergence), whereas the
nonrandomly. Although the GluR1 subunit of -amino-3- efferent output of a single neuron contacts multiple postsy-
hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptors naptic target neurons (divergence). A hippocampal neuron
seems fairly evenly dispersed (Molnar et al., 1993), GluR2 may receive several thousand synaptic contacts that originate
subunits are lacking in interneurons (Leranth et al., 1996), from distant brain areas (e.g., the septum or entorhinal cor-
giving interneuronal AMPA receptors a signicant permeabil- tex) or, conversely, arise from neighboring excitatory and/or
ity for calcium (Burnashev et al., 1992) (see Chapters 6 , 7, and inhibitory local-circuit neurons (Buhl et al., 1994a; Freund
10). Metabotropic glutamate receptors also show a highly spe- and Buzsaki 1996). In turn, hippocampal principal and
cic distribution. For example, mGluR1a-containing recep- interneurons give rise to extensive axonal systems that
tors are enriched in somatostatin-containing O-LM cells arborize locally and/or project to distant, extrahippocampal
(Baude et al., 1993). These cells, along with other horizontal targets.
interneurons with cell bodies in the stratum oriens, appear to Using light-microscopic bouton counts and correlated
be the only class of hippocampal interneurons that exclusively electron microscopy, it is possible to determine the degree of
mediate feedback inhibition (Blasco-Ibanez and Freund, divergence of individual hippocampal neurons (Halasy et al.,
1995; Maccaferri and McBain, 1995). 1996). Such estimates indicate that most hippocampal
Just as for GABAA and glutamate receptors, there is a sim- interneurons contact well in excess of 500 postsynaptic target
ilar degree of specicity in the expression of receptors that neurons. In addition, it has been estimated that approximately
mediate responses of ascending modulatory ber tracts, 25 basket cells may innervate a single pyramidal neuron (Buhl
among them the cholinergic and serotoninergic systems. et al., 1994a). This massive divergence of inhibitory axons and
With respect to the latter, the subpopulation of GABAergic convergence of multiple inhibitory neuronal inputs onto sin-
CCK-positive basket cells express high levels of ionotropic gle pyramidal cells underpins the ability of GABAergic neu-
serotonin (5HT)-3 receptors (Morales et al., 1998). Likewise, rons to synchronize hippocampal population activity.
metabotropic 5HT-2 receptors are enriched in GABAergic
interneurons (Morilak et al., 1994). A similar picture is begin-
ning to emerge with respect to the distribution of cholinergic
receptor subtypes. For example, fast postsynaptic nicotinic 8.2 Dentate Gyrus
responses have been recorded in GABAergic interneurons but
are absent in pyramidal neurons (Jones and Yakel, 1997; Granule cells are the principal neuronal type of the dentate
Alkondon et al., 1998; Frazier et al., 1998), suggesting that gyrus. Their sole efferent projection, the mossy ber pathway,
nicotinic cholinergic responses are mediated indirectly via forms the second link of the so-called tri-synaptic loop by tar-
changes in interneuronal activity. However, more recent evi- geting CA3 pyramidal cells. Granule cells also form connec-
dence has shown that nicotinic receptors on glutamatergic tions with interneurons, mossy cells (excitatory neurons in the
terminals may serve to modulate plastic changes at excitatory hilus) and, in some epilepsy models, other granule cells
synapses (Fabian-Fine et al., 2001; Ge and Danni, 2005). (Dashtipour et al., 2002). Thus, within the dentate gyrus,
Muscarinic acetylcholine receptors may also show a high granule cells may function as elements in circuits generating
degree of specicity in their expression patterns. Postsynaptic feedforward and feedback inhibition as well as homo- and
m2 receptors, for example, are predominantly found in a sub- heterogeneous recurrent excitatory circuits.
set of GABAergic interneurons, whereas at the presynaptic
level m2 receptors are enriched in the perisomatic terminals of 8.2.1 Inputs to the Dentate Gyrus
basket and axo-axonic cells (Hajos et al., 1998).
As yet, too little is known about the precise cellular and There is, as yet, little evidence for target cell specicity of exci-
subcellular distribution of most hippocampal neurotransmit- tatory inputs to the dentate granule cells. However, there is
ter receptor systems to extract any general functional and/or abundant evidence for subcellular domain specicity. The
organizational principles. The examples listed above provide main perforant path input to the dentate gyrus from entorhi-
no more than a sketchy picture. Hopefully, with the use of nal stellate cells targets the distal apical dendrites of granule
single-cell polymerase chain reaction (PCR) and subunit- cells, whereas commissural/associational bers innervate the
302 The Hippocampus Book

Figure 82. Basic organization of interneuron output elds in the tion pathway-associated interneuron; MOPP, molecular layer
dentate gyrus. Subcellular domain specicity of dendrite-targeting perforant path-associated interneuron; HIPP, hilar perforant
interneurons co-organizes with afferent inputs. M.l., molecular path-associated interneuron. Shaded regions mark the axonal
layer; G.c.l., granule cell layer; HICAP, hilar commissural/associa- arbors for each cell type.

proximal dendrites (Blackstad 1958; Buhl and Dann, 1991; with septal projections). Most transmitter-containing bou-
Deller et al., 1996). This target domain specicity is precisely tons in this part of the raphe projection (> 75%) do not
reected in the preferred interneuron targets for these dentate appear to make synaptic specializations and show no obvious
gyrus inputs (Fig. 82). MOPP (molecular layer perforant domain-specic innervation patterns or association with par-
path-associated) cells receive excitatory inputs only from the ticular target cells (Oleskevich et al., 1991).
perforant path projection; there is no recurrent feedback
from innervated granule cells. HIPP (hilar perforant path- 8.2.2 Granule Cell Projection to Area CA3
associated) cells also receive inputs from the perforant path,
whereas HICAP (hilar commisural/association pathway- Granule cell axons aggregate in a distinct bundle of unmyeli-
associated) cells receive external inputs from the commis- nated axons above the CA3 pyramidal cell layer, termed the
sural/association pathway (Han et al., 1993). stratum lucidum because of its characteristic light-micro-
Other afferent input pathways also show some degree of scopic appearance. Here they form three distinct types of
specicity. The submammillary projection forms a dense synaptic terminal: the large en passant mossy boutons (410
plexus of axons in the border regions of the granule and m), lopodial extensions of large mossy boutons (0.5-2.0
molecular layer (Magloczky et al., 1994). In addition, the m), and smaller en passant-type terminals (0.52.0 m)
noradrenergic inputs from the locus coeruleus appear to tar- (Acsady et al., 1998). The characteristic mossy terminals
get areas receiving mossy ber inputthe hilus and stratum exclusively innervate hilar mossy cells and CA3 pyramidal
lucidum of area CA3 (Loy et al., 1980; Oleskevich et al., 1989). neurons, forming a multitude (3040) of synaptic junctions
In terms of neuronal subtypes, the serotoninergic raphehip- (release sites) with a single thorny excrescence on each inner-
pocampal projection demonstrates the highest degree of tar- vated CA3 or hilar neuron (Chicurel and Harris, 1992).
get cell specicity. One part forms large synaptic boutons that Conversely, lopodial extensions and en passant-type termi-
form synapses predominantly with GABAergic interneurons, nals selectively target hilar and CA3 area interneurons (Acsady
avoiding principal cells (Halasy et al., 1992). Within this group et al., 1998). The total number of mossy terminals along the
of interneurons there is specicity for distinct subsets. trajectory of completely labeled granule cell axons is relatively
The serotoninergic input preferentially targets calbindin-, small (10-18), and it has been suggested that a single granule
calretinin-, somatostatin-, and neuropeptide Y-containing cell may contact no more than 11 to 15 CA3 pyramidal cells
cells (i.e., those with an efferent target prole for the dendritic and a somewhat smaller number of hilar mossy cells (Acsady
domain of principal neurons) (Miettinen and Freund, 1992). et al., 1998). Thus, in contrast to other cortical areas, there
In contrast, the other part of the raphe projection conforms to appears to be only a modest degree of excitatory convergence
the notion of modulatory inputs exerting a diffuse action (as and divergence in the mossy ber output. Moreover, due to
 Local Circuits 303

the low ring probability of granule cells in vivo (Jung and 8.2.4 InterneuronGranule Cell Connections
McNaughton, 1993) the statistical likelihood of coincident
activity among afferent inputs to a single target neuron seems Dentate interneurons all release the amino acid transmitter
relatively low. In addition, single spike transmission at the GABA. However, beyond this unifying characteristic, they
mossy ber synapse is also rare, whereas transmission in constitute a heterogeneous class of neurons. Dentate interneu-
response to spike trains in granule cells is robust (Geiger and rons differ with respect to their content of neuropeptides,
Jonas, 2000; Henze et al., 2002). calcium-binding proteins, and somatodendritic architecture
as well as their axonal arborization pattern (Freund and
8.2.3 Granule CellInterneuron Connections Buzsaki, 1996). The latter feature in particular has been a
useful guide to subdividing hippocampal interneurons into
In the CA3 area granule axons do not appear to display a pref- distinct subclasses.
erence for particular subtypes of GABAergic interneurons The advent of intracellular labeling techniques has facili-
(Acsady et al., 1998). Thus, they make contact with substance tated the near-complete visualization of interneurons, reveal-
P receptor-containing interneurons, interneurons positive for ing axonal arbors with highly complex but distinct
mGluR1a (a marker for somatostatin-containing interneurons ramication patterns. As in all other hippocampal areas, sev-
with a dendritic innervation pattern, among them O-LM cells eral classes of GABAergic interneurons target distinct
in the CA areas), and parvalbumin-labeled interneurons, domains of their postsynaptic target neurons. Thus, axo-
which innervate the perisomatic region of their target cells. axonic cells, frequently referred to as chandelier cells, have an
They do, however, contact, via en passant or lopodial termi- absolute target preference for the initial segment of granule
nals, interneurons in considerably larger numbers than pyram- and mossy cells (Somogyi et al., 1985; Han et al., 1993; Halasy
idal neurons; and it has been estimated that, compared to other and Somogyi, 1993; Buhl et al., 1994b). A specic example of
cortical principal cells, granule cells innervate GABAergic tar- this target preference for granule cells is shown in Figure
gets ~10 times more frequently. The smaller mossy terminals 83C. In contrast, basket cells direct their efferent output
on interneurons typically establish no more than a single toward the somata (78%) and proximal dendrites (22%) of
synaptic release site (Acsady et al., 1998). Therefore, the sub- granule cells (Halasy and Somogyi, 1993). Whereas these two
stantially greater degree of mossy ber convergence onto types of interneuron appear to account for most of the
GABAergic neurons appears to be counterbalanced by a com- inhibitory synapses in the perisomatic region, the axonal
paratively lower strength of unitary connections. It must be arbors of several other interneuron classes occupy distinct
emphasized, however, that in general the absolute strength (in strata in the dentate molecular layer, innervating either prox-
terms of amplitude and/or conductance) and kinetics of uni- imal or distal segments of postsynaptic dendrites (Halasy and
tary granule cell-to-interneuron connections are comparable Somogyi, 1993; Han et al., 1993). Interestingly, as we have seen
to pyramidal-to-interneuron connections in other hippocam- (Fig. 82), the target zones of these GABAergic axons overlap
pal regions (Miles, 1990; Scharfman et al, 1990; Gulyas et al., with the laminar distribution of the major excitatory path-
1993a,b; Buhl et al., 1994a,b; Geiger et al., 1997; Ali and ways. Axons of the HICAP cells co-stratify with the commis-
Thomson, 1998; Ali et al., 1998). sural/associational afferents in the inner third of the dentate
As yet, relatively little is known with respect to putative dif- molecular layer, as demonstrated in Figure 84 (lower neu-
ferences in the innervation of dentate interneurons. Such dif- ron), whereas two additional classes of dendrite-targeting
ferences are apparent in the hippocampal CA1 area (Ali and neuron, HIPP and MOPP cells, innervate the outer two-thirds
Thomson, 1998; Ali et al., 1998); and based on their dendritic of the molecular layer, co-laminating with the perforant path
architecture it seems reasonable to assume that certain sub- input, as shown in Figure 84 (Han et al., 1993). It thus
types of dentate gyrus interneurons, such as those with their appears that distinct types of neuron have evolved that selec-
dendritic arbor restricted to the outer two-thirds of the tively and locally modulate the dendritic integration of a dis-
molecular layer (MOPP cells) (Han et al., 1993), are unlikely tinct set of glutamatergic afferents. Interestingly, this division
to receive a signicant degree of recurrent excitatory granule of labor is even more rened, as those interneuron classes with
cell input. In contrast, others, such as a class of interneurons overlapping terminal elds, such as HICAP and axo-axonic
that have their dendrites largely conned to the hilar region cells (Fig. 83), are further specialized to participate in feed-
(HIPP cells) (Han et al., 1993), are predisposed to mediate forward or feedback inhibitory microcircuits, respectively
recurrent input from granule cell collaterals. Indeed, their (Han et al., 1993). However, other types of dendrite-targeting
neurochemical equivalent, mGluR1a-containing neurons, neurons show no apparent laminar preference (Soriano et al.,
have been shown to receive granule cell synapses, whereas it 1993).
has been demonstrated in the CA1 area that their counter- Although only a small number of dentate interneurons,
parts, mGluR1a/somatostatin-containing O-LM cells, are such as somatostatin-positive cells, have a long-range com-
exclusively innervated by recurrent collaterals of CA1 pyram- missural projection (Bakst et al., 1986; Schwerdtfeger and
idal neurons (Blasco-Ibanez and Freund, 1995; Maccaferri Buhl, 1986; Sik et al., 1997), virtually all of those that do have
and McBain, 1995). extensive axonal arbors (Han et al., 1993; Buckmaster and
304 The Hippocampus Book

Figure 83. Pattern of axo-axonic cell axonal arborization. granule cell layer, the hilus, and the tip of area CA3c. C. Location of
A. Action potential in an axo-axonic cell (Aa) generates a short- eight synaptic junctions on the initial axonal segments of a granule
latency IPSP in a granule cell (Ab) with reversal potential around cell whose responses are shown in Ab-e. (Source: Adapted from Buhl
-80 mV (Ace). B. Axon of the axo-axonic cell innervates the et al., 1994a, with permission.)

Schwartzkroin, 1995a,b; Sik et al., 1997). Individual axons The relatively low number of interneurons and the modest
have been shown to ramify extensively, lling both dentate number of release sites that characterize unitary connections
blades and extending in excess of 3 mm along the septo- are in stark contrast to estimates of the total bouton number
temporal axis (Sik et al., 1997). Although it appears that established by individual GABAergic axons, ranging from
interneurons generally establish more than a single synaptic approximately 10,000 to 75,000 (Sik et al., 1997). Accordingly,
junction with a target neuron, the total number of release these data suggest a large degree of synaptic divergence, with
sites appears to be relatively small. In one instance, correlated individual interneurons occupying a large volume fraction of
light and electron microscopic analysis of a synaptically con- an entire hippocampus, perhaps in the range of 25%, and
nected axo-axonic-to-granule cell pair revealed a total of eight innervating up to 10,000 postsynaptic target neurons.
sites of membrane apposition with an equivalent number of Conversely, unitary inhibitory interactions appear to be rela-
synaptic release sites that mediated the unitary interaction tively weak (Scharfman et al., 1990; Buhl et al., 1994a).
(Buhl et al., 1994a), with other light microscopic estimates Electron microscopic studies of individual biocytin-labeled
being in a similar range. Because the granule cell axon initial axons consistently show a greater number and/or density of
segment received an additional 40 unlabeled synapses, it is unlabeled inhibitory terminals converging on postsynaptic
in view of the virtually exclusive innervation of this target target structures, suggesting that several neurons of a given
domain by axo-axonic cellsreasonable to assume a conver- type converge on a postsynaptic neuron. It therefore appears
gence factor in the range of six. Clearly, however, there are that at least some of the GABAergic circuitry in the dentate
too few data to extract any general rules as, for example, in gyrus is laid out to affect information processing at highly
one instance 36 VIP-positive boutons originating from a sin- specic sites, such as dendritic segments, but generally at the
gle axon collateral were counted on a single spiny substance level of neuronal populations it simultaneously affects several
P-positive dendrite in the hilus (Sik et al., 1997). thousands of target cells.
 Local Circuits 305

viding the spatial proximity of dendrites and recurrent axons

that might be expected to encourage monosynaptic granule
cellgranule cell synapses (Seress and Mrzljak, 1987; Buhl and
Dann, 1991). However, whether and to what extent primate
granule cells are interconnectedand if so, what conse-
quences it would have for their processing capabilities
remains to be established. Moreover, it is, as yet, unclear
whether there is an archetypical mammalian blueprint for the
dentate gyrus microcircuit and whether primates or rodents
conform more closely to this ancestral blueprint.

8.2.6. InterneuronInterneuron Connections

With the notable exception of axo-axonic cells, which exclu-

sively innervate excitatory principal cells, the other interneu-
ron types described above appear to have no particularly
bias for a specic type of target neuron, innervating both
principal neurons and interneurons (Sik et al., 1997).
Immunocytochemical data, however, suggest an additional
class of interneurons that are specialized to innervate other
interneurons. A subpopulation of VIP-positive nonprincipal
cells innervate substance P receptor-immunopositive hilar
interneurons as their major postsynaptic target, and an
additional subclass of VIP neurons with a projection to the
molecular layer innervate several types of GABAergic neurons
(Hajos et al., 1996). The functional role of these interneuron-
Figure 84. Domain specicity of axons in the dentate gyrus. targeting interneurons, which may include disinhibition or
Upper neuron: Reconstruction of a HIPP cell illustrating axonal the synchronization of GABAergic networks, awaits experi-
elds closely associated with the distal granule cell dendritic termi- mental investigation.
nation zone for perforant path inputs from the entorhinal cortex.
Bar  0.1 mm. Lower neuron: Reconstruction of a HICAP cell
demonstrating an axonal eld distribution different from that of

the HIPP cell. In this case the HICAP cell axonal eld was associ-
ated with the proximal dendritic termination zone associated with
8.3 Areas CA3 and CA1
commissural and associational pathways. Bar  0.1 mm. (Source:
Adapted from Han et al., 1993, with permission.) The cornu ammonis is usually delineated into areas CA3 and
CA1, with some researchers also distinguishing area CA2
(between area CA3 and CA1) (see Chapter 3). In the classic
8.2.5 Granule CellLocal Excitatory hippocampal trisynaptic circuit, activity is projected from the
Neuron Connections dentate gyrus (see Section 8.2) to CA3 and then from CA3 via
Schaffer collaterals to CA1. However, direct input to areas CA3
Apart from dentate and CA3 interneurons, granule cells also and CA1 also originate from the entorhinal cortex, nucleus
innervate hilar mossy cells, both with large mossy terminals, reuniens, and neuromodulatory regions (see below).
impinging on mossy cell thorny excrescences as well as smaller
terminals, contacting conventional spines on mossy cell den- 8.3.1 Inputs to CA3 and CA1
drites (Frotscher et al., 1991; Acsady et al., 1998). However, in
contrast to principal cells in other hippocampal and cortical With one exception, the stratum radiatum giant cells (Gulyas
areas, granule cells in rodents do not establish recurrent exci- et al., 1998), the principal neurons in the hippocampal sub-
tatory feedback to themselves, perhaps owing to the spatial elds, constitute a relatively homogeneous population of glu-
segregation of granule cell dendrites (rodent granule cells lack tamate-releasing pyramid-shaped neurons. In the CA1 and
basal dendrites) and their axons, which do not enter the CA3 areas pyramidal neurons bear numerous spines, with
molecular layer. Only certain pathological conditions, such as estimates being as high as 30,000 (Bannister and Larkman,
temporal lobe epilepsy, favor aberrant sprouting of mossy 1995; Trommald et al., 1995), and most of these spines receive
bers into the dentate molecular layer and the de novo for- a single excitatory synapse. Although single-cell-labeling stud-
mation of recurrent feedback loops. Interestingly, in adult ies indicate that all pyramidal neurons have a local axonal
rhesus monkeys and humans a sizable proportion of granule arbor, most excitatory synapses in the hippocampal subelds
cells extend basal dendrites into the hilar region, thereby pro- are of extraneous origin, arriving from a multitude of sources,
306 The Hippocampus Book

such as a different hippocampal subeld, the contralateral that in such unitary interactions the probability of spike
hippocampus, the entorhinal cortex, the submammillary transduction can be as high as 60%, with mean EPSP ampli-
body (Magloczky et al., 1994), and the nucleus reuniens thal- tudes varying between 0.2 and 4.0 mV (Miles, 1990; Gulyas et
ami (Wouterlood et al., 1990). More diffuse neuromodulatory al., 1993b). Similar values were reported for the CA1 area
inputs are seen from septal cholinergic axons (Frotscher and (Buhl et al., 1994a; Ali and Thomson, 1998; Ali et al., 1998).
Lenrath, 1985; Lenrath and Frotsher, 1987) and interneuron- Multidisciplinary studies using paired recordings and post
targeting raphe axons (Freund et al., 1990). hoc correlated light and electron microscopic analysis demon-
In contrast to the mossy ber synapse(s) on the thorny strate that pyramidal cell-to-interneuron connections in the
excrescences of CA3 pyramidal neurons, the amplitude of CA3 and CA1 regions are generally established by one presy-
unitary excitatory inputs to area CA1 is weak ( 2 mV) (Miles naptic bouton establishing no more than a single synaptic
and Wong, 1986; Sayer et al., 1990; Malinow, 1991) and well junction (Gulyas et al., 1993b; Buhl et al., 1994a). Quantal
below spike threshold. Although there are, as yet, no accurate analysis of such interactions revealed a relatively high release
estimates of the number of release sites at unitary connections probability (0.76) and, not surprisingly in view of the pre-
established by an afferent excitatory axon, a wealth of indirect sumed small number of release sites, a sizable quantal ampli-
evidence suggests that the number of contact sites is small (see tude (0.70 mV). Thus, it appears that a small number of
Chapters 3 and 6). First, the trajectory of incoming axons is synaptic junctions with a large quantal amplitude and a low
often perpendicular to that of pyramidal dendrites, not dis- failure rate, perhaps boosted by active dendritic conductance
similar to the geometrical arrangement of parallel bers and (Traub and Miles, 1995; Martina et al., 2000), is sufficient to
cerebellar Purkinje cells, thus decreasing the statistical likeli- guarantee relatively reliable spike transmission at pyramidal-
hood of synaptic encounters on an individual target cell. to-interneuron connections.
Second, so-called single-ber stimulation experiments gener- Using paired recordings in conjunction with intracellular
ally yield low-amplitude responses, which are commensurate labeling techniques, at least three classes of interneurons have
with the release of relatively few quanta of transmitter. These been conclusively identied as postsynaptic targets of local
experiments involving paired recordings of CA3-to-CA1 axon collaterals of pyramidal neurons: basket, bistratied, and
pyramidal pairs (i.e., activation of a single Schaffer collateral) O-LM cells (Gulyas et al., 1993b; Buhl et al., 1994a; Ali and
have revealed small-amplitude unitary excitatory responses, Thomson, 1998; Ali et al., 1998). At least in the CA1 area, the
with quantal analysis providing evidence that the low quantal overall probability of obtaining such a connection with paired
content was due to a small number of synaptic release sites recordings is substantially higher than that for pyramidal cell
rather than a low release probability (Sayer et al., 1990; pairs. Moreover, there are marked differences depending on
Larkman et al., 1991; Malinow, 1991; Bolshakov and the type of postsynaptic interneuron. Thus, the probability of
Siegelbaum, 1995). Synaptic CA3CA1 responses have been obtaining a pyramidal-to-basket cell connection is in the
attributed to as little as a single release site (e.g., Turner et al., range of 1:22; that of nding a pyramidal-to-bistratied cell
1997). Ultimately, however, correlated physiological and elec- pair is 1:7; and pyramidal cell-to-O-LM cell interactions are
tron microscopic approaches are needed to resolve this most readily encountered (1:3) (Ali et al., 1998). Although, as
important issue. a general rule, physiological estimates from paired recordings
are prone to some degree of experimental bias, these differ-
8.3.2 Pyramidal CellInterneuron Connections ences appear to be sufficiently clear-cut to suggest not only
that pyramidal neurons in the CA1 area have a marked target
Local connections between pyramidal neurons and interneu- preference for inhibitory interneurons, but that there is a dis-
rons form the physiological substrate for feedback inhibition tinct bias toward favoring different subclasses of interneurons.
in the hippocampal network. The efficacy of recurrent inhibi- Hippocampal interneurons receive fewer excitatory
tion can be readily demonstrated, as action potentials in CA3 synapses than principal neurons, with the numbers varying
pyramidal neurons can frequently elicit IPSPs in neighboring considerably depending on the class of cell (Gulyas et al.,
pyramidal cells. These IPSPs are generated by the local net- 1999). Parvalbumin-positive cells with a perisomatic efferent
work of inhibitory cells, which in turn have been synaptically target prole (axo-axonic and basket cells) receive, on average,
activated by the excitatory feedback of the presynaptic pyram- approximately 15,000 excitatory synapses, whereas the group
idal neuron. Accordingly, both GABAA and AMPA/kainate of dendrite-targeting calbindin-positive interneurons (e.g.,
receptor antagonists are effective in abolishing disynaptic bistratied cells) receive in the range of 2600 excitatory
IPSPs (Miles, 1990). Apart from demonstrating the presence inputs. The class of interneuron-targeting calretinin-positive
of inhibitory local feedback circuits, these data show that exci- interneurons are targeted by even fewer, in the order of 1700
tatory postsynaptic potentials (EPSPs) evoked by individual excitatory synapses (Gulyas et al., 1999). Because it is
pyramidal neurons must be sufficiently strong to trigger unknown how many of the excitatory inputs to interneurons
action potentials reliably in postsynaptic interneurons. More are from neighboring pyramidal neurons, it is not possible to
direct evidence in support of this notion has been obtained calculate convergence factors for feedforward and recurrent
with intracellular recordings of synaptically coupled pyrami- pyramidal cells inputs. However, even if the proportion of
dal cell-to-interneuron pairs (Miles, 1990). These data show recurrent inputs were as low as 10%, it (e.g., in the case of par-
 Local Circuits 307

valbumin-positive cells) suggests that approximately 1500 ties of the target cell may also affect the integration of unitary
neighboring pyramidal cells converge on a single interneuron. synaptic potentials, determining, for example, whether suc-
Accordingly, this conservative estimate suggests that the con- cessive events will show response summation, thereby leading
comitant activity of no more than 1% of the local pyramidal to an overall amplitude increase of the compound potential
cell population, with an average quantal amplitude of 0.7 mV, despite the fact that the input may show frequency-dependent
would be required to elicit 10 mV recurrent EPSPs in an depression. Because of their brief membrane time constants,
interneuron. Thus, it seems likely that only a small fraction of unitary EPSPs in hippocampal interneurons generally show
feedforward and/or recurrent excitatory inputs is required to modest, if any, response summation, even with brief inter-
be simultaneously active (see below) to drive interneurons to event intervals, whereas EPSPs evoked in pyramidal cells
their ring threshold. exhibit an incremental increase of the compound potential
It appears to be a general rule that EPSPs on hippocampal (Miles, 1990). Accordingly, pyramidal cells are often viewed as
interneurons, regardless of whether they originate locally or in integrators (i.e., their output reects the sum of inputs),
a different (hippocampal) brain area, have faster kinetic prop- whereas interneurons have been termed coincidence detec-
erties than excitatory inputs on principal neurons (Miles, tors, as their activation is strongly dependent on the temporal
1990; Lacaille, 1991; Gulyas et al., 1993b; Buhl et al., 1994a, coherence of their inputs, showing only a signicant degree of
1996; Ali and Thomson, 1998; Ali et al., 1998). There are sev- response summation with (near) synchronous inputs (Konig
eral reasons for this: First, interneurons have comparatively et al., 1996).
short membrane time constants, in part due to their expres-
sion of Kv3 potassium channel subtypes (Lien and Jonas, 8.3.3 InterneuronPyramidal Cell Connections
2003); second, active dendritic conductance may counterbal-
ance the electrotonic slowing of synaptic potentials; third, the As is the case in the dentate gyrus, a variety of intracellular
molecular composition of interneuronal AMPA receptors labeling studies have provided evidence that GABAergic
endows them with faster single-channel kinetics (Geiger et al. interneurons in the CA1 and CA3 regions are morphologically
1997); fourth, the contribution of an NMDA receptor-medi- diverse and can be classied with respect to their laminar posi-
ated component, which would slow EPSP kinetics, appears to tion, somatodendritic morphology, and most importantly
be relatively small (Buhl et al., 1996). It is conceivable that all efferent connectivity (Gulyas et al., 1993a,b; Buhl et al.,
these factors act in unison to accelerate spiking in interneu- 1994a,b, 1995; McBain et al., 1994; Sik et al., 1994, 1995; Halasy
rons and thereby minimize the delay of disynaptic IPSPs, also et al., 1996; Vida et al., 1998; Vida and Frotscher, 2000; but see
explaining the well known observation that recurrent inhibi- Parra et al., 1998). A general scheme for interneuronal con-
tion is effective in curtailing EPSPs evoked using electrical nectivity is illustrated in Figure 85 (compare with Fig. 82).
stimulation of afferent fibers. Moreover, fast EPSPs in Although there are region-specic variations, it appears that
interneurons improve the precision of spike timing by the fundamental principles in the organization of interneuron
decreasing the temporal jitter of action potentials, a factor that microcircuitry are similar in all hippocampal regions and
is of considerable importance for the network activity during indeed in all cortical areas (Somogyi et al., 1998). Thus, axo-
inhibition-based brain rhythms. axonic and basket cells in the CA1 and CA3 subelds target the
Excitatory unitary interactions in the adult hippocampus perisomatic domain, with the former almost exclusively inner-
may show both frequency-dependent depression and facilita- vating the axon initial segment of principal cells and the latter
tion of the postsynaptic response magnitude; that is, during forming synapses on somata and proximal dendrites (Gulyas
repetitive presynaptic activity, successive action potentials et al., 1993a,b; Buhl et al., 1994a,b, 1995). Likewise, several
elicit EPSPs with a gradually decremental or incremental classes of dendrite-targeting interneurons have their axonal
amplitude (Miles and Wong, 1986; Miles, 1990; Deuchars and arbor co-laminating with afferent excitatory inputs. The axons
Thomson, 1996; Ali and Thomson, 1998; Ali et al., 1998). In of bistratied cells in the CA1 area co-stratify with the Schaffer
general, unitary EPSPs in interneurons of the CA3 area show collateral/commissural input (Buhl et al., 1994b). The remark-
a greater tendency to facilitate, perhaps related to the greater able delineation of axonal elds between perisomatic targeting
propensity of CA3 neurons to re bursts of action potentials. (basket cell) and dendrite targeting (e.g., bistratified)
Moreover, evidence suggests that short-term plasticity of uni- interneurons is illustrated in Figure 86. OLM cells in both
tary excitatory connections may be target-dependent (i.e., hippocampal subelds selectively innervate the perforant path
determined by the postsynaptic neuron) (Ali and Thomson, termination zone in the stratum lacunosum-moleculare
1998; Ali et al., 1998). The nding is particularly intriguing (Gulyas et al., 1993a,b; McBain et al., 1994); and mossy ber-
because what in essence is considered a presynaptic phenom- associated interneurons have their axons restricted to the stra-
enon is determined by the type of postsynaptic neuron. Thus, tum lucidum of the CA3 region and, as the name suggests,
upon repetitive activation, the unitary responses evoked by have been proposed to modulate selectively the granule cell
CA1 pyramidal cells in CA1 basket and bistratied cells output to the CA3 region (Vida and Frotscher, 2000). As in the
invariably exhibit gradual depression of their amplitudes, dentate gyrus, more than a single type of interneuron may be
whereas those evoked in O-LM cells strongly facilitate (Ali associated with a particular pathway, such as OL-M cells and
and Thomson, 1998; Ali et al., 1998). The biophysical proper- perforant-path associated interneurons at the stratum radia-
308 The Hippocampus Book

Figure 85. Basic anatomical organization of interneuron output elds in area CA1. As with
the dentate gyrus, dendrite-targeting interneurons show a great deal of specicity when associ-
ating with termination elds for afferent inputs of differing anatomical origin. S.l.m., stratum
lacunosum moeculare; S.rad., stratum radiatum; Pc.l., pyramidal cell layer; S.o., stratum oriens;
O-LM cell, oriens-lacunosum moleculare cell.

Figure 86. Compartmental target specicity for interneurons in commissural afferents). Right-hand neuron: Axonal and dendritic
area CA1. Left-hand neuron: Axonal and dendritic arborizations of arborizations for a basket cell illustrating the high precision of
a bistratied neuron demonstrating the near absence of input to axons for the perisomatic compartments and complete absence of
perisomatic compartments but dense arborization in the stratum projections to dendritic elds. Bar  0.1 mm.
radiatum and the stratum oriens (associated with Schaffer collateral
 Local Circuits 309

tum/lacunosum-moleculare border (Vida et al., 1998). few are associated with individual postsynaptic target neu-
However, their dendritic architecture suggests that interneu- rons, providing further anatomical evidence for the divergent
rons with a similar efferent target prole may be preferentially nature of inhibition (see Section 8.1.6). Correlated light and
involved in either feedforward or recurrent inhibitory micro- electron microscopic analysis of unitary inhibitory interac-
circuits. Indeed, for O-LM cells compelling anatomical and tions revealed that hippocampal basket cells establish 2 to 12
physiological data show that they exclusively mediate recur- synaptic junctions, whereas dendrite-targeting interneurons
rent inhibition onto pyramidal cells (Blasco-Ibanez and (e.g., bistratied cells) form 5 to 17 contacts with an individ-
Freund, 1995; Maccaferri and McBain, 1995). Again, as in the ual pyramidal neuron (Buhl et al., 1994a,b; Miles et al., 1996;
dentate gyrus, there may be input selectivity for at least some Tamas et al., 1997). Although there are relatively few published
of those interneurons involved in feedforward inhibitory cir- examples with electron microscopic counts of unitary release
cuits. For example, in view of their dendritic architecture, bis- site numbers, several light microscopic estimates obtained
tratied cells presumably receive commissural/associational from O-LM, bistratied, Schaffer collateral-associated, and
afferents and are less likely to process a signicant proportion axo-axonic cells conrm the above data (Gulyas et al.,
of direct entorhinal input (Halasy et al., 1996). 1993a,b; Vida et al., 1998; Maccaferri et al., 2000). Thus, it
Interneurons in the CA3 and CA1 subelds have axonal seems reasonable to assume that the efferent output of most
arbors of varying size. Neurogliaform and O-LM cells have the hippocampal interneurons distributes onto approximately
most compact axonal arbors (Gulyas et al., 1993a,b; Sik et al., 1000 to 2000 target neurons.
1995; Ali and Thomson, 1998; Vida et al., 1998), generally Connectivity estimates of hippocampal neurons have been
spanning less than 1 mm, whereas axo-axonic, basket, bistrat- obtained with paired recordings. On average, CA1 basket cells
ied, and trilaminar cells have considerably larger axonal contact approximately 22% of pyramidal neurons within the
trees, with in vivo estimates for basket cell axon arbors being connes of their axonal arbor, with the connection probabil-
in the range of 1 mm, approximately 2 mm for a bistratied ity dropping from 54% for immediate neighbors to 5% for
cell, and about 2.5 mm for a trilaminar interneuron (Sik et al., longer-range connections (Ali et al., 1999). Although there is
1995). The available evidence suggests that axonal arbors tend an inevitable experimental bias with this approach, earlier
to be relatively symmetrical in all hippocampal planes. quantitative anatomical data showed a surprising degree of
Moreover, apart from having extensive local axonal arbors, agreement with estimates obtained from paired recording.
several types of hippocampal interneuron have extra-areal Calculations based on pyramidal cell and basket cell terminal
projections. Axons of perforant path-associated interneurons densities, taking an average number of synaptic release sites
in the CA1 area cross the hippocampal ssure and spread into into account, suggest that CA1 basket cells contact approxi-
the outer molecular layer of the dentate gyrus (Vida et al., mately 28% of nearby target cells, with the connection proba-
1998). Back-projecting cells in the CA1 region appear to have bility plummeting to 4% toward the peripheral extent of the
even larger axonal arbors, extending throughout all hip- axonal arbor (Halasy et al., 1996).
pocampal subfields (Sik et al., 1994); and a subset of Although it is probable that thousands of pyramidal cells
GABAergic calbindin-positive interneurons with an as yet converge on each inhibitory interneuron, for a number of rea-
unknown morphology project to the medial septum (Toth sons the convergence factors in the reciprocal direction are
and Freund, 1992). substantially smaller. First, as pyramidal cells are innervated
Electron microscopic data show that individual terminal by many classes of GABAergic neurons, each subpopulation
boutons of most hippocampal interneurons usually establish supplies no more than a fraction of the total GABAergic
a single synaptic junction per bouton, although occasionally input. Accordingly, cell class-specic convergence factors are
a double junction is seen (Buhl et al., 1994a,b, 1995; Halasy inversely related to the number of interneuron types. Second,
et al., 1996; Vida et al., 1998). Thus, light-microscopic bouton pyramidal cells receive fewer inhibitory than excitatory
counts are generally thought to be a reasonable approxima- inputs, although the exact ratio is unknown. Although there
tion of the total number of synaptic contact sites established are currently no data on the number of inhibitory inputs onto
by a single inhibitory axon. Such estimates for a variety of dendrites, electron microscopic estimates suggest that pyram-
interneurons labeled in vitro, among them basket, axo-axonic, idal cell somata receive in the range of 120 synapses and axon
neurogliaformand bistratied cells as well as Schaffer collat- initial segments in the range of 100 to 200 (Kosaka 1980; Buhl
eral and perforant path-associated interneurons, are consis- et al., 1994). Third, unitary inhibitory interactions are gener-
tently above 5000 and can be as high as ~13,000 (Gulyas et al., ally mediated by several synaptic release sites, as illustrated in
1993a,b) (Halasy et al., 1996, Miles et al., 1996; Vida et al., Figure 87. Accordingly, current estimates suggest that
1998). Not surprisingly, in vivo estimates are substantially approximately 25 basket cells and even fewer axo-axonic cells
higher, approximately twofold, with basket cell axons having (210), converge on a single pyramidal neuron (Li et al., 1992;
bouton counts up to 12,000 (Sik et al., 1995). Likewise, bou- Buhl et al., 1994). It therefore follows that, in contrast to
ton counts for individual bistratied, trilaminar, and O-LM- presynaptic pyramidal cell input, a disproportionately large
cells were in the range of 15,000 (Sik et al., 1995). However, fraction of inhibitory neurons (at least 10%) must be active
despite the large total number of boutons per axon, relatively during neuronal population activity. Rhythmic network oscil-
310 The Hippocampus Book

Figure 87. Basket cell morphology and effects on principal cells in depolarizing membrane potential episode (Bc) that was sufficient
area CA1. A. Single action potentials in basket cells (Aa,d) elicit to generate action potentials (Be,f). C. Axonal and dendritic elds
short latency IPSPs in principal cells. Average of > 100 single IPSPs of a single basket cell overlaid with a single target pyramidal
is shown in Ab, but individual IPSPs showed a great deal of ampli- neuron. Axon collaterals were limited to the perisomatic region.
tude variability (Ac). Hyperpolarization of pyramidal cells resulted D. Expanded view of the pyramidal neuron somatic region showing
in depolarizing IPSPs with a reversal potential, indicating a chlo- 10 inhibitory synaptic connections. (Source: Adapted from Buhl
ride-mediated event (Ae). B. Some IPSPs were followed by a distinct et al., 1994a, with permission.)

lations, for example, require sustained GABAergic input from data illustrating the behavior of identied interneurons in the
a specic source, such as fast perisomatic inhibition originat- intact, behaving animal, showing, for example, phase-locked
ing from basket or axo-axonic cells. Indeed, simultaneous discharge of basket cells and O-LM cells during theta EEG
multielectrode recordings of pyramidal cells and interneurons activity (Buzsaki et al., 1983; Skaggs et al., 1996; Klausberger
during rhythmic electroencephalographic (EEG) activity in et al., 2003). A number of studies have attempted to address
the intact, behaving animal show that at least some interneu- the functional role of interneurons at the cellular level. Paired
rons re considerably and consistently faster than neighbor- interneuron-to-pyramidal cell recordings have revealed that
ing pyramidal neurons, whereas others (e.g., the so-called all unitary inhibitory interactions have relatively fast kinetics
anti-theta cells) may actually decrease their ring rate (Freund and are (as yet without exception) mediated by GABAA recep-
and Buzsaki, 1996). tors (Miles and Wong, 1983; Lacaille et al., 1987; Buhl et al.,
Currently, our appreciation of the amazing morphological, 1994a; 1995; Ouardouz and Lacaille, 1997; Vida et al., 1998;
neurochemical, and physiological complexity of GABAergic Thomson et al., 2000). In addition, the activation of GABAB
interneurons at the reductionist level contrasts with our rela- receptors may either require high-frequency tetanic stimula-
tive lack of understanding of their functional role in the hip- tion of a unitary inhibitory connection or, alternatively, the
pocampal network as a whole. There is a modest amount of concomitant activation of multiple GABAergic inputs (Mody
 Local Circuits 311

et al., 1994; Thomson and Destexhe, 1999). Although this may Moreover, as pyramidal neurons have a propensity to re dur-
suggest overall uniformity in function, it is also clear that uni- ing the most depolarized part of the oscillatory waveform
tary hippocampal interactions differ with respect to the (Buhl et al., 1995), rhythmic unitary IPSPs are therefore also
molecular composition and associated pharmacological strikingly effective in pacing the discharge of postsynaptic
properties of differentially distributed and domain-specic pyramidal neuronswithout, however, necessarily affecting
GABAA receptor subunits (Thomson et al., 2000). IPSPs the neuronal ring rate (Cobb et al., 1997). These authors
evoked by fast-spiking basket cells are more potentiated by demonstrated that the output from an individual interneuron
the 1 subunit-selective benzodiazepine zolpidem than are not only may pace spontaneous action potential generation in
IPSPs elicited by regular-spiking basket cells. It has been pro- multiple pyramidal cells but also synchronize these paced out-
posed that fast-spiking basket cells are parvalbumin-positive puts, as shown in Figure 88. In view of the widespread diver-
and associated with postsynaptic receptors carrying an 1 gence of basket and axo-axonic cell axons, it is reasonable to
subunit. Conversely, regular spiking basket cells may be the assume that the temporally coherent discharge of relatively
electrophysiological signature of CCK-positive basket cells few interneurons with a perisomatic termination pattern is
and target perisomatic synapses with a predominance of sufficient to synchronize hippocampal population activity.
GABAA receptors containing 2 subunits (Thomson et al., Although they do not prove a causal relation, in vivo studies
2000). In addition, bistratied cells may be associated with
dendritically located 5 subunits, and indeed bistratied cell-
mediated unitary IPSPs are diazepam-sensitive but not zolpi-
Figure 88. Single IPSPs can synchronize ongoing pyramidal cell
dem-sensitive (Thomson et al., 2000). Recent data from outputs. A. Overlaid responses from two concurrently recorded
transgenic animals indicate that the molecular heterogeneity principal cells (PC1, PC2) receiving a single IPSP at the time illus-
of GABAA receptors is indeed reected in different functional trated by the arrow. Rebound from the IPSP generates action poten-
roles, such as anxiety (Low et al., 2000). Moreover, work in tials in a time-dependent manner such that both pyramidal cells
intact animals suggests that neuroactive steroids may act as discharge within a small time window. B. Entrainment of theta-fre-
endogenous modulators to regulate the function of GABAA quency pyramidal cell output by successive IPSPs from a single
receptors with different subunit compositions differentially interneuron (Bi). Two pyramidal cells (PC1, PC2) were concur-
(Smith et al., 1998). rently recorded and action potentials elicited in a single interneuron
Apart from differences in their molecular pharmacology, at times indicated by the arrows.. Cross correlations between the
the function of GABAA receptors may differ with their subcel- two pyramidal cells are shown before interneuron ring (Bii) and
during (right-hand side) single, repetitive interneuron ring (Biii).
lular location. Because of the shunt and/or membrane polar-
(Source: From Cobb et al., 1995, with permission.)
ization resulting from the opening of GABAA receptor
channels, other domain-specic conductance may also be
affected. Thus, unitary interactions mediated by dendrite-
targeting interneurons, such as bistratied and O-LM cells,
have slower kinetics and smaller amplitudes/conductance lev-
els when measured at the level of the cell body. Some of this
effect has been attributed to dendritic ltering (Buhl et al.,
1994a,b; Miles and Freund, 1996; Maccaferri et al., 2000).
Although there are, as yet, no compelling hippocampal data to
demonstrate differences in the local versus somatic integra-
tion of compartmentalized excitatory and inhibitory inputs,
there is at least evidence suggesting that domain-specic
GABAergic inputs interact with local conductance. In the hip-
pocampal CA3 area, microstimulation experiments have
revealed that inhibitory inputs onto the dendrites of pyrami-
dal neurons are effective in suppressing calcium-dependent
spikes, and perisomatic inhibitory cells inhibit the repetitive
discharge of sodium-dependent action potentials (Miles et al.,
1996). However, the local interaction of unitary IPSPs and
voltage-gated conductance can be even more complex. Basket
and axo-axonic cells (i.e., interneurons with a perisomatic
innervation pattern) have been shown to interact with intrin-
sic subthreshold membrane oscillations (Cobb et al., 1996). By
resetting their phase, repetitive IPSPs originating from a sin-
gle presynaptic interneuron are effective in phase-locking sub-
threshold oscillations, with the entrainment of oscillatory
activity being most effective in the theta frequency band.
312 The Hippocampus Book

are clearly consistent with this notion, as they demonstrate a boring pyramidal cells and interneurons (Buhl et al., 1994a,b;
close temporal correlation of interneuron and principal cell Deuchars and Thomson, 1996; Ali and Thomson, 1998; Ali et
activity during theta and gamma EEG activity (Bragin et al., al., 1998). The degree of pyramidal cell interconnectivity is
1995; Ylinen et al., 1995). approximately 1% and thus markedly lower than in the CA3
area (Deuchars and Thomson, 1996), which may also explain
8.3.4 PyramidPyramid Local Connections why the isolated CA1 area is less prone to develop seizure-like
activity in the presence of GABAA receptor antagonists.
In contrast to the dentate gyrus, principal neurons in the CA1 Unitary responses tend to be relatively small (0.171.5 mV)
and CA3 subelds are interconnected with recurrent excita- and show paired-pulse depression, suggesting that a small
tory axon collaterals, thus increasing the computational com- number of synaptic release sites with a high release probabil-
plexity of local information processing. Pyramidal cells in the ity underlies the modest unitary responses. Indeed, correlated
CA3 area are not only the source of major efferent ber tracts, light and electron microscopic analysis of a single pyramidal-
such as the Schaffer collateral pathway, they also give rise to to-pyramidal cell pair showed that a unitary EPSP with a
several local axon collaterals, which emanate from the main mean amplitude of 1.5 mV was mediated by a total of two
axon in the stratum oriens. Axonal side branches branch fur- synaptic release sites (Deuchars and Thomson, 1986). Both
ther, with most of them remaining in the stratum oriens and synaptic contacts were established on two third-order basal
the remainder penetrating the pyramidal cell layer and dendrites, suggesting that despite some degree of dendritic l-
arborizing in the stratum radiatum (Gulyas et al., 1993a,b; Sik tering even relatively distal excitatory quantal inputs can
et al., 1993; Li et al., 1994). The terminal arbors in the CA3 evoke measurable somatic responses.
area are generally large, frequently occupying several millime-
ters in the transverse and septo-temporal axis, indicating a 8.3.5 InterneuronInterneuron Connections
large degree of divergence. Bouton counts of individual axons
suggest that the number of hippocampal synaptic terminals is Reciprocally connected GABAergic interneurons comprise
in the range of 10,000 to 60,000, with a sizable proportion one of the key substrates for the generation of gamma fre-
remaining in the CA3 area. Quantitative studies suggest that quency network oscillations. This nding has underscored the
CA3 pyramidal cells contact at least some of their postsynap- pivotal role of GABAergic neurons for the generation of EEG
tic targets randomly (i.e., in proportion to their occurrence in rhythms (Whittington et al., 1995). In principle, GABAergic
the neuropil) (Sik et al., 1993). Although there are no anatom- inputs on hippocampal GABAergic neurons can be subdi-
ical studies to provide reliable estimates with respect to vided into three major categories.
pyramidal-to-pyramidal cell convergence, divergence, and
Extrinsic sources, particularly the septo-hippocampal
number of synaptic release sites, physiological data provide
GABAergic projection
evidence that CA3 pyramidal cells may be synaptically inter-
Output of interneurons, such as basket cells, which
connected (Miles and Wong, 1986; Miles et al., 1996), with the
innervate principal cells and other interneurons in
connection probability of neighboring neurons being in the
approximately random proportions
range of 2% (Miles and Wong, 1986). In general, unitary
Several subpopulations of interneurons that appear to
pyramidal-to-pyramidal cell EPSPs have small amplitudes,
have special, exclusive innervation of other interneurons
ranging between 1 and 2 mV, which uctuate in amplitude
and show apparent postsynaptic response failures, suggesting Regarding the septo-hippocampal GABAergic input, com-
that they may be mediated by a small number of release sites bined immunocytochemical and tracing studies have shown
(Miles and Wong, 1986). Thus, only a small proportion of that, in contrast to the cholinergic input, this projection is
unitary EPSPs (4%) are of sufficient strength to trigger action highly selective in terms of targeting GABAergic interneurons,
potentials in a postsynaptic neuron. Interestingly, however, albeit with no apparent preference for particular subtypes (see
when presynaptic pyramidal cells re a salvo of several action Section 8.1.2) (Freund and Antal, 1988; Miettinen and
potentials so as to mimic their physiological burst ring pat- Freund, 1992). The earlier in vivo observation that stimula-
tern, postsynaptic responses show a substantial degree of aug- tion of the medial septum facilitates granule cell activity
mentation, and the incidence of burst-to-spike transduction (Bilkey and Goddard, 1985) can be most parsimoniously
can be as high as 35% (Miles and Wong, 1986). Indeed, exper- explained by the inhibition of inhibitory interneurons (i.e., a
iments in disinhibited slices show that a single CA3 pyramidal disinhibitory mechanism). More direct evidence in support of
neuron can trigger synchronous population discharges (Miles this notion was obtained with a series of elegant in vitro
and Wong, 1983). Thus, despite the connection probability experiments using a combined septo-hippocampal slice
being relatively low, there is a sufficient degree of divergence preparation (Toth et al., 1997), demonstrating a disinhibitory
and convergence to allow rapid, reliable synchronization of effect of septal stimulation on CA3 pyramidal cell activity.
network activity. Thus, theta frequency activity in the medial septum diagonal
As in area CA3, pyramidal neurons in the CA1 area emit band complex is likely to entrain hippocampal population
local axon collaterals. The latter, however, establish smaller activity by rhythmically releasing pyramidal neurons from a
local axonal arbors and remain conned to the strata oriens background of tonic inhibitory activity, thereby allowing
and alveus, where they establish synaptic contacts with neigh- them to re while inhibition is maximally suppressed.
 Local Circuits 313

In contrast to the septal GABAergic projection, many statin/mGluR-positive class of O-LM cells (Acsady et al.,
interneurons in areas CA3 and CA1 target specic subcellular 1996a,b). It thus appears that distinct subsets of GABAergic
domains, but they appear to have no apparent specicity for interneurons selectively modulate the activity of several
innervating either principal cells or interneurons, contacting other types of interneuron, such as O-LM cells, which in turn
them in what has been referred to as a quasi-random fash- have an efferent projection to the termination zone of
ion (i.e., based on the statistical distribution of postsynaptic entorhinal afferents in the stratum lacunosum-moleculare.
elements in the neuropil) (Freund and Buzsaki, 1996). In the Although a number of suggestions have been made as to their
case of parvalbumin-positive basket cells, it has been esti- functional relevance there are, as yet, no experimental or
mated that they contact not only approximately 1500 pyram- modeling data to attribute specific functional roles to
idal cells (see above) but also a proportionally smaller number interneuron-selective interneurons during hippocampal
(in the range of 60) parvalbumin-positive neurons (Sik et al., information processing.
1995). These anatomical data are corroborated by paired
recordings of synaptically coupled and anatomically identied 8.3.6 Gap Junction Connections
interneurons, revealing their divergent output to both excita-
tory and inhibitory neurons (Cobb et al., 1997; Ali et al., As already mentioned, during oscillatory network activity at
1999). In addition, these studies not only demonstrate directly gamma frequency, fast perisomatic IPSPs have a major role in
that basket cells are interconnected, they established the exis- pacing and synchronizing population activity. In view of their
tence of synaptic coupling across interneuronal classes, such kinetics, however, experimental and computational data sug-
as between basket cells and trilaminar interneurons. Thus, the gest that the upper limit of fast inhibition-based rhythms may
efferent output of basket cells not only directly (i.e., monosy- not exceed ~80 Hz. Yet, in awake animals hippocampal sharp
naptically) affect inhibitory and excitatory hippocampal neu- waves are accompanied by a high-frequency (~200 Hz) EEG
rons, the output is likely to occur almost synchronously in all oscillation with a high degree of spatiotemporal coherence
target neurons, an assumption with important implications (Buzsaki et al., 1992; Ylinen et al., 1995). As this ultrafast
for their putative role in hippocampal rhythmogenesis. rhythm, frequently referred to as ripples, is unlikely to be
Experimental ndings and computational simulations pro- generated by phasic inhibition, it seems likely that a funda-
vide compelling evidence for a synchronizing role of periso- mentally different synchronizing mechanism is at work.
matic fast IPSPs during gamma frequency network Indeed, hippocampal slices maintain a fast ripple-like oscilla-
oscillations (Whittington et al., 1995; Wang and Buzsaki, tory activity pattern that remains after blockade of synaptic
1996; Traub et al., 1997; Fisahn et al., 1998). It follows that the transmission but is abolished by a variety of gap junction
discharge of interneurons such as basket cells, with a widely uncoupling agents (Draguhn et al., 1998). Interestingly, gap
divergent output, is phase-locked with the population junction blockers are also effective in abolishing pharmaco-
rhythm, as they are the source of the phasic IPSPs (Andersen logically induced gamma oscillations (Traub et al., 2000). It
and Eccles, 1962). Indeed, in the awake rat the ring of puta- therefore appears that gap junction-mediated electrical cou-
tive interneurons is phase-locked to the underlying gamma pling may have a strong role in the synchronization of gamma
rhythm (Bragin et al., 1995), and the discharge of identied frequency oscillations as well as the generation of ultrafast
basket cells was found to be temporally correlated with both neuronal population activity.
gamma and theta activity (Ylinen et al., 1995). Despite the Gap junctions are the morphological correlate of electrical
importance of GABAergic mechanisms for the generation of synapses, and anatomical studies have clearly demonstrated
synchronous network behavior, it should be stressed that their presence in the adult mammalian hippocampus (Kosaka,
other synaptic as well as nonsynaptic factors, such as fast exci- 1983; Fukuda and Kosaka, 2000). Interestingly, however, these
tation and electrical coupling (see below), appear to have dif- data have so far provided evidence for gap junctions only
ferent but equally important roles in hippocampal between the dendrites of GABAergic interneurons. There is
rhythmogenesis (Traub et al., 1996, 2000; Fisahn et al., 1998; compelling evidence for electrically coupled interneurons in
for a comprehensive review see Traub et al., 1999). the neocortex; and, intriguingly, these data also point to spe-
Unlike basket cells, a second major group of GABAergic cic coupling between members of the same neuronal class
interneurons, comprising at least three neurochemically dis- (Galarreta and Hestrin, 1999; Gibson et al., 1999; Tamas et al.,
tinct subclasses, exclusively innervate other GABAergic 2000). This may also apply for the hippocampus, where gap
interneurons (for detailed review, see Freund and Buzsaki, junctions have been identied between the dendrites of par-
1996). These cells demonstrate that, in addition to subcellular valbumin- and calretinin-positive interneurons (Gulyas et al.,
domain specicity, interneurons may exhibit a high degree 1996; Fukuda and Kosaka, 2000), although these ndings do
of target cell selectivity. The three subtypes are (1) a class of not rule out the occurrence of heterologous connections.
calretinin-positive neuron that innervates calbindin-positive Modeling studies have suggested that parameter hetero-
dendrite-targeting cells, CCK/VIP-positive basket cells, and geneity in interneuronal circuits (e.g., differences in ring
other calretinin-positive cells (Gulyas et al., 1996); (2) an rates and/or kinetics of inhibitory conductance) destabilizes
interneuron-selective GABAergic cell that contains VIP and synchronous population activity (Wang and Buzsaki, 1996;
contacts calbindin-positive cells; and (3) a group of VIP- White et al. 1998). However, when interneurons in such net-
positive interneurons that show a preference for the somato- works are interconnected with both chemical and electrical
314 The Hippocampus Book

synapses (i.e., GABAergic synapses and gap junctions), their

conjoint action tends to improve population coherence
(White et al., 1998). Indeed, data from connexin36-decient
mice (a recently cloned neuronal gap junction-forming pro-
tein) (Condorelli et al., 1998) not only showed that
GABAergic interneurons are electrically uncoupled, they
demonstrated that the presumed lack of gap junctions
between interneurons results in impaired gamma frequency
oscillatory network activity (Hormuzdi et al., 2001). With
respect to the underlying mechanism, recordings of neocorti-
cal basket cell pairs showed an apparent synergistic effect of
concurrent synaptic and nonsynaptic communication
between interneurons (Fig. 89). With both mechanisms
in place, action potential generation in one interneuron
produced a biphasic membrane potential change in the cou-
pled neuron: initial transient depolarization followed by
IPSP-mediated hyperpolarization. Repeated action potential
generation provided a powerful subthreshold membrane
potential oscillation that inuenced spike ring (Tamas et al.,
2000).
Although this mechanism may account for establishing
synchrony in the interneuronal network at gamma frequen-
cies, it is unlikely to explain the generation of ultrafast ripple
activity. Modeling studies suggest that when pyramidal cells
are coupled by gap junctions in soma or dendrites the soma-
todendritic compartment, in conjunction with the gap junc-
tion, acts as a low-pass lter. The effect is to slow the kinetics
of spikelets to such an extent that it is difficult, based on our
current understanding of the biophysical properties of
pyramidal neurons, to explain the emergence of ultrafast syn- Figure 89. Synchrony via both gap junctional and IPSP-mediated
chronous network activity (Draguhn et al., 1998; Traub and interactions. A. Trains of presynaptic action potentials in one inter-
Bibbig, 2000). Moreover, despite numerous electron micro- neuron (top trace) generate a decrementing IPSP and spikelets in a
scopic studies, there is, as yet, no ultrastructural evidence of connected interneuron (lower trace). B. Presynaptic action potential
gap junctions between the somatodendritic domain of hip- generation reduced the postsynaptic ring frequency and synchro-
pocampal pyramidal neurons. However, putative axo-axonal nized these action potentials with those from the presynaptic cell
(bottom trace). C. Temporal correlation of postsynaptic ring prob-
gap junctions between pyramidal cells can allow spikes to pro-
ability in the postsynaptic cell during one action potential cycle in
duce either a spikelet in the postjunctional axon or, if above the presynaptic cell. Control condition is with no drive to either cell
threshold, an action potential that may antidromically invade pair. Entrainment condition is as illustrated in B. (Source: Adapted
the parent cell body and/or rapidly propagate (provided there from Tamas et al., 2000, with permission).
is a sufficient degree of interconnectivity) through the axonal
plexus, thereby prompting the generation of coherent, ultra-
fast oscillatory activity as an emergent property of the axonal electron microscopy of the sites of interactionis needed to
pyramidal network (Traub et al., 1999). These simulations resolve this intriguing issue.
indicate that as few as 2.5 to 3.0 gap junctions per axon may
be sufficient, perhaps explaining why possible axo-axonal gap
junctions have also gone unnoticed by anatomists. Indeed,
intracellular labeling studies using time-lapse confocal 8.4 Summary
microscopy have shown dye coupling between axons of hip-
pocampal pyramidal neurons, thus providing experimental Local circuits in the hippocampus are comprised of excitatory
evidence for this notion (Schmitz et al., 2001). Moreover, elec- connections from principal cells to interneurons and mainly
trophysiological studies have revealed the occurrence of inhibitory interneuronal connections onto principal cells. A
spikelets in principal neurons following, for example, the smaller number of connections are made up of mutual
antidromic activation of CA1 pyramidal cell axons (Kandel inhibitory interneuronal interconnectivity and recurrent exci-
and Spencer, 1961; Taylor and Dudek, 1982; Schmitz et al., tatory connectivity between principal cells. The pattern of
2001). Ultimately, however, a more direct approachsuch as local connectivity is diverse and represents a rich tapestry of
paired recordings in conjunction with post hoc correlated prospective strategies for modifying initial responses to inputs
 Local Circuits 315

and post hoc responses to continued input activity on the REFERENCES

basis of previously generated outputs.
The diverse prole of anatomical, pharmacological, and Acsady L, Arabadzisz D, Freund TF (1996a) Correlated morphologi-
electrophysiological properties of local circuit connections cal and neurochemical features identify different subsets of
can, to some extent, be reduced to a few basic properties. First, vasoactive intestinal polypeptide-immunoreactive interneurons
afferent activity invariably co-activates not only local princi- in rat hippocampus. Neuroscience 73:299315.
Acsady L, Gorcs TJ, Freund TF (1996b) Different populations of
pal cells but also local inhibitory interneurons. This organiz-
vasoactive intestinal polypeptide-immunoreactive interneurons
ing principle offers a mechanism by which any initial input
are specialized to control pyramidal cells or interneurons in the
can organize the response to any subsequent input, with the hippocampus. Neuroscience 73:317334.
specic temporal prole being a consequence of the kinetics Acsady L, Kamondi A, Sik A, Freund T, Buzsaki G (1998) GABAergic
of interneuronal synaptic postsynaptic responses and their cells are the major postsynaptic targets of mossy bers in the rat
compartmental localization onto principal cells. Second, hippocampus. J Neurosci 18:33863403.
efferent activity also co-activates local inhibitory interneu- Ali AB, Thomson AM (1998) Facilitating pyramid to horizontal
rons. There is some (but by no means complete) overlap oriens-alveus interneurone inputs: dual intracellular recordings
between interneurons receiving feedforward and feedback sig- in slices of rat hippocampus. J Physiol (Lond) 507:185199.
nals. This organizing principle allows local circuits to modify Ali AB, Deuchars J, Pawelzik H, Thomson AM (1998) CA1
responses to afferent activity further on the basis of whether pyramidal to basket and bistratied cell EPSPs: dual intracellu-
lar recordings in rat hippocampal slices. J Physiol (Lond)
an output is generated. Third, interconnectivity within either
507:201217.
principal cell populations or local interneuron populations
Ali AB, Bannister AP, Thomson AM (1999) IPSPs elicited in CA1
provides a mechanism by which spatially discrete patterns of pyramidal cells by putative basket cells in slices of adult rat hip-
input (and subsequent output) may inuence populations of pocampus. Eur J Neurosci 11:17411753.
neurons outside the active area. Alkondon M, Pereira EFR, Albuquerque EX (1998) alpha-
Taken together these three principles provide a form of Bungarotoxin- and methyl lycaconitine-sensitive nicotinic
architecture that seems unsuited to the classic notion of receptors mediate fast synaptic transmission in interneurons of
point-to-point neuronal communication. Instead, time rat hippocampal slices. Brain Res 810:257263.
appears to play a more critical role in governing neuronal out- Andersen P, Eccles JC (1962) Inhibitory phasing of neuronal dis-
put than space: Activation of feedforward circuits favors charge. Nature 195:645647.
responses to inputs after a period of quiescence (governed by Bakst I, Avendano C, Morrison JH, Amaral DG (1986) An experi-
mental analysis of the origins of somatostatin-like immuno-
the duration of inuence of feedforward inhibitory postsy-
reactivity in the dentate gyrus of the rat. J Neurosci 6:
naptic potentials). Activation of feedback circuits provides a
14521462.
further temporal constraint on subsequent outputs, effectively Bannister NJ, Larkman AU (1995) Dendritic morphology of CA1
providing a short-term history of a local circuits output. pyramidal neurones from the rat hippocampus. II. Spine distri-
Convergence, divergence, and recurrent connectivity effec- butions. J Comp Neurol 360:161171.
tively spread these temporal inuences spatially within the Baude A, Nusser Z, Roberts JDB, Mulvihill E, McIlhinney RAJ,
target neuronal population. If one considers the property of Somogyi P (1993) The metabotropic glutamate receptor
bias in principal neuronal responses to input (i.e., differ- (mGluR1a) is concentrated at perisynaptic membrane of neu-
ences in afferent excitatory synaptic weights and/or tonic ronal subpopulations as detected by immunogold reaction.
membrane potential heterogeneity), local circuit architecture Neuron 11:771787.
can be seen to serve as an ideal mechanism by which to read Bilkey DK, Goddard GV (1985) Medial septal facilitation of hip-
pocampal granule cell activity is mediated by inhibition of
out this bias as a specic pattern of outputs.
inhibitory interneurones. Brain Res 361:99106.
Local circuits of the type described in this chapter there-
Blackstad TW (1958) On the termination of some afferents to the
fore provide a substrate for converting blanket (relatively hippocampus and fascia dentata: an experimental study in the
unstructured) inputs into specic spatiotemporal proles of rat. Acta Anat (Basel) 35:202214.
output on the basis of the history of each principal neuron Blasco-Ibanez JM, Freund TF (1995) Synaptic input of horizontal
in the target population. Thus, a deeper insight into the role interneurons in stratum oriens of the hippocampal CA1 sub-
of the hippocampus as a region involved in aspects of learn- eld: structural basis of feed-back activation. Eur J Neurosci
ing and memory can be obtained by taking into account 7:21702180.
the organization of local patterns of neuronal intercon- Bolshakov VY, Siegelbaum SA (1995) Regulation of hippocampal
nectivity. transmitter release during development and long-term potenti-
ation. Science 269:17301734.
Bragin A, Jando G, Nadasdy Z, Hetke J, Wise K, Buzsaki G (1995)
Gamma (40100 Hz) oscillation in the hippocampus of the
ACKNOWLEDGMENTS behaving rat. J Neurosci 15:4760.
Buckmaster PS, Schwartzkroin PA (1995a) Physiological and
This work was conceived, and the rst draft entirely written, by the morphological heterogeneity of dentate gyrus-hilus interneu-
late E.H. Buhl. Professor Buhl died during the revision of this chap- rons in the gerbil hippocampus in vivo. Eur J Neurosci
ter, and we dedicate it to his memory. 7:13931402.
316 The Hippocampus Book

Buckmaster PS, Schwartzkroin PA (1995b) Interneurons and inhibi- Fabian-Fine R, Skehel P, Errington ML, Davies HA, Sher E, Stewart
tion in the dentate gyrus of the rat in vivo. J Neurosci 15:774789. MG, Fine A (2001) Ultrastructural distribution of the alpha7
Buckmaster PS, Soltesz I (1996) Neurobiology of hippocampal nicotinic receptor subunit in rat hippocampus. J Neurosci 21:
interneurons: a workshop review. Hippocampus 6:330339. 79938003.
Buhl EH, Dann JF (1991) Cytoarchitecture, neuronal composition, Fisahn A, Pike FG, Buhl EH, Paulsen O (1998) Cholinergic induction
and entorhinal afferents of the flying fox hippocampus. of network oscillations at 40 Hz in the hippocampus in vitro.
Hippocampus 1:131152. Nature 394:186189.
Buhl EH, Halasy K, Somogyi P (1994a) Diverse sources of hip- Frazier CJ, Rollins YD, Breese CR, Leonard S, Freedman R,
pocampal unitary inhibitory postsynaptic potentials and the Dunwiddie TV (1998) Acetylcholine activates an alpha-
number of synaptic release sites. Nature 368:823828. bungarotoxin-sensitive nicotinic current in rat hippocampal
Buhl EH, Han Z-S, Lorinczi Z, Stezhka VV, Karnup SV, Somogyi P interneurons, but not pyramidal cells. J Neurosci 18:1187
(1994b) Physiological properties of anatomically identied 1195.
axo-axonic cells in the rat hippocampus. J Neurophysiol 71: Freund TF, Antal M (1988) GABA-containing neurons in the septum
12891307. control inhibitory interneurons in the hippocampus. Nature
Buhl EH, Cobb SR, Halasy K, Somogyi P (1995) Properties of unitary 336:170173.
IPSPs evoked by anatomically identied basket cells in the rat Freund TF, Buzsaki G (1996) Interneurons of the hippocampus.
hippocampus. Eur J Neurosci 7:19892004. Hippocampus 6:347470.
Buhl EH, Szilagyi T, Halasy K, Somogyi P (1996) Physiological prop- Freund TF, Gulyas AI, Acsady L, Gorcs L, Toth K (1990) Serotonergic
erties of anatomically identied basket and bistratied cells in control of the hippocampus via local inhibitory interneurons.
the CA1 area of the rat hippocampus in vitro. Hippocampus Proc Natl Acad Sci USA 87:85018505.
6:294305. Frotscher M, Leranth C (1985) Cholinergic innervation of the rat
Burnashev N, Schoepfer R, Monyer H, Ruppersberg JP, Gunther W, hippocampus as revealed by choline acetyltransferase immuno-
Seeburg PH, Sakmann B (1992) Control by asparagine residues cytochemistry: a combined light and electron microscopic
of calcium permeability and magnesium blockade in the study. J Comp Neurol 239:237246.
NMDA receptor. Science 257:14151419. Frotscher M, Leranth C, Lubbers K, Oertel WH (1984) Commissural
Buzsaki G, Leung L-WS, Vanderwolf CH (1983) Cellular bases of hip- afferents innervate glutamate decarboxylase immunoreactive
pocampal EEG in the behaving rat. Brain Res Rev 6:139171. non-pyramidal neurons in the guinea pig hippocampus.
Buzsaki G, Horvath Z, Urioste R, Hetke J, Wise K (1992) High- Neurosci Lett 46:137143.
frequency network oscillation in the hippocampus. Science Frotscher M, Seress L, Schwerdtfeger WK, Buhl E (1991) The mossy
256:10251027. cells of the fascia dentata: a comparative study of their ne
Cherubini E, Gaiarsa JL, Ben-Ari Y (1991) GABA: an excitatory trans- structure and synaptic connections in rodents and primates.
mitter in early postnatal life. Trends Neurosci 14:515519. J Comp Neurol 312:145163.
Chicurel ME, Harris KM (1992) Three-dimensional analysis of the Fukuda T, Kosaka T (2000) Gap junctions linking the dendritic
structure and composition of CA3 branched dendritic spines network of GABAergic interneurons in the hippocampus.
and their synaptic relationships with mossy ber boutons in the J Neurosci 20:15191528.
rat hippocampus. J Comp Neurol 325:169182. Galarreta M, Hestrin S (1999) A network of fast-spiking cells in
Cobb SR, Buhl EH, Halasy K, Paulsen O, Somogyi P (1996) the neocortex connected by electrical synapses. Nature
Synchronization of neuronal activity in hippocampus by indi- 402:7275.
vidual GABAergic interneurons. Nature 378:7578. Gao B, Fritschy JM (1994) Selective allocation of GABAA receptors
Cobb SR, Halasy K, Vida I, Nyiri G, Tamas G, Buhl EH, Somogyi P containing the a1 subunit to neurochemically distinct subpopu-
(1997) Synaptic effects of identied interneurons innervating lations of rat hippocampal interneurons. Eur J Neurosci
both interneurons and pyramidal cells in the rat hippocampus. 6:837853.
Neuroscience 79:629648. Gasbarri A, Verney C, Innocenzi R, Campana E, Pacitti C (1994)
Condorelli DF, Parenti R, Spinella F, Salinaro AT, Belluardo N, Cardile Mesolimbic dopaminergic neurons innervating the hippocam-
V, Cicirata F (1998) Coning of a new gap junction gene (C36) pal formation in the rat: a combined retrograde tracing and
highly expressed in mammalian brain neurons. Eur J Neurosci immunohistochemical study. Brain Res 668:7179.
10:12021208. Gasbarri A, Sulli A, Packard MG (1997) The dopaminergic mesen-
Dashtipour K, Yan XX, Dinh TT, Okazaki MM, Nadler JV, Ribak CE cephalic projections to the hippocampal formation in the rat.
(2002) Quantitative and morphological analysis of dentate Prog Neuropsychopharmacol Biol Psychiatry 21:122.
granule cells with recurrent basal dendrites from normal and Ge S, Dani JA (2005) Nicotinic acetylcholine receptors at glutamate
epileptic rats. Hippocampus 12:235244. synapses facilitate long-term depression or potentiation. J
Deller T, Martinez A, Nitsch R, Frotscher M (1996) A novel entorhi- Neurosci 25:60846091.
nal projection to the rat dentate gyrus: direct innervation of Geiger JR, Jonas P (2000) Dynamic control of presynaptic Ca(2)
proximal dendrites and cell bodies of granule cells and inow by fast-inactivating K() channels in hippocampal
GABAergic neurons. J Neurosci 16:33223333. mossy ber boutons. Neuron 28:927939.
Deuchars J, Thomson AM (1996) CA1 pyramid-pyramid connec- Geiger JRP, Lubke J, Roth A, Frotscher M, Jonas P (1997)
tions in rat hippocampus in vitro: dual intracellular recordings Submillisecond AMPA receptor-mediated signaling at a princi-
with biocytin lling. Neuroscience 74:10091018. pal neuron-interneuron synapse. Neuron 18:10091023.
Draguhn A, Traub RD, Schmitz D, Jefferys JGR (1998) Electrical cou- Gibson JR, Beierlein M, Connors BW (1999) Two networks of elec-
pling underlies high-frequency oscillations in the hippocampus trically coupled inhibitory neurons in neocortex. Nature
in vitro. Nature 394:189192. 402:7579.
 Local Circuits 317

Gulyas AI, Miettinen R, Jacobowitz DM, Freund TF (1992) Calretinin Kandel ER, Spencer WA (1961) Electophysiological properties of an
is present in non-pyramidal cells of the rat hippocampus. I. A archicortical neuron. Ann NY Acad Sci 94;570603.
new type of neuron specically associated with the mossy bre Klausberger T, Magill PJ, Marton LF, Roberts JD, Cobden PM,
system. Neuroscience 48:127. Buzsaki G, Somogyi P (2003) Brain-state and cell-type-specic
Gulyas AI, Miles R, Hajos N, Freund TF (1993a) Precision and vari- firing of hippocampal interneurons in vivo. Nature
ability in postsynaptic target selection of inhibitory cells in the 421:844848.
hippocampal CA3 region. Eur J Neurosci 5:17291751. Konig P, Engel AK, Singer W (1996) Integrator or coincidence detec-
Gulyas AI, Miles R, Sik A, Toth K, Tamamaki N, Freund TF (1993b) tor? The role of the cortical neuron revisited. Trends Neurosci
Hippocampal pyramidal cells excite inhibitory neurons through 19:130137.
a single release site. Nature 366:683687. Kosaka T (1980) The axon initial segment as a synaptic site: ultra-
Gulyas AI, Hajos N, Freund TF (1996) Interneurons containing cal- structure and synaptology of the initial segment of the pyrami-
retinin are specialized to control other interneurons in the rat dal cell in the rat hippocampus (CA3 region). J Neurocytol
hippocampus. J Neurosci 16:33973411. 9:861882.
Gulyas AI, Toth K, McBain CJ, Freund TF (1998) Stratum radiatum Kosaka T (1983) Gap junctions between non-pyramidal cell den-
giant cells: a type of principal cell in the rat hippocampus. Eur J drites in the rat hippocampus (CA1 and CA3 regions). Brain Res
Neurosci 10:38133822. 271:157161.
Gulyas AI, Megias M, Emri Z, Freund TF (1999) Total number and Kosaka T, Katsumaru H, Hama K, Wu J-Y, Heizmann CW (1987)
ratio of excitatory and inhibitory synapses converging onto sin- GABAergic neurons containing the Ca2-binding protein par-
gle interneurons of different types in the CA1 area of the rat valbumin in the rat hippocampus and dentate gyrus. Brain Res
hippocampus. J Neurosci 19:1008210097. 419:119130.
Gupta A, Wang Y, Markram H (2000) Organizing principles for a Lacaille J-C (1991) Postsynaptic potentials mediated by excitatory
diversity of GABAergic interneurons and synapses in the neo- and inhibitory amino acids in interneurons of stratum pyrami-
cortex. Science 287:273278. dale of the CA1 region of rat hippocampal slices in vitro. J
Hajos N, Acsady L, Freund TF (1996) Target selectivity and neuro- Neurophysiol 66:14411454.
chemical characteristics of VIP-immunoreactive interneurons Lacaille J-C, Williams S (1990) Membrane properties of interneurons
in the rat dentate gyrus. Eur J Neurosci 8:14151431. in stratum oriens-alveus of the CA1 region of rat hippocampus
Hajos N, Papp EC, Acsady L, Levey AI, Freund TF (1998) Distinct in vitro. Neuroscience 36:349359.
interneuron types express M2 muscarinic receptor immunore- Lacaille J-C, Mueller AL, Kunkel DD, Schwartzkroin PA (1987) Local
activity on their dendrites or axon terminals in the hippocam- circuit interactions between oriens/alveus interneurons and
pus. Neuroscience 82:355376. CA1 pyramidal cells in hippocampal slices: electrophysiology
Halasy K, Somogyi P (1993) Subdivisions in the multiple GABAergic and morphology. J Neurosci 7:19791993.
innervation of granule cells in the dentate gyrus of the rat hip- Larkman A, Stratford K, Jack J (1991) Quantal analysis of excitatory
pocampus. Eur J Neurosci 5:411429. synaptic action and depression in hippocampal slices. Nature
Halasy K, Miettinen R, Szabat E, Freund TF (1992) GABAergic 350:344347.
interneurons are the major postsynaptic targets of median Leranth C, Frotscher M (1987) Cholinergic innervation of hip-
raphe afferents in the rat dentate gyrus. Eur J Neurosci pocampal GAD- and somatostatin-immunoreactive commis-
4:144153. sural neurons. J Comp Neurol 261:3347.
Halasy K, Buhl EH, Lorinczi Z, Tamas G, Somogyi P (1996) Synaptic Leranth C, Szeidemann Z, Hsu M, Buzsaki G (1996) AMPA receptors
target selectivity and input of GABAergic basket and bistratied in the rat and primate hippocampus: a possible absence of
interneurons in the CA1 area of the rat hippocampus. GluR2/3 subunits in most interneurons. Neuroscience
Hippocampus 6:306329. 70:631652.
Han ZS, Buhl EH, Lorinczi Z, Somogyi P (1993) A high degree of Li X-G, Somogyi P, Tepper JM, Buzsaki G (1992) Axonal and den-
spatial selectivity in the axonal and dendritic domains of physi- dritic arborization of an intracellularly labeled chandelier cell in
ologically identied local-circuit neurons in the dentate gyrus of the CA1 region of rat hippocampus. Exp Brain Res 90:519525.
the rat hippocampus. Eur J Neurosci 5:395410. Li X-G, Somogyi P, Ylinen A, Buzsaki G (1994) The hippocampal
Henze DA, Wittner L, Buzsaki G (2002) Single granule cells reliably CA3 network: an in vivo intracellular labeling study. J Comp
discharge targets in the hippocampal CA3 network. Nat Neurol 339:181208.
Neurosci 5:790795 Lien CC, Jonas P (2003) Kv3 potassium conductance is necessary and
Hormuzdi SG, Pais I, LeBeau FEN, Towers SK, Rozov A, Buhl EH, kinetically optimized for high-frequency action potential gener-
Whittington MA, Monyer H (2001) Impaired electrical signal- ation in hippocampal interneurons. J Neurosci 23:20552068.
ing disrupts gamma frequency oscillations in connexin 36 de- Low K, Crestani F, Keist R, Benke D, Brunig I, Benson JA, Fritschy JM,
cient mice. Neuron 31:487495. Rulicke T, Bluethmann H, Mohler H, Rudolph U (2000)
Ishida Y, Shirokawa T, Miyaishi O, Komatsu Y, Isobe K (2000) Age- Molecular and neuronal substrate for the selective attenuation
dependent changes in projections from locus coeruleus to hip- of anxiety. Science 290:131134.
pocampus dentate gyrus and frontal cortex. Eur J Neurosci Loy R, Koziell DA, Lindsey JD, Moore RY (1980) Noradrenergic
12:12631270. innervation of the adult rat hippocampal formation. J Comp
Jones S, Yakel JL (1997) Functional nicotinic ACh receptors on Neurol 189:699710.
interneurones in the rat hippocampus. J Physiol (Lond) Maccaferri G, McBain CJ (1995) Passive propagation of LTD to stra-
504:603610. tum oriens-alveus inhibitory neurons modulates the tem-
Jung MW, McNaughton BL (1993) Spatial selectivity of unit activity poroammonic input to the hippocampal CA1 region. Neuron
in the hippocampal granular layer. Hippocampus 3:165182. 15:137145.
318 The Hippocampus Book

Maccaferri G, Roberts JDB, Szucs P, Cottingham CA, Somogyi P geting of 2-subunit-containing GABAA receptors in synapses
(2000) Cell surface domain specic postsynaptic currents of hippocampal pyramidal cells of the rat. Eur J Neurosci
evoked by identied GABAergic neurones in rat hippocampus 13:428442.
in vitro. J Physiol (Lond) 524:91116. Oleskevich S, Descarries L, Lacaille J-C (1989) Quantied distribu-
Magee JC (1998) Dendritic hyperpolarisation-activated currents tion of the noradrenaline innervation in the hippocampus of
modify the integrative properties of hippocampal CA1 pyrami- adult rat. J Neurosci 9:38033815.
dal neurons. J Neurosci 18:76137624. Oleskevich S, Descarries L, Watkins KC, Seguela P, Daszuta A (1991)
Magloczky Z, Acsady L, Freund TF (1994) Principal cells are the post- Ultrastructural features of the serotonin innervation in adult rat
synaptic targets of supramammillary afferents in the hippocam- hippocampus: an immunocytochemical description in single
pus of the rat. Hippocampus 4:322334. and serial thin sections. Neuroscience 42:777791.
Malinow R (1991) Transmission between pairs of hippocampal slice Ouardouz M, Lacaille J-C (1997) Properties of unitary IPSCs in hip-
neurons: quantal levels, oscillations, and LTP. Science pocampal pyramidal cells originating from different types of
252:722724. interneurons in young rats. J Neurophysiol 77:19391949.
Martina M, Vida I, Jonas P (2000) Distal initiation and active propa- Parra P, Gulyas AI, Miles R (1998) How many subtypes of inhibitory
gation of action potentials in interneuron dendrites. Science cells in the hippocampus? Neuron 20:983993.
287:295300. Samson Y, Wu JJ, Friedman AH, Davis JN (1990) Catecholaminergic
Martinez A, Lubke J, Del Rio JA, Soriano E, Frotscher M (1996) innervation of the hippocampus in the cynomolgus monkey.
Regional variability and postsynaptic targets of chandelier cells J Comp Neurol 298:250263.
in the hippocampal formation of the rat. J Comp Neurol Sayer RJ, Friedlander MJ, Redman SJ (1990) The time course and
376:2844. amplitude of EPSPs evoked at synapses between pairs of
McBain CJ, DiChiara TJ, Kauer JA (1994) Activation of metabotropic CA3/CA1 neurons in the hippocampal slice. J Neurosci
glutamate receptors differentially affects two classes of hip- 10:826836.
pocampal interneurons and potentiates excitatory synaptic Scharfman HE, Kunkel DD, Schwartzkroin PA (1990) Synaptic con-
transmission. J Neurosci 14:44334445. nections of dentate granule cells and hilar neurons: results of
Miettinen R, Freund TF (1992) Convergence and segregation of sep- paired intracellular recordings and intracellular horseradish
tal and median raphe inputs onto different subsets of hip- peroxidase injections. Neuroscience 37:693707.
pocampal inhibitory interneurons. Brain Res 594:263272. Schmitz D, Schuchmann S, Fisahn A, Draguhn A, Buhl EH, Petrasch-
Miles R (1990) Synaptic excitation of inhibitory cells by single CA3 Parwez RE, Dermietzel R, Heinemann U, Traub RD (2001) Axo-
hippocampal pyramidal cells of the guinea-pig in vitro. J Physiol axonal coupling: a new mechanism for ultrafast neuronal
(Lond) 428:6177. communication. Neuron 31:669671.
Miles R, Wong RKS (1983) Single neurones can initiate synchronized Schwerdtfeger WK, Buhl E (1986) Various types of non-pyramidal
population discharge in the hippocampus. Nature 306:371373. hippocampal neurons project to the septem and contralateral
Miles R, Wong RKS (1986) Excitatory synaptic interactions between hippocampus. Brain Res 386:146154.
CA3 neurones in the guinea-pig hippocampus. J Physiol (Lond) Seress L, Mrzljak L (1987) Basal dendrites of granule cells are normal
373:397418. features of the fetal and adult dentate gyrus of both monkey and
Miles R, Toth K, Gulyas AI, Hajos N, Freund TF (1996) Differences human hippocampal formations. Brain Res 405:169174.
between somatic and dendritic inhibition in the hippocampus. Sik A, Tamamaki N, Freund TF (1993) Complete axon arborization
Neuron 16:815823. of a single CA3 pyramidal cell in the rat hippocampus, and
Molnar E, Baude A, Richmond SA, Patel PB, Somogyi P, McIlhinney its relationship with postsynaptic parvalbumin-containing
RAJ (1993) Biochemical and immunocytochemical characteri- interneurons. Eur J Neurosci 5:17191728.
zation of antipeptide antibodies to a cloned GluR1 glutamate Sik A, Ylinen A, Penttonen M, Buzsaki G (1994) Inhibitory CA1-
receptor subunit: cellular and subcellular distribution in the rat CA3-hilar region feedback in the hippocampus. Science
forebrain. Neuroscience 53:307326. 265:17221724.
Morales R, Battenberg E, Bloom FE (1998) Distribution of neu- Sik A, Penttonen M, Ylinen A, Buzsaki G (1995) Hippocampal CA1
rons expressing immunoreactivity for the 5HT(3) receptor interneurons: an in vivo intracellular labeling study. J Neurosci
subtype in the rat brain and spinal cord. J Comp Neurol 15:66516665.
402:385401. Sik A, Penttonen M, Buzsaki G (1997) Interneurons in the hip-
Morilak DA, Somogyi P, Lujan-Miras R, Ciaranello RD (1994) pocampal dentate gyrus: an in vivo intracellular study. Eur
Neurons expressing 5-HT2 receptors in the rat brain: neuro- J Neurosci 9:573588.
chemical identication of cell types by immunocytochemistry. Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta
Neuropsychopharmacology 11:157166. phase precession in hippocampal neuronal populations and the
Mott DD, Turner DA, Okazaki NM, Lewis DV (1997) Interneurons compression of temporal sequences. Hippocampus 6:149172.
of the dentate-hilus border of the rat dentate gyrus: morpho- Smith SS, Gong QH, Hsu FC, Markowitz RS, ffrenchMullen JMH, Li
logical and electrophysiological heterogeneity. J Neurosci HS (1998) GABA(A) receptor alpha 4 subunit suppression pre-
17:39904005. vents withdrawal properties of an endogenous steroid. Nature
Nusser Z, Sieghart W, Benke D, Fritschy J-M, Somogyi P (1996) 392:926930.
Differential synaptic localization of two major -aminobutyric Somogyi P, Freund TF, Hodgson AJ, Somogyi J, Beroukas D, Chubb
acid type A receptor a subunits on hippocampal pyramidal cells. IW (1985) Identied axo-axonic cells are immunoreactive for
Proc Natl Acad Sci USA 93:1193911944. GABA in the hippocampus and visual cortex of the cat. Brain
Nyiri G, Freund TF, Somogyi P (2001) Input-dependent synaptic tar- Res 332:143149.
 Local Circuits 319

Somogyi P, Tamas G, Lujan R, Buhl EH (1998) Salient features of (1996) Analysis of gamma rhythms in the rat hippocampus in
synaptic organisation in the cerebral cortex. Brain Res Rev vitro and in vivo. J Physiol (Lond) 493:471484.
26:113135. Traub RD, Jefferys JGR, Whittington MA (1997) Simulation of
Soriano E, Martinez A, Farinas I, Fortscher M (1993) Chandelier cells gamma rhythms in networks of interneurons and pyramidal
in the hippocampal formation of the rat: the entorhinal area cells. J Comput Neurosci 4:141150.
and subicular complex. J Comp Neurol 337:151167. Traub RD, Schmitz D, Jefferys JGR, Draguhn A (1999) High-
Sperk G, Schwarzer C, Tsunashima K, Fuchs K, Sieghart W (1997) frequency population oscillations are predicted to occur in
GABAA receptor subunits in the rat hippocampus. 1. Immuno- hippocampal pyramidal neuronal networks interconnected
cytochemical distribution of 13 subunits. Neuroscience 80: by axoaxonal gap junctions. Neuroscience 92:407426.
9871000. Traub RD, Bibbig A, Fisahn A, LeBeau FEN, Whittington MA,
Swanson LW, Hartman BK (1975) The central noradrenergic system: Buhl EH (2000) A model of gamma-frequency network oscilla-
an immunouorescence study of the location of the cell bodies tions induced in the rat CA3 region by carbachol in vitro. Eur
and their efferent connections in the rat utilizing dopamine- J Neurosci 12:40934106.
beta-hydroxylase. J Comp Neurol 163:467507. Trommald M, Jensen V, Andersen P (1995) Analysis of dendritic
Tamas G, Buhl EH, Somogyi P (1997) Fast IPSPs elicited via multiple spines in rat CA1 pyramidal cells intracellularly lled with a u-
synaptic release sites by different types of GABAergic neurone in orescent dye. J Comp Neurol 353:260274.
the cat visual cortex. J Physiol (Lond) 500:715738. Turner DA, Chen Y, Isaac JT, West M, Wheal HV (1997) Excitatory
Tamas G, Somogyi P, Buhl EH (1998) Differentially interconnected synaptic site heterogeneity during paired pulse plasticity in CA1
networks of GABAergic interneurons in the visual cortex of the pyramidal cells in rat hippocampus in vitro. J Physiol (Lond)
cat. J Neurosci 18:42554270. 500:441461.
Tamas G, Buhl EH, Lorincz A, Somogyi P (2000) Proximally targeted Vida I, Frotscher M (2000) A hippocampal interneuron associated
GABAergic synapses and gap junctions synchronize cortical with the mossy fiber system. Proc Natl Acad Sci USA
interneurons. Nat Neurosci 3:366371. 97:12751280.
Taylor CP, Dudek (1982) A physiological test for electrotonic cou- Vida I, Halasy K, Szinyei C, Somogyi P, Buhl EH (1998) Unitary IPSPs
pling between CA1 pyramidal cells in rat hippocampal slices. evoked by interneurons at the stratum radiatum stratum
Brain Res 235:351357. lacunosum-moleculare border in the CA1 area of the rat hip-
Thomson AM, Destexhe A (1999) Dual intracellular recordings and pocampus in vitro. J Physiol (Lond) 506:755773.
computational models of slow inhibitory postsynaptic poten- Wang X-J, Buzsaki G (1996) Gamma oscillation by synaptic inhibi-
tials in rat neocortical and hippocampal slices. Neuroscience tion in a hippocampal interneuronal network model. J Neurosci
92:11931215. 16:64026413.
Thomson AM, Bannister AP, Hughes DI, Pawelzik H (2000) Wei DS, Mei YA, Bagal A, Yao JP, Thompson SM, Tang CM
Differential sensitivity to zolpidem of IPSPs activated by mor- (2001) Compartmentalised and binary behaviour of terminal
phologically identied CA1 interneurons in slices of rat hip- dendrites in hippocampal pyramidal neurons. Science
pocampus. Eur J Neurosci 12:425436. 293:22722275.
Toth K, Freund TF (1992) Calbindin D28k-containing nonpyrami- White JA, Chow CC, Ritt J, Soto-Trevino C, Kopell N (1998)
dal cells in the rat hippocampus: their immunoreactivity for Synchronization and oscillatory dynamics in heterogeneous,
GABA and projection to the medial septum. Neuroscience 49: mutually inhibited neurons. J Comput Neurosci 5:516.
793805. Whittington MA, Traub RD, Jefferys JGR (1995) Synchronized oscil-
Toth K, Freund TF, Miles R (1997) Disinhibition of rat hippocampal lations in interneuron networks driven by metabotropic gluta-
pyramidal cells by GABAergic afferents from the septum. mate receptor activation. Nature 373:612615.
J Physiol (Lond) 500:463474. Wouterlood FG, Saldana E, Witter MP (1990) Projection from the
Traub RD, Bibbig A (2000) A model of high-frequency ripples in nucleus reuniens thalami to the hippocampal region: light and
the hippocampus based on synaptic coupling plus axon-axon electron microscopic tracing study in the rat with the antero-
gap junctions between pyramidal neurons. J Neurosci 20: grade tracer Phaseolus vulgaris-leucoagglutinin. J Comp Neurol
20862093. 296:179203.
Traub RD, Miles R (1995) Pyramidal cell to inhibitory cell spike Ylinen A, Soltesz I, Bragin A, Penttonen M, Sik A, Buzsaki G (1995)
transduction explicable by active dendritic conductances in Intracellular correlates of hippocampal theta rhythm in identi-
inhibitory cell. J Comput Neurosci 2:291298. ed pyramidal cells, granule cells, and basket cells. Hippocampus
Traub RD, Whittington MA, Colling SB, Buzsaki G, Jefferys JGR 5:7890.
This page intentionally left blank
9 Elizabeth Gould

Structural Plasticity

numerous studies have identied factors and conditions that

9.1 Overview regulate its occurrence. This information has suggested possi-
ble functions for adult-generated neurons.
The hippocampal formation has been described as a relatively This chapter begins by discussing the types of structural
late-developing brain region. Rather than having a xed, plasticity in the hippocampal formation and then concen-
immutable structure, as might be implied by the earlier trates on the evidence revealing that new neurons are formed
anatomical and physiological chapters, it is in fact known to in the dentate gyrus throughout life. It then reviews the liter-
undergo several structural changes throughout life that are ature related to the modulation of neurogenesis in the dentate
traditionally described as developmental. Numerous reports gyrus by hormones and by experience and nally focuses on
have demonstrated that the hippocampal formation is capable the role these new cells may play in hippocampal function.
not only of substantial reorganization when damaged but also The possibility that new cells participate in certain types of
dramatic structural change when intact. It appears to undergo learning and in the modulation of anxiety and stress
dynamic modications continually in the form of dendritic responses is explored. Some recent evidence suggesting
extension and retraction as well as synapse formation and unusual properties of immature neurons is considered in light
elimination. of the hypothesis that a continual inux of these unique cells
Perhaps the most basic of all structural changes is the addi- to hippocampal circuitry has important functional conse-
tion of new neurons, a phenomenon known as neurogenesis. quences.
Although this process was once believed to be restricted to the
embryonic or early postnatal period, neurogenesis is now rec-
ognized to be a substantial process in some regions of the
adult brain. In fact, the dentate gyrus of the rat adds thou-
sands of new neurons every day throughout adulthood 9.2 Dendritic and Synaptic Plasticity
(Cameron and McKay, 2001), raising intriguing questions in the Hippocampal Formation
about the regulation and functional signicance of this phe-
nomenon. 9.2.1 Naturally Occurring Structural Plasticity
The incorporation of new neurons into preexisting cir-
cuitry results in a cascade of structural changes that further For more than a century, neuroanatomists have speculated
increase the structural plasticity of this region. For example, about the capacity of the adult brain to change its structure
new neurons elaborate axons and dendrites (Hastings and (see Stanisch and Nitsch, 2002 for review). Evidence accumu-
Gould, 1999; Markakis and Gage, 1999; Cameron and McKay, lated over the past several decades has identied the hip-
2001) and undergo synaptogenesis (Kaplan and Hinds, 1977; pocampus as a region with a large degree of structural
Markakis and Gage, 1999). These progressive events are typi- plasticity. These studies together indicate that the ne, and
cally followed by a series of regressive phenomena, such as cell sometimes even the gross, structure of the hippocampal for-
death (Gould et al., 1999a, 2001; Dayer et al., 2003), which mation is constantly changing under normal conditions. Far
likely involves process retraction and synapse elimination. from being a structurally static region, the hippocampal for-
Coincident with the growing acceptance of adult neurogene- mation is a dynamic area whose synapses and dendrites are
sis as a signicant phenomenon in the mammalian brain, undergoing continuous rearrangement.

321
322 The Hippocampus Book

9.2.2 Hormones and Dendritic Architecture 9.2.3 Experience and Dendritic Architecture

Ovarian Steroids Stress

A continual uctuation in the numbers of dendritic spines on Dendritic architecture in the hippocampus is also inuenced
the pyramidal cells of the CA1 region has been reported in by experience. Chronic stress has generally been shown to
adult female rats and monkeys. This uctuation occurs over a have negative effects on the structure of dendrites in the hip-
period of hours to days and is controlled by the levels of cir- pocampus. As observed with chronic glucocorticoid treat-
culating ovarian steroids (Woolley et al., 1990b, 1997; Shors et ment, the dendritic tree of the CA3 pyramidal neurons in
al., 2001a; Hao et al., 2003; Cooke and Woolley, 2005). When adult male rats and tree shrews decreases in size following
ovarian steroids are at a high level, the number of dendritic repeated restraint stress (Watanabe et al., 1992; Magarinos et
spines is high. When the hormones are low, the number of al., 1996; McEwen, 1999). These effects appear to be reversible
spines is low. These changes in the number of dendritic spines in unstressed conditions. In addition, changes in the dendritic
are paralleled by changes in the number of excitatory synapses tree with stress and recovery parallel performance on tasks
in this region (Woolley and McEwen, 1992; Woolley et al., that require the hippocampus (Luine et al., 1994).
1996). The extent to which these structural changes reect Acute stress has also been shown to alter the number of
functional changes remains a matter of debate (Woolley, dendritic spines on CA1 pyramidal cells of adult rats, but this
1998). For example, conicting literature exists on the inu- effect seems to be dependent on the sex of the animal and, in
ence of estrogen on behaviors associated with the hippocam- the case of females, the stage of estrous. Brief, intermittent tail
pus. Some reports indicate that estrogen improves shocks have been shown to increase the number of dendritic
performance on hippocampus-dependent tasks (Daniel et al., spines on pyramidal cells of males (Shors et al., 2001a). By
1997; Gibbs, 1999; Daniel and Dohanich, 2001; Li et al., contrast, female rats showed the opposite effecta decrease in
2004; Sandstrom and Williams, 2004); and in some of these the number of dendritic spines on pyramidal cells when
studies, the time course of behavioral improvement corre- shocked during diestrus but not when shocked during estrus
sponds to changes in dendritic spines. Other reports have (Shors et al., 2001a). Although it remains unclear what the
failed to demonstrate improvements in hippocampus- functional signicance of these effects is, it is worth noting
dependent tasks with estrogen treatment or across the that opposite effects of stress on learning have been reported
estrous cycle (Berry et al., 1997; Warren and Juraska, 1997; in males and females (Wood et al., 2001). These behavioral
Chesler and Juraska, 2000). Thus, the relation between behav- effects appear to parallel the changes in dendritic spines sug-
ior and dendritic spine number in the hippocampus remains gesting a functional relation.
unclear.
It is worth noting, however, that continuous estrous cycles, Environmental Complexity and Learning
a common hormonal prole for female laboratory rodents,
are probably uncommon in rodents living in the wild. A more Beginning with the early work of Rosenzweig and Diamond
likely endocrine prole would include postpubertal preg- (Rosenzweig et al., 1962; reviewed in Will et al., 2004), a large
nancy followed by bouts of lactation, single estrous cycles body of literature has amassed regarding the effects of living
just after weaning, and subsequent pregnancy. Thus, the rele- in an enriched laboratory environment on brain structure.
vance of estrogen-associated changes in dendritic spines on These studies have demonstrated repeatedly that, compared to
hippocampal function may lie in events such as pregnancy laboratory control animals, animals living in enriched envi-
and the postpartum period (see Woolley, 1998, 1999 for com- ronments during either development or adulthood have larger
mentary). brains along multiple measures. Most of this work has focused
on the neocortex, but the effects are generally similar in the
Adrenal Steroids hippocampus. Studies have demonstrated that living in
enriched environments increases the size of the dendritic tree
High levels of circulating glucocorticoids have been associated (Juraska et al., 1985; Faherty et al., 2003), the number of den-
with atrophy of dendrites in the CA3 region. Chronic treat- dritic spines (Moser et al., 1994), and the number of synapses
ment with corticosterone, the main glucocorticoid in rodents, (Altschuler, 1979; Briones et al., 2005) in the hippocampus
decreases the complexity and size of the apical dendritic tree compared to controls. Other studies have shown that rearing
of CA3 pyramidal neurons (Woolley et al., 1990a). Although in an enriched environment is associated with improved per-
it is possible that these changes are an impending sign of cell formance on hippocampus-dependent learning tasks (Juraska
death, which has been reported with chronic elevated gluco- et al., 1984; van Praag et al., 2000; Teather et al., 2002).
corticoids (Sapolsky et al., 1985), it has been suggested that Enriched laboratory environments typically included a
retraction of dendrites may play an adaptive role in protecting multitude of stimuli that increase stress, physical activity, and
the hippocampus from excessive glutamate (Luine et al., learning. Numerous studies have sought to identify the spe-
1994). cic experiential cues that mediate the brain changes; but to
 Structural Plasticity 323

date, questions remain. Several studies appear to have ruled could be discovered, it might be possible to harness them in
out stress as a chief mediator of the effects of the enriched the service of brain repair. The phenomenon is not specic to
environment on the hippocampus either by demonstrating just one set of axons in the hippocampus: Uninjured axons
comparable stress hormone levels between groups or differen- that sprout when hippocampal afferents are severed include
tial effects on brain structure following specic stressors. commissural and crossed entorhinal afferents (Deller and
Some studies have suggested that physical activity is a con- Frotscher, 1997), cholinergic septo-hippocampal fibers
tributing factor to the changes in hippocampal structure asso- (Forster et al., 1997), serotonergic axons (Zhou et al., 1995),
ciated with living in an enriched environment. For instance, and noradrenergic bers (Peterson, 1994). Other axons how-
Eadie et al. (2005) reported an increased density of dendritic ever, including perforant path bers, do not appear to sprout.
spine density on dentate granule cells in rodents that had In some studies, structural regeneration has been associ-
access to a running wheel (but see Faherty et al., 2003). This ated with functional recovery. Time-dependent changes in
nding is similar to that reported for animals living in open eld activity after unilateral entorhinal lesions are corre-
enriched environments (Faherty et al., 2003). The extent to lated with the growth of the crossed pathway from the con-
which changes in hippocampal structure induced by living in tralateral entorhinal region (Steward et al., 1977). Whereas
a complex environment are the result of one particular cue animals with unilateral entorhinal lesions are sometimes
versus a constellation of cues remains unknown. The contri- unable to perform hippocampus-dependent tasks shortly
bution of learning and synaptic plasticity to changes in hip- after the damage has occurred, recovery of function has been
pocampal structure has also been investigated. noted following regeneration of sprouted axons (Leonard et
Several studies have reported structural changes in the hip- al., 1995). This is particularly intriguing given the fact that it
pocampus as a result of learning, or long-term potentiation is often a different population of cells whose axons repopulate
(LTP), a form of synaptic plasticity often associated with the absent target sites after the lesion. However, it is not
learning. These effects of learning and LTP range from absolutely clear that axonal regeneration is the cause of func-
changes in the shape of dendritic spines (reviewed by Yuste tional recovery, even if it takes place over the same time
and Bonhoeffer, 2001), increases in spine number (Moser et period. For example, studies by Ramirez and Stein (1984)
al., 1994; Trommald et al., 1996; OMalley et al., 2000; Leuner indicated that sectioning the dorsal psalterium (through
et al., 2003), and changes in the number and distribution of which the growing entorhinal bers of the crossed pathway
synapses along dendrites (Chang et al., 1991; Rusakov et al., ordinarily travel) does not alter the time course of functional
1997; Andersen and Soleng, 1998; Toni et al., 1999). These recovery seen after unilateral entorhinal lesions. Other mech-
ndings mirror some of the changes associated with living in anisms have been associated with functional recovery after
enriched environments. However, the extent to which these axotomy, including hyperactivity of the remaining axons
changes contribute to learning remains unknown. (Gage et al., 1983a) and the supersensitivity of postsynaptic
neurotransmitter receptors (Patel et al., 1996). In some cases,
9.2.4 Structural Plasticity Following Damage these additional mechanisms appear to subside once axonal
ingrowth is complete (Gage et al., 1983b).
Perhaps not surprisingly, given the large degree of structural Other examples of structural plasticity in the hippocampal
change that can occur in the intact hippocampal formation formation have been observed in studies using animal models
even in the adult animal, this region also has a remarkable of disease states such as epilepsy and stroke. In these cases,
capacity for regeneration after injury. During the early 1970s, many cells in the hippocampal formation die. With time fol-
Raisman and Field (1973) reported evidence for extensive, lowing seizure or stroke, there is considerable structural reor-
predictable reorganization of synaptic connections in the sep- ganization in the form of axon sprouting, which could be
tal nuclei following transection of the mbria. This seminal viewed as regenerative (Sloviter, 1999). Studies examining
study set the stage for work designed to explore related struc- human brains of epileptic individuals and those with
tures, and shortly thereafter similar results were observed for Alzheimers disease have detected evidence of structural reor-
the deafferented hippocampus. Since then, the phenomenon ganization (Cotman et al., 1990; Nadler, 2003). The extent to
of axonal sprouting has been examined extensively following which these compensatory structural changes slow functional
transection of perforant path axons. After this manipulation, deterioration remains unknown. In addition, it is possible that
nondamaged axons sprout and occupy the now devoid target experience can enhance or potentially exacerbate damage-
spaces in a layer-specic manner (Lynch et al., 1976; Scheff et induced reorganization (Icanco and Greenough, 2000).
al., 1977; Deller and Frotscher, 1997; Frotscher et al., 1997). The brain is protected from mechanical injury by the skull
This remarkable phenomenon has had a large impact on neu- and from chemical injury by the blood-brain barrier. Thus, it
roscience as it provided denitive evidence that regeneration is not obvious that the regenerative capacity inherent to the
in the adult brain was possiblea phenomenon suggested by hippocampus evolved to correct brain damage, particularly
the work of Ramon y Cajal and other early neuroanatomists that which occurs in old age. However, axonal sprouting and
(Stahnisch and Nitsch, 2002). The therapeutic implications of synaptogenesis may be developmental phenomena that are
this work are clear: If the mechanisms underlying sprouting reinstated in the presence of certain environmental cues (e.g.,
324 The Hippocampus Book

the formation of a large number of target sites following cell with an exogenous source of afferents or progenitor cells.
death). Alternatively, or in addition, axonal sprouting and Future studies using information gleaned from investigations
synaptogenesis may be occurring under normal circum- of naturally occurring adult neurogenesis may facilitate the
stances at a low level, and the magnitude of their occurrence successful replacement of damaged neurons and complete
increases with damage-associated cues. Given the large recovery of function. This approach, however, may be useful
amount of evidence suggesting that the hippocampus is struc- only under certain conditions where the primary neuropatho-
turally plastic under normal conditions, the latter possibility logical hallmark is cell death, such as occurs with stroke. Other
seems especially likely. conditions, such as Alzheimers disease, which result in the
accumulation of abnormal proteins and the formation of
9.2.5 Transplantation senile plaques and neurofibrillary tangles, may be less
amenable to transplantation.
The remarkable plasticity of the hippocampus has also been
demonstrated through experiments designed to induce
regrowth of the damaged hippocampus via transplantation.
Studies have shown that the hippocampus of adult rats incor- 9.3 Adult Neurogenesis
porates fetal tissue from various brain regions (Bjorklund and
Stenevi, 1984; Bjorklund and Gage, 1985). These studies have The rst evidence for adult neurogenesis in the intact mam-
demonstrated much greater growth and survival of trans- malian brain was published more than 40 years ago by Joseph
planted tissue when the host hippocampus has been damaged Altman and colleagues (Altman, 1962, 1963, 1969; Altman
either directly or via axotomy or mbria-fornix destruction and Das, 1965a,b). Over the ensuing decades, the ndings
(Low et al., 1982; Gage and Bjorklund, 1986). Grafting of sub- were corroborated (Kaplan and Hinds, 1977; Kaplan, 1981;
cortical noradrenergic or cholinergic fetal tissue into the den- Kaplan and Bell, 1984; but see Rakic 1985), but only during
ervated adult hippocampus results in ingrowth of the the last decade has the phenomenon received intensive
transplanted tissue into the unoccupied target areas (Nilsson scrutiny and investigation by the neuroscience community
et al., 1988), restoration of neurotransmitter levels (Bjorklund (Boxes 91 and 92). The dentate gyrus is one of two brain
et al., 1990; Kalen et al., 1991), emergence of relatively normal regions in which adult neurogenesis has been widely recog-
electrophysiological responses (Buzsaki et al., 1987), and at nized (Box 92), and the basic steps in the process of forming
least partial recovery of function (Low et al., 1982). However, new neurons in this area are well understood.
there appears to be an advantage of homotypic tissue in terms In the dentate gyrus, new cells originate from progenitor or
of restoring a denervated area; cholinergic fetal transplants are precursor cells located in the dentate gyrus itself. These cells
more likely to restore the structural, biochemical, and behav- have the capacity to proliferate and are more restricted in their
ioral decits associated with selective cholinergic lesions than progeny than stem cells. The progenitor cells have some char-
nonspecic lesions affecting numerous afferent types (Leanza acteristics of astroglia (Seri et al., 2001) and are primarily con-
et al., 1998). These studies suggest possible approaches for centrated in the subgranular zone, the region between the
treatment of human conditions associated with diminished granule cell layer and the hilus. The subgranular zone is not a
hippocampal function. However, there are serious limitations distinct layer but, rather, a sporadic collection of progenitor
to this approach. Apart from the concern of immune rejection cell clusters lining the deep aspect of the granule cell layer.
(Lawrence et al., 1990), the grafted tissue is rarely incorpo- Cells in this region divide, presumably asymmetrically, to pro-
rated completely, and its survival and continued efficacy are duce some daughter cells that retain the ability to divide and
not guaranteed. These limitations, coupled with the recogni- other cells that become either neurons or glia. The new cells
tion that certain parts of the brain support neurogenesis even migrate the short distance to the granule cell layer and differ-
during adulthood, have led to the development of transplant entiate into neurons (Fig. 91).
strategies using fetal stem cells. Adult neurogenesis has been observed in the hippocampus
Transplantation of expanded neural stem cells into the of virtually every mammalian species examined, including
damaged hippocampus has revealed successful incorporation mice, voles, rats, cats, rabbits, guinea pigs, tree shrews, mar-
of new neurons into the CA elds. These studies have shown mosets, macaques, and humans (Altman and Das, 1965a,b;
that new neurons can arise from progenitors, migrate to their Gould et al., 1997, 1998, 1999b; Kaplan and Hinds, 1977;
appropriate locations where they form long distance axons, Eriksson et al., 1998; Galea and McEwen, 1999; Kornack and
make appropriate connections, and exihibit normal electro- Rakic, 1999). There is some debate about the relative signi-
physiological responses (Englund et al., 2002). Studies suggest cance of this phenomenon across vertebrate species, and the
that the hippocampus, particularly the dentate gyrus, is con- argument has been made that an inverse relation exists
ducive to the acceptance of neural stem cell transplants between the amount of adult neurogenesis and cognitive
because the environment of this brain region supports neuro- complexity (Rakic, 1985a,b). However, the robustness of adult
genesis under normal conditions (Fricker et al., 1999). The neurogenesis in the brains of adult avians (Nottebohm, 2002),
eld of transplantation research has elucidated the capacity of creatures with extensive cognitive capabilities, argues against
the hippocampus to reorganize structurally when provided this interpretation. Although it is difficult to assess the relative
 Structural Plasticity 325

Box 91 Box 92
Why Was Structural Change in the Adult Adult Neurogenesis in Non-neurogenic
Hippocampal Formation Not Detected Earlier? Brain Regions?

Given the magnitude of new neuron addition to the dentate Around the time of the initial reports of adult neurogenesis
gyrus, the fact that early neuroanatomists did not report evi- in the dentate gyrus and olfactory bulb (Altman, 1962, 1969;
dence for structural change in this intensively studied brain Altman and Das, 1965a,b), Altman reported similar evidence
region is puzzling. If ~9000 new cells are added to the den- for new neurons in the neocortex of adult rats and cats
tate gyrus every day and a signicant proportion of these (Altman, 1962, 1963). This article was corroborated by a later
cells extend axons, elaborate dendrites, and undergo synapto- study claiming new cells with ultrastructural characteristics
genesis, evidence of massive structural change should have of neurons in the neocortex of adult rats (Kaplan, 1981).
been noted. There are probably several reasons for this major This work was met with skepticism, as were the ndings in
oversight. the dentate gyrus and olfactory bulb (Rakic, 1985a,b; Kaplan,
First, there is stability during the turnover. Neuron num- 2001). With the application of a different set of techniques
ber in the dentate gyrus appears to be tightly regulated. In (see Box 93), adult neurogenesis was rediscovered in the
addition to massive cell production, neurite extension, and dentate gyrus and olfactory bulb during the 1990s and is
synaptogenesis, there is massive cell death, neurite retraction, now a relatively well established phenomenon. Both of these
and synapse elimination. Traditional histological methods regions have been termed neurogenic in that they are
would generally have failed to reveal these dynamic processes widely believed to support adult neurogenesis. However, the
because they do not result in changes in the size or shape case for the neocortex, as well as other brain regions, remains
of the brain region or in the number of cells, dendritic ele- unresolved. Several studies have reported positive evidence in
ments, or synapses. the form of BrdU-labeled cells with neuronal staining char-
Second, because the prevailing view has been that struc- acteristics in the neocortex (Gould et al., 1999c, 2001;
tural change is a developmental, not adult, phenomenon, evi- Bernier et al., 2002; Dayer et al., 2005), striatum (Bedard et
dence for an adult change may have been interpreted al., 2002; Dayer et al., 2005), substantia nigra (Zhao et al.,
otherwise. One possible example of this may be seen with 2003), amygdala (Bernier et al., 2002; Fowler et al., 2005),
descriptions of mossy ber terminals in the CA3 region. and hypothalamus (Rankin et al., 2003; Fowler et al., 2005;
Detailed neuroanatomical descriptions of the specialized Xu et al., 2005). Numerous other studies employing the same
synapses of mossy bers on CA3 pyramidal cells, called general methodologies reported no evidence for adult neuro-
thorny excrescences, have described numerous morpholog- genesis in these regions or evidence only for adult neurogen-
ical types of these terminals. Previously, these various types esis following damage (Magavi et al., 2000; Benraiss et al.,
of terminal were presented as evidence that mossy ber 2001; Kornack and Rakic, 2001; Koketsu et al., 2003; Kodama
synapses were unusually complex. Given our current under- et al., 2004; Yoshimi et al., 2005). The reasons for these dis-
standing of neurogenesis in the region, it seems reasonable crepancies remain unclear, although they may be related to
to consider the possibility that these various types of synapse the fact that in the positive reports the relative number of
reect different maturational stages of connections formed new neurons reported in each of these regions is much lower
by newly generated granule cells. Structural variability in than that reported in the olfactory bulb and dentate gyrus
the adult hippocampus may reect the dynamic changes (see Gould et al., 2001 for a quantitative comparison of
occurring. double-labeled cells between dentate gyrus and neocortex).
If the phenomenon exists but at a much lower rate, it is pos-
sible that slight technical differences could make the differ-
Figure 91. Adult neurogenesis in the dentate gyrus. 1. Precursor
ence between detecting and failing to detect a small number
cells, expressing the characteristics of astrocytes, reside in the sub- of new neurons. Alternatively, it is possible that the reports
granular zone (SGZ). 2. Precursor cells divide asymmetrically at a of adult neurogenesis outside the hippocampus and olfactory
rate that is modiable by neural activity and experience. 3. bulb are examples of false positivesnew cells incorrectly
Following division, one cell remains in the SGZ and retains the identied as neurons or developmentally generated neurons
capacity to proliferate, and the other cell enters the granule cell layer incorrectly identied as new (Rakic, 2002). Future experi-
(GCL) of the dentate gyrus. Most of the newly born cells that enter mentation and the use of more denitive techniques should
the granule cell layer acquire neuronal characteristics. resolve these issues.

numbers of new neurons across species given that different

methodologies are often utilized and there may be species dif-
ferences in the efficacy of certain techniques, it seems clear
that rodents exhibit a substantial amount of adult neurogene-
sis in the dentate gyrus.
In the rat dentate gyrus, the evidence that new cells differ-
entiate into neurons is strong. New cells have been shown to
receive synaptic input (Kaplan and Hinds, 1977; Markakis and
326 The Hippocampus Book

large number of new neurons during adulthood raises the

issue of whether the hippocampal formation continues to
grow throughout life. It is now known that, alongside neuro-
genesis, there is substantial death of granule cells, especially of
newborn neurons in adult rodents and primates under con-
trol conditions (Gould et al., 2001; Dayer et al., 2003). This cell
death may offset the continual addition of new cells, resulting
in a near-static total population of neurons (Box 91). The
presence of pyknotic or degenerating cells in the dentate gyrus
supports this view.
Several questions arise in connection with adult neurogen-
esis. First, do the new cells replace neurons that were born
during development, or might they selectively replace those
born during adulthood? Evidence suggests that both of these
possibilities may be the case. Cells generated during the rst
postnatal week of life as well as those generated during adult-
hood decrease in number with time following mitosis (Dayer
et al., 2003). However, most of the dying cells appear to be
those that have been produced during adulthood. Pyknotic
cells in the dentate gyrus are located primarily in the subgran-
ular zone or deep in the granule cell layer, regions that contain
progenitor cells and immature neurons. In addition, BrdU-
Figure 92. Most cells generated in the dentate gyrus of adults and 3H-thymidine-labeling studies in adults suggest that the
express neuronal characteristics. A. Photomicrograph of a originally labeled population expands owing to mitosis and
BrdU-positive cell (red) in the adult rat dentate gyrus, expressing
then drops off in number. Coincident with the decrease in
TUJ-1 (green), a marker of immature and mature neurons.
labeled cell number is the appearance of BrdU-labeled
B. Photomicrograph of a BrdU-positive cell (red) in the adult rat
dentate gyrus, not expressing the astrocytic protein GFAP (green). pyknotic cells (Gould et al., 2001). However, in the intact
C. Photomicrograph of a segment of dentate gyrus granule cell brain, environmental factors have been shown to inuence the
layer, triple-labeled with Hoechst, a DNA dye (blue), BrdU (red), survival of adult-generated granule cells. This raises the issue
and NeuN (green). Many but not all BrdU-positive cells express of whether information on the turnover of newly produced
NeuN, a marker of mature neurons. granule cells is confounded by the abnormal conditions under
which laboratory control animals exist. Thus, the natural life-
Gage, 1999), extend axons into the CA3 region (Staneld and span of adult-generated neurons in the dentate gyrus remains
Trice, 1988; Hastings and Gould, 1999; Markakis and Gage, unknown.
1999), and express several biochemical markers of granule
neurons, including NeuN, NSE, calbindin, MAP-2, and Tuj1 9.3.2 Hormones and Adult Neurogenesis
(Kuhn et al., 1996; Gould et al., 2001) (Fig. 92, see color insert;
Box 93). Adult-generated cells in the mouse dentate gyrus A separate set of questions concerning neurogenesis has to do
have been shown to generate action potentials and exhibit with its regulation. One potential regulatory mechanism is
other electrophysiological characteristics associated with gran- endocrine, as the hippocampal formation is known to be
ule cells (van Praag et al., 2002). Some newly generated cells in richly endowed with hormone receptors. Many cell types in
the dentate gyrus of adult rats may have the functional prop- the hippocampal formation contain hormone receptors and,
erties of inhibitory neurons (Liu et al., 2003). Although these as discussed above, this area responds to experimental hor-
ndings suggest that two subpopulations of new neurons exist mone manipulations during adulthood with dramatic struc-
in this brain region (one with excitatory properties and the tural changes (McEwen, 1999). Several studies have shown
other with inhibitory properties), an alternative possibility is that the production of new neurons in the dentate gyrus is
that new neurons temporarily exhibit characteristics of sensitive to the levels of circulating steroid hormones. This
inhibitory neurons as part of a natural process of maturation. regulation can be in a negative direction, with the net effect
being a decrease in the number of immature granule neurons,
9.3.1 Turnover of Dentate Gyrus Granule Cells or in a positive direction, with the net effect being an increase
in the number of new neurons.
Several thousand new neurons are generated in the hip-
pocampus every day in the young adult rat (Cameron and Negative Regulation: Glucocorticoids
McKay, 2001) (Box 94). Although this number declines with
age during adulthood, the phenomenon of adult neurogenesis Several studies have indicated that glucocorticoids inhibit the
appears to remain robust until animals become aged (Kuhn et production of new neurons by decreasing the proliferation of
al., 1996; Cameron and McKay, 1999). The addition of such a granule cell precursors. During development and adulthood,
 Structural Plasticity 327

Box 93
Methods for Visualizing New Neurons

To show that a new neuron has been produced in the adult brain, cell division and differentia-
tion of the progeny into neurons must be demonstrated. The traditional method for demon-
strating mitosis in the brain is 3H-thymidine autoradiography. For in vivo studies, this
technique requires injecting 3H-thymidine into live animals (usually intraperitoneally) and
then examining the brains at various survival times after injection. 3H-thymidine is taken up by
cells that are undergoing DNA synthesis in preparation for division. 3H-thymidine can be used
as a marker of proliferating cells (at short survival times) and their daughter cells (at longer
survival times). 3H-thymidine is a powerful method, but it has drawbacks, including lack of
compatibility with stereological methods for cell counting and inability to determine with cer-
tainty whether new cells are co-labeled with cell type-specic markers.
A newer method, labeling with the thymidine analogue bromodeoxyuridine (BrdU), has
been used more recently to investigate adult neurogenesis. BrdU is also incorporated into cells
during S phase and can be used to label proliferating cells or their progeny. BrdU can be visual-
ized using immunohistochemistry and has the advantage of being amenable to stereological
estimates of the total number of new cells and the unequivocal demonstration that new cells
are co-labeled with cell type-specic markers.
Evidence that new cells have differentiated into neurons can be achieved by combining 3H-
thymidine or BrdU with the following.
1. Electron microscopy: 3H-thymidine autoradiography has been combined with electron
microscopy to demonstrate synapses on cell bodies and dendrites of new cells in the den-
tate gyrus (Kaplan and Bell, 1984). BrdU labeling is less compatible with electron micro-
scopic methods due to the harsh pretreatments necessary for staining.
2. Retrograde tracing: 3H-thymidine autoradiography and BrdU labeling have been com-
bined with retrograde tracer injections to show that new cells in the dentate gyrus extend
axons into the CA3 region (Staneld and Trice, 1988; Hastings and Gould, 1999; Markakis
and Gage, 1999). This method is particularly useful when combined with BrdU because
confocal microscopy can be used to verify double labeling in orthogonal planes.
3. Immunohistochemistry for neuronal markers: 3H-thymidine autoradiography and BrdU
labeling have been combined with immunolabeling for cell type-specic markers. These
studies have shown that most cells made in the dentate gyrus of adult animals express
neuronal markers (Kuhn et al., 1996; Cameron and McKay, 2001; Gould et al., 2001). This
method is particularly useful when combined with BrdU because confocal microscopy
can be used to assess double labeling throughout the extent of the cell.
4. Electrophysiology: There is evidence that new cells have the electrophysiological charac-
teristics of neurons. This has been gained by identifying new cells with BrdU staining after
electrophysiological characterization (Liu et al., 2003), by labeling new cells with GFP-
tagged retrovirus (van Praag et al., 2002), and by recording from cells that are likely to be
adult-generated on the basis of their location in the granule cell layer and staining for
markers of immature neurons (Wang et al., 2000, Snyder et al., 2001). All of these
approaches have revealed that new cells produced in the dentate gyrus during adulthood
have the electrophysiological properties of neurons (see also Schmidt-Hieber et al., 2004
for the electrophysiological signature of newly born cells).

treating rat pups with corticosterone decreases the production may reect naturally occurring processes relevant to hip-
of new granule cells (Gould et al., 1991; Cameron and Gould, pocampal function. A good example may be in observed in
1994; Ambrogini et al., 2002). During adulthood, removal of the dentate gyrus of the meadow vole. Meadow voles are sea-
circulating adrenal steroids by adrenalectomy increases the sonal breeders, and the breeding season is associated with
proliferation of granule cell precursors and ultimately the higher levels of circulating glucocorticoids in females; cell
production of new neurons (Gould et al., 1992; Cameron and proliferation is suppressed at this time (Galea and McEwen,
Gould, 1994). This phenomenon also occurs in aged animals 1999).
in which granule cell production has diminished. Removal of
adrenal steroids restores young adult levels of cell prolifera- Positive Regulation: Estrogen
tion in the dentate gyrus of old rats (Cameron and McKay,
1999; Montaron et al., 2005). The ovarian steroid estradiol has a positive effect on the pro-
The levels of circulating glucocorticoids change at various duction of new cells in the dentate gyrus of adult rats
times of development and with experience, presenting the (Tanapat et al., 1999, 2005; Banasr et al., 2001; Ormerod et al.,
possibility that glucocorticoid inhibition of cell proliferation 2003). Removal of estrogen by ovariectomy decreases the pro-
328 The Hippocampus Book

9.3.3. Experience and Adult Neurogenesis

Box 94
Methodological Issues Related
A second major regulatory factor governing neurogenesis is
to the BrdU Technique
experience. Environmental experience takes many forms: sen-
The introduction of BrdU labeling has signicantly advanced sory and motor experience, changing cognitive demands, and
our understanding of adult neurogenesis. However, it has stress.
become clear that our understanding of this technique was
based on erroneous assumptions derived from earlier devel- Negative Regulation: Stress
opmental studies and thus, the phenomenon of neurogenesis
continues to be underestimated and, perhaps, misinterpreted. The catabolic actions of stress and stress hormones have long
Cameron and McKay (2001) have shown that been recognized. A major component of the ght or ight
the modal dose of BrdU for studies of adult neurogenesis reaction is the mobilization of energy stores to facilitate reac-
(50 mg/kg) labels fewer than 50% of the total cells in S
tion to imminent danger (Selye, 1976). This response inhibits
phase. In contrast, when used during development, this
costly processes (e.g., growth) that are only important for the
dose appears to label all cells in S phase. Although BrdU is
a small molecule, differences in metabolism or blood-brain future. It should come as no surprise, then, that stress inhibits
barrier (BBB) permeability prevent the adult animal from the production of new granule cells during development and
incorporating it at the same rate as the developing animal. in adulthood (Gould and Tanapat, 1999). This phenomenon
Higher doses of BrdU (up to 600 mg/kg) label more prolifer- has been demonstrated in three mammalian species using a
ating cells, do not label nonspecically, and are not toxic to range of stressors. Exposure of postnatal and adult rats to the
the cells or animals. Using higher doses of BrdU, it appears odors of natural predators, electric shock, or restraint
that approximately 9000 new cells are generated every day decreases the rate of cell proliferation in the dentate gyrus
in the dentate gyrus of the young adult rat. Divisions of (Tanapat et al., 2001; Malberg and Duman, 2003; Pham et al.,
the progenitor cells are presumed to be asymmetrical. Thus, 2003; Westenbroek et al., 2004). Similar studies have been car-
~4500 new cells are produced daily; and approximately
ried out in adult marmosets and adult and aged tree shrews
80% of these cells (~3600) exhibit characteristics of imma-
and marmosets using stress paradigms of resident-intruder
ture neurons. Over the course of 1 month (30 days), the
number of immature neurons generated would be ~108,000. (Gould et al., 1998) and social subordination (Gould et al.,
Estimates of the number of granule cells in the adult rat 1997; Czeh et al., 2001; Simon et al., 2005). In all cases, stress
range from 1.5 million to 2.0 million. Given these numbers, has been shown to decrease the proliferation of granule cell
a conservative estimate of the percentage of immature neu- precursors and ultimately the production of immature neu-
rons produced in a month of a young adult rats life is rons in the dentate gyrus. This effect appears to be region-
about 5%. specic in that changes in the number of BrdU-labeled cells
These ndings have two important implications. First, were not observed in the subventricular zone (another site of
the number of new neurons produced during adulthood is neurogenesis) following stress.
substantially higher than was originally predicted. Second, Interestingly, prenatal stress has a persistent impact on the
studies using nonsaturating doses of BrdU to investigate the
later production of granule cells. In both rodents and mon-
effects of experimental manipulations on neurogenesis may
keys, prenatal stress results in persistent dampening of the
be detecting changes in BrdU availability (e.g., due to alter-
ations in blood ow or changes in the BBB) in contrast to proliferation of granule cell precursors during adulthood (in
changes in the actual number of new cells (Gould and rats: Lemaire et al., 2000) and the late juvenile period (in
Gross, 2002). monkeys: Coe et al., 2003). Furthermore, the postnatal stress
of maternal separation results in persistent suppression of cell
proliferation and immature neuron production in adult rats
liferation of new cells, and replacement of estradiol to ovariec- (Mirescu et al., 2004). Adult rats subjected to maternal sepa-
tomized animals restores the rate of cell proliferation. ration early in life not only exhibit impaired neurogenesis but
Moreover, a natural uctuation in the proliferation of granule they also exhibit no changes in adult neurogenesis when
cell precursors is evident across the estrous cycle; cell prolifer- exposed to stress during adulthood. As indicated earlier, expo-
ation is highest during proestrus, the time of maximal estro- sure to predator odor decreases adult neurogenesis in control
gen levels. The stimulation of cell proliferation by estrogen animals (this effect is absent in maternally separated animals).
results in a sex difference that favors females in the production Prenatal and early postnatal stress are associated with persist-
of immature neurons. However, estrogen-related increases in ent changes in the hypothalamic-pituitary-adrenal (HPA)
the production of new cells are transient: By a few weeks after axis, including the development of an inefficient shut-down
BrdU injection, no difference in the number of new cells can mechanism for glucocorticoids. These ndings present the
be detected. As with estrogen-related differences in spine den- possibility that new cell production is persistently inhibited in
sity, similar differences in new neuron production may be animals exposed to early life stressors because these animals
confounded by the manner in which laboratory animals are either experience higher than normal levels of glucocorticoids
housed. Estrogen may further increase the production of new or are hypersensitive to normal levels of these hormones.
neurons by enhancing their survival (Ormerod et al., 2004). Indeed, lowering glucocorticoid levels below control values by
 Structural Plasticity 329

adrenalectomy and low-dose corticosterone replacement has environment paradigm; juvenile, young adult, and old
been shown to restore the levels of cell production in the rodents living in enriched environments maintained more
dentate gyrus of maternally separated animals to control val- new granule cells than those living in standard, deprived
ues (Mirescu et al., 2004). This is likely due to the fact that laboratory cages (Kempermann et al., 1997, 1998, 2002;
maternal separation results in lowered levels of corticosteroid- Nilsson et al., 1999; Brown et al., 2003). These studies may
binding globulin, which modulates the amount of free corti- indicate that standard control animals, be they birds living in
costerone that enters the brain. aviaries or rodents living in laboratory cages, are living in an
Collectively, these ndings support the view that the pro- extremely deprived state (Box 95). Given the elaborate social
duction of immature neurons in the dentate gyrus varies and cognitive abilities of primates, it is likely that monkeys
depending on exposure to stressful experiences. In fact, the living in standard laboratory cages are experiencing an even
pool of immature neurons in the dentate gyrus may be an more deprived set of experiences than less complex verte-
important substrate by which stress exerts long-lasting effects brates.
on hippocampal function. However, to understand the poten- There are many variables that differ between living in the
tial impact of stress on hippocampal structure and function, it wild and living in captivity or, in captivity, between living in
may be necessary to examine these issues in animals living in an enriched environment and living in a standard laboratory
more natural environments. Animals living in laboratory- cage. Among them are stress, social interaction, physical activ-
controlled conditions appear to lose a larger proportion of ity, and learning. In addition to reporting that birds living in
adult-generated cells than those living in more complex envi- complex environments have more new hippocampal neurons,
ronments (Kempermann et al., 1998; Nilsson et al., 1999; a seasonal difference in new neuron number correlates with a
Tanapat et al., 2001). Thus far, the effects of chronic stress seasonal difference in seed caching and retrieval. Because
while living in naturalistic, relatively complex environments these behaviors are likely to include spatial navigation, a func-
have not been explored. tion attributed to the hippocampal formation (see Chapter
13), a causative link between learning and new neurons has
Positive Regulation: Environmental Complexity, been proposed. It has been shown that the number of new
Learning, and Physical Activity hippocampal neurons in the avian hippocampus can be
increased by engaging birds in learning tasks that involve this
The rst evidence that environmental complexity inuences
new hippocampal neurons in the adult was published by
Barnea and Nottebohm (1994). This study showed that black-
capped chickadees living in the wild maintained a higher
Box 95
number of new hippocampal neurons than those living in
Environmental Enrichment Versus
captivity. Work done on rodents also indicates a greater
Environmental Deprivation
amount of cell proliferation and neurogenesis in the hip-
pocampus of animals living in the wild compared to those liv- Numerous studies have investigated structural changes in the
ing in the laboratory (Amrein et al., 2004a,b). hippocampal formation associated with rearing or living in
Adult neurogenesis as well as the entire volume of the hip- enriched environments. These studies have found several
pocampal formation also changes over the seasons in some differences in the cortex, hippocampus, and other brain
vertebrates. For example, the volume of the avian hippocam- regions of animals living in the more complex environment
pal formation uctuates across the seasons in seed-caching compared to what is observed in animals living in standard
birds such as the black-capped chickadee (Clayton, 1995, laboratory cages (Rosenzweig et al., 1962; Icanco and
Greenough, 2000; van Praag et al., 2000). In the case of
1998; Smulders et al., 2000) and may be related to behaviors
rodents, the enriched environment typically consists of a
specic to these animals (see Chapter 13). Indeed, the lack of
larger living space, more conspecics, toys, tunnels, exercise
a change in the volume of this structure in birds that do not equipment, and often a more varied diet than the controls.
store food seasonally supports this view (Lee et al., 2001). This Living in this type of environment enhances performance on
volumetric difference seems to be attributable to changes in certain hippocampus-dependent learning tasks and increases
the number of hippocampal neurons, with peak numbers the hippocampal volume, number of dendrites of hippocam-
occurring in the fall. Although sex differences in hippocampal pal neurons, number of glial cells, and number of new gran-
volume have been reported in some species of food-storing ule cells compared to controls. The most commonly applied
mammals, such as gray squirrels, no seasonal changes have yet interpretation of these ndings is that hippocampal structure
been observed (Lavenex et al., 2000). It may be that the phe- and function are responsive to environmental complexity. A
nomenon does not occur in mammals, or that naturally living converse interpretation is that they are deprivation effects;
that is, the brain and behavior of control animals in the stan-
species that occupy ecological niches with large changes in
dard cages are abnormal because of the absence of normal
seasonal temperature and food availability need to be studied stimulation. This puzzle could be solved by providing evi-
more closely. dence for differences in animals living in super-complex ver-
Similar ndings linking adult neurogenesis to experience sus complex environments.
were reported in the mouse and rat using an enriched
330 The Hippocampus Book

brain region (Patel et al., 1997; Clayton, 1998). Studies have fact that running causes activation of the HPA axis (Droste et
shown a similar phenomenon in rodents; learning tasks al., 2003).
thought to depend on the integrity of the hippocampal Although voluntary running might be considered a posi-
formation enhance the number of new granule cells (Gould et tive stressor (Selye, 1976), forced running is likely to have a
al., 1999a; Ambrogini et al., 2000; Lemaire et al., 2000; more aversive component. Yet, both types of experience ulti-
Dobrossy et al., 2003; Leuner et al., 2004; Hairston et al., mately enhance adult neurogenesis. A few studies have
2005; Olariu et al., 2005), whereas other types of learning may demonstrated that high levels of running inhibit adult neuro-
not have such an effect. For example, trace but not delay eye- genesis, but these negative effects appear to be specic to cer-
blink conditioning increases the number of new granule cells tain strains of animals (Naylor et al., 2004). The ability of this
in the dentate gyrus (Gould et al., 1999a; Leuner et al., 2004). stimulus to override the negative actions of glucocorticoids is
This increase in the number of new neurons persists until intriguing, and future experimentation is necessary to eluci-
at least 2 months after mitosis (Leuner et al., 2004) suggest- date the mechanisms that enable running-induced neurogen-
ing that the trophic effects of learning on new cells esis despite activation of the HPA axis. It should be noted that
are not eeting. Early training on tasks that require the hip- there is an obvious hedonic component to running in rodents.
pocampal formation also enhances the number of new neu- Rodents are highly motivated to run and do so without addi-
rons, provided that learning has occurred; a signicant tional incentives. In addition, rodents accustomed to running
correlation exists between the number of BrdU-labeled cells exhibit physiological and behavioral symptoms of withdrawal
in the dentate gyrus and the percentage of conditioned when access to a running wheel is denied. The extent to which
responses in animals that have not yet fully acquired the con- these ndings can be generalized to other animals that do not
ditioning task (Leuner et al., 2004). However, some studies have an internal drive to run remains to be determined.
have failed to nd a signicant increase in the number of new Some work suggests that the increase in neurogenesis asso-
neurons following hippocampus-dependent learning (van ciated with running leads to enhanced LTP as well as
Praag et al., 1999b; Snyder et al., 2005; Van der Borght et al., improved performance on hippocampus-dependent learning
2005), whereas others have reported decreased numbers of tasks (van Praag et al., 1999b). Although this evidence suggests
newly generated cells with learning (Dobrossy et al., 2003; a positive correlation between adult neurogenesis and learn-
Olariu et al., 2005; Pham et al., 2005). The reasons for these ing, it is important to emphasize that running also alters other
discrepancies remain unresolved; possible factors include neural parameters in the hippocampus (Eadie et al., 2005).
variations in learning paradigms, different time courses Thus, like the case for stress and enriched environments,
between labeling new cells and training, and species and sex changes in adult neurogenesis may contribute to altered hip-
differences (reviewed by Abrous et al., 2005). pocampal function, but they may not be the only cellular
Because the hippocampus-dependent versions of the mediators of these effects.
watermaze and eyeblink conditioning take more trials to
acquire than the hippocampus-independent versions of these Individual Differences in Adult Neurogenesis
same tasks, it is possible that new cells respond especially to
tasks that are more difficult. Although no study has yet Individual differences in the acquisition of learning tasks
matched the task difficulty of hippocampus-dependent and abound, even among animals of the same strain housed under
hippocampus-independent learning paradigms and examined the same conditions. Differences in the rate of acquisition of
new cell numbers, it is potentially instructive that place learn- the trace eyeblink conditioned response are correlated posi-
ing in the standard reference memory version of the water- tively with the number of new neurons in the dentate gyrus
maze is considerably easier (in terms of number of trials (Leuner et al., 2004).
required to reach the criterion) than delay eyeblink condi- Individual differences also exist in reaction to novelty. Rats
tioning. Thus, a simple relation to task difficultywhere this that are highly reactive to novelty exhibit lower levels of cell
is operationally dened in terms of time or extent of training proliferation, and therefore of neurogenesis or survival, in the
required for learningdoes not seem to prevail. dentate gyrus than their less reactive counterparts (Lemaire et
Unlike the unsettled case for a link between learning and al., 1999). Moreover, individual differences in aggressive
neurogenesis, there is now unambiguous evidence that physi- behavior exist and appear to correlate with the rate of adult
cal activity increases cell proliferation and ultimately the pro- neurogenesis. When adult rodents are group-housed in a rel-
duction of new neurons in the dentate gyrus of adult rodents. atively large, complex enclosure, a dominance hierarchy
Numerous studies have demonstrated that voluntary activity forms. Dominant animals, characterized by the amount of
in a running wheel increases adult neurogenesis in rats and offensive or aggressive behavior displayed, have substantially
mice (van Praag et al., 1999ab; Fabel et al., 2003; Rhodes et al., more new neurons than subordinate animals. Once the dom-
2003; Farmer et al., 2004; Holmes et al., 2004; Naylor et al., inance hierarchy is formed, the difference in the number of
2004; Persson et al., 2004; Bjornebekk et al., 2005). Even new neurons persists regardless of whether animals continue
forced exercise, via running on a treadmill, has been shown to to live in the complex environment (Kozorovitskiy and Gould,
induce this effect (Trejo et al., 2001; Kim et al., 2004). This 2004). These results again raise the issue of whether the natu-
robust phenomenon is particularly surprising in light of the ral expression of structural plasticity in a given species is
 Structural Plasticity 331

thwarted under articial laboratory conditions. Differences in et al., 2001; Kee et al., 2001; Kokaia and Lindvall, 2003;
cell proliferation in the hippocampus of mountain chickadees Scharfman, 2004). Damage-induced neurogenesis can also be
depending on their dominance status have also been reported. observed in the CA elds of the hippocampal formation;
Subordinate chickadees make fewer new cells in the hip- ischemic damage induces the production of new hippocampal
pocampus than dominant chickadees (Pravosudov and pyramidal cells, cells not normally thought to undergo adult
Omanska, 2005), again suggesting that individual experiences neurogenesis (Rietze et al., 2000; Nakatomi et al., 2002).
may increase variability on this measure. The extent to which damage-induced neurogenesis con-
tributes to functional recovery remains unknown (Scharfman,
9.3.4 Neurogenesis Following Damage 2004), but some evidence suggests that neurogenesis is upreg-
ulated in neurodegenerative conditions such as Alzheimers
Another question raised by the phenomenon of neurogenesis disease (Jim et al., 2004) and Parkinsons disease (Yoshimi
is whether the ongoing processes of cell death and cell prolif- et al., 2005). One report suggested that there is a functional
eration in the dentate gyrus are causally linked. There are sev- relation between regenerative adult neurogenesis and re-
eral grounds to suspect that they are. First, endocrine covery insofar as blocking cell proliferation with irradiation
manipulations that alter survival of granule cells also affect prevents the formation of new synapses after ischemia (Wang
the production of new cells in a direction that tends to keep et al., 2005).
the total number of cells constant. For example, removal of Damage-induced neurogenesis may occur via at least two,
circulating adrenal steroids by adrenalectomy results in mas- not mutually exclusive, mechanisms.
sive death of granule cells (Sloviter et al., 1989; Gould et al.,
1. Dying neurons send signals to progenitors to divide at
1990) as well as enhanced proliferation of granule cell precur-
a faster rate.
sors (Gould et al., 1992; Cameron and Gould, 1994). Second,
2. Death of mature neurons releases progenitor cells from
a specic lesion of the granule cell layer, by mechanical or
inhibition by eliminating a neurogenesis-inhibitory
excitotoxic means, results in a compensatory increase in the
signal.
proliferation of granule cell precursors and ultimately the
production of new neurons (Gould and Tanapat, 1997). Both of these mechanisms have been identied in other
Third, experimental conditions associated with granule cell systems (Hastings and Gould, 2003) (Fig. 93), but the rele-
death, such as ischemia and seizures, result in increased gran- vance of these results to the adult hippocampal formation
ule cell neurogenesis (Parent et al., 1997; Scott et al., 2000; Jim remains unknown.

Figure 93. Cell proliferation is

enhanced with neuronal damage: alter-
native mechanisms. A. In the intact adult
brain, differentiated neurons do not
affect the rate of cell proliferation; how-
ever, dying neurons can release mole-
cules that stimulate nearby precursors
to divide. B. In the intact adult brain,
differentiated neurons release molecules
that suppress cell proliferation. Neuronal
death removes this inhibition, permitting
a higher rate of cell proliferation after
damage.
332 The Hippocampus Book

9.3.5 Unusual Features of larizing potentials and respond to -aminobutyric acid

Adult-Generated Neurons (GABA) with depolarization instead of hyperpolarization.
These data from hippocampal slices taken at different stages of
All cells are not created equal. Not only are there the well development raise the possibility that adult-generated cells are
known differences between different cell types, there are also transiently developmental in nature, at least for a short time
developmental differences between individual cells of a com- after their generation. Indeed, it has been reported that imma-
mon cell type. The granule cell layer consists of neurons rang- ture granule cells in the dentate gyrus of the adult rat have
ing in age from hours to years. Although many thousands of more robust LTP than mature granule cells (Wang et al., 2000;
new neurons are added to the dentate gyrus every day, this is Schmidt-Hieber et al., 2004). Moreover, this enhanced LTP
a relatively small proportion of the large number of mature cannot be inhibited by GABA agonists, unlike LTP elicited
granule neurons produced during development and that sur- from mature granule cells (Snyder et al., 2001; Schmidt-
vive through to adulthood (the total number of granule neu- Hieber et al., 2004). These ndings support the view that a rel-
rons is approximately one to two million in the adult rat). For atively small population of immature granule cells could exert
the newly born cells to have an impact on adult hippocampal a substantial functional inuence over the hippocampal for-
function, they must have unique functional properties that mation. However, research into this issue remains in its
empower them in the face of the numerical superiority of infancy; and much work is needed to determine the functional
mature granule cells. Growing evidence suggests that the characteristics of new neurons in the dentate gyrus.
newly born cells may be functionally stronger.

Structural Characteristics
of Adult-Generated Neurons 9.4 Possible Functions of New Neurons

Adult-generated neurons are likely to have structural charac- The daily production of thousands of new granule cells and
teristics that differ from developmentally generated neurons. their incorporation into the existing circuitry are costly in
These structural differences may enable a relatively small terms of energy expenditure. It is likely that these new cells are
number of immature neurons to exert a disproportionate not produced merely to replace older cells that die because a
effect on hippocampal physiology. For example, adult- substantial amount of cell death occurring in this region
generated granule cells are capable of undergoing rapid struc- involves the adult-generated population. The most intuitive
tural change, as evidenced by the fact that they have axons in explanation for the continual inux of new cells is that it pro-
the distal CA3 within 4 to 10 days after mitosis (Hastings and vides an important mechanism for some aspect of the func-
Gould, 1999). This suggests a great capacity to form new con- tion of the dentate gyrus that cannot be obtained with a
nections under the appropriate environmental conditions. structure comprising mature neurons exclusively.
Retrograde tracer studies have also suggested differences in the
axon terminal elds of granule neurons produced at different 9.4.1 A Possible Role in Learning?
developmental time points. Some granule cells produced dur-
ing development have axons that diverge longitudinally in the The existence of a large pool of immature neurons in a brain
CA3 region, whereas no such adult-generated granule cells region that is important for certain types of learning and
have been observed (Hastings et al., 2002). Detailed informa- memory raises the possibility that these new cells participate
tion about the sequence of synaptogenesis onto newly gener- in these functions. Although most theories of learning do not
ated granule cells in the dentate gyrus is not yet available. involve new neurons, many ideas about hippocampal function
However, because adult-generated granule cells are produced are compatible with such a possibility. Chapter 13 discusses
in the deep part of the granule cell layer, their dendritic trees the wide range of current theories of hippocampal function.
do not extend as far into the molecular layer as those generated One theory is that the hippocampal formation acts to associ-
during development. This presents the possibility that adult- ate discontiguous events (i.e., stimuli that are separated by
generated granule cells, because of their location in the layer, space or time) (Wallenstein et al., 1998). A continual inux of
receive a different proportion of synaptic inputs from various new neurons accompanied by rapid synapse formation could
sources. Clearly, more research is needed to determine the play a role in connecting stimuli separated spatially or tempo-
extent to which cells produced at different time points form rally. A widely held view is that the hippocampal formation
different types of circuits in the hippocampal formation. plays a temporary role in storing new memories. This theory
is consistent with data showing that recent, but not remote,
Functional Characteristics memories for certain information are eliminated by lesioning
of Adult-Generated Neurons the hippocampal formation (Kim et al., 1995; Zola and Squire,
2001). A rapidly changing population of adult-generated neu-
Immature neurons have electrophysiological characteristics rons with unique functional properties may serve as a sub-
which differ from those of mature neurons. For example, strate for this temporary role of the hippocampal formation
developing neurons have been shown to exhibit giant depo- in information storage. Neurons produced during adulthood
 Structural Plasticity 333

might play a role in memory processing for a short time after ing only when imposed upon a particular neural background
their generation. The new cells might degenerate or undergo and under certain learning circumstances. For example, a
changes in connectivity, gene expression, or both around the decline in the new neuron number may fail to correlate with
time the hippocampal formation no longer plays a role in the learning impairment because other age-associated changes
storage of that particular memory. A temporary role for adult- dominate hippocampal function, such as diminished num-
generated granule neurons in learning has been suggested in bers of synapses (Morrison and Hof, 2002). When examining
canaries, in which seasonal changes in song correlate with the data related to the issue of adult neurogenesis and learning, it
transient recruitment of more new neurons into the relevant is important to note that new cells are likely to require some
circuitry (Barnea and Nottebohm, 1996; Nottebohm, 2002). time for integration into the existing circuitry. Dendrites must
However, a problem with taking this view of hippocampal be grown and functional synaptic connections formed on
memory is that it requires a cellular rather than a distributed them; axons must grow and nd their targets. Thus, acute
synaptic view of memory storage. Neural network models of changes in cell proliferation are not likely to inuence hip-
memory formation (see Chapter 14) support the distributed pocampal function until a certain amount of time has passed.
perspective. In addition, a temporary role for new neurons in Chronic changes in cell proliferation are those most likely to
learning implies that the new cells would persist only for as have functional consequences because their effects may be
long as the hippocampus remains necessary for retention of additive. This point is well illustrated by considering the acute
that information. The observation that learning enhances the and chronic effects of stress. Acute stress and chronic stress
number of new neurons for at least 2 months (Leuner et al., diminish the proliferation of granule cell precursors in the
2004)well beyond the time when the rodent hippocampal dentate gyrus. However, acute stress has been shown (with
formation is necessary for the maintenance of those memo- certain paradigms) to enhance learning, whereas chronic
riesargues against a clear role for new neurons in the tem- stress typically impairs learning (McEwen, 1999; Shors, 2001).
porary aspects of information storage. A direct link between new neurons and learning remains
Thus, although several theories of hippocampal learning controversial. Thus far, two approaches have been used to
are consistent with a possible role for new neurons, there are deplete new cells: antimitotic agents and irradiation. Focal
many details to be worked out. Do any data exist to support irradiation has been shown to decrease the number of new
such a possibility? There is now some evidence for a positive neurons almost completely, whereas treatment with the
correlation between the number of new hippocampal neurons antimitotic agent methylazoxymethanol acetate (MAM) for 2
and learning. For example, many factors and conditions that weeks has been shown to diminish the pool of new neurons by
increase the number of new granule neurons have been shown about 80% (Shors et al., 2000b, 2002; Bruel-Jungerman et al.,
to be associated with improved performance on hippocam- 2005). The rst obvious question is whether the depletion of
pus-dependent tasks, although the causal relation remains new neurons produced during adulthood would be sufficient
unproven (Gould et al., 1999a). Examples of factors that to inuence hippocampal function. Cameron and McKay
enhance both the number of new neurons and hippocampus- (2001) have shown that approximately 9000 new cells are pro-
dependent learning are estrogen treatment and living in an duced every day in the rat dentate gyrus. If we assume that cell
enriched environment. Similarly, situations that decrease the division in this population is asymmetrical, a maximum of
number of new hippocampal neurons, such as glucocorticoids 4500 new cells would be available to differentiate into neu-
and stress, are associated with impaired performance on these rons. Of this population, about 80% express markers of
tasks. immature neurons (3600). Assuming a 30-day month, with
However, some recent studies have found either no relation approximately 108,000 immature neurons generated during
between learning and the number of new neurons or, in fact, this time, complete depletion would eliminate about 100,800
an inverse relation. With regard to the rst case, aging is asso- immature neurons (> 5% of the total number of granule cells,
ciated with a decline in new cell production in the hippocam- which is about 1.52.0 million). Although this is not a pro-
pus, and yet this decrease does not appear to be always portionately large number, if, as indicated above, the granule
associated with impaired performance on hippocampus- cells produced during adulthood are functionally different
dependent tasks (Merrill et al., 2003). With regard to the sec- less likely to be inhibited by GABA and more likely to display
ond case, chronic stress has been shown to inhibit synaptic plasticitythey could be especially influential.
neurogenesis persistently in tree shrews; yet, this experience is Despite the fact that involvement of new neurons in learning
associated with improved, not impaired, performance on a is plausible, the available evidence supporting this view is
specic spatial navigation learning task (Bartolomucci et al., incomplete and mixed.
2002). Do these negative ndings indicate that new neurons Treatment with the antimitotic agent MAM has been
are not involved in learning? Perhaps, but they may indicate shown to impair performance on trace eyeblink condition-
that modulation of the rate at which new neurons accumulate ing (Shors et al., 2001b). Delay eyeblink conditioning, the
affects hippocampal function only under certain conditions. hippocampus-independent version of the same task, is unaf-
Because stress, aging, and environmental complexity inuence fected by MAM treatment. Replenishment of immature neu-
multiple brain measures in addition to neurogenesis, it is pos- rons after cessation of MAM treatment results in the recovery
sible that changes in the new neuron number relate to learn- of hippocampal function as indexed by normal trace eyeblink
334 The Hippocampus Book

conditioning. Additional studies have shown that focal irradi- health of the animals at the start of the study and the degree
ation (which stops cell proliferation and hence neurogenesis) to which they experience stress during treatment. However,
of the hippocampal formation of adult rodents impairs per- there is general agreement across groups that at high doses
formance of a place recognition task in a T maze but not of MAM is generally toxic. Nonspecic effects of treatment are
an object recognition task (Madsen et al., 2003); it also also a feature of irradiation. Although the possibility that spu-
impairs performance in a spatial version of the Barnes maze rious changes, in contrast to specic effects on adult neuroge-
(Raber et al., 2004). In addition, a role for new neurons in nesis, contribute to the learning decits with MAM treatment
enhanced learning associated with living in an enriched envi- cannot be completely ruled out, there is some evidence to sug-
ronment has been reported, using MAM to decrease the pro- gest this is not the case. The same regimen of MAM treatment
duction of neurons (Bruel-Jungerman et al., 2005). These that produces learning impairment does not cause weight
results are surprising given that studies of behavioral changes loss, liver damage, or structural changes to the brain (includ-
following lesions of the hippocampus rarely uncover a learn- ing the hippocampus and cerebellum) (Shors et al, 2000;
ing impairment unless a large percentage of the total hip- Bruel-Jungerman et al., 2005). In addition, MAM treatment
pocampal formation is destroyed (Moser et al., 1995). This does not prevent the hippocampus from having some normal
again raises the issue of whether the increased efficacy of electrophysiological responses, including induction of LTP in
adult-generated neurons compared to their developmentally the CA1 region.
generated counterparts imparts a relatively small number of Collectively, these ndings suggest that new neurons may
new neurons with substantial power to inuence hippocam- play a role in certain types of learning but not others.
pal function. Denitive information awaits the development of better
It is important to note that contradictory data exist. One methods for controlling neuronal proliferation and survival in
study revealed that other forms of hippocampus-dependent the absence of nonspecic effects that are functionally delete-
learning tasks are insensitive to manipulations that decrease rious. Transgenic mice designed to knock out new neurons
the number of new neurons. For example, spatial learning in a during specic times of adulthood would be useful tools for
watermaze is not impaired by MAM treatment or irradiation investigating the possible role of new neurons in learning. If
(Shors et al., 2002; Snyder et al., 2005). Moreover, a simple these animals exhibit normal learning and memory on a mul-
form of contextual fear conditioning (a learning paradigm titude of tasks, the function of new neurons will remain a
that involves the hippocampal formation) (Fanselow, 2000) is completely open question. One possibility is that the new neu-
unaffected by MAM treatment (Shors et al., 2002). Because rons participate in learning but only in a redundant fashion.
both spatial learning and context fear conditioning are both That is, the new neurons may be necessary for learning only
much more rapidly acquired than trace eyeblink conditioning, when other aspects of hippocampal function are compro-
the possibility that new cells are important only for training mised (e.g., following irradiation or treatment with an antim-
on difficult tasks arises, as previously indicated. This suggests itotic agent).
that either new dentate granule cells participate in only certain
types of hippocampus-dependent learning (e.g., those that 9.4.2 A Possible Role in Endocrine Regulation?
involve associating stimuli separated in time) or that some
action of the drug other than its effects on neuron production In addition to its widely recognized role as a brain region
is impairing trace eyeblink learning. However, the results from involved in learning and memory, the hippocampal formation
irradiation studies suggest that spatial memory, if not learn- plays a less publicized role in the regulation of endocrine
ing, is affected by depletion of new neurons. Prevention of function. As indicated earlier, the hippocampal formation is
neurogenesis results in an inability to retain spatial informa- known to have a relatively high concentration of steroid hor-
tion pertaining to watermaze learning, whereas learning itself mone receptors, in particular adrenal steroid receptors
remains unaffected (Snyder et al., 2005). (McEwen, 1999). The hippocampal formation has been linked
All methods utilized thus far to block adult neurogenesis in functionally to regulation of the HPA axis. In particular, it is
the dentate gyrus are likely to have some nonspecic effects. purported to play a role in limiting the levels of circulating
MAM is a nonspecic antimitotic agent (sometimes used for glucocorticoids following stress. For example, lesions of the
chemotherapy). Impairment of hippocampus-dependent hippocampal formation prevent the efficient shut-off of the
learning could be the result of the generalized toxic effects of HPA axis and the return of glucocorticoid levels to their basal
the drug. The efficacy of MAM appears to vary across studies, state (Herman et al., 1995).
with some investigators reporting larger depletions in new Research into the effects of early stressors on the develop-
neuron number than others (compare Shors et al., 2000 and ment of the HPA axis and adult neurogenesis has revealed
Bruel-Jungerman et al., 2005 with Dupret et al., 2005). In potential links between adult-generated neurons and hip-
addition, variability in the degree to which the drug is pocampal neuroendocrine function. That is, prenatal and
reported to be toxic at doses sufficient to block neurogenesis postnatal stress have been shown to result in persistent
exists (compare Shors et al., 2000 and Bruel-Jungerman et al., changes in the HPA axis in the form of inefficient shut-off of
2005 with Dupret et al., 2005). The reasons for these differ- the stress response during adulthood. Early stressors have
ences remain unknown but may be related to the overall additionally been shown to diminish cell proliferation in an
 Structural Plasticity 335

acute manner and to dampen permanently the production of REFERENCES

new neurons even during adulthood (Lemaire et al., 2000; Coe
et al., 2003). These ndings raise the as yet untested possibil- Abrous DN, Koehl M, LeMoal M (2005) Adult neurogenesis:
ity that adult-generated neurons play an important role in from precursors to network and physiology. Physiol Rev
regulating the HPA axis. Studies designed to examine the 85:523569.
effects of specic depletion of adult-generated neurons on the Altman J (1962) Are new neurons formed in the brains of adult
mammals? Science 135:11271128.
shut-off of the stress response might reveal evidence that
Altman J (1963) Autoradiographic investigation of cell proliferation
would add support to this view.
in the brains of rats and cats. Anat Rec 145:573591.
Observations of individuals suffering from psychiatric Altman J (1969) Autoradiographic and histological studies of post-
conditions, such as depression and posttraumatic stress natal neurogenesis. IV. Cell proliferation and migration in the
disorder, have revealed signs of HPA axis dysregulation and anterior forebrain, with special reference to persisting neuroge-
diminished hippocampal volume in magnetic resonance nesis in the olfactory bulb. J Comp Neurol 137:433457.
imaging (MRI) studies (Vythilingam et al., 2002). Although Altman J, Das GD (1965a) Autoradiographic and histological evi-
the mechanisms that underlie diminished hippocampal vol- dence of postnatal hippocampal neurogenesis in rats. J Comp
ume in these conditions are unknown, it is possible that Neurol 124:319335.
persistently decreased neurogenesis plays a role. Interestingly, Altman J, Das GD (1965b) Post-natal origin of microneurones in the
the psychiatric conditions are also associated with impaired rat brain. Nature 207:953956.
Altschuler RA (1979) Morphometry of the effect of increased experi-
learning, but the mechanisms that underlie these cognitive
ence and training on synaptic density in area CA3 of the rat hip-
changes remain unknown.
pocampus. J Histochem Cytochem 27:15481550.
Ambrogini P, Cuppini R, Cuppini C, Ciaroni S, Cecchini T, Ferri P,
9.4.3 A Possible Role in the Etiology Sartini S, Del Grade P (2000) Spatial learning affects immature
and Treatment of Depression? granule cell survival in adult rat dentate gyrus. Neurosci Lett
286:2124.
A growing body of evidence points to a role for new granule Ambrogini P, Orsini L, Mancini C, Ferri P, Barbanti I, Cuppini R
neurons in the etiology of depression. This idea grew out (2002) Persistently high corticosterone levels but not normal
of three ndings: (1) hippocampal volume is smaller in circadian uctuations of the hormone affect cell proliferation
depressed patients; (2) stress, which predisposes individuals in the adult rat dentate gyrus. Neuroendocrinology 76:366
to develop depression, inhibits hippocampal neurogenesis; 372.
and (3) antidepressants enhance hippocampal neurogenesis. Amrein I, Slomianka L, Lipp HP (2004a) Granule cell number, cell
These observations have led to the suggestion that some of death and cell proliferation in the dentate gyrus of wild-living
rodents. Eur J Neurosci 20:33423350.
the symptoms associated with depression may result from a
Amrein I, Slomianka L, Poletaeva II, Bologova NV, Lipp HP (2004b)
persistent decrease in new neuron formation. This idea has
Marked species and age-dependent differences in cell prolifera-
been tested more directly in a study examining the effects of tion and neurogenesis in the hippocampus of wild-living
focal irradiation on the behavioral effects of antidepressants rodents. Hippocampus 14:10001010.
in mice (Santarelli et al., 2003). Prevention of an antidepres- Andersen P, Soleng AF (1998) Long-term potentiation and spatial
sant-induced increase in adult neurogenesis blocks the training are both associated with the generation of new excita-
ability of this therapy to alter behavioral signs of anxiety. tory synapses. Brain Res Rev 26:353359.
These ndings suggest that the efficacy of antidepressants Banasr M, Hery M, Brezun JM, Daszuta A (2001) Serotonin mediates
may depend on their ability to restore neurogenesis to normal oestrogen stimulation of cell proliferation in the adult dentate
levels. To establish these links more conclusively, evidence gyrus. Eur J Neurosci 14:14171424.
that the hippocampus does indeed participate in the sympto- Barnea A, Nottebohm F (1994) Seasonal recruitment of hippocampal
neurons in adult free-ranging black-capped chickadees. Proc
matology of depression would be necessary. Furthermore,
Natl Acad Sci USA 91:1121711221.
conrmation that diminished hippocampal volumewhich
Barnea A, Nottebohm F (1996) Recruitment and replacement of hip-
has not been reported in all studies of depressed patients pocampal neurons in young and adult chickadees: an addition
and is extremely small even when it has been established to the theory of hippocampal learning. Proc Natl Acad Sci USA
(about a 5% reduction)is the result of reduced neurogene- 93:714718.
sis seems critical. Bartolomucci A, de Biurrun G, Czeh B, van Kampen M, Fuchs E
(2002) Selective enhancement of spatial learning under chronic
psychosocial stress. Eur J Neurosci 15:18631866.
ACKNOWLEDGMENTS Bedard A, Cossette M, Levesque M, Parent A (2002) Proliferating cells
can differentiate into neurons in the striatum of normal adult
Special thanks go to the editors for constructive criticism on several monkey. Neurosci Lett 328:213216.
drafts of the manuscript. Also, my thanks go to past and present Benraiss A, Chmielnicki E, Lerner K, Roh D, Goldman SA (2001)
members of my laboratory for their hard work and dedication, espe- Adenoviral brain-derived neurotrophic factor induces both
cially Yevgenia Kozorovitskiy for help with the chapter and providing neostriatal and olfactory neuronal recruitment from endoge-
gures. Finally, thanks to the NIH and NARSAD for funding some of nous progenitor cells in the adult forebrain. J Neurosci
the research described this chapter. 21:67186731.
336 The Hippocampus Book

Bernier PJ, Bedard A, Vinet J, Levesque M, Parent A (2002) Newly rodent and Alzheimers disease brain: a reactivation of develop-
generated neurons in the amygdala and adjoining cortex of mental mechanisms? Prog Brain Res 83:427434.
adult primates. Proc Natl Acad Sci USA 99:1146411469. Czeh B, Michaelis T, Watanabe T, Frahm J, de Biurrun G, van
Berry B, McMahan R, Gallagher M (1997) Spatial learning and mem- Kampen M, Bartolomucci A, Fuchs E (2001) Stress-induced
ory at dened points of the estrous cycle: effects on perform- changes in cerebral metabolites, hippocampal volume, and cell
ance of a hippocampal-dependent task. Behav Neurosci proliferation are prevented by antidepressant treatment with
111:267274. tianeptine. Proc Natl Acad Sci USA 98:1279612801.
Bjorklund A, Gage FH (1985) Neural grafting in animal models of Daniel JM, Dohanich GP (2001) Acetylcholine mediates the estro-
neurodegenerative diseases. Ann NY Acad Sci 457:5381. gen-induced increase in NMDA receptor binding in CA1 of the
Bjorklund A, Stenevi U (1984) Intracerebral neural implants: neu- hippocampus and the associated improvement in working
ronal replacement and reconstruction of damaged circuitries. memory. J Neurosci 21:69496956.
Annu Rev Neurosci 7:279308. Daniel JM, Fader AJ, Spencer AL, Dohanich GP (1997) Estrogen
Bjorklund A, Nilsson OG, Kalen P (1990) Reafferentation of the sub- enhances performance of female rats during acquisition of a
cortically denervated hippocampus as a model for transplant- radial arm maze. Horm Behav 32:217225.
induced functional recovery in the CNS. Prog Brain Res Dayer AG, Ford AA, Cleaver KM, Yassaee M, Cameron HA (2003)
83:411426. Short-term and long-term survival of new neurons in the rat
Bjornebekk A, Mathe AA, Brene S (2005) The antidepressant effect of dentate gyrus. J Comp Neurol 460:563572.
running is associated with increased hippocampal cell prolifer- Dayer AG, Cleaver KM, Abouantoun T, Cameron HA (2005) New
ation. n
I t J Neuropsychopharmacol 8:112. GABAergic interneurons in the adult neocortex and striatum
Briones TL, Suh E, Jozsa L, Rogozinska M, Woods J, Wadowska M are generated from different precursors. J Cell Biol 168:415427.
(2005) Changes in number of synapses and mitochondria in Deller T, Frotscher M (1997) Lesion-induced plasticity of central
presynaptic terminals in the dentate gyrus following cerebral neurons: sprouting of single bers in the rat hippocampus after
ischemia and rehabilitation training. Brain Res 1033:5157. unilateral entorhinal cortex lesion. Prog Neurobiol 53:687727.
Brown J, Cooper-Kuhn CM, Kempermann G, Van Praag H, Winkler Dobrossy MD, Drapeau E, Aurousseau C, LeMoal M, Piazza PV,
J, Gage FH, Kuhn HG (2003) Enriched environment and physi- Abrous DN (2003) Differential effects of learning on neuro-gen-
cal activity stimulate hippocampal but not olfactory bulb neu- esis: learning increases or decreases the number of newly born
rogenesis. Eur J Neurosci 17:20422046. cells depending on their birth date. oMl Psychiatry 8:974982.
Bruel-Jungerman E, Laroche S, Rampon C (2005) New neurons in Droste SK, Gesing A, Ulbricht S, Muller MB, Linthorst ACE, Reul JM
the dentate gyrus are involved in the expression of enhanced (2003) Effects of long-term voluntary exercise on the mouse
long-term memory following environmental enrichment. Eur J hypothalamic-pituitary-adrenocortical axis. Endocrinology 144:
Neurosci 21:513521. 3013023.
Buzsaki G, Czopf J, Kondakor I, Bjorklund A, Gage FH (1987) Dupret D, Montaron MF, Drapeau E, Aurousseau C, Le Moal M,
Cellular activity of intracerebrally transplanted fetal hippocam- Piazza PV, Abrous DN (2005) Methylazoxymethanol acetate
pus during behavior. Neuroscience 22:871883. does not fully block cell genesis in the young and aged dentate
Cameron HA, Gould E (1994) Adult neurogenesis is regulated gyrus. Eur J Neurosci 22:778783.
by adrenal steroids in the dentate gyrus. Neuroscience 61: Eadie BD, Redila VA, Christie BR (2005) Voluntary exercise alters the
203209. cytoarchitecture of the adult dentate gyrus by increasing cellu-
Cameron HA, McKay RD (1999) Restoring production of hippocam- lar proliferation, dendritic complexity, and spine density.
pal neurons in old age. Nat Neurosci 2:894897. J Comp Neurol 486:3947.
Cameron HA, McKay RD (2001) Adult neurogenesis produces a large Englund U, Bjorklund A, Wictorin K, Lindvall O, Kokaia M (2002)
pool of new granule cells in the dentate gyrus. J Comp Neurol Grafted neural stem cells develop into functional pyramidal
435:406417. neurons and integrate into host cortical circuitry. Proc Natl Acad
Chang PL, Isaacs KR, Greenough WT (1991) Synapse formation Sci 99:1708917094.
occurs in association with the induction of long-term potentia- Eriksson PS, Perlieva E, Bjork-Eriksson T, Alborn AM, Nordborg C,
tion in two-year-old rat hippocampus in vitro. Neurobiol Aging Peterson DA, Gage FH (1998) Neurogenesis in the adult human
12:517522. hippocampus. Nat eM d 4:13131317.
Chesler EJ, Juraska JM (2000) Acute administration of estrogen and Fabel K, Fabel K, Palmer TD (2003) VEGF is necessary for exercise-
progesterone impairs the acquisition of the spatial Morris induced neurogenesis. Eur J Neurosci 18:28032812.
watermaze in ovariectomized rats. Horm Behav 28:234242. Faherty CJ, Kerley D, Smeyne RJ (2003) A Golgi-Cox morphological
Clayton NS (1995) The neuroethological development of food- analysis of neuronal changes induced by environmental enrich-
storing memory: a case of use it, or lose it! Behav Brain Res ment. Dev Brain Res 141:5561.
70:95102. Fanselow MS (2000) Contextual fear, gestalt memories, and the hip-
Clayton NS (1998) Memory and the hippocampus in food-storing pocampus. Behav Brain Res 110:7371.
birds: a comparative approach. Neuropharmacology 37:441452. Farmer J, Zhao X, Gage FH, Christie BR (2004) Effects of voluntary
Coe CL, Kramer M, Czeh B, Gould E, Reeves AJ, Kirschbaum C, exercise on synaptic plasticity and gene expression in the den-
Fuchs E (2003) Prenatal stress diminishes neurogenesis in the tate gyrus of adult male Sprague-Dawley rats in vivo.
dentate gyrus of juvenile rhesus monkeys. Biol Psychiatry 54: Neuroscience 124:7179.
10251034. Forster E, Naumann T, Deller T, Straube A, Nitsch R, Frotscher M
Cooke BM, Woolley CS (2005) Gonadal hormone modulation of (1997) Cholinergic sprouting in the rat fascia dentata after
dendrites in the mammalian CNS. J Neurobiol 64:3446. entorhinal lesion is not linked to early changes in neurotrophin
Cotman CW, Geddes JW, Kahle JS (1990) Axon sprouting in the messenger RNA expression. Neuroscience 80:731739.
 Structural Plasticity 337

Fowler CD, Johnaon F, Wang Z (2005) Estrogen regulation of cell pocampal and neocortical neurons in macaques have a transient
proliferation and distribution of estrogen receptor-alpha in the existence. Proc Natl Acad Sci USA 98:1091010917.
brains of adult female prairie and meadow voles. J Comp Neurol Hairston IS, Little MT, Scanlon MD, Barakat MT, Palmer TD,
489:166179. Sapolsky RM, Heller HC (2005) Sleep restriction suppresses
Fricker RA, Carpenter MK, Winkler C, Greco C, Gates MA, neurogenesis induced by hippocampus-dependent learning. J
Bjorklund A (1999) Site-specic migration and neuronal differ- Neurophysiol 94:42244233.
entiation of human neural progenitor cells after transplantation Hao J, Janssen WG, Tang Y, Roberts JA, McKay H, Lasley B, Allen PB,
in the adult rat brain. J Neurosci 19:59906005. Greengard P, Rapp PR, Kordower JH, Hof PR, Morrison JH
Frotscher M, Heimrich B, Deller T (1997) Sprouting in the hip- (2003) Estrogen increases the number of spinophilin-
pocampus is layer-specic. Trends Neurosci 20:218223. immunoreactive spines in the hippocampus of young and aged
Gage FH, Bjorklund A (1986) Enhanced graft survival in the hip- female rhesus monkeys. J Comp Neurol 465:540550.
pocampus following selective denervation. Neuroscience 17: Hastings NB, Gould E (1999) Rapid extension of axons into the CA3
8998. region by adult-generated granule cells. J Comp Neurol 413:
Gage FH, Bjorklund A, Stenevi U (1983a) Local regulation of com- 146154.
pensatory noradrenergic hyperactivity in the partially dener- Hastings NB, Gould E (2003) Neurons inhibit neurogenesis. Nat eM d
vated hippocampus. Nature 303:819821. 9:264266.
Gage FH, Bjorklund A, Stenevi U (1983b) Reinnervation of the par- Hastings NB, Seth MI, Tanapat P, Rydel TA, Gould E (2002) Granule
tially deafferented hippocampus by compensatory collateral neurons generated during development extend divergent axon
sprouting from spared cholinergic and noradrenergic afferents. collaterals to hippocampal area CA3. J Comp Neurol
Brain Res 268:2737. 452:324333.
Galea LA, McEwen BS (1999) Sex and seasonal differences in the rate Herman JP, Cullinan WE, Morano MI, Akil H, Watson SJ (1995)
of cell proliferation in the dentate gyrus of adult wild meadow Contribution of the ventral subiculum to inhibitory regulation
voles. Neuroscience 89:955964. of the hypothalamo-pituitary-adrenocortical axis. J Neuroen-
Gibbs RB (1999) Estrogen replacement enhances acquisition of a docrinol 7:475482.
spatial memory task and reduces decits associated with hip- Holmes MM, Galea LA, Mistlberger RE, Kempermann G (2004) Adult
pocampal muscarinic receptor inhibition. Horm Behav 36: hippocampal neurogenesis and voluntary running activity: cir-
222233. cadian and dose-dependent effects. J Neurosci Res 76:216222.
Gould E, Gross CG (2002) Adult neurogenesis: some progress and Icanco TL, Greenough WT (2000) Physiological consequences of
problems. J Neurosci 22:619623. morphologically detectable synaptic plasticity: potential uses
Gould E, Tanapat P (1997) Lesion-induced proliferation of neuronal for examining recovery following damage. Neuropharmacology
progenitors in the dentate gyrus of the adult rat. Neuroscience 39:765776.
80:427436. Jim K, Minami M, Lan JQ, Mao XO, Batteur S, Simon RP, Greenberg
Gould E, Tanapat P (1999) Stress and hippocampal neurogenesis. DA (2001) Neurogenesis in the dentate subgranular zone and
Biol Psychiatry 46:14721479. rostral subventricular zone after focal cerebral ischemia in the
Gould E, Woolley CS, McEwen BS (1990) Short-term glucocorticoid rat. Proc Natl Acad Sci USA 98:47104715.
manipulations affect neuronal morphology and survival in the Jim K, Peel AL, Mao XO, Xie L, Cottrell BA, Henshall DC, Greenberg
adult dentate gyrus. Neuroscience 37:367375. DA (2004) Increased hippocampal neurogenesis in Alzheimers
Gould E, Woolley CS, Cameron HA, Daniels DC, McEwen BS (1991) disease. Proc Natl Acad Sci USA 101:343347.
Adrenal steroids regulate postnatal development of the rat den- Juraska JM, Henderson C, Muller J (1984) Differential rearing expe-
tate gyrus. II. Effects of glucocorticoids and mineralocorticoids rience, gender, and radial maze performance. Dev Psychobiol
on cell birth. J Comp Neurol 313:486493. 17:209215.
Gould E, Cameron HA, Daniels DC, Woolley CS, McEwen BS (1992) Juraska JM, Fitch JM, Henderson C, Rivers N (1985) Sex differences
Adrenal hormones suppress cell division in the adult rat dentate in the dendritic branching of dentate granule cells following dif-
gyrus. J Neurosci 12:36423650. ferential experience. Brain Res 333:7380.
Gould E, McEwen BS, Tanapat P, Galea LA, Fuchs E (1997) Kalen P, Cenci MA, Lindvall O, Bjorklund A (1991) Host brain regu-
Neurogenesis in the dentate gyrus of the adult tree shrew is lation of fetal locus coeruleus neurons grafted to the hippocam-
regulated by psychosocial stress and NMDA receptor activation. pus in 6-hydroxydopamine-treated rats: an intracerebral
J Neurosci 17:24922498. microdialysis study. Eur J Neurosci 3:905918.
Gould E, Tanapat P, McEwen BS, Flugge G, Fuchs E (1998) Kaplan MS (1981) Neurogenesis in the 3-month old rat visual cortex
Proliferation of granule cell precursors in the dentate gyrus of J Comp Neurol 195:323338.
adult monkeys is diminished by stress. Proc Natl Acad Sci USA Kaplan MS (2001) Environmental complexity stimulates visual cor-
95:31683171. tex neurogenesis: death of a dogma and a research career. Trends
Gould E, Beylin A, Tanapat P, Reeves A, Shors TJ (1999a) Learning Neurosci 24:617620.
enhances adult neurogenesis in the hippocampal formation. Kaplan MS, Bell DH (1984) Mitotic neuroblasts in the 9-day old and
Nat Neurosci 2:260265. 11- month old rodent hippocampus. J Neurosci 4:14291441.
Gould E, Reeves AJ, Fallah M, Tanapat P, Gross CG, Fuchs E (1999b) Kaplan MS, Hinds JW (1977) Neurogenesis in the adult rat: electron
Hippocampal neurogenesis in adult Old World primates. Proc microscopic analysis of light radioautographs. Science
Natl Acad Sci USA 96:52635267. 197:10921094.
Gould E, Reeves AJ, Graziano MS, Gross CG (1999c) Neurogenesis in Kee NJ, Preston E, Wojtowicz JM (2001) Enhanced neurogenesis after
the neocortex of adult primates. Science 286:548552. transient global ischemia in the dentate gyrus of the rat. Exp
Gould E, Vail N, Wagers M, Gross CG (2001) Adult-generated hip- Brain Res 136:313320.
338 The Hippocampus Book

Kempermann G, Kuhn HG, Gage FH (1997) More hippocampal neu- Leuner B, Falduto J, Shors TJ (2003) Associative memory formation
rons in adult mice living in an enriched environment. Nature increases the observation of dendritic spines in the hippocam-
386:493495. pus. J Neurosci 23:659665.
Kempermann G, Kuhn HG, Gage FH (1998) Experience-induced Leuner B, Mendolia-Loffredo S, Kozorovitskiy Y, Samburg D, Gould
neurogenesis in the senescent dentate gyrus. J Neurosci 18: E, Shors TJ (2004) Learning enhances the survival of new neu-
32063212. rons beyond the time when the hippocampus is required for
Kempermann G, Gast D, Gage FH (2002) Neuroplasticity in old memory. J Neurosci 24:74777481.
age: sustained vefold induction of hippocampal neurogenesis Li C, Brake WG, Romeo RD, Dunlop JC, Gordon M, Buzescu R,
by long-term environmental enrichment. Ann Neurol 52: Magarinos AM, Allen PB, Greengard P, Luine V, McEwen BS
135143. (2004) Estrogen alters hippocampal dendritic spine shape and
Kim JJ, Clark RE, Thompson RF (1995) Hippocampectomy enhances synaptic protein immunoreactivity and spatial mem-
impairs the memory of recently, but not remotely acquired, ory in female mice. Proc Natl Acad Sci USA 101:21852190.
trace eyeblink conditioned responses. Behav Neurosci 109: Liu S, Wang J, Zhu D, Fu Y, Lukowiak K, Lu YM (2003) Generation of
195203. functional inhibitory neurons in the adult rat hippocampus. J
Kim Y-P, Kim H, Shin MS, Chang H-K, Jang M-H, Shin MC, Lee SJ, Neurosci 23:732736.
Lee HH, Yoon JH, Jeong IG, Kim CJ (2004) Age-dependence of Low WC, Lewis PR, Bunch ST, Dunnett SB, Thomas SR, Iversen SD,
the effect of treadmill exercise on cell proliferation in the den- Bjorklund A, Stenevi U (1982) Function recovery following
tate gyrus of rats. Neurosci Lett 355:152154. neural transplantation of embryonic septal nuclei in adult rats
Kodama M, Fujioka T, Duman RS (2004) Chronic olanzapine or u- with septohippocampal lesions. Nature 300:260262.
oxetine administration increases cell proliferation in hippocam- Luine V, Villegas M, Martinez C, McEwen BS (1994) Repeated stress
pus and prefrontal cortex of adult rat. Biol Psychiatry 56: causes reversible impairments of spatial memory performance.
570580. Brain Res 639:167170.
Kokaia Z, Lindvall O (2003) Neurogenesis after ischaemic brain Lynch G, Gall C, Rose G, Cotman C (1976) Changes in the distribu-
insults. Curr Opin Neurobiol 13:127132. tion of the dentate gyrus associational system following unilat-
Koketsu D, Mikami A, Miyamoto Y, Hisatsune T (2003) Nonrenewal eral or bilateral entorhinal lesions in the adult rat. Brain Res
of neurons in the cerebral cortex of adult macaque monkeys. J 110:5771.
Neurosci 23:937942. Madsen TM, Kristjansen PEG, Bolwig TG, Wortwein G (2003)
Kornack DR, Rakic P (1999) Continuation of neurogenesis in the Arrested neuronal proliferation and impaired hippocampal
hippocampus of the adult macaque monkeys. Proc Natl Acad Sci function following fractionated brain irradiation in the adult
USA 96:57685773. rat. Neuroscience 119:635642.
Kornack DR, Rakic P (2001) Cell proliferation without neurogenesis Magarinos AM, McEwen BS, Flugge G, Fuchs E (1996) Chronic psy-
in the adult primate neocortex. Science 294:21272130. chosocial stress causes apical dendritic atrophy of hippocampal
Kozorovitskiy Y, Gould E (2004) Dominance hierarchy inuences CA3 pyramidal neurons in subordinate tree shrews. J Neurosci
adult neurogenesis in the dentate gyrus. J Neurosci 24: 16:35343540.
67556759. Magavi SS, Leavitt BR, Macklis JD (2000) Induction of neurogenesis
Kuhn HG, Dickinson-Anson H, Gage FH (1996) Neurogenesis in the in the neocortex of adult mice. Nature 405:951955.
dentate gyrus of the adult rat: age-related decrease of neuronal Malberg JE, Duman RS (2003) Cell proliferation in adult hippocam-
progenitor proliferation. J Neurosci 16:20272033. pus is decreased by inescapable stress: reversal by uoxetine
Lavenex P, Steele MA, Jacobs LF (2000) Sex differences, but no sea- treatment. Neuropsychopharmacology 28:15621571.
sonal variations in the hippocampus of food-caching squirrels: Markakis E, Gage F (1999) Adult-generated neurons in the dentate
a stereological study. J Comp Neurol 425:152166. gyrus send axonal projections to eld CA3 and are surrounded
Lawrence JM, Morris RJ, Wilson DJ, Raisman G (1990) Mechanisms by synaptic vesicles. J Comp Neurol 406:449460.
of allograft rejection in the rat brain. Neuroscience 37:431462. McEwen BS (1999) Stress and hippocampal plasticity. Annu Rev
Leanza G, Martinez-Serrano A, Bjorklund A (1998) Amelioration of Neurosci 22:105122.
spatial navigation and short-term memory decits by grafts of Merrill DA, Karim R, Darraq M, Chiba AA, Tuszynski MH (2003)
foetal basal forebrain tissue placed into the hippocampus and Hippocampal cell genesis does not correlate with spatial learn-
cortex of rats with selective cholinergic lesions. Eur J Neurosci ing ability in aged rats. J Comp Neurol 459:201207.
10:23532370. Mirescu C, Peters JD, Gould E (2004) Early life experience alters
Lee DW, Smith GT, Tramontin AD, Soma KK, Brenowitz EA, Clayton response of adult neurogenesis to stress. Nat Neurosci 7:
NS (2001) Hippocampal volume does not change seasonally in 841846.
a non food-storing songbird. Neuroreport 12:19251928. Montaron MF, Drapeau E, Dupret D, Kitchener P, Aurousseau C, Le
Lemaire V, Aurousseau C, Le Moal M, Abrous DN (1999) Behavioural Moal M, Piazza PV, Abrous DN (2006) Lifelong corticosterone
trait of reactivity to novelty is related to hippocampal neuroge- level determines age-related decline in neurogenesis and mem-
nesis. Eur J Neurosci 11:40064014. ory. Neurobiol Aging 27:645654.
Lemaire V, Koehl M, Le Moal M, Abrous DN (2000) Prenatal stress Morrison JH, Hof PR (2002) Selective vulnerability of corticocortical
produces learning decits associated with an inhibition of neu- and hippocampal circuits in aging and Alzheimers disease. Prog
rogenesis in the hippocampus. Proc Natl Acad Sci USA 97: Brain Res 136:467486.
1103211037. Moser MB, Trommald M, Andersen P (1994) An increase in dendritic
Leonard BW, Amaral DG, Squire LR, Zola-Morgan S (1995) spine density on hippocampal CA1 pyramidal cells following
Transient memory impairment in monkeys with bilateral spatial learning in adult rats suggests the formation of new
lesions of the entorhinal cortex. J Neurosci 15:56375659. synapses. Proc Natl Acad Sci USA 91:1267312675.
 Structural Plasticity 339

Moser MB, Moser EI, Forrest E, Andersen P, Morris RG (1995) Spatial Pravosudov VV, Omanska A (2005) Dominance-related changes in
learning with a minislab in the dorsal hippocampus. Proc Natl spatial memory are associated with changes in hippocampal cell
Acad Sci USA 92:96979701. proliferation rates in mountain chickadees. J Neurobiol 62:3141.
Nadler JV (2003) The recurrent mossy ber pathway of the epileptic Raber J, Rola R, LeFevour A, Morhardt D, Curley J, Mizumatsu S,
brain. Neurochem Res 28:16491658. VandenBerg SR, Fike JR (2004) Radiation-induced cognitive
Nakatomi H, Kuriu T, Okabe S, Yamamoto S, Hatano O, Kawahara N, impairments are associated with changes in indicators of hip-
Tamura A, Kirino T, Nakafuku M (2002) Regeneration of pocampal neurogenesis. Radiat Res 162:3947.
hippocampal pyramidal neurons after ischemic brain injury Raisman G, Field P (1973) A quantitative investigation of the devel-
by recruitment of endogenous neural progenitors. Cell 110: opment of collateral reinnervation after partial deafferentation
429441. of the septal nuclei. Brain Res 50:241264.
Naylor AS, Persson AI, Eriksson PS, Jonsdottir IH, Thorlin T (2004) Rakic P (2002) Adult neurogenesis in mammals: an identity crisis.
Extended voluntary running inhibits exercise induced adult J Neurosci 22:614618.
hippocampal progenitor proliferation in the spontaneously Rakic P (1985a) Limits of neurogenesis in primates. Science 227:
hypertensive rat. J Neurophysiol 93:24062414. 10541056.
Nilsson M, Perlieva E, Johansson U, Orwar O, Eriksson PS (1999) Rakic P (1985b) DNA synthesis and cell division in the adult primate
Enriched environment increases neurogenesis in the adult rat brain. Ann NY Acad Sci 457:193211.
dentate gyrus and improves spatial memory. J Neurobiol 39: Ramirez JJ, Stein DG (1984) Sparing and recovery of spatial alterna-
569578. tion performance after entorhinal cortex lesions in rats. Behav
Nilsson OG, Brundin P, WIdner H, Strecker RE, Bjorklund A (1988) Brain Res 13:5361.
Human fetal basal forebrain neurons grafted to the denervated Rankin SL, Partlow GD, McCurdy RD, Giles ED, Fisher KR (2003)
rat hippocampus produce an organotypic cholinergic reinner- Postnatal neurogenesis in the vasopressin and oxytocin-
vation pattern. Brain Res 456:193198. containing nucleus of the pig hypothalamus. Brain Res 971:
Nottebohm F (2002) Neuronal replacement in adult brain. Brain Res 189196.
Bull 57:737749. Rhodes JS, van Praag H, Garland T, Gage FH (2003) Exercise
Olariu A, Shore LE, Brewer MD, Cameron HA (2005) A natural form increases hippocampal neurogenesis to high levels but does not
of learning can increase and decrease survival of new neurons in improve spatial learning in mice bred for increased voluntary
the dentate gyrus. Hippocampus 15:750762. wheel-running. Behav Neurosci 117:10061016.
OMalley A, OConnell C, Murphy KJ, Regan CM (2000) Transient Rietze R, Poulin P, Weiss S (2000) Mitotically active cells that gener-
spine density increases in the mid-molecular layer of hip- ate neurons and astrocytes are present in multiple regions of the
pocampal dentate gyrus accompany consolidation of a spatial adult mouse hippocampus. J Comp Neurol 424:397408.
learning task in the rodent. Neuroscience 99:229232. Rosenzweig MR, Krech D, Bennett EL, Diamond MC (1962) Effects
Ormerod BK, Lee TT, Galea LA (2003) Estradiol initially enhances of environmental complexity and training on brain chemistry
but subsequently suppresses (via adrenal steroids) granule cell and anatomy: a replication and extension. J Comp Physiol
proliferation in the dentate gyrus of adult female rats. Psychol 55:429437.
J Neurobiol 55:247260. Rusakov DA, Richter-Levin G, Stewart MG, Bliss TV (1997)
Ormerod BK, Lee TT, Galea LA (2004) Estradiol enhances neurogen- Reduction in spine density associated with long-term potentia-
esis in the dentate gyri of adult male meadow voles by increas- tion in the dentate gyrus suggests a spine fusion-and-branching
ing the survival of young granule neurons. Neuroscience model of potentiation. Hippocampus 7:489500.
128:645654. Sandstrom NJ, Williams CL (2004) Spatial memory retention is
Parent JM, Yu TW, Leibowitz RT, Geschwind DH, Sloviter RS, enhanced by acute and continuous estradiol replacement. Horm
Lowenstein DH (1997) Dentate granule cell neurogenesis is Behav 45:128135.
increased by seizures and contributes to aberrant network reor- Santarelli L, Saxe M, Gross C, Surget A, Battaglia F, Dulawa S,
ganization in the adult rat hippocampus. J Neurosci 17: Weisstaub N, Lee J, Duman R, Arancio O, Belzung C, Hen R
37273728. (2003) Requirement of hippocampal neurogenesis for the
Patel SN, Clayton NS, Krebs JR (1997) Spatial learning induces neu- behavioral effects of antidepressants. Science 301:805809.
rogenesis in the avian brain. Behav Brain Res 89:115128. Sapolsky RM, Krey LC, McEwen BS (1985) Prolonged glucocorticoid
Patel TD, Azmitia EC, Zhou FC (1996) Increased 5-HT1A receptor exposure reduces hippocampal neuron number: implications
immunoreactivity in the rat hippocampus following 5,7- for aging. J Neurosci 5:12221227.
dihydroxytryptamine lesions in the cingulum bundle and m- Scharfman HE (2004) Functional implications of seizure-induced
bria-fornix. Behav Brain Res 73:319323. neurogenesis. Adv Exp eM d Biol 548:192212.
Persson AI, Naylor AS, Jonsdottir IH, Nyberg F, Eriksson PS, Thorlin Scheff S, Benardo I, Cotman C (1977) Progressive brain damage
T (2004) Differential regulation of hippocampal cell prolifera- accelerates axon sprouting in the adult rat. Science 197:795797.
tion by opioid receptor antagonists in running and non-running Schmidt-Hieber C, Jonas P, Bischofberger J (2004) Enhanced synap-
spontaneously hypertensive rats. Eur J Neurosci 19:18471855. tic plasticity in newly generated granule cells in the adult hip-
Peterson GM (1994) Sprouting of central noradrenergic bers in the pocampus. Nature 429:184187.
dentate gyrus following combined lesions of its entorhinal and Scott BW, Wojtowicz JM, Burnham WM (2000) Neurogenesis in the
septal afferents. Hippocampus 4:635648. dentate gyrus of the rat following electroconvulsive shock
Pham K, Nacher J, Hof PR, McEwen BS (2003) Repeated restraint seizures. Exp Neurol 165:231236.
stress suppresses neurogenesis and induces biphasic PSA- Seri B, Garcia-Verdugo JM, McEwen BS, Alvarez-Buylla A (2001)
NCAM expression in the adult rat dentate gyrus. Eur J Neurosci Astrocytes give rise to new neurons in the adult mammalian
17:879886. hippocampus. J Neurosci 21:71537160.
340 The Hippocampus Book

Selye H (1976) The stress oflife . New York: McGraw-Hill. Trejo JL, Carro E, Torres-Aleman I (2001) Circulating IGF-1 mediates
Shors TJ (2001) Acute stress rapidly and persistently enhances exercise-induced increases in the number of new neurons in the
memory formation in the male rat. Neurobiol Learn eM m adult hippocampus. J Neurosci 21:16281634.
75:1029. Trommald M, Hulleberg G, Andersen P (1996) Long-term potentia-
Shors TJ, Chua C, Falduto J (2001a) Sex differences and opposite tion is associated with new excitatory spine synapses on rat den-
effects of stress on dendritic spine density in the male versus tate granule cells. Learn eM
m 3:218228.
female hippocampus. J Neurosci 21:62926297. Van der Borght K, Wallinga AE, Luiten PG, Eggen BJ, Van der Zee EA
Shors TJ, Miesegaes G, Beylin A, Zhao M, Rydel T, Gould E (2001b) (2005) Morris watermaze learning in two rat strains increases
Neurogenesis in the adult is involved in the formation of trace the expression of the polysialylated form of the neural cell adhe-
memories. Nature 410:372376. sion molecule in the dentate gyrus but has no effect on hip-
Shors TJ, Townsend DA, Zhao M, Kozorovitskiy Y, Gould E (2002) pocampal neurogenesis. Behav Neurosci 119:926932.
Neurogenesis may relate to some but not all types of hippocam- Van Praag H, Christie BR, Sejnowski TJ, Gage FH (1999a) Running
pal-dependent learning. Hippocampus 12:578584 enhances neurogenesis, learning, and long-term potentiation in
Simon M, Czeh B, Fuchs E (2005) Age-dependent susceptibility of mice. Proc Natl Acad Sci USA 96:1342713431.
adult hippocampal cell proliferation to chronic psychosocial Van Praag H, Kempermann G, Gage FH (1999b) Running increases
stress. Brain Res 1049:244248. cell proliferation and neurogenesis in the adult mouse dentate
Sloviter RS (1999) Status epilepticus-induced neuronal injury and gyrus. Nat Neurosci 2:266270.
network reorganization. Epilepsia 1:S34S39. Van Praag H, Kempermann G, Gage FH (2000) Neural consequences
Sloviter RS, Valiquette G, Abrams GM, Ronk EC, Sollas AL, Paul LA, of environmental enrichment. Nat Rev Neurosci 1:191198.
Neubort S (1989) Selective loss of hippocampal granule cells in Van Praag H, Schinder AF, Christis BR, Toni N, Palmer N, Gage FH
the mature rat brain after adrenalectomy. Science 243:535538. (2002) Functional neurogenesis in the adult hippocampus.
Smulders TV, Shiett MW, Sperling AJ, DeVoogd TJ (2000) Seasonal Nature 415:10301034.
changes in neuron numbers in the hippocampal formation of a Vythilingam M, Heim C, Newport J, Miller AH, Anderson E, Bronen
food-hoarding bird: the black-capped chickadee. J Neurobiol R, Brummer M, Staib L, Vermetten E, Charney DS, Nemeroff
44:414422. CB, Bremner JD (2002) Childhood trauma associated with
Snyder JS, Kee N, Wojtowicz JM (2001) Effects of adult neurogenesis smaller hippocampal volume in women with major depression.
on adult synaptic plasticity in the rat dentate gyrus. J Am J Psychiatry 159:20722080.
Neurophysiol 85:24232431. Wallenstein GV, Eichenbaum H, Hasselmo ME (1998) The hip-
Snyder JS, Hong NS, McDonald RJ, Wojtowicz JM (2005) A role for pocampus as an associator of discontiguous events. Trends
adult neurogenesis in spatial long-term memory. Neuroscience Neurosci 21:317323.
130:843852. Wang S, Scott BW, Wojtowicz JM (2000) Heterogeneous properties of
Stahnisch FW, Nitsch R (2002) Santiago Ramon y Cajals concept of dentate granule neurons in the adult rat. J Neurobiol
neuronal plasticity: the ambiguity lives on. Trends Neurosci 42:248257.
25:589591. Wang S, Kee N, Preston E, Wojtowicz JM (2005) Electrophysiological
Staneld BB, Trice JE (1988) Evidence that granule cells generated in correlates of neural plasticity compensating for ischemia-
the dentate gyrus of adult rats extend axonal projections. Exp induced damage in the hippocampus. Exp Brain Res
Brain Res 72:399406. 165:250260.
Steward O, Loesche J, Horton WC (1977) Behavioral correlates of Warren SG, Juraska JM (1997) Spatial and nonspatial learning across
denervation and reinnervation of the hippocampal formation the rat estrous cycle. Behav Neurosci 111:259266.
of the rat: open eld activity and cue utilization following bilat- Watanabe Y, Gould E, McEwen BS (1992) Stress induces atrophy of
eral entorhinal cortex lesions. Brain Res Bull 2:4148 apical dendrites of hippocampal CA3 pyramidal neurons. Brain
Tanapat P, Hastings NB, Reeves AJ, Gould E (1999) Estrogen Res 588:341345.
stimulates a transient increase in the number of new neurons in Westenbroek C, Den Boer JA, Veenhuis M, Ter Horst G (2004)
the dentate gyrus of the adult female rat. J Neurosci 19: Chronic stress and social housing differentially affect neurogen-
57925801. esis in male and female rats. Brain Res Bull 64:303308.
Tanapat P, Hastings NB, Rydel TA, Galea LA, Gould E (2001) Will B, Galani R, Kelche C, Rosenzweig MR (2004) Recovery from
Exposure to fox odor inhibits cell proliferation in the hip- brain injury in animals; relative efficacy of environmental
pocampus of adult rats via an adrenal hormone-dependent enrichment, physical exercise or formal training (19902002).
mechanism. J Comp Neurol 437:496504. Prog Neurobiol 72:167182
Tanapat P, Hastings NB, Gould E (2005) Ovarian steroids inuence Wood GE, Beylin AV, Shors TJ (2001) The contribution of adrenal
cell proliferation in the dentate gyrus of the adult female rat in and reproductive hormones to the opposing effects of stress on
a dose- and time-dependent manner. J Comp Neurol trace conditioning in males versus females. Behav Neurosci
481:252265. 115:175187.
Teather LA, Magnusson JE, Chow CM, Wurtman RJ (2002) Woolley CS (1998) Estrogen-mediated structural and functional
Environmental conditions inuence hippocampus-dependent synaptic plasticity in the female rat hippocampus. Horm Behav
behaviours and brain levels of amyloid precursor protein in rats. 34:140148.
Eur J Neurosci 16:2405415. Woolley CS, Gould E, McEwen BS (1990a) Exposure to excess gluco-
Toni N, Buchs PA, Nikonenko I, Bron CR, Muller D (1999) LTP pro- corticoids alters dendritic morphology of adult hippocampal
motes formation of multiple spine synapses between a single pyramidal neurons. Brain Res 531:225231.
axon terminal and a dendrite. Nature 402:421425. Woolley CS, Gould E, Frankfurt M, McEwen BS (1990b) Naturally
 Structural Plasticity 341

occurring uctuation in dendritic spine density on adult hip- Yoshimi K, Ren YR, Seki T, Yamada M, Ooizumi H, Onodera M, Saito
pocampal pyramidal neurons. J Neurosci 10:40354039. Y, Murayama S, Okano H, Mizuno Y, Mochizuki H (2005)
Woolley CS, McEwen BS (1992) Estradiol mediates uctuation in Possibility for neurogenesis in substantia nigra of parkinsonian
hippocampal synapse density during the estrous cycle in the brain. Ann Neurol 58:3140.
adult rat. J Neurosci 12:25492554. Yuste R, Bonhoeffer T (2001) Morphological changes in dendritic
Woolley CS, Wenzel HJ, Schwartzkroin PA (1996) Estradiol increases spines associated with long-term synaptic plasticity. Annu Rev
the frequency of multiple synapse boutons in the hippocampal Neurosci 24:10711089.
CA1 region of the adult female rat. J Comp Neurol 373:108117. Zhao M, Momma S, Delfani K, Carlen M, Cassidy RM, Johansson CB,
Woolley CS, Weiland NG, McEwen BS, Schwartzkroin PA (1997) Brismar H, Shupliakov O, Frisen J, Janson AM (2003) Evidence
Estradiol increases the sensitivity of hippocampal CA1 pyrami- for neurogenesis in the adult mammalian substantia nigra. Proc
dal cells to NMDA receptor-mediated synaptic input: correla- Natl Acad Sci USA 100:79257930.
tion with dendritic spine density. J Neurosci 17:18481859. Zhou FC, Azmitia EC, Bledsoe S (1995) Rapid serotonergic ber
Woolley CS (1999) Effects of estrogen in the CNS. Curr O pin sprouting in response to ibotenic acid lesion in the striatum and
Neurobiol 9:349354. hippocampus. Dev Brain Res 84:8998.
Xu Y, Tamamaki N, Noda T, Kimura K, Itokazu Y, Matsumoto N, Zola SM, Squire LR (2001) Relationship between magnitude of dam-
Dezawa M, Ide C (2005) Neurogenesis in the ependymal layer of age to the hippocampus and impaired recognition memory in
the adult rat 3rd ventricle. Exp Neurol 192:251264. monkeys. Hippocampus 11:9298.
This page intentionally left blank
10 Tim Bliss, Graham Collingridge, and Richard Morris

Synaptic Plasticity in the Hippocampus

cols that produce it, and the role of the NMDA receptor. Other
10.1 Overview aspects of NMDA receptor-dependent synaptic plasticity in
the hippocampus are introduced, including EPSP-spike (E-S)
Of all the properties of hippocampal synapses, perhaps the potentiation, synaptic scaling, and metaplasticity. In the next
most beguiling and consequential, and certainly the most section, 10.4, we turn to the mechanisms that support the
enthusiastically studied, is their ability to respond to specic expression of LTP. In contrast to the consensus that exists con-
patterns of activation with long-lasting increases or decreases cerning the NMDA receptor-dependent events that mediate
in synaptic efficacy. Although it is now clear that synaptic the induction of LTP, the mechanisms underlying its expres-
plasticity is a property of many, perhaps most, excitatory sionthat is, the pre- or postsynaptic changes that directly
synapses in the brain, much of what is known about long- result in an increase in synaptic efficacyare more complex
term potentiation (LTP) and long-term depression (LTD) has and less well understood. Attention has mostly focused on
been worked out in the structure where they were discovered. postsynaptic changes involving the phosphorylation or inser-
When it comes to synaptic plasticity, the hippocampus is the tion of AMPA receptor subunits. We assess the evidence that
fons et origo, the onlie begetter, the home team. presynaptic changes also contribute to the increase in synaptic
The chapter opens with a brief historical account of the efficacy during the rst few hours after induction. The study of
rst two decades of research into LTP and LTD. During this morphological changes goes back to the very early days of LTP
period, from 1966 to the mid-1980s, the major characteristics research, and it has often been proposed that a growth in the
of LTP, including its persistence, input specicity, and associa- number or size of synapses is a correlate of LTP. Modern opti-
tivity, were established, the critical role of the N-methyl-D- cal techniques allow spine growth to be imaged in real time,
aspartate (NMDA) receptor in the induction of LTP identied, and there is evidence that this may occur in LTP. Additional
and the rst steps taken to link LTP to hippocampus-depend- mechanisms, involving gene transcription and protein synthe-
ent learning. At the end of this period, it was clear that at most sis, are required to sustain LTP for longer periods, giving rise
hippocampal synapses the conditions required for the induc- to the distinction between early LTP (E-LTP) and protein syn-
tion of LTP are identical to the conditions required for the thesis-dependent late LTP (L-LTP). Late LTP, like early LTP, is
activation of the NMDA receptor, and that the cellular event input-specic, which raises the question of how somatically
that triggers the induction of LTP is the inux of Ca2 through generated transcripts and proteins are targeted specically to
the activated NMDA receptor. We then, in Section 10.2, begin recently activated inputs, a discussion that leads to the idea of
our survey of synaptic plasticity in the hippocampus with a synaptic tagging.
discussion of the various short-term forms of plasticity that Around the time that the NMDA receptor was being hailed
hippocampal synapses share with most if not all synapses: for the elegant molecular explanation it offered for input
paired-pulse facilitation and depression and post-tetanic specicity and associativity, examples of activity-dependent
potentiation. We devote the next two sections, 10.3 and 10.4, plasticity that did not require the NMDA receptor were being
to the mechanisms supporting the induction and expression uncovered. In particular, the mossy ber projection from
of NMDA receptor-dependent LTP, the form of LTP displayed granule cells to CA3 pyramidal cells supports a very different
by, for example, perforant path-granule cell synapses and by form of LTP that is not blocked by NMDA receptor antago-
commissural and Schaffer collateral synapses on CA1 pyrami- nists. Both the induction and expression of mossy ber LTP
dal cells. Section 10.3 covers the properties of LTP, the proto- appear to be predominantly presynaptic. The discussion of

343
344 The Hippocampus Book

mossy ber LTP receives its own section, 10.5, and is followed, frequency stimulation of the perforant path lasting a few sec-
in 10.6, by a survey of neuromodulatory systems that affect onds the responses evoked by subsequent stimuli could
LTP, such as the cholinergic input, catecholamines, and neu- remain enhanced for many tens of minutes. He reported his
rotrophins. In section 10.7 we explore the world of long-term observations at the 1966 meeting of the Scandinavian
depression, and the depotentiation of LTP. Although more Physiological Society in bo, Finland (Lmo, 1966). In 1968,
difficult to obtain than LTP, and more difficult in adult than Lmo and Bliss, a postdoctoral fellow from the National
juvenile animals, a substantial literature, both in vitro and in Institute for Medical Research in London with an interest in
vitro, supports the idea that LTD has both NMDA receptor- the neural basis of learning, embarked on a systematic inves-
dependent and NMDA receptor-independent forms. Synaptic tigation of what they at rst called long-lasting potentiation
plasticity is not conned to excitatory connections between (later renamed long-term potentiation by Graham Goddard)
principal neurons; in Section 10.8 we examine the evidence (Douglas and Goddard, 1975), a name that stuck no doubt
for LTP and LTD at excitatory synapses onto interneurons and because its acronym, LTP, tripped more lightly off the tongue
inhibitory synapses made by interneurons. Next, in section than LLP. Bliss and Lmo delivered test stimuli at intervals of
10.9, we turn to synaptic plasticity in development, and in the 2 to 3 seconds to the perforant path on each side of the brain
aging animal, and discuss how LTP and LTD are affected in and recorded the evoked response from either the granule
animal models of cognitive decline. cell layer of the dentate gyrus or the synaptic terminal region
Finally, whats it all for? Why do hippocampal synapses in the molecular layer. Tetanic stimulation of 100 Hz for 3 to
display these striking forms of plasticity, and does the intact 4 seconds or, more usually, 15 Hz for 10 to 20 seconds was
animal exploit them to acquire and store memories? In given to the perforant path on one side, the other acting as a
Section 10.10 we survey and assess the evidence relating to this nontetanized control. An example of the potentiation they
central question. The readers will probably not be surprised to observed following repeated episodes of tetanic stimulation is
learn that we are able to provide only partial answers. shown in Figure 10-1A. In many cases, potentiation of the
eld EPSP (fEPSP) was accompanied by an increase in the
10.1.1 LTP: The First Two Decades amplitude and a decrease in the latency of the population
spike (see traces in Fig. 10-1A) (for an introduction to eld
The term synaptic plasticity was introduced by the Polish potentials and their interpretation, see Chapter 2, Box 21).
psychologist Jerzy Konorski to describe the persistent, activity- The potentiation was apparent within 30 to 60 seconds of the
driven changes in synaptic efficacy that he assumed, in com- tetanus and could be followed, with little decrement, for sev-
mon with most neuroscientists since Santiago Ramon y Cajal, eral hours.
to be the basis of information storage in the brain (Konorski, A more extensive account by Bliss and Lmo of the back-
1948). A formal hypothesis embodying this idea was advanced ground to their experiments can be found at http://www.
by the Canadian psychologist Donald Hebb in 1949. ergito.com/main-lcd.jsp:bcsEXP.13.8; other recollections of
According to Hebbs celebrated postulate (Hebb, 1949), the early days of LTP by Andersen, Bliss, Lmo, Graham
Collingridge, Gary Lynch, Bruce McNaughton, and Richard
When an axon of cell A is near enough to excite a cell
Morris are gathered in LTP: Enhancing Neuroscience for 30
B and repeatedly or persistently takes part in ring it,
Years (Oxford: Oxford University Press; 2004. Reprinted from
some growth process or metabolic change takes place
Bliss et al., 2003).
in one or both cells such that As efficiency, as one of
Subsequently, Bliss and Tony Gardner-Medwin (1973)
the cells ring B, is increased.
found that tetanus-induced potentiation could persist for
Nearly 25 years were to pass before the rst description of many days in the unanesthetized rabbit with chronically
a Hebb-like synapse in the mammalian brain. In 1973, Tim implanted electrodes (Fig. 101B). Potentiation was not seen
Bliss and Terje Lmo reported that long-lasting changes in with low-intensity tetanization but required a stimulation
synaptic efficacy at perforant path-granule cell synapses in the strength that exceeded a threshold approximately equal to that
hippocampus could be induced by brief tetanic stimulation required to activate a population spike. Analysis of input-
(Bliss and Lmo, 1973). The discovery of what has come to be output curves suggested that only the synaptic inputs that had
known as long-term potentiation (or LTP) emerged from actually been tetanized were potentiated, a phenomenon called
experiments that Per Andersen and Lmo, then a PhD stu- input specicity (Bliss et al., 1973). Direct proof was deliv-
dent, were conducting at the University of Oslo during the ered some years later following the introduction of the hip-
mid-1960s on the phenomenon of frequency potentiation in pocampal slice preparation, which made it easier to activate
excitatory hippocampal pathways. Working with anesthetized two separate pathways converging on the same population of
rabbits, Andersen had earlier observed that eld potentials cells. LTP of the fEPSP was induced in the tetanized pathway,
evoked by stimulation at frequencies of 1 Hz or less were sta- whereas the response evoked in the second, untetanized path-
ble, whereas the response to successive stimuli grew steadily way remained at baseline levels (Andersen et al. 1977). In some
larger when the stimulus frequency was stepped up to 5 to 10 cases, LTP of the population spike occurred without a corre-
Hz (Andersen, 1960). While exploring this phenomenon in sponding change in the fEPSP (Bliss and Lmo, 1973). This
the dentate gyrus, Lmo noticed that after episodes of high- result implies the existence of a second component of LTP,
 Synaptic Plasticity in the Hippocampus 345

A
Before conditioning After conditioning
Exp Cont Exp Cont

+
2_mV

Amplitude of fEPSP (% change)

% 10 msec

200

100

0 1 2 3 4 5 6
Time (h)

Figure 101. Tetanic stimulation induces long-lasting

B
potentiation in the dentate gyrus of the intact rabbit.
Spike ampl. (mV)

A. Brief tetanic stimulation (arrows) of the perforant 4

path induces a rapid, persistent increase in the amplitude 3
of the eld EPSP. Further episodes of tetanic stimulation 2 a
(15 Hz for 15 seconds) produce more potentiation until 1
eventually, as illustrated in Figure 102A, the response 0
saturates. The sample responses above reveal that the 10 V 30 V 60 V 10 V 30 V 60 V
amplitude of the population spike is also potentiated and 15/sec, 15 sec trains
Spike ampl. (mV)

5 b
its latency reduced. (Source: Bliss and Lmo, 1973.) B. 30 min
4
Potentiation of the population spike lasting for many c
3
weeks in a freely moving rabbit with chronically
2
implanted electrodes given high-frequency trains of
1
increasing stimulus intensity. Note the lack of effect
0
of weak trains. The upper and lower part of the gure 12 h 24 hr 6 days 16 weeks +
are continuous. Following the second train at 60 V, the 2 mV
response was monitored for 45 minutes and the animal 5 msec
then returned to its home cage. Twelve hours later, a b c

potentiation had declined to about half its maximum

value and remained at that level for the following 16
weeks. The traces at the bottom of the panel were
obtained at times indicated by ac in the spike amplitude
plots. (Source: Bliss and Gardner-Medwin, 1973.)

reecting an increase in the coupling between the synaptic (McNaughton et al., 1978). Goddard and his colleagues, prin-
response and the population spike. The effect was subse- cipally Rob Douglas, Carol Barnes, and Bruce McNaughton,
quently given the name EPSP-spike (E-S) potentiation were also the rst to nd a way to block the induction of LTP
(Andersen et al., 1980). by delivering a coincident high-frequency train to the com-
Other important early contributions to the study of LTP in missural input to the dentate gyrus (Douglas et al., 1982).
the intact animal came from the laboratory of Graham Because commissural stimulation suppressed the ring of
Goddard in Halifax, Nova Scotia. These studies established granule cells, this last result suggested that the locus of con-
that briefer, more physiologically plausible tetani (7 stimuli at trol for induction was postsynaptic. From Goddards labora-
400 Hz) could produce LTP in the perforant path of the anes- tory too came the rst description of the property of
thetized rat (Douglas and Goddard, 1975) and formalized associativity, whereby a weak input could be potentiated if
the notion of cooperativity between afferent bers to explain its activation coincided with a tetanus to another input
the existence of an intensity threshold for LTP induction (McNaughton et al., 1978). Around the same time, associativ-
346 The Hippocampus Book

ity was also documented in the weak pathway made by con- showed that D-2-amino-5-phosphonopentanoic acid (D-
tralateral perforant path bers to the molecular layer of the AP5), a highly specic antagonist of the NMDA subtype of
dentate gyrus (Levy and Steward, 1979). glutamate receptor that had been developed by Jeff Watkins
The experiments described so far were all carried out on and coworkers (Davies et al., 1981), blocks the induction of
intact animals. Introduction of the living brain slice as a phys- LTP in area CA1 while having little or no effect on synaptic
iological preparation by Henry McIlwain during the 1960s responses recorded at low rates of stimulation (Collingridge et
(see Chapter 2) stimulated the development of the transverse al., 1983). Then Gary Lynch and his collaborators reported
hippocampal slice preparation in the Oslo laboratory (Skrede that injection of a calcium chelator, ethyleneglycol-bis-
and Westgaard, 1971). This preparation allows ready access to ( -aminoethylether)-N,N,N,N,-tetraacetic acid (EGTA),
all pathways of the dentate gyrus and hippocampus, provides into CA1 pyramidal cells blocks the induction of LTP (Lynch
mechanical stability for intracellular recording, and makes et al., 1983). This experiment conrmed the critical impor-
possible rapid pharmacological manipulation of the extracel- tance of the postsynaptic cell in the induction process and
lular environment (see Chapter 5). Not surprisingly, it has established an essential role for calcium.
achieved enormous popularity with investigators of synaptic As discussed in Chapter 6, excitatory synapses express up to
function. Using this preparation, LTP was revealed at two four types of postsynaptic glutamate receptor: -amino-3-
other hippocampal pathways: the mossy ber projection to hydroxy-5-methyl-4-isoxazolepropionate (AMPA), kainate,
CA3 pyramidal cells and the interleaved commissural and NMDA, and metabotropic glutamate (mGlu) receptors. The
Schaffer collateral bers from CA3 to CA1 pyramidal cells excitatory postsynaptic currents (EPSCs) evoked at low rates
(Schwartzkroin and Wester, 1975). We refer to these bers as of stimulation are largely mediated by activation of AMPA
the Schaffer-commissural projection. As we have seen, the receptors (Davies and Collingridge, 1989). NMDA receptors
transverse slice also allowed the input specicity of LTP to be contribute only a small component of these EPSCs but are
demonstrated in a more direct way than had been possible in critical for synaptic plasticity. Two unexpected properties of
the dentate gyrus in vivo (Andersen et al., 1977; Lynch et al., the NMDA receptor were identied during the 1980s. First,
1977). the NMDA channel is unusual for a ligand-conducted
In contrast to in vivo preparations where the dentate gyrus ionophore in that its conductance is both ligand- and voltage-
offers the most stable recording conditions, the region most dependent. At near-resting membrane potentials, the channel
commonly studied in vitro has been area CA1. This is partly is blocked by Mg2, and substantial depolarization is required
because LTP in the dentate gyrus in vitro is difficult to obtain to drive Mg2 from the channel ( Ault et al., 1980; Mayer et al.,
without reducing tonic inhibition by perfusing a GABAA 1984; Nowak et al., 1984). Thus, activation of the NMDA
receptor antagonist (Wigstrm and Gustafsson, 1983) and channel requires the temporal coincidence of activity in presy-
perhaps partly because of a feeling of unease engendered by naptic terminals to release transmitter plus adequate depolar-
the fact that in the slice preparation the perforant path bers ization of the postsynaptic membrane. The properties of
are unattached to their parent cell bodies. It is well to remem- cooperativity, input specificity, and associativity follow
ber that the hippocampal slice is a heavily reduced prepara- directly from the dual ligand and voltage dependence of the
tion. Although it makes possible a range of physiological and NMDA receptor. As a result of these properties, the NMDA
pharmacological experiments that would not otherwise be receptor can be thought of as a molecular coincidence detec-
feasible, a full understanding of the mechanisms and signi- tor, endowing pyramidal and granule cells with the capacity to
cance of hippocampal synaptic plasticity cannot be reached by detect coincident activity at multiple excitatory inputs.
in vitro studies alone. Second, the ionophore of the activated NMDA receptor is per-
Experimental analysis of the properties and mechanisms of meable to Ca2 ions (MacDermott et al., 1986; Jahr and
LTP has concentrated on the Hebbian form of synaptic plas- Stevens, 1987; Ascher and Nowak, 1988). This led to the idea
ticity exhibited by the perforant path projection to granule that the trigger for the cellular processes leading to LTP is the
cells of the dentate gyrus and by Schaffer-commissural affer- entry of Ca2 ions through the ionophore associated with the
ents to area CA1. It is this classical form of LTP that we dis- NMDA receptor and the subsequent activation of Ca2-
cuss in most detail. Tetanic stimulation of both these pathways dependent enzymes in the dendritic spine.
results in a rapid, persistent increase in synaptic responses, Thus, by the second half of the 1980s studies had matured
whether recorded intracellularly or extracellularly. LTP can be to the point where a simple, compelling molecular explana-
followed for many hours in both the slice preparation and the tion could be given for three dening properties of LTP: coop-
anesthetized animal and for days, weeks, or months in animals erativity, input specicity, and associativity. The hypothesis
with implanted electrodes. LTP generally decays to baseline that the necessary and sufficient conditions for inducing LTP
over a period of several days in rats with electrodes chronically are just those required to activate the NMDA receptor received
implanted in the dentate gyrus, though potentiation lasting further support in 1986. Holger Wigstrm, Bengt Gustafsson,
more than a year has been reported (Abraham, 2003). and others showed that even low-frequency stimuli could
Two important clues to the mechanisms underlying induce LTP in single CA1 pyramidal cells if each stimulus was
the induction of LTP were discovered in 1983. The rst clue given in conjunction with a strong depolarizing pulse
was provided by Graham Collingridge and colleagues, who (Wigstrm et al., 1986). This dramatic result showed that the
 Synaptic Plasticity in the Hippocampus 347

induction of LTP did not absolutely depend on the strong, Facilitation

high-frequency trains that had been largely used up until then
and the articial nature of which had called into question the Facilitation can be attributed to the transient increase in the
physiological relevance of LTP; on the contrary, weak low- concentration of intraterminal calcium produced by an
frequency stimulation was equally effective provided the tar- invading action potential. The concentration declines to basal
get cell was sufficiently depolarized by neighboring activity. values over a few hundred milliseconds, but the calcium inux
The realization of the Hebbian nature of the induction of LTP at the time of the second stimulus adds to the residual calcium
took root at the same time as Richard Morris and his col- from the rst, resulting in an enhanced calcium concentration
leagues were demonstrating that animals without the capacity and hence to an increase in the probability of release (Wu and
for hippocampal LTP owing to blockade of the NMDA recep- Saggau, 1994). The magnitude of facilitation is inversely
tor were severely impaired in the acquisition and retention of related to the initial probability of release. Analysis of elemen-
spatial learning in the watermaze (Morris et al., 1986). The tary (quantal) synaptic events suggests that facilitation is
mid-1980s was thus a time when LTP seemed well on the way largely due to presynaptic mechanisms (Zucker and Regehr,
to gaining acceptance as a physiological mechanism underly- 2002). Note, however, that a postsynaptic component of facil-
ing the cognitive faculties of learning and memory and mark itation has been described in both cortical pyramidal cells
an appropriate point at which to bring this brief historical (Rozov and Burnashev, 1999) and CA1 pyramidal neurons
survey to a conclusion. (Bagal et al., 2005). A widely used test for discriminating
presynaptic from postsynaptic mechanisms in LTP was intro-
duced by McNaughton (1982), based on the fact that the
higher the probability of release, the less scope there is for
10.2 Transient Activity-dependent presynaptically mediated facilitation, and so the smaller is the
Plasticity in Hippocampal Synapses observed degree of paired-pulse facilitation. A reduction in
paired-pulse facilitation following the expression of LTP is
10.2.1 Short-term Activity-dependent therefore often taken as evidence for an increase in the proba-
Changes in Synaptic Efficacy in the bility of release. Initial studies found that whereas paired-
Hippocampal Formation pulse facilitation was decreased in post-tetanic potentiation
(PTP), there was no change in LTP, implying that the expres-
Synaptic transmission at the level of the single synapse is an sion mechanism was more likely to be postsynaptic (see Fig.
inherently stochastic process (see Box 101; see also Chapter 1010A). As with most tests for the locus of expression of LTP,
6, Box 61) that can be modulated by prior activity in the the paired-pulse facilitation test is suggestive but not conclu-
presynaptic or postsynaptic neuron and by a wide variety of sive: For example, no change in facilitation would be observed
neuromodulators acting both pre- and postsynaptically. Some if LTP resulted from recruitment of a population of silent
of the activity-dependent changes in synaptic efficacy that boutons (that is, boutons with zero probability of release
occur in the hippocampal formation, such as facilitation and before the induction of LTP) that, following induction,
depression, augmentation, and (post-tetanic) potentiation, assumed a distribution of release probabilities similar to that
are relatively short-lasting. It is with these short-lived types of of the population activated before induction. (See below and
plasticity that we begin our discussion (reviewed by Zucker Box 101 for a discussion of this and other tests to determine
and Regehr, 2002) (see Chapter 62). the locus of expression of LTP.)

10.2.2 Single Stimuli in Hippocampal Depression

Pathways Produce Two Transient
Aftereffects: Facilitation and Depression Transient depression of synaptic efficacy also occurs to the
second of a pair of stimuli. To a large extent, this can be
The aftereffects on synaptic transmission of a single stimulus explained by a process of vesicle depletion, although receptor
can be tested by delivering a second stimulus at a variable time desensitization may also make a contribution. Evidence that
after the rst. When pairs of stimuli are delivered to the per- transmitter depletion contributes to depression comes from
forant path, the amplitude of the second response recorded in the observation that in situations where transmitter release is
the molecular layer of the dentate gyrus in vivo is typically articially reduced, as when extracellular calcium concentra-
facilitated at interstimulus intervals of less than 200 to 300 ms tion is lowered, depression is attenuated. This allows the full
and depressed at longer intervals of up to a few seconds time course of facilitation to be revealed. Transmitter released
(McNaughton, 1980, 1982). Similar effects are seen at by an action potential can also dampen the response produced
Schaffer-commissural synapses in area CA1 (Andersen, 1960). by a subsequent stimulus via negative feedback involving acti-
Paired-pulse facilitation and depression were rst studied in vation of presynaptic glutamate receptors. In a similar man-
the neuromuscular junction and appear to be a common fea- ner, presynaptic GABAB autoreceptors on inhibitory terminals
ture of all chemically transmitting synapses (Magleby, 1979; are thought to be responsible for the reduced inhibition that
Nicholls et al., 2001). underlies the effectiveness of primed burst stimulation in
 Box 101
Pitfalls of Paired-Pulse Facilitation and Quantal Analysis as Techniques for
Studying the Locus of Change in Synaptic Plasticity

PAIRED-PULSE FACILITATION

Paired-pulse facilitation (PPF) is due to an activity-dependent presynaptic modulation of

transmitter release (Zucker and Regehr, 2002). It reects the fact that at many synapses, such as
those made by Schaffer commissural axons, an invading action potential has a greater chance of
evoking exocytosis of neurotransmitter when it arrives within a few tens of milliseconds of a
preceding action potential. When studying PPF, two stimuli are delivered with an interval of,
say, 50 ms, and the amplitude of the rst and second synaptic responses are compared. The
amplitude of the second response relative to the rst (facilitation ratio) is then a reection of
the increase in the probability of transmitter release (Pr). Facilitation is a function of the mean
Pr of the synapses under study. Clearly if Pr is at or close to 1, there is little or no scope for
facilitation. The standard argument, rst advanced in the context of LTP by McNaughton
(1982), is that modulation of synaptic efficacy due to a change in Pr is reected in a converse
change in facilitation; that is, an increase in Pr results in a decrease in facilitation and vice
versa. During the rst few minutes after a tetanus, the rapidly declining phases of PTP and
short-term potentiation (STP) are mirrored by a decrease in facilitation; this is as expected for
PTP, which is presynaptically mediated (see Section 10.2.3) and suggests that the early kinase-
independent phenomenon of STP is also presynaptic. Thereafter, during LTP, facilitation is at
baseline levels, arguing against a presynaptic change in LTP (see Fig. 1011A). However, the
evidence provided by a PPF analysis is suggestive rather than conclusive, and a change in PPF is
not a wholly reliable guide to a presynaptic locus. To take a hypothetical example, suppose PPF
increases after the induction of LTD. The conventional interpretation would be that this reects
a reduction in Pr. However, there are plausible postsynaptic changes that could give rise to an
increase in PPF. Thus, if LTD involves the internalization of receptors, and the probability of
internalization is dependent on the binding of transmitter, receptors at those synapses with the
highest Pr would be internalized rst. In this model, the mean Pr of the remaining active
synapses would steadily fall, and PPF would increase as a result of a purely postsynaptic change.
Evidence for this model is found in mGluR-dependent LTD (see Section 10.7.5)

QUANTAL ANALYSIS

When the quantal variable n, the number of release sites, is large (e.g., when stimulating many
Schaffer-commissural bers) the stochastic nature of neurotransmitter release is seen as trial-
by-trial uctuations in the amplitude of the evoked synaptic response. A measure of this vari-
ability is the coefficient of variance (CV  SD/mean), which is sensitive to changes in Pr and n
but also, in some circumstances, to postsynaptic factors (Auger and Marty, 2000). For example,
if during LTP Pr tends toward 1 (i.e., fewer failures), there is less variability due to trial-by-trial
uctuations in neurotransmitter release, and CV decreases. In the case of minimal stimulation,
where one or at most a few axons are stimulated, there may be random failures of response due
to the failure of the invading action potential to cause exocytosis of transmitter. A decrease in
failures is classically equated with an increase in Pr and vice versa.
When a package of neurotransmitter is released, it evokes a postsynaptic response, the size of
which is referred to as the quantal amplitude (q). The magnitude of this response varies from
trial to trial as a result of the stochastic nature of postsynaptic channel opening and uctua-
tions in the amount of L-glutamate released. When a single release site is activated (n  1), the
variability in the magnitude of successful responses is due solely to this intra-site quantal vari-
ance. When multiple sites are activated (n  1) there is an additional variance due to the
between-site variation in quantal size. This inter-site variability could be large, reecting, for
example, differences in receptor number and electrotonic distance from the recording site.
The study of the relative contributions of n, Pr, and q to changes in synaptic efficacy is the
subject of quantal analysis. Typically, quantal analysis involves collecting many responses to
plot an amplitude frequency histogram. The mean amplitude of an evoked synaptic response is
simply n.Pr.q. The spacing between peaks in the histogram provides an estimate of q. A change
in q alters the interpeak spacing without altering the amplitudes of the peaks, whereas a change
in n or Pr modies the amplitudes of the peaks but not the spacing between them
Although these approaches are powerful in theory, they are often difficult to interpret in
practice. There are many reasons for this situation, which can be considered in two main cate-
gories. The rst major problem is the difficulty of making accurate measurements. This
requires high quality voltage-clamp recordings, so events are measured (as far as possible) at
(Continued)

348
 Synaptic Plasticity in the Hippocampus 349

the same membrane voltage. It also requires the elimination of as many confounding variables
(e.g., voltage-dependent conductances that may be imperfectly clamped, and synaptic inhibi-
tion) as is practical. Assuming that this has been achieved, a remaining major problem is that
unitary EPSCs in hippocampal neurons are very small. As a result, it is difficult to distinguish
all of the EPSCs above the background noise. Consequently, estimates of the frequency and
amplitude distributions of miniature EPSCs can be inaccurate, and small evoked EPSCs may be
misclassied as failures. The technically demanding technique of dendritic recording close to
the site of synaptic activation improves the signal-to-noise ratio; using this approach it has
been shown that EPSCs at Schaffer commissural synapses can be generated by fewer than ve
AMPA receptors (Isaac et al., 1998). These very small responses would probably be undetectable
with standard somatic recording techniques and so would be misclassied as failures. A second
problem is that quantal variance (both inter- and intra-site) is often large, making it difficult to
construct clean histograms for quantal analysis.
The second class of problem is how to interpret the parameters that are altered during LTP
or LTD. The classical interpretation is that changes in n or Pr are due to presynaptic changes,
and changes, in q are due to postsynaptic changes. However, even at the neuromuscular junc-
tion where quantal analysis was invented and the application used to most powerful effect, it
was recognized that these parameters could be interpreted in different ways. For example, an
increase in q could be due to more neurotransmitter contained in a synaptic vesicle. In hip-
pocampal neurons, LTP is often associated with a decrease in the failure rate (Malinow and
Tsien, 1990). This was initially interpreted in the classical tradition as an increase in Pr.
However, the failure rate is determined by the product of Pr and n, the number of competent
synaptic sites. In was subsequently postulated (Kullmann, 1994) and then demonstrated (Isaac
et al., 1995; Liao et al., 1995) that a decrease in failures may be due to the unsilencing of silent
synapses (see Fig. 1021A). The idea is that a stimulated afferent ber may fail to generate a
response either because Pr  0 or because there are no postsynaptic receptors to which trans-
mitter can bind. By awakening these synapses, LTP adds new synaptic sites (n) to the equation.
There has been much debate as to whether the increase in n is presynaptic (more functional
release sites) or postsynaptic (more synapses with functional receptors), as discussed in Section
10.4; but either way, this alters radically the interpretation of changes in failure rates. The unsi-
lencing of silent synapses during LTP and, conversely, the silencing of synapses during LTD is
now a widely documented phenomenon (see Sections 10.4 and 10.7) and needs to be taken into
account during any interpretation of analyses of the quantal parameters of LTP. For example,
alterations in the frequency of spontaneous mini-EPSCs could reect changes in Pr and/or n,
where changes in n reect pre- or postsynaptic unsilencing of silent synapses.
Alterations in n, whether pre- or postsynaptic, can also explain changes in PPF, as in the
hypothetical example given above in which a postsynaptic change gives rise to a change in
PPF. Conversely, presynaptic changes could occur without a change in PPF (e.g., if LTP were
achieved by unsilencing of a population of new release sites with a similar distribution of
Pr to existing sites). Therefore changes in each of the basic electrophysiological parameters of
neurotransmission (n, Pr, q, PPF) can have more than one, radically different interpretation.

inducing LTP (Davies et al., 1991) (see Section 10.3.1 and Fig. the case in vitro at synapses of the medial but, interestingly,
106). not the lateral perforant path (McNaughton, 1980; Allen et al.,
Facilitation and depression necessarily interact. The condi- 2000). Another consequence of the interplay between facilita-
tions for facilitation are set in place by the arrival of an action tion and depression is the difficulty of predicting how the
potential at the presynaptic terminal and are independent of induction of LTP affects a cells response to bursts of stimuli.
whether transmitter release occurs. The same is not true for The increased response to single test stimuli conventionally
depression, which is a consequence of transmitter release. used to monitor synaptic efficacy in LTP experiments is not
Thus, if a pair of action potentials invades the terminal within necessarily a useful predictor of the response that a postsy-
a few tens of milliseconds, the response to the second action naptic cell might make to a burst of afferent impulses. For
potential is determined by both facilitation and inhibition if example, an increase in the probability of release as the result
the rst action potential resulted in transmitter release but is of the induction of LTP might, because of paired-pulse
inuenced by facilitation alone if the rst action potential fails depression, lead to a reduction in the probability of release to
to release transmitter. On the other hand, where release prob- the second or following stimuli in the burst (Lisman, 1997).
ability is relatively high, thus limiting the scope for facilitation, However, although Markram and Tsodyks (1996) reported
depression may predominate at all intervals; this appears to be effects of this kind at potentiated synapses in visual cortex,
350 The Hippocampus Book

Selig et al. (1999) did not nd a similar redistribution of computer-based system that calculates various measures of
synaptic efficacy when bursts of stimuli were delivered to eld potentials, EPSPs or EPSCs, as appropriate (Bortolotto et
potentiated hippocampal synapses; instead, the response to al., 2002). Once this preliminary work is complete, the exper-
each stimulus of the burst was potentiated. iment proper progresses through three phases. In the rst
Paired-pulse facilitation and depression of the population phase baseline responses are elicited using low-frequency test
spike can give useful information about feedforward and feed- pulses (e.g., at 30-second intervals) for a period sufficient to
back inhibitory circuits in the hippocampus. If the rst of a ensure stability (the duration of which depends on the type of
pair of stimuli is strong enough to evoke a sizable population experiment to be performed but typically is in the range of
spike, a brief period of intense GABAA-mediated feedback 1060 minutes). Sometimes an input-output curve is
inhibition lasting 10 to 20 ms is generated, during which time obtained in which the strength of the stimulation is systemat-
a second stimulus of equal or lesser intensity does not evoke a ically varied to provide information on the relation between
population spike; this gives way to a much longer period of the ber volley (reecting the number of bers activated), the
spike facilitation that can be explained by suppression of feed- synaptic component of the evoked response (the fEPSP), and
forward inhibition mediated by presynaptic GABAB autore- the population spike (reecting the number of cells dis-
ceptors; this effect peaks at an interstimulus interval of 100 to charged). The second phase is brief but critical: It is the point
200 ms and can last for 1000 ms or longer (Davies et al., 1991). at which the stimulated pathway is tetanized or, in the case of
most intracellular recordings, the moment when afferent
10.2.3 Post-tetanic Potentiation Is stimulation is paired with postsynaptic depolarization.
the Sum of Two Exponential Components: Common tetanization protocols include one or more trains of
Augmentation and Potentiation pulses at 100 Hz for 1 second per train or patterns consisting
of shorter trains arranged in a pattern that contains a theta
Repetitive stimulation of peripheral and central synapses frequency component (see below). The third phase involves a
brings into play another transient facilitatory process called return to low-frequency stimulation at the same frequency
post-tetanic potentiation (PTP). At perforant path synapses, a and intensity as was used to measure the baseline; data for fur-
weak high-frequency train leads to a substantial elevation of ther input-output curves may be collected at intervals after
test responses immediately after the train, which declines back tetanization.
to baseline within a few minutes (Fig. 102A). With longer In a typical LTP experiment, the response is dramatically
stimulus trains, PTP is superimposed on a background of LTP, amplied following the tetanus. The initial slope of the EPSP
and this complicates the analysis of the time course of both rises more steeply, the onset latency of the population spike (if
phenomena. When suitable controls are conducted, PTP is present) is reduced, and its amplitude is often enhanced sev-
seen to be the sum of two exponential processes known as eral-fold. So startling is the transformation that Bliss and
augmentation and potentiation that in the medial perforant Marina Lynch (1988) were moved to confess no matter how
path have time constants of 7 s and 2 to 3 minutes, respectively often one has witnessed the phenomenon, it is impossible not
(McNaughton, 1982). The concentration of Ca2 in terminal to retain a sense of amazement that such modest stimulation
boutons rises during PTP, suggesting that, like facilitation, it is can produce so immediate, so profound and so persistent an
a presynaptic process (Wu and Saggau, 1994; Tang and effect. The basic phenomenon of LTP is easily elicited and
Zucker, 1997). gratifyingly immediate.
It is a common convention to refer to synaptic potentiation
as LTP if the potentiated response is maintained without
appreciable downward drift for longer than 30 to 60 minutes.
10.3 NMDA Receptor-dependent Long-term This covers only the early phase of LTP, however, and allows
Potentiation: Properties and Determinants the experimenter to say nothing about later phases. In eld
potential studies, three phases of potentiation can be dened
We begin by discussing the varieties of stimulus protocol that on the basis of susceptibility to drugs. As discussed in more
lead to the induction of NMDA receptor-dependent LTP and detail below, LTP decays to baseline over a period of 3 to 5
then explore more fully its properties and characteristics. hours when induction takes place in the presence of inhibitors
of gene transcription or protein synthesis. Frey and coworkers
10.3.1 Long-term Potentiation: Tetanic assigned the name late LTP (L-LTP) to the persistent, pro-
Stimulation Induces a Persistent tein synthesis-dependent phase of LTP; the protein synthesis-
Increase in Synaptic Efficacy independent phase was termed early LTP (E-LTP) (Frey et
al., 1993). A third, earlier phase that survives protein kinase
Whether conducted in vivo or in vitro, a typical experiment inhibitors and decays to baseline usually within 30-60 minutes
on LTP takes the following form. The animal or brain slice is is referred to as short-term potentiation (STP).
prepared, electrodes are located at the appropriate sites for Originally observed in anesthetized rabbits, hippocampal
stimulating and recording, and observations are made to LTP has since been described in numerous vertebrate species,
ensure that electrophysiological responses are stable, via a including rats, guinea pigs, mice, and chicks (Margrie et al.,
 Synaptic Plasticity in the Hippocampus 351

A B

fEPSP slope (% change)

fEPSP slope (% change)
200
150

100

0
0 1 2 3 4 -15 0 15 30 45 60
Time (h) Time (min)

C D
20 M Anisomycin
170
Pop spike (% change)

150

fEPSP slope (% change)

60
100 40

20

50 0

-20 BDNF

0 -40
0 1 2 3 4 5 6 7 8 0 2 4 6 8
Time (h) Time (h)
E F G
5
Latency of pop spike (msec)

25
4
Spike height (mV)

20

3
15
a b
2
10

1 5

0 0
0 2 4 6 0 2 4 6 8 10
Amplitude of fEPSP (mV) fEPSP slope (mV/ms)

Figure 102. Types of activity-dependent potentiation in the hip- population spike without potentiation of the fEPSP. Field potentials
pocampus. A. Post-tetanic potentiation (PTP), lasting for a few min- recorded in the pyramidal cell layer (upper traces) and stratum
utes, is always induced in the dentate gyrus of the intact animal, radiatum (lower traces) of area CA1 before (left) and 25 minures
even when longer-lasting forms of potentiation are saturated by after the induction of LTP by an orthodromic/antidromic pairing
repeated tetanization. Tetani of increasing intensity were given; note protocol. Calibrationz: 10 ms, 2 mV. (Source: Jester et al., 1995.) F.
that at the lowest intensitiy LTP is not induced though PTP is still The latency of the population spike plotted as a function of the
present. (Source: Bliss et al., 1983) B. Short-term potentiation (STP) amplitude of the EPSP before (solid circles) and 8 hours after (open
in area CA1 in vitro. In the presence of a broad-spectrum protein circles) induction of LTP by tetanic stimulation of the perforant
kinase inhibitor, tetanus-induced potentiation returns to baseline path (same experiment as in Figure 101A). The curve was plotted
within 30 to 60 minutes (open triangles). After initial PTP, LTP in from responses evoked by a series of stimuli generated at different
control medium (lled triangles) persists without further decrement stimulus strengths. Note that an fEPSP of a given amplitude evokes a
for 60 minutes. (Source: Malenka et al., 1989.) C. The protein syn- population spike with a shorter latency after the induction of LTP
thesis inhibitor anisomycin curtails the duration of LTP. The ampli- than before, demonstrating that LTP consists not only of an increase
tude of the population spike in area CA1 in vitro was monitored in the synaptic drive but enhancement of the coupling between
after tetanic stimulation in control medium (triangles) or in the synaptic input and cell ring. This latter component is called EPSP-
presence of anisomycin (circles). E-LTP, dened as the component of spike (E-S) potentiation. (Source: Bliss and Lmo, 1973.) G. Another
LTP that is protein synthesis-independent, returns to baseline within experiment in which E-S curves were generated before and after the
5 to 8 hours. Late LTP is the more persistent component that is induction of LTP. Here, the amplitude of the population spike is
blocked by protein synthesis inhibitors. (Source: Frey et al., 1988.) D. plotted against the EPSP. The curve shifts to the left, reecting E-S
Transient exposure to brain-derived neurotrophic factor (BDNF) potentiation. (Source: Bliss et al., 1983.) In F, G points indicated by
induces a slow-onset late LTP in the dentate gyrus of the anes- lled and open arrow heads were obtained before and after induc-
thetized rat. (Source: Bramham and Messaoudi, 2005.) E. LTP of the tion of LTP respectively.
352 The Hippocampus Book

1998), macaque monkeys (Urban et al., 1996), and humans or in terms of local dendritic depolarization, does exist and
(Beck et al., 2000; reviewed by Cooke and Bliss, 2006). has important consequences for network behavior. First, net-
Long-term potentiation has been subjected to more work stability is less at risk from unrestricted increases in exci-
intense study in the hippocampal formation (particularly the tatory drive; second, the cooperativity threshold may act as
dentate gyrus and CA1 and CA3 pyramidal subelds) than lter, restricting access to long-term storage to those neural
elsewhere in the brain. In addition to the synapses of the tri- signals with a sufficiently high information content.
synaptic loop (i.e., perforant path to granule cells; mossy The magnitude and duration of LTP is also a function of
bers to CA3 cells; Schaffer collaterals to CA1 cells), LTP has tetanus intensity. If a tetanus that is suprathreshold for LTP is
been observed at associational connections between CA3 cells, repeated, it may induce a further increment of LTP when
the perforant path projections to area CA3 (Do et al., 2002) repeated some time later (Bliss and Lmo, 1973), suggesting
and area CA1 (Colbert and Levy, 1993), the projections from that at the level of the single synapse LTP may be incremental
CA1 to the subiculum (OMara et al., 2000), in some but not rather than all-or-none. Eventually, however, LTP becomes
all excitatory connections onto hippocampal interneurons saturated, and tetanic stimulation produces no further effect.
(Lamsa et al., 2005) and even at synaptic connections made by Clearly, the response to identical episodes of tetanic stimula-
hippocampal CA1 neurons on a class of macroglia-like cells tion depends on the history of the network, and this has given
(Ge et al., 2006). However, the fact that LTP has been observed rise to the concept of metaplasticity (see below). The maximal
in birds and reptiles lacking a laminated hippocampal forma- level of synaptic efficacy a given pathway can reach is not nec-
tion raises the possibility that its discovery in the mammalian essarily attained when LTP is apparently saturated. In the rst
hippocampus was fortuitous. Although the laminated struc- place, PTP can transiently elevate the response above the
ture of the hippocampus makes it particularly convenient for existing level. Furthermore, if an interval of several hours is
the study of LTP, there is no reason to suppose that hip- allowed to elapse after saturating tetanic stimulation, addi-
pocampal synapses occupy a privileged position in the hierar- tional potentiation can be elicited with a further tetanus, even
chy of plasticity. LTP has been described in many other brain if there has been no decrement in the response during the
regions (Racine et al., 1983), including cerebellum (Crpel intervening period (Frey et al., 1995).
and Jaillard, 1991), amygdala (Chapman et al., 1990) , sensory The stimulus patterns used to elicit LTP have varied widely
cortex (Artola and Singer, 1987), motor cortex (Sakamoto et among laboratories, ranging from brief trains at 400 Hz
al., 1987), prefrontal cortex (Laroche et al., 1990), and nucleus (Douglas and Goddard, 1975) to single stimuli of high inten-
accumbens (Kombian and Malenka, 1994). Long-term sity repeated at 1 Hz in the presence of picrotoxin (Abraham
changes in synaptic efficacy have also been found in sympa- et al., 1986). Probably the most commonly used protocol is a
thetic ganglia (Brown and McAfee, 1982) and in the nocicep- single train of 100 Hz for 1 second. Two other widely used
tive circuitry of the spinal cord (Ikeda et al., 2003). protocols are primed-burst stimulation (PBS) and theta-burst
stimulation (TBS), in which the common feature is an inter-
A Wide Range of Stimulus Protocols Can Induce LTP val of 200 ms, either between a priming stimulus and a brief
burst of stimuli in the case of PBS (Larson and Lynch, 1986),
LTP can be induced by a broad range of stimulus parameters. or between a succession of brief trains in TBS (Rose and
As rst noted by Bliss and Gardner-Medwin (1973) in their Dunwiddie, 1986). It may not be coincidental that 200 ms is
study of LTP in awake rabbits, the stimulus intensity during close to the periodicity of the theta rhythm (an endogenous
the tetanus must be sufficiently strong. In a systematic study hippocamapal rhythm generated during movementsee
of this effect, McNaughton et al. (1978) gradually increased Chapter 11). In most cases, it is probably true to say that the
the intensity of tetanization to reveal that LTP occurred only choice of protocol favored by a particular laboratory relies
above what they termed the cooperativity threshold. PTP more on tradition than any clearly demonstrated superiority.
was observed at lower intensities with the response amplitude In a parametric study of induction protocols in which theta-
declining to baseline between successive tetanizations (see burst stimulation was compared with simple trains of high-
example in Fig. 102A). Their supposition was that some frequency stimulation, Hernandez et al. (2005) concluded that
minimum number of afferent bers had to be activated for the major factor controlling the magnitude of LTP is the num-
LTP to occur and that thereafter the magnitude of LTP could ber of stimuli in a train rather than the pattern of stimulation.
increase until an asymptote was reached. However, it is now High-frequency trains can be used to induce LTP in cell pop-
known that activation of some minimum number of bers is ulations or single cells. In the latter case, two other protocols
neither a necessary nor a sufficient condition for inducing for inducing LTP are available: pairing, and spike timing. With
LTP. Although not indisputably established, it is likely that the former, single stimuli repeated at low frequency are paired
LTP can be induced by activating a single axon if a target cell with depolarizing pulses that induce brisk ring of the
of that axon is sufficiently depolarized. In the behaving ani- recorded cell (Abraham et al., 1986; Kelso et al., 1986). With
mal, therefore, there may be circumstances in which individ- spike timing-dependent potentiation (STDP), the pairing is
ual synapses can be independently potentiated. Nevertheless, between the afferent stimulus and a brief depolarizing pulse
cooperativity serves as a reminder that a threshold for LTP, that res the target cell only once. The latter protocol allows
whether expressed in terms of the number of activated axons the timing between pre- and postsynaptic ring to be con-
 Synaptic Plasticity in the Hippocampus 353

trolled with some precision and has led to the important has been claimed that LTP in this region is sometimes
observation that LTP occurs only when the presynaptic nondecremental (Staubli and Lynch, 1987). Later, we dis-
impulse precedes the ring of the target cell; when the oppo- cuss mechanisms responsible for modulating the persistence
site relation obtains, the result is LTD rather than LTP (Bi and of LTP. Intriguingly, the decay of LTP may be an active pro-
Poo, 1998) (see Section 10.3.8). cess, as in the dentate gyrus LTP is prolonged if an NMDA
The reason for the effectiveness of the prime-burst or receptor antagonist is given after induction (Villarreal et al.,
theta-burst procedure has been claried (Davies et al., 1991). 2001).
The priming stimulus (or initial brief train) activates feedfor- Can all excitatory synapses onto principal cells in the hip-
ward GABA interneurons, leading to GABAA- and GABAB- pocampus be potentiated? Not all attempts to induce LTP
mediated hyperpolarization in the pyramidal cell but, between pairs of cells using a pairing induction protocol or
importantly, also to activation of presynaptic GABAB autore- between a presumed single axon and a target cell have been
ceptors. The latter produce a transient reduction in GABA successful. However, this may reect an unhealthy or dialyzed
release that is maximal at around 100 to 200 ms. Thus the sec- cell or an inappropriate stimulus protocol for that synapse.
ond train produces much less GABA-mediated hyperpolariza- Based on results of attempting to induce LTP by a pairing pro-
tion with consequent enhancement of the voltage-dependent tocol using minimal stimulation to activate one or at most a
NMDA receptor-mediated current. Minimal patterns of stim- few synapses in area CA1, Petersen et al. (1998) estimated that
ulation of this kind are far more likely to occur naturally than between 45% and 75% of excitatory synapses in area CA1 can
the longer trains of tens or hundreds of stimuli (e.g., 100 Hz be potentiated, a proportion low enough to raise suspicions
for 1 second) that are frequently used for reasons of conven- that not all such synapses are potentiable. A similar conclusion
ience and custom. Hippocampal pyramidal neurons can and was reached by Debanne et al. (1999), working with pairs of
do re in high-frequency bursts, although it remains to be interconnected cells in organotypic hippocampal cultures. In
established whether LTP is induced by naturally occurring their sample, 24% of unitary EPSPs (a term here used to cover
patterns of activity in the freely moving animal (Buzsaki, the multiple connections that a CA3 cell in organotypic cul-
1987) (see Chapter 8, Section 8.4 for a further discussion of ture may make with a target CA3 or CA1 cell) could not be
this point). potentiated. Unpotentiable connections may represent
synapses at which LTP has been saturated and that might
10.3.2 Time Course of LTP: Rapid therefore be more susceptible to LTD-inducing stimulation,
Onset and Variable Duration but 63% of nonpotentiable connections in the experiments of
Debanne et al. (1999) could not be depressed either. These
How quickly after a tetanus is LTP induced? The time course studies suggest that there is a small proportion of excitatory
of the onset of LTP is difficult to measure because it is con- synapses on hippocampal pyramidal cells, perhaps 10% to
founded by the extremely rapid onset of PTP. By studying PTP 20% of the total, that do not exhibit synaptic plasticity.
in isolation using trains subthreshold for LTP or in the pres-
ence of AP5, which blocks the induction of LTP without 10.3.3 Three Distinct Temporal Components
affecting PTP, and by making the assumption that LTP and of Potentiation: STP, Early LTP, Late LTP
PTP are independent processes, it is possible to dissect out the
two processes (McNaughton, 1980; Gustafsson et al., 1989). In Broad-spectrum protein kinase inhibitors curtail the duration
CA1 in vitro, LTP begins within 2 to 3 seconds of the tetanus, of potentiation to an hour or so (e.g., Lovinger et al., 1987;
peaks within 30 seconds, and then declines for 10 to 20 min- Malenka et al., 1989; Malinow et al., 1989 Matthies and
utes to a lower, more stable value. In the dentate gyrus, LTP Reymann, 1993; see also reviews by Soderling and Derkach,
takes longer (12 minutes) to reach its peak (Hanse and 2000; Lisman et al., 2002; Nguyen and Woo, 2003) (Fig 102B
Gustafsson, 1992). Estimates are confounded, however, by the and Section 10.4.3). In the presence of protein synthesis
possibility that onset kinetics are modulated by the test stim- inhibitors such as emetine or anisomycin, or transcription
uli given to sample synaptic responses. inhibitors such as actinomycin, the potentiated response in
How long does LTP last? This is a question of obvious sig- area CA1 of hippocampal slices returns to baseline within 5 to
nicance but one to which no simple answer can be given. A 6 hours (Frey et al., 1988) (Fig. 102C). From these ndings
short answer is that LTP shows variable persistence. The dura- emerged the dening characteristics of early and late LTP (E-
tion of the effect depends on the parameters of the tetanus, LTP and L-LTP) as protein kinase-dependent and protein syn-
the type of preparation, and the region of the hippocampus. thesis-dependent, respectively. Note that this leaves a
In both hippocampal slices and anesthetized animals, the component of LTP, usually lasting less than an hour, referred
maximal duration of LTP is obviously limited to the lifetime to as short-term potentiation (STP). STP is NMDA receptor-
of the preparation. In the intact animal, time constants as dependent but depends neither on protein synthesis nor, in
short as a few days (Racine et al., 1983) and as long as 73 days general, on protein kinase activity (Fig. 102B). Weak stimu-
(Doyre and Laroche, 1992) to more than a year (Abraham, lus trains sometimes produce E-LTP without L-LTP, but no
2003) have been reported for the dentate gyrus. Fewer stimulus protocols have been identied that produce L-LTP
experiments have been carried out in area CA1 in vivo, but it without the preceding phase of E-LTP. However, as as we shall
354 The Hippocampus Book

A C
S1 S2

comm CA1
Sch

mf
CA3 pp 2 mV

fEPSP, S1 (% change)
DG 50
fim 10 mV

B -50
0 1 2
5 mV 100
10 mS

fEPSP, S2 (% change)
100 50
fEPSP (% change)

50
0
0

-50 -50
0 1 2 3 4 0 1 2
Time (h) Time (h)

Figure 103. Induction properties of LTP: coooperativity, input slope of the fEPSP evoked by test stimuli to S1 and S2 are plotted
specicity, and associativity. A. Diagram of a transverse section in the two panels below the diagram of the recording arrangement.
through hippocampus showing principal excitatory connections. The intensity of the tetanus delivered via S1 is set below the cooper-
B. LTP induced by a tetanus (250 Hz for 200 ms) to the perforant ativity threshold for LTP, so only PTP is elicited (upper panel).
path in an anesthetized rat. Field responses recorded in the granule At 1 hour a strong tetanus is given to S2, inducing robust LTP.
cell layer before and after the induction of LTP are shown above. Note that the response to S1 is unchanged: LTP is input-specic.
C. Input specicity and associativity demonstrated using a two- However, when a weak tetanus to S1, which by itself is ineffective,
pathway protocol in area CA1 in vitro. Stimulating electrodes S1 is combined with a strong tetanus to S1, LTP is induced in both
and S2 are placed in the stratum radiatum on either side of a inputs. This is the property of associativity. (Source: Bliss and
recording electrode to excite two independent sets of axons. The Collingridge, 1993.)

see in Section 10.4.9, late-onset, protein synthesis-dependent of the slice. Thus, stimulating electrodes positioned on either
potentation can be induced by applying activators of PKA to side of the target neuron (in single-cell studies) or of a popu-
area CAI, or brain-derived neurotrophic factor (BDNF) to the lation of neurons (in eld recordings) activate a nonoverlap-
dentate gyrus (Fig. 102D). ping set of axons projecting to the same target cell(s) (Fig.
104).
10.3.4 Input-Specicity of LTP: Potentiation Input specicity at the relatively crude level of two-pathway
Occurs Only at Active Synapses experiments does not imply specicity at the level of single
synapses. To demonstrate the latter, it will be necessary to
Compelling evidence for input specicity was rst obtained in monitor the activation of neighboring synapses. Optical meth-
area CA1 of the hippocampal slice in experiments using two ods offer the best chance of achieving this, but the relevant
stimulating electrodes positioned to activate two sets of bers experiments have not yet been done. Even so, experiments can
converging on the same cell population (Andersen et al., 1977; be conducted to investigate whether the expression of LTP is
Lynch et al., 1977). Tetanization of either pathway resulted in restricted to the synapses at which it is induced. Two ingenious
potentiation specic to that pathway. This outcome is true electrophysiological studies, one by Haley et al. (1996) and the
whether one stimulating pathway is in the stratum oriens and other by Engert and Bonhoeffer (1997), suggest that synapse
the other in the stratum radiatum, or both pathways are in the specicity breaks down within a radius of a few tens of
same stratum. The two-pathway design, which has made an microns of a site of potentiation, regardless of whether
important contribution to physiological studies of LTP, synapses are on the same target cell or are active at the time of
exploits the fact that most bers travel at an angle to the plane potentiation.
 Synaptic Plasticity in the Hippocampus 355

A B to the molecular layer of the contralateral dentate gyrus.

Tetanic stimulation of the weak crossed pathway was inca-
REC
(I.C. STIM) input 1 pable of inducing LTP when given alone; but if given at the
same time as a tetanus to the strong ipsilateral perforant path-
STIM1 STIM2 way, LTP was induced in both ipsilateral and contralateral
10 ms 10 mv pathways (see Fig. 1017A). This experiment demonstrated
input 2
that activity at one input has the potential for promoting plas-
REC
ticity at another active inputin other words that LTP is asso-
ciative. Levy and Steward argued that the asociative property
C of LTP was a desirable feature of memory elements in the
hippocampal formation. These two sets of experiments also
input 1 mV ms-1
4 made it clear that the distinction between cooperativity and
3
2
associativity is essentially semantic. Cooperativity can be
1 thought of as an associative interaction between the afferent
0
input 2 mV ms-1 bers accessible to a single stimulating electrode. Associativity
4 was subsequently observed at the intracellular level in the hip-
3
2 pocampal slice using a two-pathway design (Barrionuevo and
1 Brown, 1983).
0
0 10 20 30 40 These early experiments did not distinguish between asso-
conjunction ciative interactions occurring presynaptically and postsynap-
Time (min) tically. One possibility was that cooperativity and associativity
reected the requirement for sufficient depolarization of the
Figure 104. LTP as a Hebbian mechanism. The repeated pairing
of single stimuli with depolarization of the membrane potential is target cells. As already noted, experiments demonstrating that
sufficient to induce LTP. A. Experimental arrangement, with two inhibition of granule cells by commissural afferents could
independent afferent pathways and an intracellular electrode to suppress the induction of LTP, suggested to Goddard and his
record evoked EPSPs and through which current can be passed to colleagues that the postsynaptic cell provided the locus of
depolarize the membrane potential. B. Examples of EPSPs before control for LTP (Douglas et al., 1982). In 1986 Wigstrm,
and after the induction of LTP in input 1. Note that there is no Gustafsson and their coworkers (Wigstrm et al., 1986), and
change in input 2. C. Plots of the response evoked by test stimuli independently two other groups (Kelso et al., 1986; Sastry et
delivered alternately at an interval of 3.5 seconds to S1 and S2. al., 1986), conrmed in spectacular fashion that LTP in area
During the period indicated by the bar (conjunction), the stimu- CA1 could be induced by the simple expedient of repeatedly
lus to S1 was followed 7 ms later by a 400-ms depolarizing pulse
pairing single stimuli to Schaffer-commissural bers with a
that elicted a burst of cell ring. The pairing of single stimuli and
depolarizing pulse generating multiple spikes in the impaled
depolarization induced an increase in the EPSP that reached a max-
imum within 20 to 30 pairings and persisted, with some initial cell (Fig. 10-4). The explanation for the efficacy of this proce-
decay, for the rest of the recording session. The response to the sec- dure became clear when it was shown that it enabled
ond pathway was unaffected. (Source: Wigstrm et al., 1986.) NMDAR-mediated responses to be evoked by single stimuli
(Collingridge et al., 1988). Pairing single stimuli with ion-
tophoretic application of glutamate was similarly effective at
10.3.5 Associativity: Induction of LTP Is inducing LTP (Hvalby et al., 1987). By varying the time
Inuenced by Activity at Other Synapses between the stimulus and the onset of the depolarizing pulse,
the time window for associative interaction could be studied.
Associativity, a dening property of LTP, was discovered by The maximum effect occured when the stimulus was given at
McNaughton et al. (1978), and independently by Levy and the beginning of the 100-ms depolarizing pulse; the degree of
Steward (1979) who introduced the term in the context of potentiation declined to zero when the interval between this
LTP. In both cases, interactions between two afferent pathways single stimulus and the onset of the depolarizing pulse was
in the dentate gyrus of the anesthetized rat were examined. longer than 100 ms and no potentiation occurred if the affer-
McNaughton and his colleagues compared cooperativity in ent stimulus was given after the depolarizing pulse
the medial and lateral perforant path projections to the ipsi- (Gustafsson et al., 1987). Extracellular experiments in the
lateral dentate gyrus; they found that when weak tetani were dentate gyrus (Levy and Steward, 1983) and in area CA1
given simultaneously to the medial and lateral branches of the (Gustafsson and Wigstrm, 1986; Kelso et al., 1986), using a
perforant path, LTP was induced in both pathways, though two-pathway protocol with one weak and one strong input
each tetanus, when given separately, was below the coopera- also showed that associative potentiation of the weak pathway
tivity threshold. Levy and Steward investigated the properties requires near-coincidence of weak and strong inputs, with
of the crossed perforant path. The latter is a weak projection potentiation occurring if the weak immediately precedes the
containing few axons that projects from the entorhinal cortex strong input but not if the strong precedes the weak input.
356 The Hippocampus Book

The timing requirements for coincident pre- and postsynaptic molecular complex, from which it can be rapidly released by a
activity that emerged from these early experiments are some- ash of ultraviolet light, to establish that the obligatory rise in
what less constrained than those that have more recently been calcium is conned to a window of less than 2.0 to 2.5 seconds
characterized in the context of spike timing-dependent plas- following the tetanus. Second, as we have seen, LTP can be
ticity (STDP) (see Section 10.3.8). induced by the repeated conjunction of single (Wigstrm et
Because of its potential for strengthening co-active inputs, al., 1986) or multiple (Kelso et al., 1986; Sastry et al., 1986)
the property of associativity lies at the heart of the role that afferent volleys with strong depolarizing pulses applied
LTP is likely to play in memory functions of the hippocampus. through an intracellular electrode. In the latter experiments,
Note that cooperativity, associativity, and input specicity can the relatively long depolarizing pulses produced multiple
all be explained on the assumption that the induction of LTP spikes falling on either side of the single afferent stimulus.
at a given synapse requires presynaptic activity with coinci- Later work by Magee and Johnston (1997), Bi and Poo (1998),
dent activation of the postsynaptic neuron. This is very close and others established that LTP in the hippocampus, as in
to the criterion that Hebb proposed for the enhancement of neocortex (Markram et al., 1997), can be induced by repeated
synaptic strength in hypothetical neural circuits mediating pairing of an afferent stimulus with a single back-propagating
associative conditioning (Hebb, 1949). Consequently, hip- action potential, provided the EPSP precedes the action
pocampal synapses displaying associative LTP are often potential by less than ~10 ms (STDP) (reviewed in Dan and
referred to as Hebb synapses. In the next section we explore Poo, 2004) (see Section 10.3.8).
the Hebbian nature of LTP in more detail.
Is Postsynaptic Spiking Necessary for LTP?
10.3.6 Requirement for Tight Coincidence
of Presynaptic and Postsynaptic Activity Hebbs postulate implies that the postsynaptic cell must re
Implies a Hebbian Induction Rule for changes in synaptic weight to occur. There are several
types of plasticity where this is not the case (for example, as we
There is convincing evidence that the induction of LTP in the shall see, mossy ber LTP, heterosynaptic LTP and LTD,
Schaffer-commissural projection in area CA1 is indeed mGluR-dependent LTD), and ring may not be necessary
Hebbian in nature. In the rst place, the postsynaptic cell plays for LTP even at the canonical Hebbian synapse, the Schaffer-
an obligatory role in induction, as rst revealed in experi- commissural projection to hippocampal CA1 pyramidal cells.
ments showing that LTP could be blocked by injecting the cal- Many protocols for inducing LTP in this pathway do re the
cium chelator EGTA into the postsynaptic cell (Lynch et al., postsynaptic cell (for example, STDP and some though not all
1983) (Fig. 105B) or by hyperpolarizing the postsynaptic forms of pairing), but with tetanic stimulation the target cell
neuron (Malinow and Miller, 1986). In later work, Malenka et may re only to the rst stimulus of the train or not at all (for
al. (1992) used a caged form of the calcium chelator 1,2-bis review see Linden, 1999). Moreover, pairing-induced LTP is
(aminophenoxy)ethane-N,N,N,N,-tetraacetic acid (BAPTA), not blocked by intracellular injection of lidocaine derivatives
in which BAPTA is bound as an inactive but photolabile such as QX-314 that block spiking (Kelso et al., 1986; Gustafs-

Figure 105. The induction of LTP requires activation of NMDA sity was reduced between the time-course plots shown in the upper
receptors and postsynaptic Ca2 signaling. A. A plot of synaptic and lower sections of the gure. (Source: Collingridge et al., 1983.)
response versus time to show that the broad spectrum glutamate B. Intracellular recordings showing superimposed EPSPs before and
receptor antagonist -D-glutamylglycine (DGG), applied to the after induction of LTP under control conditions (left) and the block
synaptic region by microiontophoresis, depresses synaptic transmis- of LTP induction when the intracellular electrode contained the
sion and blocks induction of LTP. In contrast, D-2-amino-5- Ca2 buffer EGTA (right). Plots of the change in EPSP amplitude
phosphonopentanoate (D-AP5) has no effect on pre-established over time from a control and an EGTA-lled cell are also shown. A
LTP but blocks its induction. A single tetanus (100 Hz, 1 second) tetanus was delivered at the times indicated by small arrows.
was delivered at the points indicated by arrows. The stimulus inten- (Source: Lynch et al., 1983.)
 Synaptic Plasticity in the Hippocampus 357

son et al., 1987). According to Thomas et al. (1998), intracel- elucidation of the role of the NMDA receptor in its induction.
lular QX-314 prevents LTP induced by low-frequency theta The rst indication of the importance of this subtype of
trains but has no effect on the efficacy of high-frequency glutamate receptor came in 1983, when it was found that a
trains. Primed burst stimulation can readily induce LTP in specic NMDA receptor antagonist, D(-)aminophosphopen-
area CA1 in vitro at stimulus strengths that are below thresh- tanoic acid (D-AP5) (Davies et al., 1981), blocked induction
old for spike ring (Davies et al., 1991). In some circum- of LTP in area CA1 of the hippocampal slice (Collingridge et
stances, even low-frequency trains can induce LTP without al., 1983) (Fig. 105A). Because AP5 does not affect the
ring the postsynaptic cell (Krasteniakov et al., 2004; but see response to single stimuli, the NMDA receptor is not involved
Pike et al., 1999). LTP can also be induced by repetitive pulses in any obvious way in the mediation of synaptic transmission
of glutamate targeted by two-photon excitation of a caged at low frequencies (see Chapter 6). Moreover, AP5 does not
precursor to a single spine, a protocol that does not re the block LTP once it is induced, indicating that the effect of the
postsynaptic cell and is on the face of it entirely lacking in drug is on the induction, rather than the expression, of LTP.
associative credentials (Matsuzaki et al., 2004). In contrast, The nding that blockade of the NMDA receptor suppresses
STDP, which may or may not be the usual mechanism by the induction of LTP has been conrmed using other compet-
which synaptic weights are modied in vivo, is dened by a itive NMDA antagonists (Harris et al, 1984), the use-depend-
requirement for postsynaptic spiking. Indeed, STDP is the ent noncompetitive antagonist MK-801 (Coan et al., 1987)
very cynosure of Hebbian plasticity; not only does it require and the glycine site antagonist 7-chlorokynurenic acid (Bashir
tight temporal contiguity of presynaptic and postsynaptic r- et al., 1990). Independent conrmation that the NMDA
ing, but the back-propagating spike provides the associative receptor is necessary for LTP has been provided by studies on
signal that informs dendritic synapses that the postsynaptic genetically engineered mice. Particularly convincing evidence
cell has red. However, even here, in Hebbs heartland, locally came from experiments in which the cre-lox technique was
generated and spatially restricted spiking can, in distal den- used to delete genes in a cell-type and region-specic way (see
drites, take the place of a back-propagating action potential Section 10.10). Tsien et al. (1996), for example, created mice
(Golding et al., 2003). It seems safe to conclude that postsy- harboring a deletion of NMDAR1 (the obligatory subunit
naptic somatic action potentials are not an absolute require- of the NMDA receptor) in area CA1 but not in the dentate
ment for LTP at Schaffer-commissural synapses, at least in gyrus. Slices from these animals showed no LTP in CA1 but
vitro. Whether a strict Hebbian conjunction of pre- and post- normal LTP in the dentate gyrus. In site-directed mutagenesis
synaptic spiking is the norm in vivo remains to be seen experiments, Sprengel et al. (1998) created mice in which the
(Lisman and Spruston, 2005). intracellular carboxy terminus of the NMDA receptor was
In summary, although it is experimentally possible to deleted. The inability of these mutants to display LTP despite
bypass the postsynaptic action potential, it may be that in a unaltered Ca2 current through the mutant receptor indicates
physiological context the only way that sufficient depolariza- that the NMDA receptor plays a further role in the genesis
tion can be obtained is by postsynaptic ring, as postulated by of LTP, downstream of its function as a Ca2-permeable
Hebb. To the extent that it is the idea of sufficient depolariza- ionophore.
tion, however produced, that is critical for NMDA receptor An explanation of how NMDA receptors trigger the induc-
activation, the strict form of the Hebbian coincidence rule can tion of LTP quickly followed the discovery of their critical role
be relaxed slightly without compromising its spirit: It is the in the process. It was already known that NMDA receptor-
coincidence, within a narrow time window of ring in the mediated responses are potently blocked by Mg2 in spinal
presynaptic cell with sufficient depolarization in the postsy- cord neurons (Evans et al., 1977). In addition, MacDonald and
naptic cell, that is the criterion for the induction of LTP. These Wojtowicz (1980) had discovered that, in contrast to all other
considerations allow us to formulate a Hebbian induction rule known ligand-gated ion uxes, NMDA receptor-mediated
for associative synaptic potentiation at excitatory synapses on responses are strongly voltage-dependent. The key observa-
CA1 pyramidal and dentate granule cells: tion that made everything clear was the discovery in 1984 that
A synapse will be potentiated if, and only if, it is active at a the voltage dependence was caused by the Mg2 block (Mayer
time when its dendritic spine is sufficiently depolarized. et al., 1984; Nowak et al., 1984). At resting membrane poten-
It is easy to seeand is an instructive exercise for the tial, glutamate or other NMDA receptor ligands induce negli-
reader to conrmthat this rule accounts for the properties gible inward current through the channel associated with the
of cooperativity, input specicity, and associativity. NMDA receptor. With increasing depolarization, Mg2 ions
are driven from the channel, and an inward current carried by
10.3.7 Molecular Basis for the Hebbian Na and Ca2 develops.
Induction Rule: Voltage Dependence of It was quickly realized that the dual ligand and voltage
the NMDA Receptor Explains Cooperativity, dependence of the NMDA receptor response offers an elegant
Input Specicity, and Associativity molecular mechanism for both the induction of LTP
(Collingridge, 1985) and the properties of associativity and
Perhaps the most remarkable advance to date in understand- input specicity (Wigstrm and Gustafsson, 1985). Two con-
ing the cellular mechanisms responsible for LTP has been ditions must be met for an NMDA receptor-mediated response
358 The Hippocampus Book

to be generated: presynaptic activity to release glutamate and four subunits is rather narrow (Feng et al., 2004). NR2B
strong postsynaptic depolarization to relieve the Mg2 block of antagonists are much more selective, but highly variable
the NMDA receptor channel. These are just the conditions results have been reported ranging from complete block of
that, as we saw in the previous section, are required for the LTP (Barria and Malinow, 2005) to no effect on LTP (Liu et al.,
induction of LTP. Because Ca2 is required for the induction of 2004a). Further work is clearly required to establish the rela-
LTP (Lynch et al., 1983), and because activated NMDA recep- tive roles of the different NR2 subunits in LTP.
tor channels, unlike most ionotropic AMPA receptor channels,
are highly permeable to calcium (MacDermott et al., 1986; Jahr 10.3.8 Spike Timing-dependent Plasticity (STDP)
and Stevens, 1987, 1993), it seems reasonable to conclude that
the trigger for the induction of LTP is the entry of calcium ions Explicit in Hebbs postulate is the notion of causality:
through the channel associated with the NMDA receptor. Presynaptic cell A must repeatedly or persistently take part
Although activation of NMDA receptors is a necessary in ring postsynaptic cell B. In a Hebb synapse, therefore, an
condition for the induction of a major component of LTP at increase in synaptic weight occurs only when the presynaptic
Schaffer-commissural and perforant path synapses, it is not, cell res shortly before the postsynaptic cell. If induction is
or at least not always, a sufficient condition. NMDA alone can mediated by the postsynaptic cell, this condition may be
induce short-term potentiation Collingridge et al. (1983), but expressed as a requirement that the synaptically generated
neither NMDA (Kauer et al., 1988) nor glutamate (Hvalby et current should begin before, but overlap with, the ring of the
al., 1987), when applied alone to hippocampal slices, readily postsynaptic cell. In practice, given the time course of excita-
induces long-lasting potentiation. This result does not sit eas- tory EPSCs (and, in particular, of the NMDA receptor-
ily with a purely postsynaptic model of the induction and mediated component of the EPSC), this interval may extend
expression of LTP but is consistent with models in which the to a few tens of milliseconds. Early experiments on associative
expression of LTP is dependent, in part, on presynaptic mech- LTP using weak and strong inputs to the dentate gyrus
anisms. Alternatively, other postsynaptic signaling pathways, suggested that LTP satises these requirements for temporal
such as mGluRs, may be necessary co-triggers. In addition, the contiguity (Levy and Steward, 1983). The weak pathway was
way in which NMDA receptors are activated may be a critical potentiated only if it was active within 20 ms of activation of
factor, as more prolonged application of NMDA receptors the strong pathway. Moreover, if the order of activation was
tends to induce LTD or depotentiation, which could mask or reversed, with the weak input stimulated up to 20 ms after the
obliterate any underlying potentiation. Photolysis of caged strong input, LTD was induced (Levy and Steward, 1983). As
glutamate combined with depolarization does, however, pro- discussed above, other two-pathway studies using intracellular
duce long-lasting potentiation of the responses of single recording conrmed the importance of temporal contiguity
spines in organotypic culture to test pulses of uncaged gluta- in pre- and postsynaptic spiking for the induction of LTP
mate (Bagal, et al., 2005). Thus, in this preparation, NMDA (Gustafsson and Wigstrm, 1986; Kelso and Brown, 1986;
receptor activation is sufficient to induce a postsynaptic com- reviewed in Bi and Poo, 2001).
ponent of LTP. In two of the conventional ways to induce LTP tetanic
stimulation and low frequency pairing of presynaptic stimuli
Role of NMDAR subtypes in LTP. with trains of postsynaptic spikes induced by a depolarizing
pulsethe precise temporal relation between pre- and post-
There is evidence that specic NMDAR subtypes are involved synaptic spiking is not straightforward. During a tetanus at
in LTP. A role for NR2A-containing NMDARs was suggested 100 Hz, ring of the postsynaptic cell rapidly accommodates,
in experiments in which the NR2A subunit was knocked out and the number of coincident action potentials in pre- and
(Sakimura et al., 1995). These animals still exhibited LTP but postsynaptic cells is limited. During pairing, there is generally
its magnitude was reduced to approximately half of that seen only one presynaptic spike and multiple ring of the postsy-
in wildtypes. This may may reect developmental conse- naptic cell. In 1994, Stuart and Sakmann demonstrated
quences of the knockout but subsequent pharmacological unequivocally that the action potential in cortical pyramidal
experiments have also suggested a role for different NR2 sub- cells propagates back into the dendritic tree (Stuart and
types in LTP. For example, LTP was found to be more sensitive Sakmann, 1994). This discovery identied an associative sig-
to NR2A/2B selective antagonists, whilst LTD was more sensi- nal at the synapse that could link presynaptic and postsynap-
tive to NR2C/D selective antagonists (Hrabetova et al., 2000). tic ring and initiated a period of intense study into the effects
This differential sensitivity suggested a role for NR2A and/ of varying the timing between presynaptic ring and the
or NR2B receptors in LTP. Sorting out the relative roles of back-propagating action potential. The rst synapses to be
NR2A and NR2B-containing receptors has proved difficult investigated were local excitatory connections between corti-
and may be confounded by the existence of triheteromers that cal pyramidal cells (Markram et al., 1997). Potentiation was
comprise both NR2A and NR2B subunits, in addition to the achieved when the presynaptic spike preceded the postsynap-
obligatory NR1 subunit. An NR2A selective antagonist, NVP- tic spike, with maximal potentiation at near-zero intervals,
AAM077, inhibits LTP (Liu et al., 2004; Berberich et al., 2005). and the degree of potentiation declining with longer intervals;
However, the selectivity of this compound versus the other at delays greater than 20 ms no effect was seen. When postsy-
 Synaptic Plasticity in the Hippocampus 359

naptic ring preceded presynaptic ring, the effect was revers- implicated by the nding that thapsigargin, which prevents the
ed and LTD was induced; again the maximum effect was seen relling of these stores, can block the induction but not the
at the shortest post-pre intervals, declining to zero over 20 ms. expression of LTP at CA1 synapses (Harvey and Collingridge,
STDP the term was introduced by Song et al. (2000) has 1992). Ca2 may be released from stores via the generation of
also been extensively studied in the hippocampus (Bi and Poo, inositol trisphosphate (IP3) and/or via Ca2-induced Ca2
1998; Debanne et al., 1998; reviewed by Bi and Poo, 2001 and release media by IP3 and/or ryanodine receptors, respectively.
Lisman and Spruston, 2005) (see Fig. 1019C,D, below). In Evidence that Ca2-induced Ca2 release may be important
dissociated hippocampal cultures, the STDP function relating was provided by the nding that dantrolene, an inhibitor of
magnitude and direction of plasticity to the interval betweeen ryanodine receptors, blocks induction of LTP at these synapses
the pair of pre- and postsynaptic spikes is antisymmetrical (Obenaus et al., 1989). In support of this possibility, tetanic
(Fig. 1019D), as it is in cortical pyramidal cells. In acute stimulation was shown to elicit a large NMDA receptor-
slices, however, the function is more symmetrical, with two dependent Ca2 signal at synapses that was substanti-
intervals of depression anking the region of potentiation ally inhibited by either thapsigargin or ryanodine (Alford et
(Nishiyama et al., 2000) (see Section 10.7.3). al., 1993). Collectively, these data suggest that Ca2-induced
The STDP function predicts the polarity and magnitude of Ca2 release can greatly augment the Ca2 signal that perme-
change to a single pair of action potentials, one in the presy- ates NMDA receptors and may be involved in the induction of
naptic cell and the other in the postsynaptic cell. How does the LTP. Such a signal may maintain the specicity of LTP, as
computation scale up when trains of spikes are considered? In NMDA receptor-triggered, store-derived Ca2 signals can be
the simplest case, any two trains in the two connected cells can observed within individual spines (Emptage et al., 1999). Con-
be considered as sets of pairs, with each of the n1 action poten- ceivably, however, under certain circumstances sufficient Ca2
tials in one train paired with each of the n2 action potentials may enter directly via NMDA receptors such that the boost
in the other train, forming a total of n1  n2 pairs. The STDP from intracellular stores is not required. Also conceivably,
function can be used to compute the effect on plasticity for Ca2 release from intracellular stores may be induced by other
each pair independently and the total effect calculated by mechanisms, for example via activation of voltage-gated Ca2
summing the contribution to synaptic weight contributed by channels, phospholipase C (PLC)-coupled receptors, or other
each pair. Providing the STDP function has a bias in favor of Ca2-permeable ligand-gated channels. Any of these paths
depression (i.e., the integral of the post-before-pre function could, in theory, induce NMDAR-dependent LTP without
exceeds that of the pre-before-post function), a stable network activating the normal induction trigger, the NMDA receptor
evolves in which inputs that are biased toward ring the post- itself. There is evidence that Ca2 permeation through presy-
synaptic cell are strengthened at the expense of those that are naptic kainate receptors can trigger Ca2-induced Ca2 release
not (Song et al., 2000). However, experiments in the visual to induce an NMDA receptor-independent form of LTP at
cortex suggest that spike pairs do not combine independently mossy ber synapses on CA3 pyramidal cells (Lauri et al.,
in this way, and that the rst spike in a short train has a dis- 2003), as discussed in Section 10.5.3. Again, the extent to
proportionate effect on the outcome (Froemke and Dan, which Ca2 release from intracellular stores is exploited by the
2002). A model based on these ndings proved more accurate hippocampus during NMDA receptor-dependent LTP in vivo
than the independent model for predicting the polarity and is not known.
magnitude of plasticity when naturally recorded spike trains
were used as articial stimuli for monosynaptically linked Ca2 Entry via Voltage-gated Ca2 Channels
pairs of layer 2/3 neurons in cortical slices. However, it is not Enables NMDA Receptor-independent LTP
clear that either model can account for the fact that repeated
pairing of single presynaptic spikes delivered during a depo- Although it is accepted that Ca2 entry directly through the
larization-induced spike train in the postsynaptic cell invari- NMDAR channel is the trigger for NMDAR-dependent LTP,
ably produces LTP. this pathway is not necessarily the only source of elevated
Inhibitory synapses onto hippocampal pyramidal cells intracellular Ca2. In particular, depolarization during sum-
exhibit an atypical temporal window for STDPwhatever the mated EPSPs leads to Ca2 entry via voltage-gated Ca2 chan-
order, provided the interneuron and target pyramidal cell re nels (VGCCs). Because NMDA receptors provide much of the
within 20 ms of each other, the synapse is potentiated (see depolarization during a high-frequency synaptic response
Section 10.8.2). (Herron et al., 1986), this Ca2 source is largely NMDAR-
dependent (Alford et al., 1993). Determining the relative
10.3.9 Ca2 Signaling in LTP importance of the two NMDAR-dependent sources is not a
trivial matter. In theory, the latter can be excluded in voltage-
Ca2 Release from Ca2 Stores clamp experiments, but a perfect clamp of synapses located on
Contributes to Induction of LTP dendritic spines is not achievable. Blocking VGCCs pharma-
cologically is also problematic as certain types are required for
A major source of Ca2 in neurons is via release from intracel- neurotransmitter release. An alternative strategy has been to
lular stores. A role for Ca2 stores in synaptic plasticity is activate VGCCs by intracellular depolarization, in the absence
360 The Hippocampus Book

of evoked synaptic transmission. This generally does not prevented by an NMDAR antagonist, but it occluded
induce LTP unless phosphatases are also blocked (Wyllie and NMDAR-dependent tetanus-induced LTP, suggesting conver-
Nicoll, 1994). The logical conclusion is that it may be possible gence with the mechanisms underlying the expression of clas-
to induce LTP if sufficient Ca2 enters via VGCCs because it sical LTP. ACPD-induced slow-onset potentiation was blocked
can access the sites that normally are reached only by the by thapsigargin. This observation provides a potential route
Ca2-permeating NMDARs. It follows, therefore, that under by which ACPD may access NMDAR-dependent LTP, the
most conditions VGCCs are not required for the induction of point of convergence being the intracellular Ca2 stores. Ca2
NMDAR-dependent LTP. On the other hand, stronger induc- release from these stores is normally triggered by Ca2 perme-
tion protocols would promote the activation of VGCCs. ating via NMDA receptors; but under certain circumstances,
Because the depolarization spreads to other regions of the cell, stored Ca2 may be released by activation of mGluRs in the
this could lead to partial breakdown in input specicity. It absence of NMDAR activation. Group I mGluRs (mGluR1
might also result in Ca2 entry at the cell soma, where it could and mGluR5) couple to phospholipase C, leading to the liber-
act as a nuclear signal. At the extreme, depolarization pro- ation of IP3, which acts on IP3 receptors to trigger Ca2
vided by AMPA receptors alone could result in Ca2 entry and release from intracellular stores (see Chapter 6, Section 6. 3).
LTP that appears to be NMDAR-independent but in fact Consistent with this hypothesis, (1S,3R)-ACPD activates and
engages the same mechanisms as NMDAR-dependent LTP MCPG inhibits both mGluR1 and mGluR5. An unusual
downstream of the NMDAR. Consistent with this notion, aspect of the action of (1S,3R)-ACPD in area CA1 is that it
there is evidence that higher frequencies during tetanization depends on intact connections from area CA3 (Bortolotto and
(e.g., 200 Hz) promote NMDAR-independent LTP in the Collingridge, 1995). A possible explanation is that (1S,3R)-
Schaffer-commissural pathway (Grover and Teyler, 1990). ACPD excites CA3 pyramidal cells sufficiently to depolarize
However, in this case the downstream signaling pathways may CA1 neurons, which in turn may lead to the sensitization of
be also different, pointing to the existence of a truly NMDAR- intracellular Ca2 stores to the mGluR-triggered generation of
independent form of LTP at these synapses (Cavus and Teyler, IP3. Consistent with this hypothesis, modest depolarization of
1996). This LTP develops relatively slowly, is of small ampli- CA1 neurons greatly facilitates the Ca2- mobilizing response
tude but is long-lasting, and is triggered by entry of Ca2 to group I mGluR agonists (Rae et al., 2000).
through L-type calcium channels (Grover and Teyler, 1990). Several studies have attempted to replicate the original
The L-type channel isoform has now been identied as MCPG experiments, with variable results. One camp agreed
Cav1.2, and mutant mice with inactivation of this isoform do that MCPG blocked induction of LTP (Riedel and Reymann,
not display NMDA receptor-independent LTP (Moosmang et 1993; Sergueeva et al., 1993; Richter-Levin et al., 1994),
al., 2005). Signicantly, they are also severely impaired in a whereas the second camp found no effect of this antagonist
spatial discrimination task, demonstrating that NMDA recep- (Manzoni et al., 1994; Selig et al., 1995b; Thomas and ODell,
tor-independent LTP contributes to the neural processing of 1995; Martin and Morris, 1997). The simplest conclusion to
hippocampus-dependent learning and memory. Ca2- be drawn from these ndings is that MCPG-sensitive mGluRs
dependent mechanisms contributing to protein synthesis in may be necessary for the induction of LTP under certain, but
the late form of NMDAR-independent LTP are discussed in not all, experimental conditions. A possible resolution of the
Section 10.4.9. controversy emerged when it was discovered that the activa-
tion of mGluRs did not have to occur at the same time as the
10.3.10 Metabotropic Glutamate tetanus that activated NMDA receptors (Bortolotto et al.,
Receptors Contribute to Induction 1994). Thus, the induction of LTP in one pathway prevented
of NMDA Receptor-dependent LTP MCPG from inhibiting the induction of more LTP in the same
pathway. The initial conditioning tetanus did not have to
Metabotropic glutamate receptors (mGluRs), by virtue of induce LTP, as a tetanus delivered in the presence of an
their coupling to G proteins, can affect various intracellular NMDAR antagonist still produced the conditioning. This led
signaling pathways that might be involved in LTP. The rst to the proposal that the activation of mGluRs sets a molecu-
direct evidence that they may be activated during the induc- lar switch that negates the need for subsequent activation of
tion of LTP followed the development of the rst selec- mGluRs for induction of LTP in that pathway. Once activated,
tive mGluR antagonist, -methyl-4-carboxyphenylglycine the switch remains set for at least several hours unless it is
(MCPG) (Eaton et al., 1993). MCPG had no effect on basal turned off by a depotentiating stimulus. These observations
synaptic transmission or the expression of LTP but completely suggest that the initial activation of mGluRs converts them to
inhibited the induction of LTP in a fully reversible manner a constitutively active state, although how this switch is
(Bashir et al., 1993). Unlike NMDAR antagonists, however, achieved is not known. The molecular switch is a form of
MCPG had no effect on STP. In complementary studies it was metaplasticity, which may explain some of the controversy
found that the selective activation of mGluRs with (1S,3R)-1- surrounding the effects of MCPG on the induction of
aminocyclopentane-1,3-dicarboxylic acid [(1S,3R)-ACPD] NMDAR-dependent LTP. It is likely, however, that other
could induce LTP without the STP phase (Bortolotto and experimental factors are also important (Thomas and ODell,
Collingridge, 1993). This slow-onset potentiation was not 1995; reviewed by Anwyl, 1999).
 Synaptic Plasticity in the Hippocampus 361

With the development of various mGluR knockout mice anism for this effect is well established. Normally, when an
and newer mGluR antagonists, with differing patterns of sub- EPSP is evoked by stimulation of an afferent pathway, such as
type selectivity, the role of mGluR subtypes in NMDAR- the Schaffer commissural projection, it is curtailed by a bipha-
dependent LTP has been further investigated. However, no sic inhibitory postsynaptic potential (IPSP) comprising a
clear picture has yet emerged. At Schaffer-commissural rapid GABAA receptor-mediated component and a slower
synapses there are high levels of mGluR5 postsynaptically and GABAB receptor-mediated component. The hyperpolarizing
mGluR7 presynaptically. Other subunits, including mGluR1, effect of the IPSP intensies the Mg2 block of NMDARs, and
may be expressed at lower levels. The knockout of mGluR1 this greatly limits the extent to which NMDARs contribute to
was reported either to inhibit (Aiba et al., 1994) or to have no the synaptic response. This was rst demonstrated by a simple
effect (Conquet et al., 1994) on the induction of LTP. pharmacological experiment. When GABAA receptors were
Surprisingly, although an initial study reported a small blocked pharmacologically, an NMDAR-mediated synaptic
decrease in LTP in the mGlu5 knockout (Lu et al., 1997), no response appeared during low-frequency stimulation in the
effect was found by the same group in a more detailed follow- presence of Mg2 (Herron et al., 1985; Dingledine et al.,
up investigation (Jia et al., 1998). Interestingly, although LTP 1986). Indeed, it is probably no coincidence that the kinetics
was normal in the mGluR7 knockout, there was a deciency of the GABAA receptor-mediated IPSP is such that this com-
in STP (Bushell et al., 2002). This suggests that presynaptic ponent of the biphasic IPSP coincides with the greater part of
mGluRs play a role in this initial presynaptic form of plastic- the NMDAR-mediated conductance.
ity, although developmental consequences of the loss of Blockade of GABA inhibition is not simply a pharmaco-
mGluR7 cannot be discounted. Perhaps most perplexing was logical curiosity but a key feature of LTP induction. This is
the nding that LY341495, a compound that can be used at shown by the remarkable efficacy of priming induction pro-
high concentrations to block all known mGluRs (mGluR1-8), tocols, discussed above, in which a single stimulus precedes a
had no effect on LTP (Fitzjohn et al., 1998). The difference brief high-frequency burst by about 200 ms. A primed burst
between the effects of MCPG and LY341495, which can containing typically four shocks (i.e., ve shocks in total) can
demonstrated in the same pathway, suggests that either induce LTP. The potency of the burst derives from the fact that
MCPG is acting on some receptor other than mGlu1-8 or that the priming stimulus (or the preceding burst during theta
the blanket inhibition of mGluRs has effects that are different patterns) causes a substantial reduction in the GABA recep-
from the selective inhibition of some subunits. Clearly, further tor-mediated IPSP. The mechanism of this effect has been
work is required to clarify the roles of mGluRs in NMDAR- established (Davies et al., 1991) (Fig. 106). Some of the
dependent LTP. GABA that is released in response to the priming stimulus
feeds back to activate GABAB autoreceptors, which results in
10.3.11 Role of GABA Receptors in the inhibition of subsequent GABA release. Thus, the priming
Induction of NMDAR-dependent LTP stimulus releases a normal level of GABA, but the stimuli dur-
ing the burst release considerably less GABA, which facilitates
It has been known for many years that blockade of GABA the synaptic activation of NMDA receptors on pyramidal neu-
inhibition can greatly facilitate the induction of NMDAR- rons. The kinetics of the GABAB autoreceptor mechanism
dependent LTP (Wigstrm and Gustafsson, 1983). The mech- determines the optimal timing for priming; it has a latency of

Figure 106. GABAB autoreceptors regulate the induction of LTP. a tetanus (100 Hz, 1 second) is used (right panel). The open sym-
LTP plots show that the GABAB antagonist CGP35348 (1 mM) bols plot control experiments, and the lled symbols plot experi-
blocks induction of LTP induced by primed-burst stimulation (1 ments performed in the presence of CGP35348. The primed-burst
priming stimulus followed 200 ms later by 4 shocks at 100 Hz) (left stimulus or tetanus was delivered at the times indicated by arrows.
panel). However, the antagonist is ineffective if experiments are per- The traces were obtained during baseline and 30 minutes following
formed in the presence of a GABAA antagonist (middle panel) or if the priming/tetanus. (Source: Davies et al., 1996.)
362 The Hippocampus Book

around 20 ms, a peak effect around 200 ms, and a duration of tion spike, or the probability of unit discharge, against the
around 1 second (Davies et al., 1990). Thus, theta patterns of slope of the synaptic component (Bliss et al., 1973) (Fig.
activity are tuned to the maximum depression of GABA inhi- 102E-G). Although E-S potentiation is usually observed in
bition by the autoreceptor mechanism. The importance of conjunction with synaptic LTP, pure E-S potentiation has
autoinhibition for the induction of LTP can be readily demon- been obtained in area CA1 in vitro by combining low-fre-
strated. GABAB antagonists, applied at concentrations that quency afferent stimulation with high-frequency antidromic
prevent the GABAB autoreceptor from operating and thereby trains (Jester et al., 1995; Nakanishi et al., 2001). Like synaptic
maintain GABAA receptor-mediated inhibition, completely LTP, E-S potentiation in area CA1 is NMDAR-dependent
prevent induction of LTP by priming stimulation. As one (Jester et al., 1995; Lu et al., 2000a; Daoudal et al., 2002); and
would predict for a mechanism that works by modulating the observation that the NMDAR antagonist D-AP5 blocks
GABAA receptor-mediated inhibition, GABAB antagonists do both spike and EPSP potentiation implies that this is also the
not block LTP induced in the presence of a GABAA receptor case in the dentate gyrus in vivo (Errington et al., 1987).
antagonist (Davies and Collingridge, 1993). They are also However, E-S potentiation and synaptic LTP are differently
ineffective when tested against a conventional tetanus (e.g., affected by drugs that modulate the monoaminergic input to
100 Hz for 1 second) as the GABAB autoreceptor mechanism the hippocampus. Depletion of noradrenaline (norepineph-
operates for only a few stimuli. The system is nely tuned to rine) reduces synaptic LTP in the dentate gyrus but leaves LTP
suppress GABA inhibition during theta-type patterns of activ- of the population spike unaffected (Bliss et al., 1983; Munro et
ity, which from a physiological perspective suggests that hip- al., 2001).
pocampal synapses may be particularly susceptible to Several groups have reported that E-S potentiation is
potentiation during exploration, locomotion, and other input-specic when induced by tetanic stimulation in area
behavioral states that give rise to theta activity (see Chapter CA1 in vitro (Andersen et al., 1980; Daoudal et al., 2002;
11). The importance of the GABAB autoreceptor system is Marder and Buonomano, 2003). However, this conclusion has
emphasized in the induction scheme for LTP illustrated in been disputed by Jester et al. (2003), who found that input
Figure 107. specicity was violated in more than half of their experiments
Theoretically, many neurotransmitters and neuromodula- in CA1 in vitro; there is also evidence for a lack of input speci-
tors could interact with the LTP induction process via an city in the dentate gyrus in vivo (Abraham et al., 1985).
effect on GABA inhibition. For example, it has been shown Further experiments are needed to resolve this question. The
that endocannabinoids released from pyramidal neurons can persistence of E-S potentiation has not been explored in the
depress GABA inhibition, which can result in the facilitation freely moving animal.
of LTP (see Section 10.3.12). In one series of experiments it
was shown that LTP is associated with an endocannabinoid- Mechanisms of E-S Potentiation
mediated LTD of GABA inhibition in a narrow band sur-
rounding the potentiated synapses (Chevaleyre and Castillo, Inhibitors of GABAA receptors block a substantial component
2004). The LTD of synaptic inhibition then promotes induc- of E-S potentiation produced by tetanic stimulation
tion of LTP in this surround. (Abraham et al., 1987; Chavez-Noriega et al., 1989; Daoudal et
al., 2002; Marder and Buonomano, 2003; Staff and Spruston,
10.3.12 E-S Potentiation: A Component 2003), suggesting that it reects an increase in excitability
of LTP That Reects Enhanced Coupling brought about by a relative reduction in feedforward inhibi-
Between Synaptic Drive and Cell Firing tion. Support for this hypothesis has been provided by Lu et
al. (2000a), who found that E-S potentiation in CA1 neurons
Characteristics of E-S Potentiation is accompanied by a tetanus-induced LTD of feedforward
IPSPs; both E-S potentiation and LTD of IPSPs were abolished
Conventional synaptic LTP is characterized by a persistent by NMDA antagonists, or by injecting the phosphatase cal-
increase in synaptic current and a proportionate increase in cineurin into pyramidal cells. This last result implies that the
the amplitude of the population spike or, in single unit site of induction of E-S potentiation is the inhibitory synapse
recordings, in the probability of the cell ring an action poten- on the principal cell, not the excitatory input to the
tial in response to the test stimulus. In these cases, potentia- GABAergic interneuron. Nevertheless, the latter site may con-
tion of the population spike, or an increase in the probability tribute to persistent reduction in feedforward inhibition and
of cell ring, is no more than would be predicted from the hence to E-S potentiation because tetanic stimulation can
increase in the EPSP. Often, though, the increase is much induce LTD at these synapses (Laezza et al., 1999) (see Section
greater. In extreme cases, spike potentiation can occur in the 10.8.1). Because the expression of E-S potentiation manifests
absence of EPSP potentiation (Bliss and Lmo, 1973). This as an increase in postsynaptic cell ring for a given synaptic
property of hippocampal synapses was given the name EPSP- input, the results of Lu et al. (2000a) imply that expression of
spike or E-S potentiation by Andersen et al. (1980). In general, this form of LTP, like its induction, is postsynaptic. However,
E-S potentiation appears as a leftward shift of the curve this is unlikely to be the whole story. E-S potentiation has
obtained by plotting the amplitude or latency of the popula- been induced by a pairing protocol, in which plasticity in
 Synaptic Plasticity in the Hippocampus 363

Figure 107. Activation of NMDA receptors, modulated by other activated spines become strongly depolarized with the result that
ionotropic amino acid receptors, triggers the induction of LTP in the Mg2 block of the NMDA receptor channel is relieved, allowing
principal cells of area CA1 and dentate gyrus. A. Low-frequency Ca2 to permeate through the receptor. Factors that contribute to
stimulation of Schaffer commissural axons evokes an EPSP that is sustained depolarization, including temporal summation of EPSPs
mediated by L-glutamate acting on AMPA receptors. The EPSP is and the accumulation of Cl- in the spine and K in the synaptic
followed by a biphasic IPSP, produced by glutamatergic excitation of cleft, leading to depolarizing shifts in reversal potentials. Some pro-
feedforward GABAergic interneurons. The early part of the IPSP is tocols, in particular theta burst stimulation, depend for their effi-
due to activation of GABAA receptors, which are permeable to Cl- cacy on the suppression of inhibition mediated via GABAB
ions; the slow component of the IPSP reects activation of GABAB autoreceptors on inhibitory terminals. The suppression of GABA
receptors, indirectly coupled to K channels. Because of its slow release reaches a maximum around 200 ms after a priming stimu-
kinetics and the voltage-dependent block of its channel by Mg2 lus, explaining the peculiar efficacy of priming and theta burst stim-
ions, the NMDA receptor contributes little to the EPSP at low fre- ulation (see text). (Source: After Bliss and Collingridge (1993).
quencies of stimulation. B. During high-frequency stimulation, the
364 The Hippocampus Book

inhibitory circuits presumbably does not occur (Marder and potentials (Larkum et al., 1999) or favor the shift of the action
Buonomano, 2004). There is also a component of E-S poten- potential initiation site toward the dendrites. An increase in
tiation that is not GABAA receptor-dependent; in the well doc- dendritic excitability could preserve the input specicity of
umented study by Daoudal et al. (2002), this component E-S potentiation via the generation of local Ca2 spikes, lead-
amounted to 40% of the overall effect. ing to somatic action potentials.
A novel interpretation of the idea that E-S potentiation Bidirectional changes in the E-S relation have also been
reects a persistent reduction in inhibitory input has come described (Daoudal et al., 2002; Wang et al., 2003). Like E-S
from an analysis of mGluR-dependent LTD of inhibitory potentiation, E-S depression is NMDAR-dependent and dis-
transmission between interneurons and pyramidal cells in plays input specicity when monitored in the presence of
area CA1, touched on above (Chevaleyre and Castillo, 2003). picrotoxin (Daoudal et al., 2002).
This unusual form of heterosynaptic inhibitory LTD is medi-
ated by the activation of group 1 mGluRs on CA1 pyramidal E-S Potentiation: Functional Considerations
cells by high-frequency stimulation of Schaffer-commissural
bers. There is a consequent retrograde release of the cannabi- The functional consequences of E-S potentiation and depres-
noid 2-AG (Wilson and Nicoll, 2001), leading to the activation sion have not been explored in any depth, and will depend on
of endogenous CB1 receptors on the terminals of inhibitory several factors about which there is currently little informa-
axons projecting to the same pyramidal cell and, by unknown tion, including the persistence of the effect in the behaving
mechanisms, to a suppression of the release of inhibitory animal and the degree to which input specicity is main-
transmitter. The persistent disinhibition leads to E-S potenti- tained. Reactivation of a specic network of cells in which
ation. Chevaleyre and Castillo carried out their experiments E-S potentiation has been induced by a particular input pat-
in a medium that blocked ionotropic glutamate receptors. tern will be facilitated when that pattern is re-presented. Input
High-frequency stimulation in the presence of NMDAR specicity would add the useful feature that only the original
antagonists should therefore lead to E-S potentiation without input pattern would be subject to such amplication. Clearly,
synaptic potentiation. However, both in area CA1 in vitro an increase (or decrease) in excitability will also enhance or
(Collingridge et al., 1983) and in the dentate gyrus in vivo depress the ability of the network to communicate with
(Errington et al., 1987), AP5 blocks potentiation of both the downstream targets.
population spike and the fEPSP. It remains to be established
whether the retrograde release of 2-AG provides an explana- 10.3.13 Metaplasticity: The Magnitude and
tion for E-S potentiation in area CA1 in vivo. Direction of Activity-dependent Changes in
The input specicity of E-S potentiation, whether estab- Synaptic Weight Are Inuenced by Prior Activity
lished in the presence of a GABAA receptor antagonist
(Daoudal et al., 2002) or in normal bathing medium (Marder The fact that LTP eventually reaches an asymptotic level after
and Buonomano, 2003), is something of a puzzle, as an repeated episodes of high-frequency stimulation demonstrates
increase in excitability, or the tendency to re an action poten- that plasticity itself depends on the prior history of synaptic
tial, is most readily explained in terms of the conductances activity. The term metaplasticity was coined by Abraham
controlling the threshold of cell ring at the axon hillock. and colleagues to describe this concept (Abraham and Bear,
Regardless of whether E-S potentiation is due to local disinhi- 1996; Abraham and Tate, 1997). The essential feature of meta-
bition, we may ask how local changes in the region of those plasticity is a persistent change in the plastic properties of the
inputs whose activity results in E-S potentiation can give rise synapse brought about by prior synaptic or neuromodulatory
to increased probability of cell ring at those inputs and not activity that may or may not itself alter synaptic function. In
at others. A possible mechanism is upregulation of postsynap- the case of saturation, the prior activity (that is, preceding
tic conductances that allow local spiking, as modeled by strong high-frequency trains) induces LTP. However, low-
Wathey et al. (1992). High-threshold calcium channels are intensity (priming) trains, which themselves elicit no change
potential candidates (Chetkovich et al., 1991; Wathey et al., in synaptic weight, can also inhibit the subsequent induction
1992). In their analysis of GABAA receptor-independent E-S of LTP (Huang YY et al., 1992). This effect of priming trains
potentiation in hippocampal and cortical layer V pyramidal can persist for 1 to 2 hours, is specic to the primed input, and
cells, Debanne and colleagues found evidence that the effect is mediated by a mechanism that requires the activation of
depended on the activation of group I mGluRs by the acti- NMDA receptors. Similarly, low-frequency activation of
vated bers, leading by an unknown pathway to suppression NMDA receptors in Mg2-free medium can prevent induction
of a local K conductance (Daoudal et al., 2002; Sourdet et al., of LTP (Coan et al., 1989). These studies demonstrate the need
2003). Induction of LTP in CA1 pyramidal cells results in per- for an optimal level of NMDAR activation for the induction of
sistent enhancement of local dendritic excitability, ascribed to LTP; persistent low levels of activation may oppose induction
modulation of the potassium current IA (Frick et al., 2004) or of LTP via metaplasticity. Activation of mGluR receptors
in cultured neurons to the nonspecic cation current Ih appears to facilitate the subsequent induction of LTP (Cohen
(Wang et al., 2003). An increase in dendritic excitability may and Abraham, 1996) and, as discussed above, can modify the
lead to local calcium spikes and generation of somatic action subsequent requirement for mGluR activation for LTP induc-
 Synaptic Plasticity in the Hippocampus 365

tion (Bortolotto et al., 1994). There is evidence that early stim- rules proposed by Stent (1973), Sejnowski (1977), Singer
uli in the long low-frequency trains used to induce of LTD (see (1990), and Friedlander et al. (1993) among others. Thus,
Section 10.7) perform a priming function; they do not them- although we tend to regard plasticity as a mechanim for driv-
selves induce LTD but enable the later stimuli in the train to do ing changes in synaptic function, these formal sets of rules, as
so by an NMDAR-dependent mechanism (Mockett et al., well as the phenomenon of metaplasticity, remind us that
2002). An example of state dependence, which can be regarded plasticity can also be the engine of stability. If LTD is easier to
as a form of metaplasticity, is provided by the observation that induce after LTP (and vice versa) the strength of connections
LTD cannot be induced in newly unsilenced synapses for the will tend to stabilize around a mean. Similar adjustments in
rst 30 minutes after unsilencing (Montgomery and Madison, cell excitability are discussed in Section 10.3.14.
2002) (see Section 10.7.6). Another, extreme case of metaplas- The patterns of synaptic activity that lead to metaplasticity
ticity is exhibited by synapses made by mossy bers on are in many cases comparable to those that induce LTP or
interneurons in stratum lucidum of area CA3, at these LTD. In some forms of metaplasticity the NMDA receptor
synapses, depending on prior activity, tetanic stimulation of plays an essential role, presumably reecting the plasticity of
mossy bers can produce either LTP or LTD (Pelkey et al., NMDA receptor-mediated responses themselves (see Section
2005) (see Section 10.8.1) for the mechanism underlying this 10.4), whereas in others, such as that mediated by the molec-
striking instance of bidirectional plasticity. A striking case of ular switch (Section 10.3.10), metabotropic glutamate recep-
metaplasticity is exhibited by synapses made by mossy bers tors are involved. It is likely, then, that the cellular correlates of
on interneurons in stratum lucidum of area CA3; at these metaplasticity overlap with those of LTP and LTD, and the
synapses, depending on prior activity, tetanic stimulation of point has been made, with some justice, that biochemical cor-
mossy bers can produce either LTP or LTD (Pelkey et al., relates of metaplasticity may be mistaken for biochemical cor-
2005) (see Section 10.8.1) relates of LTP or LTD (Abraham and Tate, 1997).
Some years before the rst experimental studies of meta- A possible candidate for the molecular basis of metaplas-
plasticity, the idea had been explored by Bienenstock, Cooper, ticity is the NMDA receptor. The 2B subunit of the NMDA
and Munro in a theoretical analysis of developmental plastic- receptor is phosphorylated in LTP, an effect that in the
ity in the kitten visual system (Bienenstock et al., 1982; Bear et anesthetized animal peaks between 2 and 5 hours after induc-
al., 1987). In the Bienenstock, Cooper, and Munro (BCM) tion (Rosenblum et al., 1996; Rostas et al., 1996). In freely-
model, the rate of change in synaptic weight, f, is a function of moving animals, LTP in the dentate gyrus is associated with an
postsynaptic activity; it is negative for low rates of activity and increase in NR1 and NR2B levels that are still present 48 hours
positive for high rates of activity, corresponding to LTD and after induction (Williams et al., 2003). Phosphorylation of
LTP respectively (see Box 102). The critical new feature of NR2B in expression systems enhances NMDAR-mediated
the BCM formulation is that the threshold for LTP, , is not currents by increasing the open time of the channel. This has
static but is itself a function of the average activity of the cell. little effect on the largely AMPA receptor-mediated synaptic
For maintained high rates of activity, the threshold slides to responses (and thus cannot be the mechanism for LTP itself),
the right, facilitating the subsequent induction of LTD, but it does have consequences for subsequent attempts to
whereas for low rates of activity, the threshold moves to the induce plasticity. On the face of it, phosphorylation of NR2B
left, encouraging the induction of LTP. An element of homeo- and the increased receptor subunit levels constitute a positive
stasis is thereby introduced to the system, preventing runaway feedback system. However, based on the assumption that both
potentiation and allowing situations in which activity is low to LTD and LTP are triggered by increases in intracellular Ca2,
regain synaptic strength when the input is restored. The BCM with the threshold for LTD being lower than that for LTP, both
model accounts for several features of metaplasticity in the types of plasticity are facilitated by the increase in NMDAR-
hippocampus. The facilitation of LTD (depotentiation) after mediated Ca2 current. The entire BCM curve, including , is
induction of LTP is one example (to our knowledge, there are shifted to the left; at low frequencies LTD is facilitated, and at
no data on the converse prediction that LTP should be easier intermediate and higher frequencies LTP is facilitated. Just this
to obtain after induction of LTD); saturation of LTP is combination of effects is seen in the visual cortex following
another. Other aspects are not so easily explained by the BCM enhanced NMDAR function driven by sensory deprivation
model, such as the fact that metaplastic effects commonly per- (Philpot et al., 2003). Another hint that changes in NMDA
sist for only an hour or two, and the essentially homosynaptic receptor function may underlie the expression of metaplastic-
nature of metaplasticity in the hippocampus. In the BCM the- ity comes from the nding that low-intensity trains of the sort
ory, is a global threshold applicable to all synapses and is a that impair the subsequent induction of LTP in area CA1 also
function only of postsynaptic activity, which itself reects the induce LTD of the NMDAR-mediated component of the
integrated activity of all inputs to the cell. Yet metaplastic evoked response (Selig et al., 1995a).
effects in the hippocampus are for the most part input- Whatever the mechanisms by which changes in activity
specic and do not seem to be much affected by heterosynap- lead to metaplasticity, there is evidence that it may be affected
tic activity (Abraham and Tate, 1997). by a variety of neuromodulators including corticosteroids,
The BCM rule is not the only synaptic learning rule with estrogen, catecholamines, acetylcholine, and serotonin. In
built-in homeostasis; other models are based on covariance general, these modulators lead to an increase in the threshold
366 The Hippocampus Book

Box 102
Synaptic Plasticity and the BCM Curve

The BCM function was introduced by Bienenstock, Cooper, and Munro (Bienenstock et al.,
1982) to provide a mathematical description of activity-dependent changes in synaptic weight
in the developing visual system (see Bear, 1996, for its application to hippocampal plasticity).
The BCM function, f, is a function of c, the ring rate of the postsynaptic cell; it changes sign at
a value of c called the threshold, ; this itself is a function of c and the variable c (the mean
rate of postsynaptic activity, measured over minutes or hours). Thus f is a function of both c
and c. In the BCM theory, the rate of change of synaptic efficacy, mj, at the jth synapse on a
cell, is given by the equation

dmj/dt  f(c,c)dj

where dj is the rate of activity at the jth synapse. Note that the value of the BCM function, f, is
determined only by the state of the postsynaptic cell. Synaptic plasticitythe rate of change of
synaptic weightsis a product of f and afferent activity, dj.
The function f is shown graphically in Box Fig. 102A. Note that an active synapse will be
potentiated if it is active when c  . This is likely to happen only when the cell is driven at
high frequency, as with tetanic stimulation (Box Fig. 102B) or its postsynaptic state is manipu-
lated, as by application of AP5 (Box Fig. 102C) or by depolarization (Box Fig. 102D). If a
synapse is activated alone at low frequency, c is less than , and efficacy will be depressed. In
general, is a nonlinear function of c; see Bear et al. (1987) for a discussion of the case
where  c. The magnitude of the change in synaptic weight is proportional to dj,, the
rate of afferent activity; and there is no change when afferent activity is not driven (dj  0).
Note also that the BCM formulation bequeaths inherent stability to the network: When c is
high, the threshold shifts to the right, so synaptic weights tend to be reduced, and postsynaptic
activity is correspondingly dampened; and when the mean afferent activity is low, shifts to
the left, synaptic weights increase, and activity is restored to the network. This prevents synap-
tic weights from increasing explosively when activity is high or decreasing to zero when activity
is low.

A B
Change in weight of activated synapse

+ + 30
Value of BCM function, f

20
fEPSP (% change)

10

0 0 0

-10

-20

- - -30
0.1 1 3 10 50
Firing rate, c, of postsynaptic cell Frequency of tetanus (Hz)

C D
30 3
d
Amplitude (normalized)

20
Depresse

d
fEPSP (% change)

te
2 tia
ive

10 n
te
Po
Na

-10 1

-20

-30 0 -70 -50 -30 -10 +10 +30

50 20 15 7.5 0
Concentration of AP5 (M) Vm
(Continued)
 Synaptic Plasticity in the Hippocampus 367

The curve relating tetanus frequency to the magnitude and polarity of the change in synaptic
efficacy is sometimes interpreted as if it is a direct illustration of the function, f, in the BCM
theory. However, although the two sets of curves have a similar form, passing from negative
values to positive values at a modiable threshold, the changes in synaptic efficacy are a func-
tion of postsynaptic ring rate in BCM theory and of presynaptic activity along a restricted set
of afferent bers in experimental investigations of bidirectional plasticity. The two descriptions
approach congruence if presynaptic activity in the stimulated afferent(s) is the main determi-
nant of cell ring; this may be the case in vitro where spontaneous activity is low, but is not the
case in general. Reversal from LTD to LTP can also be achieved in single cells by manipulating
the membrane potential; in strongly depolarized cells, brief trains induce LTP, whereas the
same trains given when the cell is weakly depolarized induce LTD (Cummings et al., 1996;
Ngezahayo et al., 2000) (see Box Fig. 102D), conrming that the polarity of change is under
the control of the postsynaptic cell. The reversal point for LTD/LTP ( in the BCM curve) is a
function of past activity. For example, the prior induction of LTP in vitro shifts the reversal
point (in terms of membrane depolarization) to the right, whereas the prior induction of LTD
shifts it to the left (Ngezahayo et al., 2000) (see Box Fig. 102D). Similarly, prolonged prior
antidromic or heterosynaptic activity in the dentate gyrus of the freely moving rat raises the
threshold for LTP (Abraham et al., 2001). In each case, synaptic weights are changed in a
direction that restores the output of the cell toward a non-zero mean.

Box Fig. 101. Bienenstock, Cooper, and Munro (BCM) curve and bidirectional synaptic
plasticity. A. The BCM function f is negative for low ring rates of the postsynaptic cell,
crosses zero at a threshold value () and is positive thereafter. Because changes in synaptic
weight at a given synapse are given in the BCM theory by f.d, where d is the rate of afferent
activity at that synapse, the curve for changes in synaptic weight has the same shape. (Source:
After Bienenstock et al., 1982.) B. Bidirectional changes in synaptic weight can be achieved by
manipulating the frequency of afferent stimulation. The total number of stimuli (900) was the
same in each case. The result is consistent with the BCM theory, as changes in stimulus fre-
quency affect the ring of the postsynaptic cell. (Source: Bear and Linden, 2001; after Dudek
and Bear, 1992.) C. Pharmacological manipulation of the postsynaptic NMDA receptor with
varying concentrations of AP5 elicit bidirectional plasticity. Afferent stimulation was the same
in each case (600 stimuli at 20 Hz). In control medium, this protocol induced LTP. With weak
to intermediate concentrations of AP5, sufficient to reduce the ring of the postsynaptic cell to
afferent stimulation, LTD was induced. Finally, at a concentration of 50 m, changes in synap-
tic weight were blocked. (Source: Bear and Linden, 2001; after Cummings et al., 1996.) D.
Bidirectional changes in synaptic weight produced by pairing afferent stimulation with depo-
larizing pulses to the postsynaptic cell. In each case, 100 afferent stimuli were given at 2 Hz. The
polarity and magnitude of the change in synaptic efficacy was determined by the membrane
potential to which the postsynaptic cell was clamped during pairing. In nave cells (not previ-
ously subjected to pairing), depolarization to 30 mV resulted in LTD, whereas depolarization
to 10 mV or more led to LTP. Threshold () was at 20 mV. This experiment also revealed
how prior activity can alter . Once LTD had been induced,  shifted to the left; and pairing to
30 mV, which in the nave cell had induced LTD, now resulted in no change. Conversely, if
LTP was induced, it was subsequently possible to induce LTD at a more modest depolarizing
potential, implying a rightward shift of  (dotted line, not experimentally conrmed). (Source:
Ngezahayo et al., 2000.)

for LTP and a decrease in the threshold for LTD (reviewed by (threonine 286 to aspartate), the enzyme was rendered cal-
Abraham and Tate, 1997). cium-independent. The effect was to shift the curve relating
the frequency of the conditioning train to the change in syna-
Transgenic Mice Can Exhibit Shifts in the BCM Curve ptic efficacy to the right in CaMKII(T286D) animals relative
to wild-type littermates; thus, protocols that in wild-type mice
An example of a genetic modication leading to a striking produced LTP led to LTD in the transgenic animals. A similar
shift in the frequency dependence of synaptic plasticity is the shift to the left of the frequency or BCM curve for synaptic
transgenic mouse carrying a mutated form of CaMKII engi- plasticity occurs in other mutants, including an NR2B overex-
neered by Mayford et al. (1995). By changing a single residue pressing transgenic mouse (Tang et al., 1999) and a calcineu-
368 The Hippocampus Book

rin knockout (Zeng et al., 2001). In contrast, mice lacking the Finally, the increase in excitability of neurons in hippocampal
GluR2 subunit (Jia et al., 1996) or the postsynaptic density cultures that is induced by prolonged blockade of activity
protein PSD95 (Migaud et al., 1998) showed a shift to the left appears to reect an up regulation of synaptic AMPARs,
of the frequency-response curve; thus, stimulus protocols that driven by glial release of the cytokine tumor-necrosis factor-
produced LTD in wild types induced LTP in the mutant ani- (TNF-) (Stellwagen and Malenka, 2006).
mals. The shift in baseline exhibited by these and other Activity-dependent changes in intrinsic excitability have
mutants suggests that there may be two mechanistically dis- also been documented in hippocampal, cortical, and cerebel-
tinct forms of metaplasticity: (1) homosynaptic (input-spe- lar neurons (for review see Zhang and Linden, 2003). An
cic) metaplasticity lasting not more than a few hours and important insight was gained when two groups reported that
attributable to post-translational changes in, for example, persistent changes in local dendritic excitability accompany
NMDA receptors; and (2) a more persistent, generalized form synaptic plasticity in CA1 pyramidal neurons. In one case, a
of metaplasticity that regulates baseline and requires gene theta-burst pairing protocol produced a long-lasting shift in
expression (Deisseroth et al., 1995; Abraham and Tate, 1997). the inactivation curve of the K conductance IA, leading to
local boosting of synaptic responses and back-propagating
10.3.14 Synaptic Scaling and Long-term spikes (Frick et al., 2004). In the other study, spike timing-
Changes in Intrinsic Excitabililty dependent plasticity was accompanied by persistent changes
in the cationic current Ih (Wang et al., 2003). These ndings
A neural network in which Hebbian synapses provide the only imply that the throughput of synaptic information to the
mechanism for plasticity cannot sustain a state of dynamic sta- soma is enhanced not just from the tetanized region but also
bility in which the mean rate of activity remains constant, from distal synapses on the same dendrite.
while the potential for modication of individual synapses is Another intriguing nding comes from the behavioral lit-
retained. In a Hebbian network, successful inputs are strength- erature. About 50% of CA1 pyramidal cells examined in slices
ened, becoming more successful, whereas unsuccessful inputs from rabbits trained in a trace conditioning task (eyeblink
remain unsuccessful or are depressed; the network is ulti- conditioning using a tone as the conditioned stimulus and an
mately driven to an invariant state where activity is controlled airpuff as the unconditioned stimulus) exhibited a reduction
by a subset of maximally potentiated synapses. Several theo- in spike frequency adaption and spike after-hyperpolarization
retical xes to the problem of runaway excitation due to (Moyer et al., 1996). (Although these changes in excitability
Hebbian positive feedback have been proposed. In the covari- last several days, they cannot encode the memory for the con-
ance rule proposed by Stent (1973), a synapse is potentiated ditioned response, which can last months). A similar reduc-
when activity in the pre- and postsynaptic cells are in phase tion in after-hyperpolarization is seen in rat CA1 neurons
and depressed when they are out of phase. There is some after spatial training in the watermaze (Oh et al., 2003).
experimental evidence for Stents rule (Artola and Singer, Activity-dependent changes in excitability, in contrast to the
1993; Debanne et al., 1998). As we saw above, in the inuential synaptic scaling produced by prolonged hypo- or hyperactiv-
model proposed by Bienenstock et al. (1982) and extended and ity, are in general nonhomeostatic because they act in the
updated by Shouval et al. (2002), the magnitude and polarity same direction as synaptic changes. Note also that changes in
of plastic changes are controlled by the overall activity of the local dendritic excitability introduce a potential for metaplas-
network to maintain a constant mean ring rate (see also ticity, as they may affect subsequent STDP. Exploration of
reviews by Abbott and Nelson, 2000 and Bi and Poo, 2001). long-term changes in excitability is at an early stage, and rela-
There is now considerable evidence that homeostatic tively little is known about the conductances responsible or
mechanisms operate to maintain a constant mean level of r- the second messenger systems that modulate them.
ing in neural networks. Homeostasis is achieved by overall
adjustments to synaptic weights, a process called synaptic Back-propagating Synaptic Plasticity
scaling. The phenomenon was rst studied in cultured corti- and Long-range Cytoplasmic Signaling
cal networks. Addition of a GABAA antagonist to the culture
medium produced an initial increase in spontaneous activity, Back-propagation of synaptic weight changes is the central
but activity returned to control levels over the following 2 days feature of an efficient design for unsupervised learning in
(reviewed by Turrigiano and Nelson, 2004). In cultured hip- model neural networks (Rumelhart et al., 1986). There was,
pocampal neurons, similarly, the GABAA antagonist picro- however, little evidence for back-propagation in biological
toxin produced a long-term reduction in surface AMPA nervous systems until the studies by Poo and his colleagues in
receptors (Lissin et al., 1998). Homeostasis can also operate at hippocampal cultures (Fitzsimonds et al., 1997; Tao et al.,
the level of the single hippocampal pyramidal cell embedded 2000). LTP or LTD induced at a particular connection by
in an active network of cultured hippocampal neurons. spike pairing of presynaptic and postsynaptic neurons in the
Overexpression of an inward-rectier K channel resulted in an appropriate temporal conjunctionresults in the spread of
initial decrease in the ring rate of the transfected cell, but plasticity (of the same polarity) to other synapses onto the
over time the ring rate was restored to control values by a presynaptic neuron and to other synapses made by the presy-
homeostatic increase in synaptic drive (Burrone et al., 2002). naptic neuron. However, there is no spread to other connec-
 Synaptic Plasticity in the Hippocampus 369

tions made by or onto the postsynaptic neuron (Bi and Poo, strength and decays with the same time course when afferent
2001). Thus, LTP at a given synapse is propagated only to stimulation is resumed, even when the intervening period is as
synapses made by or onto the presynaptic neuron. The presy- long as 6 hours. [As a result, it was proposed that STP may be
naptic cell was voltage-clamped except when required to re more accurately referred to as a transient form of LTP (T-
an action potential to induce the primary LTP; it is therefore LTP); however, here we retain the more widely known term,
likely that back-propagated changes are not activity-driven STP.] In addition, it was found that the magnitude of STP was
but involve some form of cytoplasmic long-range signaling. inuenced by the frequency of afferent stimulation, such that
There are as yet no reports of back-propagating plasticity in higher frequencies favored larger STP. This nding may
hippocampal slices or in the hippocampus in vivo. However, explain why the extent of STP observed in different experi-
in the developing retinotectal system of the tadpole, applica- ments is highly variable. In particular, a brief high-frequency
tion of BDNF to the tectum (where axons of retinal ganglion tetanus followed by a low rate of test stimulation (as typically
cells terminate) results in potentiation of retinotectal synapses used in eld potential recordings) would favor the generation
and a back-propagated increase in the number of glutamate of a large-magnitude STP with slow decay, whereas an inter-
receptors on dendrites of ganglion cells (Du and Poo, 2004). mediate frequency of stimulation for both the induction and
expression periods (as typically used during pairing experi-
ments) militates against STP. This may explain why pairing
induced potentiation generally lacks the STP component.
10.4 NMDA Receptor-dependent LTP:
Expression Mechanisms 10.4.3 Early LTP Involves Multiple Protein
Kinase-dependent Mechanisms
10.4.1 From Induction to Expression of LTP
Many studies have presented evidence that protein kinases are
The various properties of LTP that we have summarized involved in LTP (Fig. 108). Most of this work relates to the
above arise from the complex interplay of biophysical and study of E-LTP. With few exceptions, STP is resistant to kinase
cell-biological mechanisms operating at the cellular and local inhibitors. The role of protein kinases in L-LTP is discussed
circuit levels. In approaching this complexity, it is convenient later in this section.
to distinguish between the induction of LTP, which comprises
the collection of short-lived events that trigger changes in Protein Kinase C
synaptic weight but do not themselves affect it, and the expres-
sion of LTP, which includes all those mechanisms, whether The rst kinase to be implicated in LTP was protein kinase C
presynaptic or postsynaptic, that directly enhance or, in the (PKC) (Lovinger et al., 1987) (Fig. 109A). However, the
case of LTD, diminish synaptic efficacy. (Note that, in addition inhibitors used in these initial studies were relatively nonspe-
to expression, two other equivalent terms are found in the LTP cic, so the effects observed may have been via actions on
literature: persistence and maintenance.) Thus, events associ- other kinases. Indeed, several reports have observed no effect
ated with the activation of NMDA receptors, relief of the mag- of potent, more specic PKC inhibitors on LTP (see Muller et
nesium block, permeation of Ca2 through the NMDA al., 1992; Bortolotto and Collingridge, 2000). On the other
receptor, and activation of Ca2-dependent protein kinases hand, a role for PKC has been proposed based on the use of
can all be considered components of induction. Expression peptide inhibitors applied intracellularly (Malinow et al.,
mechanisms include increases in the probability of transmit- 1989). In addition, effects of more selective (though not
ter release, phosphorylation of AMPA receptors and their totally specic) PKC inhibitors have been described under
insertion into the postsynaptic density. certain circumstances, for example during de-depression (the
tetanus-induced restoration of the response in a pathway in
10.4.2 STP Is a Transient which LTD has previously been induced) (Daw et al., 2000).
Presynaptic Form of Plasticity The most likely explanation is that PKC is involved in some,
but not all, aspects of E-LTP. Understanding the role of PKC in
In an early experiment to probe for changes in postsynaptic LTP is complicated by the existence of multiple isoforms; the
sensitivity, it was observed that LTP was associated with a Ca2 and diacylglycerol-activated, conventional PKCs (cPKCs:
delayed increase in sensitivity to AMPAR agonists (Davies et , I, II , ); Ca2-independent, diacylglycerol-activated,
al., 1989). This suggests that whereas E-LTP correlates in time novel PKCs (nPKCs: , , , ); and the Ca2-independent,
with alterations in postsynaptic function, STP occurs without diacylglycerol-independent atypical PKCs (aPKCs: , ). The
any such changes. The simplest interpretation of this result is spectrum of subtype selectivity of the various inhibitors is
that STP is presynaptic, a conclusion supported by a detailed incomplete, and there is limited information available from
paired-pulse facilitation analysis (Volianskis and Jensen, knockout experiments. Elimination of the neuron-specic
2003). This study revealed a surprising property of STP: It isoform PKC leads to inhibition of LTP, although not if the
decays in an activity-dependent manner. If afferent stimula- LTP is preceded by low-frequency stimulation (Abeliovich et
tion is discontinued after tetenization, STP appears at full al., 1993). This suggests that PKC may play a regulatory role
370 The Hippocampus Book

Figure 108. Signal transduction mechanisms in early LTP. The postsynaptic membrane (not shown). These post-translational
trigger for the induction of LTP is Ca2 permeating through the processes contribute to E-LTP. Protein synthesis-dependent mecha-
NMDA receptor. This is amplied by Ca2 release from Ca2/IP3- nisms at the synapse and by signals passing to the nucleus are
sensitive stores. The amplied Ca2 signal together with other responsible for L-LTP. Putative retrograde messengers, Nitric oxide
activators of protein kinases (zigzag arrows) then leads to the (NO) and arachidonic acid (AA), that may stimulate presynaptic
phosphorylation of substrate proteins, including AMPA and changes are also indicated. (Source: After Bliss and Collingridge,
NMDA receptors, and the insertion of AMPA receptors into the 1993.)

in the induction of LTP. Of great interest is the nding by consists of dodecamers of  and subunits. In its inactive
Sacktor and colleagues that a peptide substrate inhibitor of state, the ATP binding site of the catalytic region of the
PKM, the autonomously active catalytic domain of PKC, enzyme is bound to an autoinhibitory domain in the regula-
blocks the expression of pre-established L-LTP (Ling et al., tory region. When calcium-loaded calmodulin binds to a
2002; Serrano et al., 2005; see Section 10.4.9). Bath application calmodulin-binding region in the 3 region of the molecule, a
of the inhibitor at the time of the tetanus has no effect on E- conformational change occurs that separates the autoin-
LTP but blocks the development of L-LTP. hibitory domain from the ATP binding site and renders the
In other experiments phorbol esters have been used to acti- enzyme active. Following a transient surge in calcium concen-
vate PKC. This treatment leads to a potentiation of synaptic tration, intramolecular autophosphorylation at threonine 286
transmission which supercially resembles LTP (Malenka et can occur, producing a conformation that allows the enzyme
al., 1986; Hu et al., 1987). However, the potentiation is to remain autonomously active when calcium levels fall,
reversible following washout of the phorbol ester and does not thereby endowing the enzyme with the properties of a molec-
occlude with tetanus-induced LTP. Moreover, phorbol esters ular switch (Miller and Kennedy, 1986). Two isoforms of the
directly interact with other proteins, such as munc-13, a pro- calcium/calmodulin-dependent protein kinase II are present
tein involved in neurotransmitter release, and so any effects of in the adult hippocampus, CaMKII and CaMKII, of which
phorbol esters may be unrelated to activation of PKC the dominant form in area CA1 is CaMKII. There is a wealth
(Kazanietz et al., 2000). of evidence to indicate that CaMKII is required for and,
The trigger for the induction of PKC-dependent potentia- when activated, capable of producing LTP in area CA1 of the
tion may be Ca2 entering via NMDA receptors. An alternative adult animal (reviewed by Lisman et al., 2002). First, specic
route is via glutamate binding to group I mGluRs (mGluR1 peptide inhibitors of CaMKII, based on the autoinhibitory
and mGluR5), the activation of which leads to production of domain of the enzyme, which suppress both calcium-depend-
diacylglycerol and release of Ca2 from intracellular stores ent and calcium-independent activity, block induction of LTP
the two triggers for activation of cPKCs. In this context, potent (Malinow et al., 1989) (Fig. 109B). Second, the membrane-
PKC inhibitors prevent mGluR-mediated metaplasticity permeable CaMKII inhibitors KN-62 and KN-93 block induc-
(Bortolotto and Collingridge, 2000). tion of LTP in a reversible manner (Ito et al., 1991). Third, LTP
is impaired in CaMKII/ mice (Silva et al., 1992b), although
CaMKII there may be a background effect contributing to this decit
(Hinds et al., 1998). Fourth, mice with a threonine to alanine
Calcium/calmodulin-dependent protein kinase II (CaMKII) is point mutation at residue 286 that prevents autophosphoryla-
a major component of the postsynaptic density, comprising 2% tion do not display LTP in area CA1 (Giese et al., 1998). In
of its protein content (Lisman et al., 2002). The holoenzyme addition, LTP is associated with the dendritic accumulation of
 Synaptic Plasticity in the Hippocampus 371

Figure 109. Role of protein kinases in the induction and expres- al., 1989.) C. The PKA inhibitor Rp-cAMPS blocks L-LTP in the
sion of LTP. A. The PKC inhibitor mellitin, injected into the dentate CA1 region of the hippocampus in vitro. STP and E-LTP are unaf-
gyrus in vivo 15 minutes prior to a tetanus, blocks E-LTP but spares fected. (Source: Matthies and Reymann, 1993.) D. The tyrosine
STP. Similar ndings were obtained with polymyxin B and H-7, kinase inhibitor genistein blocks E-LTP but not STP at CA1
implicating PKC as the kinase responsible. (Source: Lovinger et al., synapses in vitro. Similar results were obtained using lavendustin
1987.) In this and subsequent panels the slope of the fEPSP (or A. (Source: ODell et al., 1991a.) E. The MAPK inhibitor PD098059
intracellular EPSP in B) is plotted versus time. Control experiments blocks induction of E-LTP but not STP at CA1 synapses in vitro.
are also shown, except in B. In each case, a tetanus was delivered at (Source: English and Sweatt, 1997.) F. ICV application of the
t  0. B. A peptide inhibitor of CaMKII [CaMKII(273-302)] loaded PI3K inhibitor wortmannin blocks induction of E-LTP at per-
into CA1 neurons in vitro blocks E-LTP but not STP. Similar results forant path-granule cell synapses in vivo. (Source: Kelly et al.,
were obtained using a peptide inhibitor of PKC. (Source: Malinow et 2000.)
372 The Hippocampus Book

activated CaMKII and an associated phosphorylation of al., 1995; Otmakhova et al., 2000). It seems that the role of PKA
AMPARs (Fukunaga et al., 1995). Finally, CaMKII mimics and is determined by a variety of factors, such as the strength and
occludes the induction of LTP (Pettit et al., 1994). However, patterns of the high-frequency trains used to induce LTP.
as compelling as the evidence is for involvement of CaMKII in Typically, PKA inhibitors are more effective when strong
LTP under some circumstances, the situation is not as strai- induction protocols are used (Nguyen and Woo, 2003). They
ghtforward as it may seem. Although there is general agree- are also more effective early in development, before the
ment that the induction of LTP is associated with a persistent CaMKII-dependent form of LTP is expressed (Yasuda et al.,
( 1 hour) phosphorylation of CaMKII, there is disagree- 2003). Interestingly, the sensitivity of E-LTP to PKA inhibitors
ment about whether the increase in enzyme activity is tran- is also dependent on the prior level of experience of the ani-
sient (Lengyel et al., 2004) or persistent (Fukunaga et al., mal, such that environmental enrichment results in a larger
1995). Also there is currently no direct evidence that a consti- LTP with the additional component being selectively sensitive
tutively active form of CaMKII is involved in the maintenance to PKA inhibition (Duffy et al., 2001).
of LTP. Inhibitors of CaMKII applied following the induction Transient bath application of the constitutively active
of LTP have no effect (Ito et al., 1991; Chen et al., 2001b). cAMP analogue Sp-cAMPS can produce a persistent, slow-
However, these inhibitors compete with the calmodulin bind- onset enhancement of synaptic transmission in area CA1 that
ing site, and hence the Ca2/calmodulin-dependent activation occludes with tetanus-induced LTP (Frey et al., 1993).
of the enzyme, but do not affect the constitutively active form. However, like the PKA-dependent component of tetanus-
Thus whilst persistent re-activation of the enzyme can be dis- induced LTP, this is not invariably observed. For example, in
counted as a mechanism it is possible that a constitutively one study cAMP analogues did not affect basal transmission
active form of CaMKII maintains LTP. Specic membrane but reversed LTD, implicating PKA in de-depression
permeable inhibitors of the constitutively active form are (Kameyama et al., 1998). Intriguingly, there is a strong strain
required to test this possibility. What is clear however is that difference in the ability of mice to express cAMP-induced
CaMKII is not ubiquitous in its involvement in NMDAR- facilitation (Nguyen et al., 2000).
dependent LTP. For example, in area CA1 at an early stage in These various observations suggest that the role of PKA is
development (P9) CaMKII inhibitors have no effect on the probably to modulate, rather than directly mediate, NMDAR-
induction of LTP (Yasuda et al., 2003), and at an intermediate dependent LTP. One idea is that PKA gates the activity of
stage of development (P14) CaMKII forms one arm of a par- the pathway leading from activation of CaMKII to the expres-
allel kinase cascade (Wikstrm et al., 2003). Thus, the impor- sion of early LTP (Blitzer et al., 1995) via inhibition of the
tance of CaMKII in LTP increases during development. Even protein phosphatase 1 (PP1), which dephosphorylates
in the adult animal, however, autophosphorylation of CaMKII at ser831 and ser845 (Blitzer et al., 1998). Whether
CaMKII is not a general requirement for NMDAR-depend- LTP-inducing stimuli induce an increase in the level of cAMP
ent LTP, since potentiation in the medial perforant path of the is disputed (Chetkovich and Sweatt, 1993; Blitzer et al., 1995;
dentate gyrus is minimally affected in T286A mutant mice; in Pokorska et al., 2003). If there is no increase, constitutive lev-
these mice autophosphorylation at residue 286 cannot occur els of cAMP must be adequate to maintain the gate in its open
(Cooke et al, 2006). However, as in juvenile CA1, there is evi- state.
dence for involvement of CaMKII as part of a parallel pathway
also involving PKA and MAPK (Cooke et al, 2006; see below). Tyrosine Kinases
Finally, it should be noted that CaMKII is involved in meta-
plasticity in area CA1 (Bortolotto & Collingridge, 1998) and Initial pharmacological evidence, using relatively nonspecic
so biochemical changes that accompany LTP may relate to PTK inhibitors, suggested that protein tyrosine kinases (PTKs)
metaplasticity rather than the plasticity per se (see Abraham are necessary for E-LTP (ODell et al., 1991a Fig. 109D).
and Tate, 1997). Various nonreceptor tyrosine kinases are expressed in the cen-
tral nervous system, including src and the src family members
Protein Kinase A fyn and lyn. Genetic elimination of fyn resulted in a decit in
LTP (ODell et al., 1991a). Although this knockout was associ-
Protein kinase A (PKA) is another kinase involved in LTP, and ated with gross structural changes in the dentate gyrus, a suc-
again its role is complex and not fully understood. In general, cessful rescue experiment suggested that the LTP decit
PKA inhibitors appear to block L-LTP but spare E-LTP reected the loss of fyn in the adult, rather than impaired neu-
(Matthies and Reymann, 1993; Huang and Kandel, 1994) (Fig. ronal development (Kojima et al., 1997). The role of fyn
109C) via an action in the postsynaptic neuron (Duffy and appears to be developmentally regulated, as no LTP decits are
Nguyen, 2003). A similar conclusion has, in general, been seen in animals at less than 14 weeks. Fyn probably acts as a
reached on the basis of various genetic experiments in mice regulatory molecule, as LTP can be induced in the fyn knock-
(Nguyen and Woo, 2003). However, LTP lasting many hours out when a strong induction protocol is used. One potential
can be induced in the presence of PKA inhibitors (Bortolotto mechanism for how fyn regulates LTP is via phosphorylation
and Collingridge, 2000). Conversely, E-LTP is affected by PKA of NMDARs (Rosenblum et al., 1996; Rostas et al., 1996) to
inhibitors in some cases (Huang and Kandel, 1994; Blitzer et increase their function. There is also evidence that src is
 Synaptic Plasticity in the Hippocampus 373

involved in the induction of LTP via upregulation of the func- and potential enhancement of LTP (Yuan et al., 2002;
tion of NMDARs (Lu et al., 1998). This pathway involves the reviewed by Sweatt, 2004).
focal adhesion kinase CAK /PyK2 upstream of src (Huang Y
et al., 2001). PI3 Kinase

MAP Kinase Phosphoinositide 3-kinase (PI3K) phosphorylates phos-

phatidylinositol 4,5-bisphosphate [PtdIns(4,5)P2, or PIP2] to
Of the several mitogen-associated protein (MAP) kinase cas- produce phosphatidylinositol 3,4,5-trisphosphate [PtdIns
cades that carry extracellular signals to the nucleus, the two (3,4,5)P3, or PIP3], which in turn can activate Ca2-inde-
that have been extensively studied in the context of LTP are pendent isoforms of PKC (such as PKC), Akt/PKB, and
inititiated by the small G proteins Ras and Rap-1. They acti- phospholipase C. PI3K was rst implicated in NMDAR-
vate the MAP kinase kinase kinases Raf-1 and B-Raf, respec- dependent LTP in the dentate gyrus, where it was shown that
tively, the rst of the kinases in the three-kinase cascade. The the inhibitor wortmannin inhibited E-LTP but not STP at per-
two pathways converge on the second kinase, MEK (MAP forant path synapses (Kelly and Lynch, 2000) (Fig. 109F).
kinase kinase), which in turn phosphorylates a pair of MAP Similarly, inhibition of E-LTP but not STP was observed at
kinases known as extracellular regulated kinases 1 and 2 CA1 synapses using a different PI3K inhibitor, LY294002
(ERK1, or p44MAPK, and ERK2, or p42MAPK). The (Raymond et al., 2002). In the dentate gyrus in vivo, this
ERK/MAP kinase pathway can be positively regulated by PKA inhibitor blocked E-LTP induced by single but not multiple
via the Rap-1, B-raf pathway and by PKC via the Ras, Raf-1 trains, suggesting that P13K is not an obligatory enzyme in
pathway. PKA can also inhibit the latter pathway. The conver- LTP induction (Horwood et al., 2006). Interestingly, PI3K may
gence of initial pathways onto MEK during the second stage of be required specically for the expression of LTP, as inhibition
the cascade, and the fact that MAPKs are the only known sub- of the enzyme led to depression of preestablished LTP and,
strates for MEK has provided a target for pharmacological following washout of the inhibitor, recovery of LTP to its ini-
intervention. Three specific inhibitors are available tial level (Sanna et al., 2002; but see Opazo et al., 2003). The
PD098059, U0126, SL327the last of which can cross the downstream processes modulated by PI3K are not known but
blood-brain barrier. may involve both MAPK-dependent and MAPK-independent
English and Sweatt were the rst to explore the role of the pathways (Opazo et al., 2003) that affect the surface expres-
MAPK pathway in LTP. They found that LTP-inducing stimuli sion of AMPARs (Man et al., 2003).
resulted in increased phosphorylation of ERK1, 2 (English
and Sweatt, 1996) and that MEK inhibitors blocked both E- LTP Involves Multiple and Parallel Kinase Cascades
LTP and L-LTP in area CA1 of hippocampal slices (English
and Sweatt, 1997) (Fig. 109E). However, MAPK seems to be The principal message to emerge from studies investigating
involved in only some forms of LTP. For example, in one study the role of kinases in NMDA receptor-dependent LTP is that
MEK inhibitors blocked induction of a form of theta-induced several kinases are involved but none is obligatory. The situa-
LTP but did not affect pairing-induced or tetanus-induced tion is complicated by the fact that the kinase cascades under-
LTP (Opazo et al., 2003). The reduction in E-LTP has also lying LTP change during development. In area CA1 of the
been questioned by Kelleher et al. (2004b), who found that adult animal, as we have seen, CaMKII is required for the
LTP in a transgenic mouse expressing a dominant negative induction of LTP, and LTP can be induced in the presence of
MEK was indistinguishable from E-LTP generated in normal PKA inhibitors. Earlier, at postnatal day 9, the lead role is
mice in the presence of protein synthesis inhibitors; they con- taken by PKA, as inhibitors of PKA, but not of CaMKII,
cluded that the main substrates of the MAP kinase pathway block induction of LTP (Yasuda et al., 2003; see Fig. 1021C).
were kinases regulating the protein translational machinery in The situation is made even more complex by the existence of
dendrites. parallel kinase cascades, as rst identied at CA1 synapses in
Nevertheless, local targets that might underlie a contribu- juvenile rats. Thus at 2 weeks of age, a combination of in-
tion of the MAP kinase cascade to E-LTP have begun to be hibitors is required to block the induction of LTP. Inhibitors of
identied. Ras, the small G protein that activates the rst CaMKII, PKA, and PKC have no effect on induction when
kinases (Raf and B-Raf) in the MAP kinase cascade, is impli- applied alone. Similarly, a PKA inhibitor plus a PKC inhibitor
cated in AMPA receptor trafficking in LTP, and another G pro- is ineffective. However, a CaMKII inhibitor applied together
tein, Rap, which activates a member of the ERK family, p38, is with either a PKA or a PKC inhibitor fully blocks the induc-
involved in the removal of AMPA receptors from synapses tion of LTP (Wikstrm et al., 2003). This suggests that two
during LTD (Zhu et al., 2002). There is also evidence that parallel kinase pathways, one involving CaMKII and the other
MAP kinases interact with scaffolding proteins and with involving both PKA and PKC, can each mediate LTP. As the
cytoskeletal proteins. A further suggestive nding is that acti- animal matures, CaMKII assumes the dominant role, and
vation of MAP kinase through PKA and PKC can downregu- PKA and PKC become more modulatory in nature. The
late a dendritic A-type K channel in CA1 pyramidal neurons, involvement of parallel kinase cascades is, however, not simply
leading to a reduction in the threshold for back-propagation a developmental issue, since a similar parallel pathway has
374 The Hippocampus Book

been observed in the medial perforant path in the dentate a larger increase in the AMPAR-mediated component (Muller
gyrus of adult mice (Cooke et al, 2006) and rats (Wu et al., et al., 1988; Kullmann et al., 1996; Dozmorov et al., 2006) to a
2006). Here we have a pathway that requires activation of similar increase in both synaptic components (Clark and
CaMKII for one limb and PKA/MAPK for the other. Since Collingridge, 1995; OConnor et al., 1995). The lack of con-
both PKA and PKC can lead to activation of MAPK it is pos- sensus plus the issues of interpretation mean that little can be
sible that in both juvenile CA1 and adult dentate gyrus the concluded from these experiments regarding the locus of
same two parallel pathways co-exist, one arm using CaMKII expression of LTP. Another conclusion to have emerged from
and the other PKA/PKC/MAPK. these studies is that the NMDAR-mediated component of
synaptic transmission is plastic. Interpretation of alterations
10.4.4 Site of Expression of Early in the NMDAR-mediated component of dual-component
LTP: Experimental Approaches synaptic transmission is complicated by the voltage escape
associated with the AMPAR-mediated component coupled
LTP research during the 1990s was enlivened by intense with the voltage dependence of the NMDAR-mediated con-
debate as to whether NMDAR-dependent LTP is expressed ductance. However, LTP has been observed by several groups
presynaptically or postsynaptically. Many experiments following blockade of AMPAR-mediated synaptic transmis-
relied on classical electrophysiological approaches, such as sion (Bashir et al., 1991; Berretta et al., 1991; OConnor et al.,
paired-pulse facilitation, failures analysis, or quantal analysis 1994; Grosshans et al., 2002). Plasticity of the NMDAR-medi-
(see Box 101). Unfortunately, no clear consensus has emerged ated response has obvious functional implications for meta-
owing to the difficulty of obtaining accurate measurements plasticity (see Section 10.3.13).
coupled with an incomplete understanding of the quantal
basis of synaptic transmission at central synapses (see Box Kinetics of Open Channel Blockers Do Not Indicate
101). an Increase in Transmitter Release in LTP
Over the years considerable evidence has accumulated
in favor of both presynaptic (Fig. 1011) and postsynaptic Another approach has been to use open-channel blockers to
(Fig. 1012) mechanisms, and we examine the evidence for estimate the probability of release, Pr. The potent noncompet-
each in turn. First, though, we consider three experimental itive NMDAR antagonist MK-801 blocks NMDA receptors
approaches that were designed to detect changes in transmit- once they have been activated by synaptically released L-gluta-
ter release following the induction of LTP and that, by and mate, and once in the channel the receptor remains blocked
large, failed to do so. during the lifetime of the experiment (Hessler et al., 1993;
Rosenmund et al., 1993). During repetitive stimulation high-
Differential Potentiation of AMPA and Pr synapses are therefore blocked more quickly than low-Pr
NMDA Receptor-Mediated Responses synapses. It was found that LTP of AMPA receptor-mediated
synaptic transmission was not associated with any change in
An electrophysiological approach that has been used to the MK-801 blocking rate of NMDAR-mediated synaptic
address the locus of expression of LTP is based on a compari- transmission (Manabe and Nicoll, 1994) (Fig. 1010B).This
son of the relative potentiation of the AMPA and NMDA was taken as evidence that LTP does not involve an increase in
receptor-mediated components of synaptic transmission Pr. A more direct approach was to use an open-channel
the logic being that an increase in the probability of transmit- blocker of the AMPAR-mediated component. Philanthotoxin
ter release (Pr) should affect both components equally, while (PhTx) blocks GluR2-lacking, Ca2-permeable AMPA recep-
a postsynaptic modication of the AMPA receptor should tors in a use-dependent manner and so can be used to moni-
result in selective potentiation of the AMPA receptor- tor Pr. LTP was not associated with any change in the PhTx
mediated component. This makes the assumption that an blocking rate in the GluR2 knockout mouse (Mainen et al.,
increase in L-glutamate would increase both synaptic compo- 1998) (Fig. 1010C), again indicating no change in Pr. Similar
nents equally. This would not be the case if, for example, experiments could not be performed in wild-type animals
NMDA receptors were saturated, or nearer to being saturated, owing to the presence of GluR2 subunits in the great majority
than AMPA receptorsan issue that still has not been fully of the AMPA receptors on principal neurons.
resolved at hippocampal synapses (see Section 6.6.1). Con-
versely, a similar increase in both synaptic components could Electrogenic Uptake of Glutamate by Glia
occur by a postsynaptic mechanism that affects both AMPA Reports an Increase in Extracellular
and NMDA receptor-mediated components of synaptic trans- Glutamate During PTP but Not During LTP
mission. In addition to these theoretical caveats, there has
been no experimental consensus with respect to whether LTP Yet another approach has been to use glial currents as a moni-
involves a specic increase in the AMPAR-mediated compo- tor of neurotransmitter release. Glial cells accumulate synapti-
nent. While one early study reported a selective increase in the cally released L-glutamate by an electrogenic uptake carrier: An
AMPAR-mediated component (Kauer et al., 1988), others increase in neurotransmitter release should therefore generate
have reported an increase in both components, ranging from a larger glial current. An increase in electrogenic current was
 F

Figure 1010. Evidence that LTP involves postsynaptic changes. A. AMPA receptor-mediated currents by spermine in the GluR2
LTP is not associated with a corresponding alteration in paired- knockout mouse. This experiment was conducted in a manner
pulse facilitation in the dentate gyrus in vivo. The upper plot shows similar to that illustrated in B except AMPA receptor-mediated
the amplitude of the EPSP evoked by the rst stimulus of the pair. synaptic transmission was monitored throughout, and a spermine
The lower plot shows the amplitude of the second EPSP divided by derivative (HPP-SP) was used to block the activated receptors.
that of the rst EPSP for each pair to generate the paired-pulse (Source: Mainen et al., 1998.) D. LTP is not associated with an alter-
ratio. Note that there was a decrease in the paired-pulse ratio only ation in L-glutamate release, as detected using glial transporter cur-
during the initial phase of PTP, lasting approximately 10 minutes. rents. LTP of fEPSPs is illustrated in the upper traces and in the
(Source: McNaughton, 1982.) B. LTP is not associated with changes time-course plot. The lower traces are glia-transporter currents in
in Pr at CA1 synapses in vitro, as assessed by measuring the rate of response to paired-pulse stimulation before and following the
blockade of NMDA receptor-mediated currents by MK-801. LTP of induction of LTP. Note that the records are superimposable, indicat-
AMPA receptor-mediated EPSCs was induced (at t  0) in one ing no change in the glial current. (Source: Diamond et al., 1998.)
input (lled symbols) but not in a control input (open symbols). E. LTP is associated with an increase in postsynaptic sensitivity to
AMPA receptors were then blocked and the neurons depolarized to exogenous AMPA receptor activation. The upper records are depo-
reveal the NMDA receptor-mediated EPSC. MK-801 was applied larizations of a CA1 neuron in response to ionophoretic application
and allowed to equilibrate before stimulation was commenced. Both of an AMPA receptor ligand (quisqualate) obtained before, immedi-
inputs were blocked at the same rate, indicating that the average Pr ately following, and 30 minutes following the induction of LTP (at t
at control and potentiated inputs was similar. (Source: Manabe and  0). The graph plots the amplitude of these depolarizations with
Nicoll, 1994.) C. LTP is not associated with changes in Pr at CA1 time. Note the gradual increase in postsynaptic sensitivity, with a
synapses in vitro, as assessed by measuring the rate of blockade of time course that correlates with E-LTP. (Source: Davies et al., 1989.)
375
376 The Hippocampus Book

observed during PTP but not during LTP in area CA1, provid- glutamate stored in or released from each vesicle. Finally, there
ing evidence against increased release as an expression mecha- could be an increase in the probability that an action potential
nism for LTP (Diamond et al., 1998; Luscher et al., 1998) (Fig. in an afferent ber leads to transmitter release, either because
1010D). An analogous result was obtained in area CA3. LTP of a more efficient conduction of the impulse to the terminal
at mossy ber synapses, but not at associational-commissural or because of a higher probability of release (Pr) in response
synapses, was associated with a sustained increase in glial to the invading action potential. The studies that have meas-
uptake current (Kamamura et al., 2004). This observation is ured the gross amount of L-glutamate released cannot distin-
consistent with other evidence that LTP at mossy bers is guish between these possibilities or other alterations that
maintained by presynaptic mechanisms (see Section 10.5) and affect the extracellular concentration of L-glutamate. In this
suggests that LTP at associational-commissural synapses lacks context, however, there is little evidence that the uptake or dif-
a strong presynaptic component. fusion of L-glutamate is affected during LTP.
Several lines of evidence are consistent with an increase in
Evidence for Presynaptic and Postsynaptic the probablility of transmitter release. Using minimal stimu-
Mechanisms in LTP lation to activate a putative single release site, Stevens and
Wang (1994) showed that LTP in area CA1 was associated
Disagreements among laboratories notwithstanding, these with a decrease in failure rate without change in potency (the
three approaches have generally been seen as providing evi- mean amplitude of responses excluding failures) (Fig.
dence against presynaptic changes and hence as strengthening 1011B). Bolshakov and Siegelbaum (1995) also reported an
the case for a purely postsynaptic change. There is, neverthe- increase in Pr in the same region in neonatal rats. These nd-
less, as we discuss below, a corpus of data that provides evi- ings are subject to the usual caveats regarding interpretation
dence for both presynaptic and postsynaptic changes. of quantal observations (see Box 101), although it should be
Although there is a sufficient body of data to make a convinc- emphasized that these were not silent synapses.
ing case for presynaptic changes, relatively little insight has yet Consistent with the idea of an increase in Pr, a few studies
been gained into molecular mechanisms that support an have detected a decrease in paired-pulse facilitation (PPF)
increase in transmitter release. For the postsynaptic side, there during LTP (Schulz et al., 1994; Kleschevnikov et al., 1997).
are also direct experiments demonstrating changes in res- For example, in slices from P6 rats there is a large decrease in
ponse to applied transmitter following induction of LTP. In PPF that fully accounts for the LTP in most, but not all,
addition, there is now a large and compelling set of data that neurons (Palmer et al., 2004) (see Fig. 10-21D). This alteration
gives detailed insights into how postsynaptic changes are is not due to changes in ber failures, suggesting that an
brought about by posttranslational modications of existing increase in Pr is the most likely explanation. However, the
receptors and trafficking of receptors between the cytoplasm effect was tightly regulated developmentally and was absent by
and extrasynaptic and synaptic membranes. A cautionary P12. Moreover, many studies have not observed changes in
note should be entered here; almost all the work that has char- PPF.
acterized the molecular substrate of postsynaptically mediated Direct evidence for a presynaptic mechanism in cultured
LTP and LTD has been carried out in in vitro preparations hippocampal neurons was obtained by measuring exocytotic-
during the rst hour or so of early LTP or LTD. The mecha- endocytotic cycling with antibodies against the synaptic vesi-
nisms that support the enduring changes seen in late LTP or cle protein synaptotagmin (Malgaroli et al., 1995) (Fig.
LTD are more likely to be structural in nature, and such 1011C). A novel mechanism proposed by Tsien and col-
changes almost certainly involve both sides of the synapse. leagues involves an LTP-dependent modication of the fusion
pore such that more L-glutamate is released from a fused
10.4.5 E-LTP: Presynaptic vesicle (Choi et al., 2000). Under basal conditions the amount
Mechanisms of Expression of L-glutamate released may be so small as not to elicit a detec-
table AMPAR-mediated postsynaptic response. Glutamate
The induction of LTP in the dentate gyrus in vivo is associated concentration may, however, be high enough to activate
with a persistent increase in the efflux of L-glutamate NMDARs, thus providing a potential presynaptic explanation
(Dolphin et al., 1982) (Fig. 1011A). The increase is activity- for silent synapses. LTP then modies the release machinery to
and NMDA receptor-dependent and has been detected using enable a greater amount of L-glutamate discharge from each
both a push-pull perfusion technique (Errington et al., 1987, fused vesicle. Another direct way of estimating the probability
but see Aniksztejn et al., 1989) and a glutamate-sensitive elec- of glutamate release is to measure the rate of destaining of
trode (Errington et al., 2003). There are several ways, not synaptic vesicles loaded with the steryl dye FM1-43. Activity-
mutually exclusive, in which LTP could be expressed presy- dependent destaining was more rapid following induction of
naptically. There could be an increase in the number of release LTP, an effect that was blocked by NMDA receptor antagonists
sites, either within a presynaptic terminal or as a result of the (Fig. 1011D). An intriguing example of presynaptic unsi-
formation of new terminals. Alternatively, there could be an lencing has been provided by Ma et al. (1999), who showed
increase in the number of vesicles released per impulse. that in hippocampal cultures treated with a cAMP analogue (a
Another possibility is an alteration in the amount of L- treatment that induces a late, protein synthesis-dependent
 Synaptic Plasticity in the Hippocampus 377

form of LTP), there was a dramatic increase in the number of 10.4.6 E-LTP: Postsynaptic
active boutons taking up the dye FM1-43 (Fig. 1011E). Mechanisms of Expression
Long-term potentiation is associated with an increase in
phosphorylation of the presynaptic protein GAP-43 (Routten- There is also direct evidence for postsynaptic alterations.
berg and Lovinger, 1985; Gianotti et al., 1992). Other studies Initial experiments compared the sensitivity of neurons,
examined glutamate-induced LTP in cultured hippocampal before and after the induction of LTP, to ligands that act on
neurons and found an increase in presynaptic boutons associ- AMPA receptors. LTP was invariably associated with an
ated with clusters of the presynaptic proteins synaptophysin NMDA receptor-dependent increase in AMPA receptor sensi-
(Antonova et al., 2001) and -synuclein (Liu et al., 2004b). tivity (Davies et al., 1989) (Fig. 1010E). The effect was not
Changes in Pr at single synapses following induction of immediate but occurred after a short delay (a few minutes or
LTP in pyramidal cells of areas CA1 and CA3 has been tens of minutes depending on the precise experimental proto-
assessed using Ca2 indicators to monitor synaptic events at col used). Subsequent work has also detected an increase in
single visualised spines in organotypic hippocampal cultures sensitivity to exogenously applied AMPA receptor ligands
(Emptage et al., 2003; Fig. 10.11F). Emptage et al. argued that (Montgomery et al., 2001). A more rened approach is to
the probability of evoking a Ca2 transient in the visualised deliver L-glutamate locally by photolytic release of glutamate
spine, PCa, provides an accurate measure of the probability of from a caged precursor. Under these conditions, a rapid
transmitter release, Pr, at that synapse, since manipulations NMDAR-dependent increase in sensitivity has been observed
that reect an increase in Pr, such as paired pulse facilitation, (Matsuzaki et al., 2004; Bagal et al., 2005).
also result in an increase in PCa, whereas manipulations that Broadly speaking, there are three categories of postsynap-
decrease Pr, such as bath application of adenosine, lead to a tic mechanisms by which LTP can be expressed. The rst is
decrease in PCa. In the majority of spines examined, induc- modication of the properties of the receptors already present
tion of LTP by high-frequency stimulation was accompanied at the synapse. The second is an increase in the number of
by an increase in PCa at the imaged spine, whereas in experi- receptors available at the synapse to respond to the synaptic
ments in which LTP was blocked by D-AP5, or in which the release of L-glutamate. The third involves changes down-
threshold for LTP was not reached, no change in PCa was stream of glutamate receptors, such as alterations in voltage-
seen. The Ca2 signal is NMDA receptor-dependent, but the gated or leak channels to affect the spread of depolarization
extent to which the evoked Ca2 transient in spines reects (e.g., Fan et al., 2005). Although there is evidence for alter-
Ca2 entering through the NMDA receptor varies with cell ations in voltage-gated ion channels, most emphasis has been
type and preparation (Yuste and Denk, 1995; Kovalchuk et al., on the glutamate receptors themselves. As discussed above,
2000; Sabatini et al., 2002). Ca2 entering via the NMDA NMDA receptor-dependent LTP is expressed as changes in
receptor following a single stimulus could not be resolved by both AMPA receptor- and NMDA receptor-mediated synaptic
Emptage et al. (2003), but the trigger signal was greatly ampli- transmission. Because synaptic responses evoked at resting
ed, and the event thus made visible, by calcium-induced cal- membrane potentials by low-frequency stimulation are medi-
cium release from internal stores (Emptage et al., 1999). The ated predominantly by the activation of AMPA receptors
indirect nature of the Ca2 transient raises the possibility that (Herron et al. 1986), it is these receptors that have been at the
the observed increase in PCa following LTP may reect an center of attention.
increase in the coupling between the initial trigger Ca2 and
store release, rather than an increase in Pr. Failures, according Modication of Existing Receptors Contributes to LTP
to this account, would include not only failures of release but
also failures of postsynaptic coupling, so that pCa gives an The amount of current that passes through an AMPA recep-
underestimate of Pr. In this view, the reason that pCa is tor is determined by the probability of opening (Po) on bind-
increased in LTP is that enhanced AMPAR function leads to ing L-glutamate, the mean open time, and the single-channel
an increase in the EPSP at the spine, a consequent increase in conductance (). Alterations in mean open time of AMPARs
trigger Ca2 entering through the NMDA receptor, and hence would be reected as a change in the kinetics of EPSCs.
to a reduction in the number of failures due to failed coupling. Despite claims to the contrary (Kolta et al, 1998), it is usually
This interpretation is rendered unlikely by the observation accepted that the decay of AMPAR-mediated EPSCs is not
that bath application of a low dose of the AMPAR antagonist changed by LTP. This is most clearly seen when dendritic
CNQX, which greatly reduces the amplitude of the AMPAR- recording is used to maximize the signal-to-noise of minimal
mediated EPSP recorded at the soma, has little if any effect on EPSCs and to reduce the dendritic ltering that is unavoidable
PCa (Emptage et al., 2003). Nevertheless, a more direct with somatic recording (Benke et al., 1998). Also, treatment
method of measuring Pr would be desirable. One promising with cyclothiazide, which inhibits both deactivation and
approach is to transfect neurons with a glutamate-binding desensitization to cause pronounced slowing of the EPSC,
protein tagged with a pair of uorophores that yield a uores- does not affect the expression of LTP (Rammes et al., 1999).
cence resonance energy transfer (FRET) signal when L-gluta- Electrophysiological experiments cannot easily distinguish
mate is released from the presynaptic terminal (Okumoto et between insertion of AMPA receptors, and changes in Po or .
al., 2005). However, in a few cases  has been estimated by nonstation-
378 The Hippocampus Book
 Synaptic Plasticity in the Hippocampus 379

ary uctuation analysis. At CA1 synapses of 2-week-old rats it CaMKII-dependent phosphorylation of GluR1 to increase the
was found that in most neurons  increased sufficiently to single-channel conductance of existing AMPA receptors.
account for LTP (Benke et al. 1998) (Fig. 1012A). In other However, although this mechanism may occur, it cannot be
neurons  did not change, suggesting that any postsynaptic the whole story, as CaMKII inhibitors applied alone fail to
change was due to an increase in N, the number of receptors, inhibit LTP at this stage of development (Wikstrm et al.,
or Po. The simplest explanation for the increase in  is a mod- 2003; Yasuda et al., 2003).
ication of receptors already present at the synapse, but it
could theoretically be due to the exchange of lower for higher Insertion of AMPA Receptors Contributes to LTP
 receptors at the synapse. The estimate of  derived from
nonstationary uctuation analysis is the weighted mean of all A popular hypothesis for the expression of LTP is an increase
subconductance states (Cull-Candy and Usowicz 1987; Jahr in the number of AMPA receptors available to respond to
and Stevens, 1987). Because AMPA receptors can adopt mul- synaptically released L-glutamate. These receptors could be
tiple conductance states, it was suggested that LTP was due inserted from an intracellular compartment and/or diffuse
to an alteration in the relative time spent in these various laterally from extrasynaptic regions. In the simplest scenario,
states (Benke et al., 1998). Consistent with this idea, CaMKII- the new synaptic receptors would have conductance proper-
mediated phosphorylation of ser831 on GluR1 not only ties identical to those of the existing ones. Alternatively, the
occurs during LTP (Barria et al., 1997a,b) but leads to an in- new receptors could have a different subunit composition,
crease in precisely this parameter in HEK293 cells expressing thereby conferring different conductance properties (i.e.,
GluR1 (Derkach et al., 1999). Furthermore, synaptic potenti- single-channel conductance, rectication, Ca2 permeability)
ation induced by postsynaptic application of a constitutive to the potentiated synapses. Indeed, early in development
form of CaMKII is associated with an increase in  (Poncer et (around the rst week of life) LTP was observed in some neu-
al., 2002). Thus a simple scheme for the induction and expres- rons that was associated with a decrease in  (Palmer et al.,
sion of LTP is that Ca2 entering via NMDA receptors triggers 2004), the simplest explanation of which is the insertion of

Figure 1011. Evidence that LTP involves presynaptic changes. A. tive distribution curve in potentiated cultures compared to control
LTP in the dentate gyrus of the anesthetized rat is associated with cultures or cultures in which LTP was blocked by NMDA receptor
an increase in the efflux of 3H glutamate. At the beginning of the antagonists indicates a signicant increase in vesicular cycling after
experiment a bolus of 3H-glutamine was injected into the dentate induction of LTP. (Source: Malgaroli et al., 1995.) D. Visualization of
gyrus. The dentate gyrus was perfused using a push-pull cannula; increased transmitter release after LTP using the dye FM1-43. The
and perfusates were collected for the measurement of 3H-glutamate, rapidly recycling pool of vesicles in Schaffer-commissural terminals
newly synthesized from 3H-glutamine. In control animals, the con- was loaded in vitro by exposing slices to FM1-43 in hypertonic
centration of 3H-glutamate declined exponentially (dotted line). ACSF to stimulate transmitter release. Activity-induced destaining
Field responses were recorded at the same time (not shown). In ani- was measured 30 minutes after reperfusion with ACSF by stimulat-
mals in which LTP was induced by high-frequency stimulation ing with 10 Hz, 2-second trains at 30-second intervals. The rate of
(arrow), the concentration of 3H-glutamate was increased relative to destaining was accelerated in potentiated tissue (open circles), rela-
control values for the following hour (solid line). Abscissa, time tive to tissue treated with AP5 to block LTP (lled diamonds) or to
since injection of 3H-glutamine; ordinate, 3H-glutamate concentra- control tissue (lled circles). (Source: Stanton et al., 2005 with addi-
tion normalized to its value in the collection interval immediately tion of control data courtesy of the authors.) E. A membrane-per-
preceding the tetanus. (Source: Dolphin et al., 1982.) B. LTP at pre- meable analogue of cAMP induces an increase in the number of
sumptive single synapses in area CA1 is associated with a decrease boutons taking up FM1-43 in CA3-CA1 neuronal cultures. The bars
in failure rate (upper panel), with no change in the potency of the show the ratio of the number of stained boutons before and 30
response (i.e., the amplitude of the evoked synaptic current for tri- minutes or 2 hours after exposure to the control solution (open
als where a release event occurred) (lower panel). Responses were bars) or solution containing Sp-cAMP. Exposure to cAMP itself can
evoked by minimal stimulation to Schaffer-commissural axons in induce slow-onset late LTP (see Section 10.4.3). Both cAMP-
the acute hippocampal slice. (Source: Stevens and Wang, 1994.) C. induced LTP and the increase in the number of active boutons is
Increase in synaptic activity visualized at single synapses in cultured protein synthesis-dependent. (Source: Ma et al., 1999.) F. Tetanus-
hippocampal neurons following glutamate-induced LTP. Goat or induced LTP in organotypic hippocampal cultures is associated with
rabbit antibodies recognizing the intraluminal domain of the vesic- an increase in the probability of obtaining a synaptically generated
ular protein synaptotagmin were used to measure vesicular cycling. Ca2 transient (pCa) in single visualized spines on pyramidal neu-
Cultures were incubated with one of the antibodies for an hour. LTP rons. The graph plots pCa before and 30 minutes after the induction
was then induced by brief application of glutamate, and the second of LTP for a group of individual spines (lled circles). In other
antibody then applied for a second hour. The degree of internaliza- experiments (open circles), LTP was not induced either because no
tion of each antibody was measured using uorescence-labeled anti- tetanus was given or induction was blocked by AP5. There was a
goat and anti-rabbit antibodies. The ratio of uorescence during the signicant increase in pCa in the group of tetanized spines from cells
second hour relative to the rst (F2/F1) was measured in several in which LTP was induced relative to control spines. (Source:
hundred individual synapses. The shift to the right in the cumula- Emptage et al., 2003.)
Figure 1012. Postsynaptic mechanisms involved in the expression et al., 2000.) C. LTP in cultured neurons is associated with the inser-
of LTP. A. LTP is associated with an increase in single-channel con- tion of native AMPA receptors, as assessed using a GluR1-specic
ductance () of synaptic AMPA receptors at CA1 synapses in vitro. antibody. AMPA receptors in cultured hippocampal neurons were
The traces are the mean amplitudes before and following induction labeled with a GluR1-specic antibody, and a nonuorescent sec-
of LTP superimposed (left) and expanded and peak-scaled (right). ondary antibody was then bound to them. LTP was induced by per-
Note the lack of change in the shape of the potentiated EPSC. The fusion with glycine. The cultures were then challenged with GluR1
graph plots the amplitude of EPSCs versus time before and follow- antibody, which bound to the newly inserted GluR1 receptors and
ing pairing induced LTP. The graph below (left) plots the variance of were detected using a uorescent secondary antibody. The panels
the decay phase of the EPSC versus the amplitude of each point on show a representative portion of a dendrite from a control neuron
the EPSC decay. The initial slope of the plot provides an estimate of (upper) and a neuron following glycine-induced LTP (lower).
, which is this neuron approximately doubled following LTP. The (Source: Lu et al., 2001.) D. LTP in cultured neurons is associated
graph (right) plots the change in  associated with LTP against the with the insertion of native AMPA receptors, as assessed using a
baseline value of . Note that in most neurons  increased (and this pan-AMPA antibody. AMPA receptors in cultured hippocampal neu-
was sufficient to account for LTP), but in some neurons there was no rons were labeled with a GluR1-4-specic antibody, and their distri-
change in . The latter neurons tended to have a higher baseline . bution visualized using a uorescent secondary antibody. LTP was
(Source: Benke et al., 1998.) B. LTP involves the insertion of AMPA then induced by transient depolarization (3  1-second pulses of 50
receptors at CA1 synapses in vitro, as determined using recombinant mM K). The cultures were then challenged with pan-AMPA anti-
receptors with distinct electrophysiological properties. The graph body raised in a different species. This bound to newly inserted
plots EPSC amplitude versus time, and LTP was induced at t  0. AMPA receptors, which were detected using a uorescent secondary
The rectication of the AMPA receptor-mediated EPSC was esti- antibody directed against the second species. The panels show a rep-
mated by comparing its amplitude at positive and negative mem- resentative portion of dendrite from a neuron before (upper) and
brane potentials following blockade of NMDA receptors. LTP caused following (lower) induction of LTP. Note the appearance of new
an increase in rectication in GluR1-transfected cells (Inf), indicat- clusters of AMPA receptors at sites that had previously lacked any
ing insertion of new GluR1 homomeric receptors. (Source: Hayashi detectable AMPA receptors. (Source: Pickard et al., 2001.)
380
 Synaptic Plasticity in the Hippocampus 381

new receptors with a lower  than the existing population of and LTD (see sections 10.3.7 and 10.7.3). From this has arisen
receptors. the concept of the subunit-dependence of synaptic plasticity.
To address AMPA receptor movement in relation to LTP In broad terms, the GluR1 subunit seems to play a major role
more directly, two complementary approaches have been in LTP whilst the GluR2 subunit is the more important in
used, each with its own relative advantages and disadvantages. LTD. However, as we shall see, this is to simplify a complex sit-
In one approach, recombinant AMPA receptor subunits are uation.
constructed containing a reporter, such as green uorescent
protein (GFP). LTP in organotypic slice cultures was found to GluR2 interactors: Evidence for rapid recycling and regulation
be associated with delivery of GFP-GluR1 to dendritic spines of subunit composition of AMPARs at synapses. An initially
(Shi et al., 1999). However, because imaging could not deter- surprising observation was that N-ethylmaleimide-sensitive
mine whether these constructs were inserted into the plasma fusion protein (NSF), an ATPase involved in membrane fusion
membrane, mutated receptors were used that had a distinct events, binds directly to the GluR2 subunit (Nishimune et al.,
electrophysiological signature (Hayashi et al., 2000). The 1998; Osten et al., 1998; Song et al., 1998). Blockade of the
synaptic delivery of these receptors was then determined by GluR2-NSF interaction resulted in a rapid decrease in electri-
alterations in the rectication properties of the synaptic cur- cally evoked AMPAR-mediated synaptic transmission, sug-
rents (Fig. 1012B). Based on these studies it is clear that gesting that AMPA receptors can rapidly recycle at the synapse
recombinant AMPA receptors can be rapidly inserted into (Nishimune et al., 1998). Blockade of this interaction also led
synapses during LTP in organotypic slices. Whether the to internalisation of AMPARs, suggesting that NSF regulates
recombinant receptors are inserted selectively into potentiated the membrane insertion or stabilisation of synaptic AMPARs
synapses and, if so, whether they fully account for the poten- (Noel et al,, 1999, Luscher et al., 1999). These observations
tiated response remains to be established. raised the possibility that modication of recyling rates could
In the second method, antibodies that recognize extracellu- underlie LTP. Indeed, there is some evidence for a role of NSF
lar epitopes of AMPA receptor subunits are used, in conjunc- and soluble NSF attachment proteins (SNAPs) in AMPA
tion with protocols that induce NMDAR-dependent receptor insertion during LTP (Lledo et al., 1998).
LTP in dissociated cultured hippocampal neurons. The alter- A second region of the C-terminal tail of GluR2 is the site
ation in native AMPA receptor distribution is determined of interaction of the PDZ-containing proteins GRIP (gluta-
using uorescently conjugated secondary antibodies. Using mate receptor interacting protein), ABP (AMPA receptor
this approach, insertion of native AMPA receptors during binding protein), and PICK1 (protein interacting with C-
NMDAR-dependent LTP has been described (Lu et al., 2001; kinase 1). ABP and GRIP contain multiple PDZ domains and
Pickard et al., 2001). These studies identied the insertion probably serve to anchor AMPA receptors at both synaptic
of GluR1-containing AMPA receptors into synapses that (Osten et al., 2000) and intracellular (Daw et al., 2000) loca-
previously lacked GluR1 (Lu et al., 2001; Fig. 1012C) via a tions. Some isoforms of ABP/GRIP are palmitoylated, which
PI3K-dependent process (Man et al., 2003). They also demon- targets them to the synaptic membrane (Yamazaki et al., 2001;
strated insertion of AMPA receptors into anatomically silent DeSouza et al., 2002). These molecules presumably also local-
synapses (synapses where AMPA receptors could not be ize other proteins involved in the LTP process, close to AMPA
detected on the cell surface) (Pickard et al., 2001; Fig. 1012D). receptors. For example, GRIP binds to an associated protein,
The use of AMPA receptor antibodies has also enabled the lat- GRASP-1, which is a neuronal RasGEF and so may link AMPA
eral mobility of AMPA receptors to be studied in the plasma receptors to Ras signaling (Ye et al., 2000).
membrane. Single-molecule uorescence microscopy has PICK1 can also bind PKC, suggesting that one of its func-
revealed that there are both mobile and immobile receptors tions is to target PKC to phosphorylate AMPA receptors.
contained within spines and extrasynaptically (Tardin et al., Indeed, there is evidence that PICK1 may phosphorylate
2003). Stimulation with L-glutamate increases synaptic AMPA GluR2 on ser880 to inhibit the binding of AMPA receptors to
receptor diffusion rates, raising the possibility that LTP ABP/GRIP (Matsuda et al., 1999), and this mechanism may be
involves the mobilization of extrasynaptic AMPA receptors involved in the mobilization of AMPA receptors from their
and their targeting to the postsynaptic density. Consistent with ABP/GRIP tethers. In the context of LTP, it has been proposed
this idea, there is evidence that following exocytosis AMPA that PICK1-PKC phosphorylation of ser880 on GluR2 mobi-
receptors are diffusively distributed along dendrites before lizes AMPA receptors from intracellular sites during de-
accumulating at synaptic sites (Passafaro et al., 2001). depression (i.e., reversal of LTD by an LTP induction protocol)
(Daw et al., 2002). PICK1 is a particularly interesting molecule
Molecules that bind to AMPARs give clues to the molecular since it also regulates the GluR2 content of AMPA receptors at
mechanisms involved in LTP. Understanding the molecular synapses (Terashima et al., 2004), raising the possibility that it
mechanisms of AMPA receptor synaptic trafficking during may play a role in LTP by altering the conductance properties
LTP has been greatly aided by the identication of proteins of synaptic AMPA receptors. Indeed, it has been found that
that bind directly to, and regulate the synaptic function of, LTP at CA1 synapses in two-week-old animals involves the
AMPA receptors. Different proteins interact with the GluR1 rapid insertion of GluR2-lacking AMPARs followed by their
and GluR2 subunits, and seem to play distinct roles in LTP replacement with GluR2-containing AMPARs, over a period
382 The Hippocampus Book

of 20 minutes or so (Plant et al., 2006). The transient insertion Tarps add a new dimension. AMPA receptors also bind to
of calcium permeable AMPARs may provide a signal, or tag, to stargazin, which in turn binds to PSD95 (Chen et al., 2000).
mark recently potentiated synapses. Given that PICK1 is able Stargazin is one of a class of TARPs (transmembrane AMPAR
to regulate the GluR2-content of AMPARs, and is the only regulatory proteins) that are involved in the recruitment of
molecule known to do this, it is a prime candidate for mediat- AMPA receptors from extrasynaptic to synaptic sites during
ing this effect. LTP (Bredt and Nicoll, 2003). Consistent with this role of
TARPS and PSD95, overexpression of PSD95 selectively
GluR1 interactors: A key role in LTP? There is evidence that enhances AMPAR-mediated synaptic transmission and
GluR1 receptors are particularly important for synaptic deliv- occludes LTP (Stein et al., 2003; Ehrlich and Malinow, 2004).
ery during LTP (Hayashi et al., 2000). Malinow and his col- Indeed, the level of PSD95 at the synapse, which may be regu-
leagues have proposed that GluR1/GluR2 heteromers are lated by ubiquitination (Colledge et al., 2003) and palmitoyla-
inserted into the membrane during LTP, whilst GluR2/GluR3 tion (El Husseini et al., 2002), could directly determine the
heteromers are important for constitutive recycling of AMPA number of AMPA receptors at synapses. In addition, the inter-
receptors at synapses (Shi et al., 2001). However, the mecha- action of AMPA receptors with TARPs is dynamically regu-
nisms involved are unclear. Both spine delivery of GluR1 and lated both allosterically via ligand binding and through
LTP in organotypic slices requires activation of CaMKII, but phosphorylation of TARPs (Chetkovich et al., 2002; Tomita et
deletion of ser831, the CaMKII phosphorylation site on the C- al., 2004).
terminal tail of GluR1, did not prevent either of these effects In addition to regulating the number of AMPARs on the
(Hayashi et al., 2000). However, a point mutation on the PDZ membrane surface, TARPs also increase the apparent affinity
domain at the extreme C-terminus of GluR1 (T887A) did pre- of AMPARs for glutamate. They slow the rate of both deacti-
vent spine delivery and LTP suggesting that CaMKII may vation and desensitization and enhance , by increasing the
phosphorylate a protein that binds to this site. One such can- prevalence of high conductance substates (Tomita et al, 2005).
didate is the membrane associated guanylate kinase It has been estimated that TARPs increase the charge transfer
(MAGUK) SAP97, which binds to GluR1 via its PDZ domain of synaptic currents by approximately 30%.
and has been implicated in trafficking to the synapse (Leonard The -8 TARP isoform is highly enriched in the hippocam-
et al., 1998). SAP97 enhances AMPAR-mediated synaptic pus and the knockout of this protein leads to a profound loss
transmission in cultured neurons, an effect mediated by a of AMPAR subunits and a reduction in AMPAR-mediated
SAP97 splice variant containing a binding site for a GluR1 EPSCs. LTP in knockout animals was greatly reduced, suggest-
binding protein, 4.1, that has been implicated in linking ing that TARPs are involved in synaptic plasticity (Rouach et
AMPA receptors to the cytoskeleton (Rumbaugh et al., 2003). al, 2005). TARPs can be phosphorylated by both CaMKII and
Overexpression of SAP97 leads to an enhancement of PKC and dephosphorylated by PP1 and PP2B. Mutation of the
AMPAR-mediated transmission in hippocampal organotypic target serines to alanine resulted in block of LTP, whilst con-
cultures that partially occludes with LTP (Nakagawa et al., versely mutation of these residues to aspartic acid, to mimic
2004); the effect depends on the L27 domain of SAP97, which phosphorylation, drove AMPARs to synapses and blocked
is involved in forming multimeric assemblies of SAP97, either LTD (Tomita et al, 2005). Thus, regulation of TARPs seems
with itself or other proteins containing a similar domain. play a critical role in hippocampal synaptic plasticity.
Another pertinent nding is that CaMKII is able to phospho- In summary, a scheme is emerging whereby alterations in
rylate SAP97 and this targets SAP97 to synapses (Mauceri et AMPA receptor number are controlled by proteinprotein
al., 2004), an effect that may explain the actions of CaMKII in interactions and phosphorylation. The types of binding part-
driving AMPARs into synapses and into the PSD. Such a ners and kinases involved seem to vary depending on the
process may involve the motor protein myosin VI, since this AMPA receptor subunit and the stage of development. During
forms a complex with both GluR1 and SAP97 (Wu et al., LTP, AMPA receptors probably diffuse laterally in the synaptic
2002). However, deletion of the entire PDZ domain of GluR1 membrane from extrasynaptic sites to the PSD to increase the
does not prevent the induction of LTP in transgenic mice number of receptors available to bind to synaptically released
(Kim et al., 2005a). Further work is therefore required to L-glutamate. This process may also be associated with inser-
establish the role of the PDZ domain of GluR1 in LTP. tion of AMPA receptors from an intracellular reservoir via
There is also evidence for a role of PKA in driving GluR1- exocytosis.
containing AMPARs into synapses (Esteban et al., 2003).
Mutagenesis of ser845 has shown that this PKA site is required Locus of Expression of NMDAR-
though its phosphorylation is not sufficient for LTP, suggest- dependent LTP: a Site Inspection
ing that this is one stage in a multi-step process. One possibil-
ity is that PKA-dependent phosphorylation of ser845 is What emerges from this summary of a large literature is that
involved in the membrane targeting of AMPARs to extrasy- there is no clear consensus on the question of the locus of hip-
naptic sites (Oh et al, 2006). PKC is also implicated in the traf- pocampal LTP. Most camps were, and some remain today,
cking of GluR1 and LTP, via phosphorylation of ser818 highly entrenched in their view that LTP can be expressed only
(Boehm et al, 2006). via their own favored mechanism. Objective analysis of all of
 Synaptic Plasticity in the Hippocampus 383

the evidence, however, points to a multitude of expression municated in a retrograde direction (Murai and Pasquale,
mechanisms that may vary according to the component of 2004). In this section we review the evidence for these various
LTP under investigation (i.e., STP, E-LTP, L-LTP), the develop- candidates. Before starting, we consider a set of criteria that a
mental stage of the animal (see Fig. 1021), and even the diffusible candidate messenger should satisfy. Note that a ret-
experimental approach used to tackle the problem. More rograde messenger is required only to generate the presynap-
recently, attention has focused on trying to understand the tic component of LTP; this is designated LTPpre in the
molecular mechanisms responsible for pre- and postsynaptic following list.
changes.
The cellular machinery for synthesis of the messenger
There are several indisputable facts. First, there is strong
exists in the postsynaptic cell.
evidence in favor of both pre- and postsynaptic mechanisms,
The synthesis of the messenger is upregulated by LTP-
so NMDAR-dependent LTP can involve changes at both sides
inducing stimuli.
of the synapse. Second, there are pronounced differences in
Postsynaptic injection of drugs that block synthesis
the signalling and expression mechanisms of LTP at different
inhibits LTP induction.
stages of development (see Section 10.9). Third, there is evi-
The messenger is released into the extracellular com-
dence that the various phases of LTP may involve different
partment following induction of LTPpre.
expression mechanisms. For example, the increase in sensitiv-
Perfusion with extracellular scavengers small enough to
ity to AMPA receptor ligands correlates with E-LTP but not
access the synaptic cleft inhibit LTPpre.
STP. Fourth, the type of LTP and its expression mechanism
A target molecule in the presynaptic terminal, positively
may depend on how it is induced; a tetanus causes repetitive
linked to transmitter release, is activated by the messen-
activation of the presynaptic cell whereas pairing does not. We
ger.
conclude therefore that LTP is expressed by alterations in
Drugs that inhibit the link between messenger and exo-
either L-glutamate release and/or AMPA receptor function,
cytosis suppress LTPpre.
the extent of each component depending on several factors, of
Exogenous application of the messenger induces LTPpre
which the list above is unlikely to be exhaustive.
or lowers the threshold for its induction.

10.4.7 Retrograde Signaling Is Required for No candidate retrograde messenger has yet been shown to
Communication Between the Postsynaptic Site satisfy all the criteria on this admittedly demanding list.
of Induction and the Presynaptic Terminal
Arachidonic Acid
We have seen that the trigger for the induction of LTP in the
Schaffer-commissural and perforant pathways is the entry of Arachidonic acid (AA) is produced by the action of a calcium-
calcium through activated NMDA channels located on the dependent enzyme, phospholipase A2, on membrane phos-
postsynaptic cell. Evidence from a variety of techniques and pholipids. In 1987, Piomelli et al. (1987) suggested that AA, or
preparations points to a presynaptic component to the expres- one of its lipoxygenase metabolites, might serve as a synaptic
sion of LTP (Section 10.4.5). Early evidence for increased retrograde messenger in LTP on the basis of experiments in
transmitter release in LTP led Bliss and Dolphin (1984) to Aplysia that identied lipoxygenase metabolites of arachidonic
postulate that a diffusible retrograde signal is released from acid as a novel class of second messenger. A year later, Bockaert
the postsynaptic site of induction to interact with the presy- and his colleagues made the signicant observation that acti-
naptic terminal and in some manner stimulate transmitter vation of NMDA receptors in cultured striatal neurons led to
release. To date, there has been no unequivocal identication the release of AA into the culture medium (Dumuis et al.,
of a retrograde signaling molecule mediating LTP or LTD at 1988). The rst evidence that the AA cascade played a role in
hippocampal excitatory synapses. Most attention has focused LTP followed in the same year in a report that an inhibitor of
on two candidates, the unsaturated membrane fatty acid AA production blocked chemically induced LTP in area CA1
arachidonic acid (AA) and the gas nitric oxide (NO) synthe- and the dentate gyrus (Williams and Bliss, 1988). Subsequent
sized from L-arginine by nitric oxide synthase. Other potential studies, reviewed in Bliss et al. (1990), demonstrated that AA
retrograde messengers not considered further here that at one satises several of the criteria for a retrograde messenger: (1)
time had their champions include platelet-activating factor LTP is accompanied by an increase in the concentration of AA
(Miller et al., 1992) and carbon monoxide (Stevens and Wang, in a postsynaptic membrane fraction and an increase in the
1993). Other potential diffusible messengers are brain-derived extracellular concentration of AA; (2) inhibitors of phospholi-
trophic factor (BDNF) (see Section 10.6.1) and an endoge- pase A2 block induction of LTP; (3) the application of arachi-
nous endocannabinoid, 2-AG, discussed above in the context donic acid to active hippocampal synapses causes delayed
of E-S potentiation (Section 10.3.12), which appears to act as but persistent potentiation of evoked responses. Neverthe-
a retrograde mediator of heterosynaptic LTD at inhibitory less, these results, although pointing to a role for arachidonic
synapses in area CA1 (Chevaleyre and Castillo, 2003). acid in LTP, do not compel the conclusion that it is a retrog-
Intersynaptic communication via membrane-spanning adhe- rade messenger; there are missing gaps in the evidence (Do
sion molecules is another way in which signals could be com- extracellular scavengers of AA block induction? What is the
384 The Hippocampus Book

presynaptic mode of action of AA that leads to an increase in stimulation led to accelerated vesicle endocytosis via a cGMP-
transmitter release?). Moreover, some of the evidence can be dependent pathway (Micheva et al., 2003). An interesting
explained by other known properties of AA, including inhibi- insight from this work is that NO acts only on stimulated ter-
tion of glutamate uptake into glia (Barbour et al., 1989) and its minals, consistent with the possibility that it could selectively
facilitatory action on NMDA currents (Miller et al., 1992). For affect neighboring active terminals, in addition to the termi-
these reasons, AA has fallen out of fashion as a candidate mes- nal(s) whose activity was responsible for evoking its release
senger for LTP. However, evidence that a 12-lipoxygenase from postsynaptic cell(s). The observed changes were blocked
metabolite of AA, 12(S)-HPETE, mediates the induction of by NMDA receptor antagonists, establishing the likelihood
mGluR-dependent LTD in area CA1 of young rats has been that NO was produced postsynaptically and was thus acting as
presented by Bolshakov and colleagues (Feinmark et al., 2003). a genuine diffusible retrograde messenger. There is also evi-
It remains to be seen whether 12(S)-HPETE satises all the cri- dence that NO affects vesicle recycling in LTD (see Section
teria for a retrograde messenger, including activity-dependent 10.7.4), suggesting that the direction of change is determined
upregulation in the postsynaptic cell. by the interaction of NO with presynaptic factors (see Section
10.4.9 for discussion of a potentially similar role played by
Nitric Oxide somatically generated plasticity factors in their interaction
with synaptic tags that are set either for LTD or LTP).
Nitric oxide has had a more durable career as a candidate ret- In a provocative study, Madison and Schuman (1994)
rograde messenger than AA, though again the evidence re- made the intriguing observation that intracellular injection of
mains incomplete. NO is a small membrane-permeable NO synthase inhibitors blocked LTP when it was induced by
molecule with a short half-lifetwo desirable properties for a pairing of single stimuli with depolarization of the impaled
potential retrograde messenger. In neurons, NO is derived cell but not when it was induced by high-frequency trains.
from L-arginine in a reaction catalyzed by nitric oxide syn- Their explanation of this dissociation is as follows. When pair-
thase, of which two isoforms, nNOS and eNOS, are expressed ing is used to induce LTP, the paired cell itself is the sole source
in dendrites of hippocampal neurons. Like AA, NO is released of the putative retrograde messenger. If it is NO (or a messen-
from cultured neurons exposed to NMDA (Garthwaite et al., ger downstream of NO synthesis), intracellular application of
1988). The target of NO is soluble guanylyl cyclase, which an NO synthase inhibitor would be expected to block LTP, as
activates cyclic guanosine monophosphate (cGMP), one sub- observed. However, when LTP is induced by a tetanus, the sur-
strate for which is protein kinase G (PKG). Little is known of rounding cells are subject to LTP and themselves make and
the effector molecules linked to this second messenger system release the putative retrograde messenger. This could act on
in hippocampal synaptic terminals (see review by Hawkins et the single cell in which NO synthesis has been shut down.
al., 1998). Thus, aided and abetted by its neighbors, terminals afferent to
Attempts to block induction of LTP with inhibitors of nitric this cell can share in the potentiation induced by the high-fre-
oxide synthase have met with variable success. In vitro quency train. However, this ingenious interpretation is at odds
(Cummings et al., 1994) and in vivo (Bannerman et al., 1994) with earlier, well established reports that maneuvers that block
experiments failed to conrm initial reports that inhibition of LTP at the level of the single cell, such as intracellular injec-
NO synthase blocked the induction of LTP (Bohme et al., 1991; tions of calcium chelators (Lynch et al., 1983) or hyperpolar-
Schuman and Madison, 1991; Haley et al., 1992). In experi- ization (Kelso et al., 1986; Malinow and Miller, 1986), block
ments in hippocampal slices, inhibition of LTP by NO synthe- both pairing-induced and tetanus-induced LTP in pyramidal
sis inhibitors was seen at room temperatures but not at higher, cells. Little further work has been done on this aspect of hip-
more physiological temperatures (Williams et al., 1989). There pocampal NO signaling, and there for the moment the story
is also disagreement about whether the application of NO, in rests.
combination with weak synaptic activation, produces LTP; two
groups reported that it did (ODell et al., 1991b; Bon and 10.4.8 Membrane Spanning Molecules
Garthwaite, 2003), whereas a third found that it did not Contribute to Signaling Between Presynaptic
(Murphy et al., 1994). However, the latter nding, based on the and Postsynaptic Sides of the Synapse
release of NO by ash photolysis from an inactive caged pre-
cursor, has been shown to reect an interaction between NO The retrograde messengers we have considered so far are dif-
and ultraviolet light, leading to artifactual inhibition of NMDA fusible substances released from the postsynaptic spine to act
receptor currents (Hopper et al., 2004). Knockout by homolo- in ways that are still somewhat mysterious to affect transmit-
gous recombination of either of the two genes encoding the ter release. Another way in which communication can be
two isoforms of NO in hippocampal neurons fails to block effected between the two sides of the synapse is via the various
LTP. However, LTP is compromised, though not abolished, in a classes of membrane-spanning molecules that have partners
double knockout of both genes (Son et al., 1996). on the opposite side. In this section we summarize the little
Evidence linking postsynaptically generated NO to that is known about the way in which these interneuronal
enhanced transmitter release has emerged from a study on cul- interactions are regulated by activity and the consequent
tured hippocampal neurons, in which NO released by synaptic effects on synaptic function.
 Synaptic Plasticity in the Hippocampus 385

Ephrins and Eph Receptors we briefly review the evidence for the involvement of
immunoglobulin superfamily members, cadherins, integrins,
Members of the large Eph family of receptor tyrosine kinases and neuroligins in LTP.
and their ephrin ligands play important roles in axon guid-
ance, migration, and boundary determination in neural devel- Immunoglobulin Superfamily Members. Mice treated with
opment (Wilkinson, 2001). Evidence is emerging that Eph antibodies to the immunoglobulin superfamily member
receptors and ephrins are involved in the synaptic processes NCAM and mice with a targeted deletion of NCAM show
leading to LTP and LTD (reviewed by Murai and Pasquale, reduced LTP in area CA1 and in the mossy ber projection to
2004) (see Section 10.5.3 for a discussion of their role in area CA3 (Benson et al., 2000). PSA-NCAM, a form of NCAM
mossy ber LTP). Eph receptors bind to ephrins via their carrying multiple copies of polysialic acid linked to the fth
extracellular domains and in doing so trigger reciprocal sig- immunoglobulin domain on the extracellular part of the mol-
naling pathways inititiated via the carboxy cytoplasmic ecule, has also been implicated in LTP. Enzymatic removal of
regions of both Eph receptors and ephrins, a capability that PSA from NCAM inhibits LTP in area CA1 (Muller et al.,
for Eph receptors is referred to as forward signaling and for 1996). Similarly, LTP is impaired in area CA1 in a knockout
ephrins as reverse signaling. EphB2, a member of one of the mouse lacking one of the two polysialyltransferases that link
two major subfamilies of Eph receptors, interacts with NMDA PSA to NCAM. However, in this mouse, LTP is normal in area
receptors and also, via a PDZ binding domain, with the AMPA CA3, suggesting that whereas PSA-NCAM is required for LTP
receptor scaffolding protein GRIP. These interactions are in area CA1, NCAM itself is needed to support LTP in the
presumably the route by which Eph receptors become respon- mossy ber projection (Cremer et al., 1998). The roles of two
sive to synaptic activity, and they also provide a route for other members of the NCAM family, L1 and Thy-1, in hip-
modulating synaptic function (Henderson et al., 2001). This pocampal synaptic plasticity have also been examined. LTP in
possibility is reinforced by the demonstration by Grunwald a Thy-1 knockout mouse was normal in area CA1 but
et al. (2004) that ephrinB ligands are required for full expres- impaired in the dentate gyrus in vivo; performance on the
sion of LTP and LTD at CA3-CA1 synapses. In both sets Morris watermaze was unaffected (Nosten-Bertrand et al.,
of studies, the functionality of the Eph receptor/ephrin part- 1996; Errington et al., 1997). There is disagreement about
nership was unimpaired by removing the cytoplasmic tail whether L1 plays a role in LTP. In an in vitro study using func-
of the Eph receptor, suggesting that reverse signaling by tional antibodies against L1, Lthi et al. (1994) saw reduced
ephrins is the active route to LTP and LTD (Grunwald et al., LTP in area CA1, but Bliss et al. (2000) found no effect on LTP
2004). in area CA1 in vitro or in the dentate gyrus in vivo in an L1
Another potentially important role for Eph receptors and knockout mouse.
their ephrin partners is in the regulation of spine morphology.
EphA4 receptors are expressed on spines of cultured hip- Cadherins. Cadherins are an extensive family of cell adhesion
pocampal pyramidal cells and can be targeted by ephrinA3, molecules with cytoplasmic signaling domains that are linked
which is strongly expressed on the astroglia that envelop to the actin cytoskeleton through two accessory proteins:
spines. This interaction leads to the retraction of spines .catenin, which binds to cadherin, and -catenin, which
(Murai et al., 2003). In this way Eph receptors and ephrin binds to both .catenin and actin. Cadherin dimers form
interactions could provide a mechanism for structural changes strongly adhesive homophilic partnerships with similar cadhe-
underlying synaptic plasticity. rin dimers on the other side of the synapse. Interestingly,
although both E- and N-cadherins are strongly expressed in
Cell Adhesion Molecules the hippocampus, N-cadherins are found only at excitatory
synapses (Huntley, 2002). The importance of cadherins in the
Cell adhesion molecules (CAMs) comprise a large class of induction of LTP is beginning to emerge from studies using
diverse membrane-spanning molecules with extracellular lig- inhibitors, targeted mutations, and confocal microscopy.
and-binding domains that are important for cell-cell recogni- Tetanus-induced LTP in area CA1 is reduced by peptide block-
tion during neural development and are potential candidates ers of N-cadherin (Tang et al., 1998). Late, delayed-onset LTP,
for both anterograde and retrograde transsynaptic signaling. induced in hippocampal slices by bath application of a mem-
CAMs expressed by neurons fall into four broad types: inte- brane-permeable cAMP analogue, is accompanied by a rise
grins, cadherins, immunoglobulin superfamily CAMs, and in synaptic puncta containing both N-cadherins and the presy-
neuroligins and their ligands neurexins (reviewed by Benson naptic marker synaptophysin (Bozdagi et al., 2000). Con-
et al., 2000 and Yamagata et al., 2003). CAMs, via their extra- versely, cAMP-induced LTP is abolished by bath application of
cellular domains, can bind to the same members of the family blocking antibodies against the dimerization and homophilic
(homophilic binding) or to other members (heterophilic interaction domains of N-cadherins (Bozdagi et al., 2000).
binding). The C terminal region of many CAMs is associated The effect on spine morphology of blocking N-cadherin in
directly or via accessory proteins to the actin cytoskeleton and cultured hippocampal neurons was investigated by Togashi et
can potentially initiate changes in the morphology of the al. (2002), who found that over a period of 2 days or more
spine or bouton in which they are embedded. In this section spines became longer and more lopodia-like. The same effect
386 The Hippocampus Book

was observed in mutant mice lacking -catenin, suggesting intercellular interactions has little effect on baseline transmis-
that cadherin signaling through actin is required for spine sion but, as we have seen, can impair the magnitude and dura-
shortening. There is some information on how synaptic activ- tion of LTP. A major gap in our knowledge is how synaptic
ity modulates N-cadherin function. Strong synaptic activation activity is signaled to and detected by CAMs. It seems likely
can change the conformation of N-cadherin (Tanaka et al., that the major contribution of CAMs is as mediators of struc-
2000), and Murase et al. (2002) have made the signicant tural changes, where they can ensure that pre- and postsynap-
observation that .catenin is rapidly (within 30 minutes) tic changes can be kept in register. However, the potential for
translocated from dendrite to spine in an NMDAR-dependent two-way signaling, from cytoplasm to cell adhesion molecule
manner following synaptic activity. Finally, tyrosine dephos- and vice versa, remains a highly attractive mechanism for con-
phorylation of .catenin promotes its movement into the veying information across the synapse that is only now begin-
spine, providing a potential point for regulation by synaptic ning to be explored.
activity via a calcium-dependent tyrosine phosphatase
(Murase et al., 2002). 10.4.9 Late LTP: Persistent Potentiation Requires
Gene Transcription and Protein Synthesis
Integrins. Integrins are heterodimeric glycoproteins formed
from . and .subunits; each comprises a single membrane- Late LTP (L-LTP) is dened as the temporal component of
spanning domain linking the ligand-binding extracellular LTP that is abolished by protein synthesis inhibitors or by
domain to a cytoplasmic tail that is connected to actin via the transcriptional blockers. Because no stimulus protocol has
accessory proteins -actinin, talin, and vinculin (Benson et been found that produces L-LTP without the preceding early
al., 2000). There are multiple forms of each subunit, a number LTP, the time course of L-LTP must be estimated by a subtrac-
of which have been identied in hippocampal pyramidal tion process. This reveals slow-onset potentiation rising from
neurons, including 3, 5, 8, and 8 (Chan et al., 2003), The baseline to reach a plateau after 1 to 2 hours. The slowly grow-
primary extracellular ligands of integrins are matrix proteins, ing L-LTP induced by BDNF (Fig. 102D) or cAMP analogues
but integrins also participate in cell-cell adhesion through has a similar time course.
binding to cadherins and members of the immunoglobulin In principle, persistent changes in synaptic efficacy could
superfamily of cell adhesion molecules. Their role in the occur through self-perpetuating renewal of posttranslational
stabilization of LTP has been explored by Lynch, Gall, and changes (Crick, 1984; reviewed by Routtenberg and Rekart,
colleagues (reviewed in Gall and Lynch, 2004). Broad- 2005). Indeed, Kandel and colleagues found evidence for such
spectrum peptide inhibitors of integrins attenuate the magni- a mechanism in the translational regulator CPEB (cytoplas-
tude and duration of LTP and can reverse LTP if given within mic polyadenylation element-binding protein), a protein with
a few minutes after induction (Staubli et al., 1998). In a survey prion-like self-perpetuating properties that is involved in
of mice with heterozygous deletions of 3, 5, or 8 integrin long-lasting facilitation in Aplysia (Si et al., 2003). However,
(homozygous deletions are lethal), Chan et al. found that the same year that Cricks proposal was published, the rst
LTP was attenuated only in the 3 mutants. Complete sup- evidence appeared that protein synthesis was indeed essential
pression of LTP and a decit in watermaze performance was if LTP in the hippocampus was to persist for more than a few
only seen in mice heterozyous for all three integrins (Chan et hours. Krug et al. (1984) infused the protein synthesis inhibi-
al., 2003). tor anisomycin into the lateral ventricles of freely moving rats
and found that the potentiated fEPSP in the dentate gyrus
Neuroligins and Neurexins. Neuroligins and their ligands slowly declined to baseline over 3 to 5 hours. The baseline
neurexins are brain-specic cell adhesion molecules, with response was not affected in control experiments. This result
multiple forms arising from differential splicing. Neurexins appeared to establish an upper limit for the duration of post-
are localized presynaptically, and members of the subgroup translational changes. A similar result was obtained in the
make heterophilic connections with postsynaptic neuroligins. anesthetized rat (Otani and Abraham, 1989) and in area CA1
The intracellular C-termini of neuroligins bind to the PDZ in the hippocampal slice (Frey et al., 1988) (Fig. 102C).
region of the postsynaptic protein PSD-95, and the C-termini Transcriptional inhibitors such as actinomycin also block L-
of neurexins are linked to another PDZ-containing protein, LTP in area CA1 in vitro (Nguyen et al., 1994) and in the den-
CASK. As Benson et al. (2000) pointed out, these interactions tate gyrus in vivo (Frey et al., 1996; but see Otani et al., 1989).
make neuroligins and neurexins potentially well suited for Mossy ber L-LTP, too, as discussed in Section 10.5, is depend-
intercellular signaling. Their involvement in LTP has yet to be ent on protein synthesis and gene transcription (Huang et al.,
explored. 1994). In addition, the slowly growing L-LTP elicited by cAMP
analogues (Nguyen et al., 1994) or by BDNF (Kang and
Summary Schuman, 1996; Messaoudi et al., 2002) (see Section 10.6.1) is
susceptible to translational and transcriptional inhibitors.
It is evident from this brief survey of the large, diverse families Thus, all forms of L-LTP that have been identied in the hip-
of neural cell adhesion molecules that there is still much to pocampus require protein synthesis. Persistent LTD induced
learn about their role in synaptic plasticity. Disruption of by mGluR agonists or by paired-pulse low-frequency stimula-
 Synaptic Plasticity in the Hippocampus 387

tion is also rapidly blocked by protein synthesis inhibitors (see memories after retrieval may open up another window of pro-
Section 10.7.4). tein synthesis dependence in the hippocampus (Alberini,
The rapid effect of translational inhibitors suggests that the 2005).
initial phase of L-LTP is dependent on protein synthesis from It is not known if there are proteins that are expressed only
preexisting mRNA in the dendrites of potentiated synapses in the context of synaptic plasticity and in no other circum-
(see below). In fact, despite an earlier negative report (Frey et stance, but is important to distinguish between synthesis of
al., 1989), there is persuasive evidence that isolated dendrites proteins that are related to the changes in synaptic strength
can support L-LTP (Vickers et al., 2005). Moreover, local and synthesis of general housekeeping proteins that reect
dendritic application of a protein synthesis inhibitor reduces and service overall changes in neural activity (see discussion
L-LTP in intact slices (Bradshaw et al., 2003). Nevertheless, in Sanes and Lichtman, 1999). These two possibilities cannot
dendritic protein synthesis cannot be the whole story, given be distinguished on the basis of a general blockade of protein
the impairment of L-LTP by transcriptional inhibitors. The synthesis. However, the fact that translational and transcrip-
effect of transcriptional inhibitors is delayed, beginning 1 to 2 tional inhibitors are effective only during a narrow temporal
hours after induction. The delay is consistent with a model in window around the time of induction, well before L-LTP is
which a signal from stimulated synapses is transported to the established, implies that proteins synthesized during that time
nucleus, where it triggers transcription of new mRNA species. are not predominantly housekeeping genes.
The products of transcription, whether mRNA transcripts Does protein synthesis in presynaptic neurons contribute
themselves or the proteins they encode, are delivered to den- to L-LTP? In the transverse hipppocampal slice, axons project-
drites to stabilize LTP at potentiated synapses. ing to granule cells are severed from their cell bodies in the
We consider in more detail below the evidence that the ini- entorhinal cortex. The fact that L-LTP can be induced in the
tial phase of L-LTP is supported by dendritic protein synthe- dentate gyrus in vitro (Balschun et al., 1999) therefore
sis. We then discuss the role of transcription in generating the demonstrates that presynaptic gene transcription is not a
persistent late phase, which leads to a description of the two requirement for L-LTP in the perforant path. Presynaptic pro-
(somatopetal and somatofugal) signaling trade routes and the tein synthesis, however, cannot be excluded on this evidence
largely unidentied molecular cargoes that are exchanged alone, and an LTP-associated increase in a number of presy-
between soma and dendrite. Finally, we consider the concept naptic proteins has been reported following the induction of
of synaptic tagging and the surprising answers this gives to the LTP in perforant path-granule cell synapses (Kelly et al.,
perplexing question of how a dendritic ow of plasticity mol- 2000). Whether these changes are necessary for the full expres-
ecules generated by transcription is restricted to potentiated sion of L-LTP is not known.
synapses to preserve the input specicity of LTP. Comprehen-
sive reviews of the ways in which translation and transcription Dendritic Protein Synthesis Plays a Role
may be regulated to achieve persistent L-LTP can be found in in the Initial Phase of Late LTP
Kelleher et al. (2004a) and West et al. (2002).
Proteins required for L-LTP could be all be generated in the
Time Window for Blocking Induction soma and transported to where they are required. However,
of Late LTP by Protein Synthesis Inhibitors there is increasing evidence that this is not what happens and
that dendritic protein synthesis plays an important role in
An important and somewhat puzzling feature of the block of generating the initial phase of late LTP. Local protein synthe-
L-LTP by protein synthesis or transcriptional inhibitors, both sis has the conceptual merit of providing an immediate solu-
in vitro and in vivo, is that the drugs are effective when given tion to the problem of input specicity at the metabolic cost
during or immediately after induction of LTP (Otani et al., of transporting a population of mRNAs to the far reaches of
1989; Nguyen et al., 1994) but are ineffective once LTP has the dendritic tree. It would, on the face of it, make the cells life
become established. The existence of this time window a good deal easier if the machinery for late as well as early LTP
implies that the synthesis of the proteins necessary to support were located at the synapse, and there are good reasons for
L-LTP occurs at the time of or soon after the tetanus and that supposing that protein synthesis does in fact occur in den-
thereafter new protein synthesis is not required, even though drites. It has long been known that ribosomes and other ele-
expression of L-LTP can take some hours to reach its maxi- ments of the protein synthetic machinery are positioned
mum and is then maintained at that level. A similar prole is subsynaptically at many hippocampal synapses (Steward,
seen in behavioral experiments when protein synthesis 1983; reviewed by Steward and Schuman, 2001), and ultra-
inhibitors are given around the time of one-trial learning; structural studies have shown that polyribosomes redistribute
short-term memory is unaffected by the inhibitors, and long- from dendrites to the base of dendritic spines with enlarged
term memory is blocked (Squire and Barondes, 1972). As postsynaptic densities following the induction of LTP (Ostroff
we shall see later when considering the synaptic tag experi- et al., 2002). An estimated 400 mRNA species are found in
ments of Frey and Morris (1997), there are circumstances in hippocampal dendrites (Eberwine et al., 2001), including
which L-LTP can be induced in the presence of protein syn- components of the translation machinery as well as mRNAs
thesis inhibitors. We also note here that the reconsolidation of for several proteins known to play a role in the induction or
388 The Hippocampus Book

expression of LTP, among them CaMKII, PKM arc/arg3.1, synthesis at both the local and global levels is involved in the
BNDF and its receptor trkB, and AMPA receptor subunits conversion of early to late LTP.
GluR1 and GluR2 (Martin and Kosik, 2002; Martin, 2004).
The demonstration that isolated dendrites of CA1 pyramidal Signaling to Nucleus
cells can support BDNF-induced LTP and mGluR-dependent
LTD provides direct evidence that dendritic protein synthesis A major route for activity-dependent protein synthesis
is sufficient to support persistent synaptic plasticity (Kang and proceeds via long-range activation of nuclear transcription
Schuman, 1996; Hber et al., 2000); in both cases the effects factors, including members of the CREB family of proteins
are blocked by protein synthesis inhibitors. Tetanus-induced which bind to cAMP response elements (CREs) in the regula-
LTP lasting a number of hours also occurs in isolated den- tory regions of target genes to initiate transcription. What is
drites, and its decay is hastened by protein synthesis inhibitors the signal, and how is it conveyed to the nucleus? We have seen
(Frey et al., 1989). These experiments demonstrate that trans- that calcium inux through the NMDA receptor is the trigger
lation of existing dendritic mRNA can support a component for NMDAR-mediated LTP; and it is therefore no surprise that
of L-LTP that extends well beyond the duration of E-LTP. calcium, particularly when bound to calmodulin, is a key
Evidence that local protein synthesis is necessary for the full player in signaling to the nucleus (reviewed by Lonze and
expression of L-LTP in intact cells has come from experiments Ginty, 2002). There is disagreement, however, about how
in which local perfusion of the protein synthesis inhibitor this is achieved. One possibility is that calcium is released
emetine selectively depressed L-LTP in the perfused region from the endoplasmic reticulum as a traveling wave proceed-
without affecting its time course at distant synapses ing from dendrite to soma, where it binds to nuclear calmod-
(Bradshaw et al., 2003). Overall, these observations strongly ulin to activate target kinases (Hardingham et al., 2001). A
suggest that local protein synthesis plays an important part in second possibility is that calcium/calmodulin translocates
the early stabilization of LTP or, to put it another way, in the from synapse to nucleus (Deisseroth et al., 1998); and a third,
passage from early to late LTP. discussed below, is that Ca2 enters the nucleus via voltage-
gated calcium channels in the somatic plasma membrane.
Transcription and Translation of New mRNAs: A principal target for nuclear calcium/calmodulin is
Dialogue Between Synapse and Nucleus CaMKIV, a member of the calcium/calmodulin-dependent
kinase family whose location is predominantly nuclear (Lonze
What is the evidence that the conversion from early to late LTP and Ginty, 2002). Activation of CaMKIV by phosphorylation
requires gene transcription and/or somatic protein synthesis? at ser133 allows it to bind an accessory CRE-binding protein
First, as we have seen, transcriptional blockers such as actino- (CBP) and initiate transcription. Other Ca2-activated signal-
mycin block L-LTP in area CA1 in vitro (Nguyen et al., 1994; ing pathways are also recruited, including the Ras/MAP kinase
Frey et al., 1996; Kelleher et al., 2004b) and in the dentate and PKA pathways (for reviews see Roberson et al., 1999;
gyrus in vivo (Frey et al., 1996; but see Otani et al., 1989). Lonze and Ginty, 2002; Sweatt, 2004).
Second, upregulation of immediate early genes such as c-fos, The MAP kinase signaling cascade can be triggered by a
zif268, and arc/arg3.1 reveal a correlation between transcrip- wide variety of stimuli, including extracellular ligands binding
tion and LTP induction (reviewed by Kuhl and Skehel (1998). to receptor tyrosine kinases and G protein-coupled receptors,
The absence of L-LTP in the dentate gyrus of a mutant mouse and by synaptic activity leading to calcium entry through
with targeted inactivation of the immediate early gene and ionotropic glutamate receptors or voltage-gated calcium
transcription factor zif268 conrms a causal role for gene channels. We have seen in Section 10.4.3 that the MAP kinase
transcription in the induction of L-LTP (Jones et al., 2001), as cascade regulates both early LTP and late LTP via substrate
does the impairment of L-LTP when the same region is proteins that control translation, probably at dendritic sites
infused with antisense mRNA to arc/arg3.1 (Guzowski et al., (reviewed by Sweatt, 2001b; Thomas and Huganir, 2004).
2000). Finally, experiments on synaptic tagging, described Here, we are concerned with its role in establishing L-LTP
below, make it clear that a global signal, presumed to originate by stimulating gene expression in the nucleus. Synaptic
from the soma, is made available to the whole cell following activity leads to the phosphorylation of ERK1, 2 in both den-
the induction of L-LTP. As we shall see, this signal can be drites and soma. As with the calmodulin signal, the mecha-
intercepted and utilized for the purposes of stabilizing LTP at nism by which the activated ERK signal is translocated to the
other synapses that may have been activated up to 1 to 2 hours nucleus is not known. Although phospho-ERK does not phos-
earlier or later than the pathway whose strong activation led to phorylate CREB, inhibition of MEK, the kinase that phospho-
the initiation of transcription. Local protein synthesis alone rylates ERK, prevents phosphorylation of CREB on ser133.
cannot account for the observation that plasticity can be mod- There must therefore be intermediate kinase(s) that are
ulated in this associative manner by a geographically distant substrates for phospho-ERK and allow it control over
pathway; there has to be a transcriptional or translational sig- CREB phosphorylation; these intermediates are thought to be
nal from the soma that either supplies the required stabilizing members of the RSK or MSK families of nuclear kinases
proteins or transcripts or stimulates local protein synthesis to (Lonze and Ginty, 2002; Thomas and Huganir, 2004) (see
do so. Thus, it is difficult to avoid the conclusion that protein below).
 Synaptic Plasticity in the Hippocampus 389

Another route from synapse to soma is via the activation of can, MAPK cannot. Unexpectedly, however, it was found that
G protein-coupled receptors, such as D1/D5 dopamine recep- MEK inhibitors completely block the increase in CREB phos-
tors. Late LTP in area CA1 is blocked by dopamine antagonists phorylation caused by bath application of the cAMP activator
(Frey et al., 1990) and facilitated by D1/D5 agonists, although forskolin, thus establishing that ERK1, 2 acts downstream of
the report by Huang and Kandel (1995) that stimulation the PKA cascade (Roberson et al., 1999). The activation of
of dopamine receptors can induce late potentiation in an PKA by cAMP leads to activation of the GTPases Rap-1 and B-
activity-independent manner has been disputed by Mockett et raf and thence to phosphorylation of MEK. The kinase inter-
al. (2004). Dopamine receptor activation leads to upregula- mediaries linking MAPK to CREB in the L-LTP pathway in
tion of cAMP and presumptive translocation of activated PKA vivo are probably members of the RSK and/or MSK kinase
to the soma, where it can directly phosphorylate CREB on families. A second arm of the MAP kinase cascade is activated
ser133 (Lonze and Ginty, 2002). by ligand binding to metabotropic glutamate receptors, acti-
vating PLC and leading to the release of diacyl glycerol and
Action Potential Signaling as a Form activation of PKC. PKC activates Ras/Raf-1, the rst links in
of Synapse-to-Nucleus Signaling the three-kinase MAP cascade, leading to phosphorylation of
MEK. BDNF, acting on its receptor TrkB, also uses the
As Adams and Dudek (2005) have pointed out, there are two Ras/Raf-1 pathway, again converging on MEK, as does acetyl-
problems with the concept of synapse-to-soma signaling that choline (ACh) acting on muscarinic receptors (Sweatt 2004).
are often overlooked. First, how does the signal generated by Thus, many of the extracellular stimuli (ACh, BDNF, gluta-
one or a few potentiated synapses (in the limit, a single mate) that are known to cause CREB phosphorylation on
synapse) located at any position on the dendritic tree reach a ser133 and thence gene expression, are funneled through the
concentration in the nucleus that allows activation of tran- Ras/Raf-1 arm of the MAPK pathway (Sweatt, 2001a, b).
scription factors? Second, transcription of immediate early Our discussion of signaling pathways in L-LTP has so far
genes such as c-fos and zif268 begins within a few minutes of avoided the question of how specic genes are targeted by dif-
LTP induction. How does the signal reach the soma so ferent stimuli if the same signal cascades are engaged in each
quickly? Adams and Dudek (2005) suggested that both prob- case. It is likely that different combinations of kinases are acti-
lems can be circumvented if the signal is generated not at the vated by different stimuli and that expression of specic target
synapse but by somatic action potentials. Because action genes is achieved by combinatorial selection from the cells
potential(s) generally (though not always, see Section 10.3.6) portfolio of transcription factors. For example, transcription
occur during the induction of LTP, the nuclear signal could in of the immediate early gene c-fos can be triggered by any one
principle be generated by Ca2 entering through somatic of several stimuli (growth factors, neuromodulators, neuro-
voltage-gated channels. Such a mechanism would provide an transmitters), all of which lead to phosphorylation of CREB
economical solution to both the problems noted above. A on serine 133. Neurotransmitters, but not growth factors or
number of observations are consistent with the hypothesis. neuromodulators, increase calcium levels, either directly
Antidromically driven action potentials can induce expression through opening calcium permeable receptor channels (e.g.,
of immediate early genes, and antidromic stimulation of CA1 the NMDA receptor) or indirectly through depolarization-
axons converts E-LTP to L-LTP, suggesting that action poten- gated opening of L-type voltage-sensitive calcium channels.
tials can induce expression of soma-to-synapse plasticity fac- High Ca2 levels can lead to phosphorylation of another tran-
tors (Dudek and Fields, 2002). Moreover, it is well established scription factor, calcium response factor (CRF), and to phos-
that the synaptic activation of NMDA receptors during high- phorylation of CREB at ser141and ser143; phosphorylation at
frequency stimulation can drive somatic action potentials and the latter two serine residues reduces the ability of ser133
so provide a rapid NMDAR-dependent signal from synapse to phospho-CREB to bind to CBP. The combination of these two
soma (Herron et al, 1986). Furthermore, synaptic activation of transcription factors (CRF and CREB phosphorylated at
NMDA receptors can lead to somatic Ca2 entry via L-type ser141 and ser143 but unassociated with CBP) directs the
voltage-gated Ca2 channels (Alford et al, 1993). In addition, stimulus-dependent transcription of BDNF (West et al.,
strong induction protocols that induce NMDAR-independent 2002). Similarly, activation of at least two transcription fac-
L-LTP in area CA1 also lead to a large somatic Ca2 signal that tors, CREB and Elk1, is required to drive expression of the
is blocked by L-type Ca2 channel inhibitors (Raymond and immediate early gene zif268 (Davis et al., 2000).
Redman, 2006). Note that the somatic Ca2 hypothesis pre- The LTP connection was taken a step further when CREB
dicts an absolute requirement for postsynaptic action poten- phosphorylation induced by tetanic stimulation was also
tials to secure L-LTP by this route. shown to be blocked by MEK inhibitors, both in CA1 in vitro
and in the dentate gyrus in vivo (Impey et al., 1998; Davis et
The Nuclear Transcription Factor CREB al., 2000). In the latter case, upregulation of zif268, one of the
Is a Target for Several Kinases target genes of CREB, was also suppressed. While CREB is
rapidly phosphorylated in a Ca2 dependent manner by stim-
CREB is activated by phosphorylation on ser133. Not all active uli that induce LTP, it is also the case that CREB is activated by
kinases can directly phosphorylate CREB on this residue: PKA patterns of synaptic activity that do not elicit potentiation
390 The Hippocampus Book

(Deisseroth et al., 1966). Thus, CREB phosphorylation may be inactivation of zif268, indicating that zif268 is required for L-
required for the transcription of genes encoding effector pro- LTP, though it is not necessary for E-LTP (Jones et al., 2001).
teins whose role is to sustain synaptic activity in general, However, enhanced expression of transcription factors such as
rather than specically to promote synaptic plasticity. Initial zif268 and c-fos can be induced by procedures that do not
studies of genetically hypomorphic CREB mice supported the induce LTP, such as nonsynaptic excitation of neurons by
idea that CREB was important for LTP as well as for hip- antidromic stimulation. The mechanical disturbance involved
pocampus-dependent memory (Bourtchuladze et al., 1994). in cutting a hippocampal slice can also induce long-lasting
Later work on CREB knockout mice failed to detect a robust upregulation of IEG expression in the dentate gyrus (French
effect on LTP, though there remained the possibility that this et al., 2001). In these cases, upregulation of IEGs is not suffi-
result could be explained by the upregulation of other mem- cient to induce LTP, possibly because synaptic tags have not
bers of the CREB family. However, Balschun et al. (2003) gen- been set. Note too that although zif268 and other immediate
erated a mouse in which expression of all CREB isoforms was early genes can be activated by seizure activity in all areas of
either absent or limited, and found no effect on L-LTP. The the hippocampus they are not readily switched on by LTP-
role of CREB in generating L-LTP cannot therefore be inducing stimulation in area CA1, in contrast to their robust
regarded as securely established. upregulation in the dentate gyrus (French et al., 2001). On the
What are the target genes for CREB? Some of them at least other hand, there is an increase in the number of pyramidal
are transcription factorsthe immediate early genes such as neurons in area CA1 expressing arc/arg3.1 mRNA following
c-fos and zif268 and others of unknown function such as exposure to a novel environment (Guzowski et al., 1999).
arc/arg3.1. An upper limit on the number of genes that might Until the reasons for these puzzling and intriguing discrepan-
be activated by CREB can be estimated from genes containing cies between the pyramidal and granule cell elds are resolved,
CRE sequences in their promoter regions (Lonze and Ginty, the use of IEG expression as a marker for the induction of
2002). More than 80 have been identied, coding for: neuro- L-LTP is probably justied only in the dentate gyrus.
transmitters and peptides (n  35), growth factors and their
receptors (n  7), structure-related proteins (n  4), proteins Beyond Immediate Early Genes:
involved in cellular metabolism (n  16), transcription factors Effector Genes that Promote Late LTP
(n  13), signal transduction proteins (n  8). The presence
of transcription factors opens up the potential for a second The identities of effector genes that are activated either in the
wave of target genes. rst round of gene expression by constitutively active tran-
scription factors such as CREB, or in the subsequent round by
Markers of LTP: Immediate Early Genes IEG transcription factors such c-fos and zif268 are still largely
zif268, c-fos, and arc/arg3.1 unknown. Some clues can be gathered from the fact that not
all IEGs encode for transcription factors, and several of those
Long-term potentiation is associated with the rapid, transient that are not transcription factors are targeted to dendrites,
expression of several transcription factors, including c-fos, either as mRNA (arc/arg3.1, homer, CaMKII) or as protein
cJun, krox 20, and zif268, of which c-fos and zif268 have been (BDNF). Transcription of Arc/arg3.1 is itself regulated by
the most intensively studied (Dragunow, 1996; Lanahan and zif268, and by another member of the egr family of transcrip-
Worley, 1998). These and other genes, such as arc/arg 3.1 and tion factors, egr3 (Li et al., 2005). Differential screens for genes
homer1a, which encode for proteins that contribute to synap- up- or downregulated by the induction of late LTP in the den-
tic scaffolding complexes, are members of a group of genes tate gyrus of the freely moving animal have yielded lists of a
known as immediate early genes (IEGs) whose transcription is dozen or so transcripts that are upregulated 75 minutes after
upregulated by activity without the need for prior protein LTP induction in the freely moving rat (Matsuo et al., 2000) or
synthesis. Enhanced mRNA expression for both c-fos and 1.5 to 7.0 hours after induction in the anesthetized rat
zif268 (a member of the krox family of transcription factors; (Hevroni et al., 1998). These plasticity-related genes are in
synonyms are krox24, egr1, and NGF1) occurs within minutes each case a subset of a considerably larger group of genes
of induction and peaks around 30 minutes after induction, whose expression is stimulated by seizure activity. Another dif-
returning to baseline levels (which in the dentate gyrus is very ferential display study followed the time course of the regula-
low for zif268, c-fos, and arg/arc3.1) within 2 to 4 hours. tion of a transcript for A kinase anchoring protein
Expression of the corresponding proteins lags by 1 to 2 hours. (AKAP-150). Unusually for an activity-regulated transcript,
Transient upregulation of IEGs is a marker for LTP in the den- AKAP mRNA was not altered by seizures but was markedly
tate gyrus in the sense that whenever a stimulus is sufficiently affected by LTP-inducing stimulation, exhibiting an initial
strong to induce L-LTP immediate early gene expression is decrease at 1 hour and climbing to a peak increase at around 6
triggered. Moreover, maneuvers that block induction of LTP hours before falling back to baseline by 24 hours after induc-
also block the enhanced expression of immediate early genes tion (Gnin et al., 2003). Using probes for specic transcripts
(Wisden et al., 1990; Dragunow, 1996). As we have seen, L-LTP at different times after induction Thomas et al. (1994) were
is not expressed in the dentate gyrus of mice with a targeted able to follow the time course of tetanus-induced changes in
 Synaptic Plasticity in the Hippocampus 391

mRNA for a number of serine/threonine protein kinases in the old for triggering local or somatic protein synthesis but
dentate gyrus; CaMKII, and PKC showed transient in- nevertheless was able to set a tag that could capture the new
creases, and increases in ERK2 and B-Raf were seen only at 24 proteins synthesized as a result of the strong tetanus to the
hours. A nding that suggests a direct link to synaptic function second pathway.
was reported by Nayak et al. (1998): LTP in isolated CA1 den- Late LTD is also associated with the setting of a tag
drites is associated with a PKA- and transcription-dependent (Kauderer and Kandel, 2000). Sajikumar and Frey (2004a)
increase in the synthesis of AMPA receptors. Study of the iden- presented evidence that it is the tag, rather than the plasticity
tication and functional characterization of effector proteins factors, that determines the direction of change; thus, LTD-
that sustain maintenance of L-LTP, and the nature of their inducing stimuli generate an LTD-specic tag and LTP-
interaction with the putative synaptic tag, is likely to prove a inducing stimuli an LTP-specic tag (but see below for evi-
challenging and fertile eld for future research. Meanwhile, a dence that PKM is an LTP-specific plasticity factor).
systematic study of transcriptional changes related to LTP and Although the tag normally has a lifetime of 1 to 2 hours, it can
their time courses has now become feasible through the use of be canceled prematurely if depotentiation is induced by low-
DNA microarray screening (Lein et al., 2004 Lee et al., 2005; frequency stimulation delivered a few minutes after LTP
Wibrand et al., 2006). induction (Sajikumar and Frey, 2004b). When tags are set on
two pathways, they compete for the available supply of plas-
Return Pathway: Signaling from Nucleus ticity-related proteins, leading to a process of competitive
to Synapse and the Synaptic Tag Hypothesis maintenance in which potentiation at one pathway occurs at
the expense of a reduction in potentiation at another (Fonseca
How are somatically generated transcripts and/or proteins et al., 2004).
delivered to the appropriate synapses? Frey and Morris (1997) Although the identity of the tag (or tags) is unknown, can-
considered a number of possibilities, of which conceptually didate tags must satisfy the following criteria (Barco et al.,
the simplest was the idea that a potentiated synapse sets a 2002; Kelleher et al., 2004a): (1) tags are generated by an activ-
marker or tag that is recognized by the mRNA or protein ity-dependent, NMDAR-dependent, but protein synthesis-
products that are transported into dendrites from the soma. independent process; (2) they have a lifetime of 1 to 2 hours
This tag must have a lifetime at least long enough to engage but are susceptible to cancelation by low-frequency trains
with the newly generated mRNA or proteins. Transcriptional within the rst few minutes of being set; (3) at least two tags
inhibitors given at the time of induction can take 2 hours or can be set, one induced by activity patterns that produce early
more to begin to affect LTP (Kelleher et al., 2004a; but see or late LTP and the other by activity leading to early or late
Barco et al., 2002 for a case where the transcriptional inhibitor LTD; (4) tags are input-specic; (5) tags interact with proteins
actinomycin also acts rapidly). The relative delay in the action generated by de novo nuclear transcription (the process of
of transcriptional inhibitors presumably reects the time synaptic capture) to trigger the conversion of early to late
taken for a signal to pass from synapse to nucleus for tran- LTP or early to late LTD. What clues to the identity of the tag
scription rst of IEGs and then of genes encoding effector do these criteria offer? The rapid, reversible, protein synthesis-
proteins, followed by transport of mRNA and/or proteins into independent setting of the tag points toward a protein kinase,
dendrites and their postulated capture by tagged spines. Frey possibly PKA (Barco et al., 2002), as candidate tags for LTP
and Morris speculated that the setting of the tag itself might and protein phosphatases for LTD (Frey and Morris, 1998a;
be protein synthesis-independent. If this were the case, a Kelleher et al., 2004a). The possibilities are much broader than
tetanus to one pathway should still generate L-LTP even in the this, however; any class of synaptic molecule whose phospho-
presence of a protein synthesis inhibitor if a second pathway rylation state, conguration, or concentration is altered as a
was tetanized within a time window of 1 to 2 hours on either result of NMDAR-mediated events is in principle a candidate
side of the tetanus to the rst pathway. The tag would be set in tag. Finally, it is worth emphasizing that a tag can be set with-
the rst pathway, notwithstanding the presence of the protein out inducing L-LTP; so although the setting of a tag may or
synthesis inhibitor, and would capture the plasticity-related may not be a necessary precursor for induction of late LTP or
proteins generated by the second pathway. These predictions late LTD, it is clearly not sufficient.
were conrmed (Frey and Morris, 1997). (Note that the con- The Frey and Morris experiments revealed a complex pat-
version of E-LTP to late L-LTP begins before the protein syn- tern of temporal and spatial associativity in hippocampal
thesis inhibitor has been washed out, which suggests that both synaptic plasticity, which they encapsulated by reference to the
the tag and the plasticity factor interacting with the tag are notion of ashbulb memory. A cell is continually bom-
proteins rather than mRNA transcripts). Equally arresting was barded with events of moderate signicance to the organism;
the demonstration that E-LTP, produced by a tetanus that at inputs activated by these events, tags are set for an hour or
by itself was too weak to produce L-LTP, could be converted to two and then expire (whether tags decay over their lifetime or
L-LTP by a preceding or subsequent strong tetanus given to a behave as molecular switches with on and off modes is
second pathway within a similar time window. Again, this unknown). When an event of major signicance occurs,
result can be explained if the weak tetanus was below thresh- equivalent to a strong tetanus, late LTP is induced not only in
392 The Hippocampus Book

NMDA receptor Ribosome strong

2+ HFS
Ca
AMPA receptor L-glutamate
PKA
MAPK
Dopamine receptor Dopamine

Adenylyl Cyclase LTP tag LTP tag

set

G-protein LTD tag weak

HFS
2+
Ca
VGCC

LTP tag
set

no
stimulus

No tag
set
Modulatory input
e.g. dopamine LFS
2+
Ca

LTD tag
ATP set
cAMP

Ca 2+ Ca 2+
PKA
MAPK

IEGs
CREB

Figure 1013. Signaling pathways involved in the genesis of late soma through voltage-gated Ca2 channels as a result of action
LTP. In this scheme, LTP induced by strong tetanic stimulation is potentials generated during activity. The targets of synapse-to-soma
represented by the distal synapse (top right). The large, rapid signaling are transcription factors, including CREB. Phosphoryla-
increase in Ca2 stimulates local changes, leading to phosphoryla- tion of CREB leads to the transcription of a number of genes that
tion of receptors and other proteins (see Fig. 108); it also activates contain CRE (cAMP response elements) domains in their regula-
synapse-to-soma signaling cascades that may involve translocation tory regions, among which are the early immediate genes such as
of Ca2/calmodulin, PKA and MAPK. At the same time, a synaptic arc/arg3.1, homer, c-fos, and zif268, some of which are themselves
tag is set to identify the synapse as having been recently activated. transcription factors. The return leg of the signaling system carries
Weaker stimulation, leading only to early LTP, does not trigger a transcripts (e.g., mRNA for arc/arg3.1 and PKM) and unidentied
synapse-to-soma signal but does set a tag. Unstimulated synapses proteins from the nucleus into dendrites. According to the tag
are left in molecular peace, whereas synapses receiving prolonged hypothesis (see text), the soma-to-dendrite messengers are inter-
low-frequency stimulation (LFS) set an LTD tag and may generate cepted or captured by the tag, the nature of which has not been
similar nuclear signals (not shown). Additional dendrite-to-nucleus identied, and this triggers the changes that lead to late LTP (for an
and soma-to-nucleus signals may be contributed by modulatory LTP tag) or late LTD (for an LTD tag).
inputs such as dopaminergic innervation and by Ca2 entering the
 Synaptic Plasticity in the Hippocampus 393

that pathway, but via tagging and synaptic capture, with all study using a myristoylated form of a -pseudosubstrate
other pathways carrying information about events of minor inhibitory peptide (termed ZIP), that binds to PKM and
signicance occurring an hour or two on either side of the reconstitutes the autoinhibition in the absence of the PKC
signicant event. We can speculate with Frey and Morris that autoregulatory domain, blocked induction of L-LTP but not
through the tagging mechanism these secondary events E-LTP (Serrano et al., 2005). Indeed, there is evidence that
acquire the same mnemonic potential as the major event PKM is a plasticity factor that specically targets synapses
and provide the contextual information that allows us to exhibiting a tag for LTP. In tagging experiments in which two
remember, for example, where we were and what we were independent pathways were given weak and strong tetani off-
doing on 9/11. A cartoon of some of the concepts we have set by 30 min, ZIP reversed L-LTP in both pathways. But when
been discussing in this section on late LTP is presented in a strong tetanus to one pathway was given in conjunction with
Figure 1013. a weak LTD-inducing protocol to the other, ZIP reversed the
maintenance of L-LTP but not of L-LTD; conversely, when a
The Duration of Late LTP Is Regulated strong LTD inducing protocol to one pathway was paired with
by the NMDA Receptor a weak tetanus to the other, the maintenance of L-LTP but
again not of L-LTD was disrupted (Sajikumar et al., 2005).
We have seen that the expression of L-LTP depends on de novo Note that this result is on the face of it inconsistent with the
protein synthesis around the time of induction, give or take a suggestion by Sajikumar and Frey (2004b) that newly tran-
few hours during which a prior or later setting of the synaptic scribed plasticity factors are process-independent, with the
tag can engage with newly synthesized proteins. Thereafter, direction of change being determined solely by local LTP- or
LTP proceeds without benet of further de novo protein syn- LTD-specic tags.
thesis. Surprisingly, among the various largely uncharacterized The mechanism by which PKM-dependent L-LTP is
posttranslational molecules and mechanisms that sustain the maintained appears to be an increase in AMPAR number,
expression of LTP is the protein indelibly associated with its since potentiation of synaptic transmission induced by PKM
induction, the NMDA receptor. The time course of LTP in the increased the amplitude, but not the frequency, of AMPAR-
dentate gyrus of the freely moving rat is prolonged by many mediated mEPSCs and had no effect on their kinetics, or on
days if the NMDAR antagonist CPP is infused after the tetanus mean single channel conductance, , the latter estimated
(Villarreal et al., 2001). Infusion can be delayed for 2 days after using non-stationary uctuation analysis (Ling et al., 2006).
induction and still rescue the decay of protein synthesis- Interestingly, application of either chelerythrine or ZIP is able
dependent LTP. This startling result demonstrates that long to reverse pre-established LTP (Ling et al., 2002) when applied
after induction NMDAR-dependent signaling continues to 1 hour or later post-induction, though in another study chel-
control the fate of the potentiated synapse perhaps via the erythrine given 30 min after induction did not block E-LTP
block of NMDAR-dependent depotentiation. Here is another (Bortolotto and Collingridge, 2000), suggesting that PKM is
area ripe for further investigation. specically involved in the maintenance of L-LTP. Perhaps
most amazingly ZIP is able to reverse completely the potenti-
PKM maintains LTP ated response at perforant path synapses when injected 22 h
following the establishment of LTP in the dentate gyrus of
In the search for molecules that mediate the long-term per- freely-moving rats, paving the way for an exploration of the
sistence of LTP an interesting candidate has emerged in PKM role of LTP maintenance in memory (Pastalkova et al., 2006;
(see Section 10.4.3). This member of the PKC family was iden- see Section 10.10.4).
tied as the only isoform that was activated 30 min after the
induction of NMDAR-dependent LTP (Sacktor et al., 1993). 10.4.10 Structural Remodeling and Growth of
PKM is a constitutively active catalytic fragment of the atyp- Spines Can Be Stimulated by Induction of LTP
ical PKC, which can be produced by proteolytic cleavage by
proteases, such a calpain, or by synthesis from mRNA encod- The idea that changes in synaptic efficacy could be embodied
ing just this fragment. Subsequent work showed that LTP in changes in synaptic number or in remodeling of synaptic
involves the de novo synthesis of PKM (Hernandez et al., morphology predates the discovery of LTP by many decades
2003). Interestingly, PKM appears to be both necessary and and appears explicitly in the writings of Ramon y Cajal,
sufficient for L-LTP since PKM mimics and occludes LTP William James, and Donald Hebb, among others. More spines,
and since inhibition of PKM either by chelerythrine, which larger spines, and wider spine necks have all borne the burden
in low concentrations is selective for the isoform, or by of theories of memory, with the implicit assumption that
transfection with a dominant negative inhibitor, PKM- appropriate changes occur also in the presynaptic partner.
K281W, blocks LTP (Ling et al., 2002). The dominant negative What is the status of these enduring theories today, and specif-
inhibitor also blocked induction of E-LTP which led to the ically do they provide a mechanism for the expression of LTP?
suggestion that PKM enzymatically cleaved from PKC, per- The evidence is not straightforward, and disagreements
haps by the actions of calpain, may mediate the early protein among laboratories are as vigorous in this area as in many
synthesisindependent phase of LTP. However, a subsequent others (reviewed by Harris, 1999; Yuste and Bonhoeffer, 2001;
394 The Hippocampus Book

Nimchinsky et al., 2002; Kasai et al., 2003; Tashiro and Yuste, tural information, have strengthened the view that LTP is not
2003; Lemphrect and Le Doux, 2004). associated with an overall change in spine density, either in the
Spine numbers, even in adult animals, are not set in stone: dentate gyrus 6 hours after tetanus in vivo (Popov et al., 2004;
Many experimental and indeed physiological situations are see Trommald et al., 1996 for a counter claim) or in area CA1
associated with dramatic changes in spine number. For exam- in vitro (2 hours after tetanus (Sorra and Harris, 1998). As we
ple, spine densities in CA1 pyramidal neurons (but, interest- shall see below, this does not exclude the possibility that there
ingly, not in CA3 pyramidal cells or granule cells) vary by up are increases in the frequency of certain forms of spine at the
to 30% during the estrus cycle in female rats (Woolley et al., expense of others.
1990). Allowing adult rats a daily opportunity to explore a In an inuential series of studies, Geinisman documented
complex environment leads to an increase in spine density in an increase in the number of synapses with perforated and
CA1 pyramidal cells (Moser et al., 1994); and in organotypic segmented PSDs in axospinous synapses in the dentate gyrus
cultures spines are rapidly retracted following bath applica- after daily episodes of tetanization for 4 days (Geinisman et al.,
tion of botulinum toxin, which suppresses all transmitter 1991, 1993). Geinisman and colleagues proposed that these
release (McKinney et al., 1999). The physiologist engaged in changes marked stages in the transformation of spines with
cutting hippocampal slices no doubt prefers not to dwell too single macular, or continuous, PSDs to potentiated spines with
deeply on a report that CA1 neurons in slices from the adult fully segmented PSDs, allowing two independent sites for
rat contain 40% to 50% more synapses than similar cells in synaptic transmission: a spine splitting model in which sin-
perfused and xed tissue (Kirov et al., 1999). Comfort can gle boutons make multiple synapses on single dendrites.
perhaps be taken from the remarkable stability of spines in rat However, Harris et al. (2003) presented electron microscopic
visual cortex when imaged over a period of many months (EM) evidence that spine splitting is unlikely to occur in the
(Grutzendler et al., 2002), although in mouse barrel cortex a adult hippocampus, as the gap between two spines is invari-
more dynamic picture is seen, with only half of the imaged ably lled with axons and dendrites. They concluded that the
spines remaining stable over a period of a month (Trachten- extra spine is more likely to represent new growth. However,
berg et al., 2002). even in potentiated tissue, instances where a single bouton
makes contact with a pair of neighboring spines on the same
Ultrastructural Changes Associated with LTP dendrite comprise less than 3% of all synapses on hippocam-
pal pyramidal cells and so are unlikely to provide a structural
The rst investigation of tetanus-induced changes in spine basis for more than a fraction of the potentiated response.
morphology was published in 1975 (Van Harreveld and An ingenious method of labeling active synapses for EM
Fifkova, 1975). In this and subsequent studies, Fifkova and her was developed by Buchs and Muller (1996) based on the tran-
colleagues reported an increase in spine area and the width of sient increase in the concentration of calcium that occurs in
the spine neck in the outer part of the molecular layer of the spines following tetanic stimulation. The method depends on
dentate gyrus following tetanic stimulation of the perforant precipitating calcium in smooth endoplasmic reticulum
path. No changes were seen in spines in the unstimulated (SER) and rendering it electron-dense. Buchs and Muller
inner molecular layer. A problem with this work is that no noted an increase in the proportion of spines that contained
recordings were made of evoked responses, so it is not possi- perforated PSDs when synapses were labeled 30 minutes after
ble to be sure that LTP was in fact induced. Further studies, the induction of LTP, compared to the proportion in labeled
notably by Desmond and Levy in the dentate gyrus, and by spines in nave tissue or in spines examined 5 minutes after
Lynch and his colleagues in area CA1, followed soon thereafter LTP induction. In a follow-up study using serial section EM,
(Lee et al., 1980; Desmond and Levy, 1983; reviewed by Yuste Toni et al. (1999) observed a delayed but persistent LTP-
and Bonhoeffer, 2001). Although Desmond and Levy found associated increase in the proportion of boutons making
no consistent changes in spine density, they documented an synapses on more than one spine (multiple synapse boutons);
increase in the length of the PSD following LTP induction the proportion remained at its baseline value of around 6%
(Desmond and Levy, 1986a) and an increase in cup-shaped for the rst 30 minutes after tetanus, then rose to 15% at 45 to
spines (Desmond and Levy, 1986b). These observations sup- 120 minutes after tetanus (Fig. 1014A). Moreover, in 66% of
ported the conclusion reached by Fifkova et al. (1997) that the latter cases, the two spines arose from the same dendrite,
LTP is associated with changes in the morphology and ultra- compared to 11% in tissue xed 5 minutes after the tetanus.
structural organization of existing spines, rather than the gen- The method allows analysis only of the small proportion
eration of new spines. In area CA1, Lynchs group also saw no (15%) of spines that contain SER. Nevertheless, these striking
change in spine density but detected a substantial increase observations suggest that LTP is expressed as an increase in the
(35%) in the number of synapsespresumed inhibitoryon number of effective synapses, with input specicity conserved
dendritic shafts (Lee et al., 1980). Technical advances in elec- by the partitioning of tetanized boutons to allow for two
tron microscopy, including the introduction of the disector release sites. This is similar to the spine-splitting model pro-
technique for unbiased stereological sampling and the use of posed by Geinisman; whether multiple spine synapes make up
serial sectioning to obtain authentic three-dimensional struc- too small a proportion of the total to account fully for LTP, as
 Synaptic Plasticity in the Hippocampus 395

maintained by Harris et al. (2003), remains for future work to the induction of LTP (Hosokawa et al, 1995; Emptage et al.,
determine. 2003; Matsuzaki et al., 2004), as discussed below.
Serial section EM allows construction of three-dimensional Two other studies provide evidence that new spines can be
images of synapses in control and potentiated tissue, which is formed after the induction of LTP in area CA1 in vitro.
often the only way to resolve the ambiguities that can arise Maletic-Savatic et al. (1999) transfected CA1 pyramidal cells
with two-dimensional sampling. Sorra and Harris (1998) in organotypic culture with green uorescent protein and
examined synapses in tissue from area CA1 prepared 2 hours used two-photon confocal microscopy to image synaptic
after the induction of LTP in vitro. No change in the density or structures. Within a few minutes of applying localized tetanic
size of spines was found. However, in the absence of any stimulation, they saw a rapid growth of lopodia-like struc-
marker for potentiated synapses, it remained possible that an tures emanating from the dendrite; lopodia were not seen in
enlargement of potentiated spines was balanced by a decrease regions of the dendrite distant from the stimulating electrode
in the size of unstimulated spines. This drawback was and were not produced by low-frequency stimulation.
addressed in further experiments from the same laboratory Moreover, lopodial growth was blocked by AP5. A small pro-
that focussed on a subpopulation of spines containing polyri- portion of induced lopodia developed bulbous heads within
bosomes, the macromolecular complexes that contain the an hour of the tetanus, suggesting that the induced lopodia
machinery for protein synthesis. Given the extensive evidence are precursors to new dendritic spines.
for local protein synthesis during LTP, could polyribosomes Engert and Bonhoeffer (1999) used two-photon micros-
be used as a marker for potentiated synapses? Harris and her copy to image dendrites and spines of dye-lled CA1 pyrami-
colleagues found that the proportion of spines containing dal cells in organotypic cultures and a local perfusion
polyribosomes was appreciably enhanced, from 12% to 39%, technique to restrict synaptic activity to a small volume
after LTP induction in area CA1 (Ostroff et al., 2002). More- (approximately 30 m in diameter). Within 20 to 40 minutes
over, the surface area of the postsynaptic density in polyribo- of the tetanus new spines began to form in the activated vol-
some-associated spines was greater in potentiated tissue ume at an average frequency of 0.06 spines/m of dendrite
than in control tissue, whereas for spines without polyribo- (relative to a basal spine density of 0.40 spines/m) (Fig.
somes the size of the PSD was unchanged (Fig. 10.14B). This 1014D). New spines were not formed elsewhere, and their
result is consistent with a model in which potentiation is appearance was blocked if the tetanus was given in the pres-
driven by synaptic enlargement. The situation is different in ence of AP5. The dramatic images obtained by Engert and
the dentate gyrus, however, where a decrease in polyribosome Bonhoeffer are reminiscent of the tetanus-induced appear-
frequency occurs after LTP induction (Desmond and Levy, ance of multiple synapse boutons on single dendrites
1990). described by Toni et al. (1999). Note, however, that the forma-
Another three-dimensional EM study analyzed spines in tion of new spines, in contrast to the near-immediate induc-
the dentate gyrus 6 hours after the induction of LTP in vivo tion of LTP, was delayed by at least 20 minutes.
(Popov et al., 2004). Again, no overall change in spine density On the face of it, the studies described above make a strong
was seen, although in this case there was an increase in the case that new spines contribute to the expression of LTP in
proportion of synapses on thin spines and reductions in the organotypic cultures and in acute slices. However, they are
number of stubby spines and spines on dendritic shafts (Fig. inconsistent with other confocal studies in area CA1 in which
10.14C). There was also a signicant increase in the volume of new spines were not observed, either after the induction of
thin and mushroom spines and an increase in the area of their chemically induced LTP (Hosokawa et al., 1995) or after
PSDs. The picture that emerges from the EM studies consid- tetanus-induced LTP at visualized single spines (Emptage et
ered thus far is that LTP is associated not with an overall al., 2003; Matsuzaki et al., 2004). Transformation of the spine
increase in spine density but, rather, an increase in the size of from one displaying a continuous PSD to a perforated or seg-
certain subpopulations of spines. mented PSD would remain undetectable, as possibly would
spine splitting; but the genesis of a new spine is well within the
Longitudinal Studies: Imaging Spines resolution of the technique. These discrepancies are unlikely
Before and After Induction of LTP to reect differences in the time points chosen for study.
Emptage et al. (2003) collected their data 30 to 60 minutes
In addition to the problem of labeling potentiated synapses, an after tetanus, whereas Matsuzaki et al. (2004), using two-
inherent limitation of electron microscopy is that it is not pos- photon photolysis of caged glutamate, routinely extended
sible to image the same structure at multiple time points, so their period of observation to 100 minutes after tetanus, span-
time-dependent changes have to be compared in different ning the period of 20 to 60 minutes during which the new
preparations. The advent of laser scanning confocal micros- spines charted by Engert and Bonhoeffer (1999) and by Toni
copy has improved the resolution of light microscopy to a level et al. (1999) were being born. Once generated, the new spines
where synaptic stuctures (boutons and dendritic spines) can be survived for at least 2 hours in both sets of experiments and so
visualized in living tissue, so the same spine, bouton, or length should have been detected, if present in any number, by Sorra
of dendrite can be measured before and at various times after and Harris (1998), whose EM study in CA1 in vitro was car-
396 The Hippocampus Book

Multiple spine boutons (%) 20 *

*
15
*

10

0
0 30 60 90 120
Time (min)

B
0.09
0.08
**
Synapse size (sq )

0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
without PR with PR

C
Synapse number per 100m 3

350 100
Control
90
300 LTP
80
*
All synapses, %

Stimulation
250 70 without LTP
60 p < 0.01
200
50
*
150 40
100 30
20
50 10 * *
0 0
Control LTP Stimulation Thin Stubby Shaft Mushroom
without LTP

D
 Synaptic Plasticity in the Hippocampus 397

ried out 2 hours after tetanus (see below). The experiments of the repositories of previously stored information and are pro-
Matsuzaki et al. (2004) provided the rst evidence of an tected from further change. However, this prediction is not
increase in spine size in longitudinal studies LTP was induced supported by the data of Popov et al. (2004), which shows an
by repetitive focal photolyis of caged glutamate in low Mg2 to increase in the size of the largest subset of mushroom spines
allow activation of both AMPA and NMDA receptors. Only following the induction of LTP.
small spines showed a persistent increase in size after the
induction of LTP, and spine enlargement was accompanied by Alterations in the Cytoskeleton
an increase in sensitivity to uncaged glutamate. Contribute to LTP and LTD
Despite two decades of work on structural changes associ-
ated with LTP, it is still not possible to reach a denitive con- Dendritic spines are not static structures; rather, their shape is
clusion on such basic questions as whether synaptogenesis determined from moment to moment by the dynamics of
plays a role in LTP or whether morphological changes in exist- their actin cytoskeleton, as dramatically demonstrated in cul-
ing spines and boutons contribute to the enhanced response. tured neurons transfected with actin-GFP (Fischer et al., 1998;
In different preparations and at different times, there is evi- Matus, 2005). There is a growing body of evidence that activ-
dence for one or other or both types of change. However, even ity-dependent regulation of actin polymerization and the
where there is compelling evidence that such changes can consequent changes in synaptic morphology contribute to the
occur, there is often no demonstrated causal link with LTP. expression of synaptic plasticity. At low concentrations,
Thus, when new spines have been observed, it has not been latrunculin A, an inhibitor of actin polymerization, blocks late
demonstrated that they contribute to the enhanced response but not early LTP in area CA1 in vitro (Krucker et al., 2003); a
(Engert and Bonhoeffer, 1999; Maletic-Savatic et al., 1999; similar result has been reported in the dentate gyrus of the
Toni et al., 1999); moreover, the delayed time course of synap- freely moving rat (Fukazawa et al., 2003). In the latter study,
togenesis rules out any contribution to the rst 20-30 minutes induction of LTP resulted in an NMDAR-dependent increase
of LTP. The experiments of Matzusaki et al. (2004) provide the in F-actin (the polymerized form of actin) that was restricted
strongest evidence that morphological changes are causally to synapses of the tetanized pathway. The increase in F-actin
linked to LTP because the increase in spine size occurred was of rapid onset ( 20 minutes) and could be detected in
immediately, was blocked by the same drugs that block the tissue from animals killed as long as 5 weeks after the induc-
induction of LTP, and was associated with an increase in the tion of LTP. An EM analysis also revealed signicantly
response to uncaged glutamate. These experiments are also increased F-actin in spines in potentiated tissue; the F-actin
partly consistent with the LTP-associated increases in the vol- content was appreciably greater in large spines, consistent
ume of thin and mushroom spines observed by Popov et al. with a model in which actin drives enlargement of the spine.
(2004) in their serial section EM analysis of LTP in the dentate F-actin is polymerized from globular G-actin monomers.
gyrus. Matzusaki et al. (2004) speculated that large spines are Okamoto et al. (2004) used a uorescence resonance energy

Figure 1014. Morphological changes associated with LTP. A. LTP is structions of dendritic segments from the medial molecular layer of
accompanied by a delayed increase in the proportion of synapses in the dentate gyrus. Three groups of rats were analyzed: controls
which the presynaptic bouton makes contact with more than one (unstimulated, black bars), those with LTP (6 hours after induction
spine on a single dendrite. The graph plots the proportion of multi- of LTP by tetanic stimulation of the perforant path in anesthetized
ple spine boutons in the overall population of tetanized synapses as animals, gray bars), and animals that received the same total num-
a function of time after the induction of LTP (5 to 120 minutes). A ber of stimuli but in a pattern that did not induce LTP (white bars).
Ca2 precipitation technique was used to label tetanized spines (Source: Popov et al., 2004.) D. High-frequency stimulation leads to
(arrowheads in the electron micrograph). (Source: Toni et al., 1999.) the appearance of new spines in the potentiated region of CA1 cells
B. Spine growth following the induction of LTP is restricted to those in organotypic cultures. Slice cultures were maintained in a solution
spines that contain polyribosomes (PR), indicated by arrows in the containing Cd2 and low Ca2 to prevent transmitter release. A cell
electon micrograph. Analysis of three-dimensionsal reconstructions was impaled with a recording pipette containing uorescent dye,
showed that the mean area of the postsynaptic density in spines and a small region of its dendritic tree was made competent for
lacking polyribosomes was the same before (open bars) and 2 hours synaptic transmission by local perfusion of a solution containing
after (gray bars) induction of LTP, whereas spines that contained normal ACSF. In the example shown in the left-hand panel, the
polyribsomes were signicantly larger after induction. LTP was also extent of the perfused area is indicated by shading, with the pipette
associated with a redistribution of polyribosomes from dendritic tip at the lower end. LTP was induced by pairing (see plot of EPSP
shafts to spines. (Source: Harris et al., 2003; adapted from Ostroff et amplitude, lower right). The two-photon confocal images on the
al., 2002.) C. Spine density is not affected by the induction of LTP right were obtained 10 minutes before and 0.5, 1, and 12 hours after
(left), but there are changes in the distribution of spine types induction of LTP. The icons below indicate the times at which the
(right), with an increase in the proportion of thin spines and a two new spines (indicated by arrows) were visible. Note the delay,
decrease in the proportion of stubby and shaft spines. Data were observed in all cases, between the induction of LTP and the appear-
obtained from three-dimensional electron microscopic (EM) recon- ance of the new spines. (Source: Engert and Bonhoeffer, 1997.)
398 The Hippocampus Book

transfer (FRET) strategy to estimate the ratio between soluble 10.5.2 Basic Characteristics of NMDA Receptor-
G-actin and polymerized F-actin in the dendritic spines of independent LTP at Mossy Fiber Synapses
pyramidal cells in organotypic hippocampal cultures. Tetanic
stimulation increased the F-actin/G-actin ratio, whereas pro- Long-term potentiation in the mossy ber projection was rst
longed low-frequency trains caused a decrease in the ratio. documented by Schwartzkroin and Wester (1975). A striking
These changes were correlated with increases and decreases feature of this form of LTP is its immunity to NMDA receptor
in spine width, suggesting a model in which bidirectional, antagonists (Harris and Cotman, 1985) (Fig. 1015A), consis-
activity-dependent changes in F-actin control synaptic effi- tent with the relatively sparse distribution of NMDA receptors
cacy through changes in synaptic morphology. Because the in stratum lucidum, the mossy ber terminal zone (Monaghan
size of the presynaptic bouton is closely matched to the size of and Cotman, 1885). (Note that synaptically evoked NMDAR-
the spine (Harris and Stevens, 1989), there must be a corre- mediated currents can nevertheless be recorded from CA3 cells
sponding change in the presynaptic side. Evidence for an following mossy ber stimulation, a property that has been
activity-dependent increase in the size and number of presy- exploited to provide evidence for a presynaptic expression
naptic boutons was uncovered by Colicos et al. (2001) in a mechanism in mossy ber LTP (Weisskopf and Nicoll, 1995)
study on dissociated hippocampal neurons transfected with (see below). Although there is wide agreement that the expres-
GFP-actin. As we saw above, the EM evidence is more consis- sion of mossy ber LTP is presynaptically mediated, there has
tent with an increase in the size than the number of synapses been a persistent and unresolved debate about whether the
during LTP. Which of the possible models (no change in spine induction of LTP in this pathway is triggered by presynaptic or
size or number, or changes in one or both) more accurately postsynaptic processesa curious reversal of the controversies
describes the activity-driven changes that may occur in vivo is that invigorate discussion on induction and expression in
another goal for future research. other regions of the hippocampus. Mossy ber LTP in vitro
shows input specicity (Zalutsky and Nicoll, 1992) and asso-
ciativity (Schmitz et al., 2003). In addition, associativity
between mossy ber and A/C bers has been reported both in
10.5 LTP at Mossy Fiber Synapses vivo (Derrick and Martinez, 1994b) and in vitro (Kobayashi
and Poo, 2004); in the latter case, induction of LTP in A/C
In this section we consider the very different form that LTP synapses could be boosted by trains to a single mossy ber.
takes at the largest synapses in the hippocampus, indeed, Cooperativity (see Section 10.1.1) has been reported in vivo
among the largest in the mammalian brainthe synapses (Derrick and Martinez, 1994b) but is not seen in vitro
made by granule cell axons (the mossy bers) on CA3 pyram- (Zalutsky and Nicoll, 1992).
idal cells. These synapses also support LTD, a topic that is cov- The duration of mossy ber LTP has not been studied in
ered in see Section 10.7. Mossy bers make most of their the freely moving animal, but with multiple trains at 100 Hz
connections not on CA3 pyramidal cells but on inhibitory LTP lasting for several hours has been recorded in vivo (Henze
interneurons in the hilus and stratum radiatum. We discuss et al., 2000). Most of what follows is based on a survey of the
plasticity at these connections in Section 10.8. much larger slice literature, but the in vivo data should be
borne in mind when considering the functional consequences
10.5.1 Mossy Fiber Synapses Display of mossy ber plasticity.
Striking Short-term Plasticity
10.5.3 Induction Mechanisms of Mossy Fiber LTP
Mossy ber boutons make giant, interdigitating synapses at
complex spines (Cajals thorny excrescences) on the proxi- A Continuing Controversy: The Locus
mal dendrites of CA3 pyramidal cells. Giant synapses contain of Induction of Mossy Fiber LTP
up to 35 release sites (Henze et al., 2000) and show a greater
degree of short-term plasticity (paired-pulse facilitation and The Hebbian characteristics of LTP in the Schaffer-commis-
frequency potentiation during repetitive stimulation) than sural and perforant path projections arise from the voltage-
associational-commissural (A/C) synapses linking ipsilateral dependent properties of the NMDA receptor. It is a matter
and contralateral CA3 pyramidal cells (Salin et al., 1996). of continuing controversy as to whether induction is also
These properties are mediated, at least in part, by synaptically Hebbian in the mossy ber projection. The most straightfor-
released glutamate acting on presynaptic kainate receptors ward way of settling the issue would be to establish whether
that are positively linked to transmitter release ( Lauri et al., postsynaptic maneuvers, such as injection of a Ca2 chelator
2001; Schmitz et al., 2001). In addition, at mossy ber into the target cell or voltage-clamping the cell at hyperpolar-
synapses in vitro, PTP of several hundred percent can be ized potentials, blocks induction of LTP. Both experiments
elicited by long tetani (Langdon et al., 1995); in the anes- have been done, and different groups have come up with dif-
thetized rat, in contrast, PTP is typically not seen with similar ferent answers. According to Johnston and colleagues, mossy
tetanus protocols (Derrick and Martinez, 1994b). ber LTP is blocked by hyperpolarizing the CA3 pyramidal
 Synaptic Plasticity in the Hippocampus 399

Figure 1015. Properties of mossy ber LTP. A. Mossy ber LTP in reversible manner. A tetanus (100 Hz, 1 second, test intensity) was
vitro is readily induced during blockade of NMDA receptors. The delivered in the presence of D-APV at the times indicated by
graph plots fEPSP amplitude vs time and shows that a tetanus arrows. (Source: Bortolotto et al., 1999.) C. Inhibition of PKA pre-
(delivered at the time indicated by the arrowhead) induced LTP in vents the full expression of mossy ber LTP in vitro. The graph
the presence of D-APS (applied for 35 min. Mean baseline ampli- plots fEPSP versus time to show the effects of the PKA inhibitor
tude of fEPSPs was 0.4 mV. (Source: Harris and Cotman, 1986.) B. KT5720 and control experiments. A tetanus was delivered at t  0.
The induction of mossy ber LTP in vitro can be prevented by (Source: Weisskopf et al., 1994.) D. Mossy ber LTP in vitro is
kainate receptor antagonists. The graph plots fEPSP amplitude ver- absent in the Rab 3A knockout. The effects of a tetanus (t  0) on
sus time and shows that the GluR5-containing kainate receptor fEPSPs is compared for Rab 3A knockout and wild-type mice.
antagonist LY382884 blocks the induction of mossy ber LTP in a (Source: Castillo et al., 1997a.)

neuron (Jaffe and Johnston, 1990) or by injecting the calcium form of mossy ber LTP, the induction of which was blocked
chelator BAPTA (Williams and Johnston, 1989). A report by by a number of postsynaptic maneuvers, including hyperpo-
Zalutsky and Nicoll (1990), working with guinea pig rather larization, internal perfusion with BAPTA, and iontropic glu-
than rat hippocampus, disputed both these ndings; induc- tamate receptor blockade with kynurenate (which prevents
tion of LTP was insensitive to membrane potential and to synaptically evoked depolarization). In contrast, long trains of
intracellular injections of BAPTA. LTP in the A/C projection, high-frequency stimulation (L-HFS) (typically 100 stimuli at
however, was blocked by BAPTA. Because these synapses are 100 Hz, repeated three times at 10-second intervals), as
located more distally on CA3 cells than the complex spines on favored by Nicoll and his colleagues, produced a non-Hebbian
which mossy bers terminate, Zalutsky and Nicoll argued that form of LTP that was not affected by hyperpolarization,
the concentration of BAPTA was likely to have been sufficient kynurenate, or BAPTA.
to chelate free Ca2 in complex spines. They concluded that Johnstons laboratory reentered the arena in 1999, report-
the induction of mossy ber LTP was not controlled postsy- ing that B-HFS and L-HFS were associated with an increase in
naptically and was therefore non-Hebbian (Nicoll and dendritic Ca2 concentration; moreover, induction of both
Schmitz, 2005). forms of LTP could be blocked by postsynaptic injection of
A subsequent in vitro study offered a partial resolution of BAPTA (Yeckel et al., 1999). However, the source of the
the controversy (Urban and Barrionuevo, 1996). Brief trains increase in Ca2 was different in the two cases. LTP induced by
of high-frequency stimulation (B-HFS) (typically 8 stimuli at brief trains was associated with an inux of Ca2 through
100 Hz, repeated eight times at 5-second intervals), as nor- voltage-gated Ca2 channels because, as shown earlier (Urban
mally used by Johnston and his colleagues, led to a Hebbian and Barrionuevo, 1996), it could be blocked by kynurenate, by
400 The Hippocampus Book

hyperpolarization, and by Ca2 channel blockers (Kapur et al., ated by stimulation of mossy bers (Castillo et al., 1997b;
1998). LTP and the rise in dendritic Ca2 induced by long Vignes and Collingridge, 1997) (see Chapter 6, Section 6.3).
trains was insensitive to membrane potential and to kynure- The role of kainate receptors in the induction of mossy ber
nate (conrming the observation of Zalutsky and Nicoll, LTP has provided another opportunity for lively disagree-
1990) but could be blocked by mGluR1 antagonists and by ment. In the rst study to use a specic kainate receptor
agents such as ryanodine and thapsigargin, which compro- antagonist, Bortolotto et al. (1999) reported that induction of
mise release of Ca2 from internal stores. These results suggest mossy ber LTP was blocked by LY382884 (Fig. 1015B).
that a build-up of glutamate in the synaptic cleft leads to acti- Mossy ber LTP was also blocked by CNQX and kynurenate,
vation of type 1 mGluRs, stimulation of PLC, and a conse- reagents that block both AMPA and kainate receptors. Their
quent IP3-induced release of Ca2 from internal stores. Both results were challenged by Nicoll and colleagues, who pointed
forms of LTP could be blocked by an inhibitor of PKA, indi- to previous experiments of theirs and others in which CNQX
cating that elevation of Ca2 by the two routes converges on a and kynurenate had failed to block induction of mossy ber
PKA-dependent pathway to drive the expression of LTP. LTP (see exchange of letters between Nicoll et al. and
However, Mellor and Nicoll (2001) were unable to conrm Bortolotto et al. in Nature 2000;406:957). Subsequently,
these results and so the impasse remains unresolved. We are Collingridges group reported that the kainate receptor antag-
left with the unsatisfactory conclusion that in certain not pre- onist LY382884 acted presynaptically to reduce transmitter
cisely dened conditions the induction of mossy ber LTP release from mossy ber terminals and that this effect
requires involvement of the postsynaptic cell, whereas in oth- occluded mossy ber LTP (Lauri et al., 2001). In this view,
ers, equally imprecisely characterized, it does not. It is notable, tetanic stimulation led to activation of presynaptic kainate
however, that a number of subsequent studies have empha- receptors containing GluR5 subunits (the target of LY382884)
sized the importance of the postsynaptic cell in the induction and thence by a Ca2-dependent pathway to a persistent
(and in one case the expression) of mossy ber plasticity. increase in transmitter release. A possible reconciliation of the
Thus, a protocol in which LTP was induced by pairing mossy two positions has emerged from later studies by the two
ber stimulation with hyperpolarization of the CA3 neuron groups. Lauri et al. (2003) found that the requirement for
was developed by Sokolov et al. (2003). In contrast, Lei et al. kainate receptor activation could be negated if the extracellu-
(2003) reported that pairing low-frequency stimulation with lar Ca2 concentration was increased from 2 mM to 4 mM
sustained depolarization of the target CA3 cell led to a form of (the level normally employed in Nicolls laboratory). Lauri
mossy ber LTD in which expression as well as induction is and colleagues suggested that, at this concentration, sufficient
postsynaptically driven. Ca2 entered via L-type Ca2 channels to compensate for the
Finally, in experiments that introduced a new dimension to block of kainate receptors. However, this cannot be the entire
mossy ber plasticity, Contractor et al. (2002) established that explanation because Schmitz et al (2003) failed to block mossy
the ephrinEphB receptor tyrosine kinase signaling system ber LTP with LY382884 even in the presence of 2 mM Ca2.
(see Section 10.4.8) plays a critical role in the induction of Hence, it appears that multiple factors determine the role of
LTP. The native B-ephrin ligand is expressed in the presynap- kainate receptors in mossy ber LTP.
tic membrane where it can potentially engage with the postsy- Nicolls group ascribed a role for presynaptic kainate
naptic EphB receptor. Postsynaptic injection of antibodies receptors in mediating associative induction properties, in
that compromise proteinprotein interaction domains on the which neighboring mossy ber or associational/commissural
intracellular C terminus of the EphB receptor blocks the activity can lower the threshold for induction of mossy ber
induction of LTP, as does bath perfusion with a soluble form LTP (Schmitz et al., 2003). Independent support for a critical
of B-ephrin. These experiments offer strong evidence for a role of presynaptic kainate receptors in mossy ber plasticity
postsynaptic involvement in the induction of mossy ber LTP. has also come from the demonstration that LTP is absent in a
They also provide a potential mechanism for retrograde tran- knockout mouse lacking a functional GluR6 subunit (Con-
synaptic signaling by which the presynaptic terminal can be tractor et al., 2001). Somewhat unfortunately for the prospects
informed about critical postsynaptic induction events. The of reconciliation, the same study showed that mossy ber LTP
reduction of mossy ber LTP in NCAM knockout mice pro- was unimpaired in mice lacking the GluR5 subunit. However,
vides another indication of the potential importance of these results are not necessarily incompatible with the studies
transsynaptic signaling at these synapses (Benson et al., 2000). using LY382884 because, rst, GluR6 is probably necessary
for targeting kainate receptors and, second, there may be
Role of Kainate Receptors in the compensatory upregulation of other kainate receptor sub-
Induction of Mossy Fiber LTP units in the knockout (Kerchner et al., 2002a). Furthermore,
more recently developed antagonists of GluR5-containing
Kainate receptors are located both presynaptically and kainate receptors, such as UBP296, also block the induction of
postsynaptically at mossy ber synapses. The presynaptic mossy ber LTP (More et al., 2004). The actions of GluR5
kainate receptors regulate L-glutamate release (Vignes and antagonists can therefore safely be attributed to their ability to
Collingridge, 1997), and the postsynaptic kainate receptors inhibit this subunit. Given that LY382884 inhibits GluR5-
contribute a slow component to the synaptic current gener- containing heteromeric kainate receptors, the most likely ex-
 Synaptic Plasticity in the Hippocampus 401

planation for the discordant results is that the presynaptic ment of opioid peptides in mossy ber LTP is yet another area
kainate receptor may, but does not invariably, contain the of dispute, with opioid receptor antagonists reportedly block-
GluR5 subunit. ing mossy ber LTP in vivo (Derrick and Martinez, 1994a,
1996) but not in vitro (Salin et al., 1995). In vivo, mossy ber
Role of mGluRs in the Induction of Mossy Fiber LTP LTP can be induced by a combination of strong activation of
A/C bers, low frequency mossy ber activation, and injection
As we saw in Section 10.3.10, one of the many controversies of  receptor agonists. Any combination of two of these pro-
that enliven the study of hippocampal plasticity concerns the cedures yields mossy ber LTD rather than LTP (Derrick and
role of metabotropic glutamate receptors in LTP. In the case of Martinez, 1996).
mossy ber LTP, it has been claimed that mGluR antagonists
(Bashir et al., 1993), or the targeted deletion of mGluR1 Role of Zinc in Mossy Fiber LTP
(Conquet et al., 1994), severely impairs induction of LTP at
mossy ber synapses. However, both these ndings have been Vesicles in mossy ber glutamatergic terminals sequester large
disputed (Hsia et al., 1995; Selig et al., 1995b). In a subsequent amounts of Zn2 (Haug, 1967; Danscher et al., 1976; reviewed
study, it was reported that antagonism of both ionotropic and by Frederickson and Danscher, 1990). There is good evidence
group 1 metabotropic receptors is required before LTP that Zn2 can be released from mossy bers during synaptic
induced by long high-frequency trains (L-HFS) is blocked stimulation, where it may be co-released with L-glutamate
(Yeckel et al., 1999). Clearly, further work is needed to eluci- (Assaf and Chung, 1984; Howell et al., 1984; Quinta-Ferreira
date the involvement of mGluRs in mossy ber LTP. and Matias, 2004). Zn2 can modulate the function of a vari-
ety of ligand-gated and voltage-gated ions channels (Harrison
Mossy Fiber LTP Requires Activation and Gibbons, 1994), via which it may modulate synaptic plas-
of the cAMP/PKA Pathway ticity. For example, Zn2 inhibits NMDA receptors, providing
a route by which it could affect the induction of NMDAR-
There is convincing evidence that the cAMP and cAMP- dependent LTP (Xie and Smart, 1994; Ueno et al., 2002).
dependent protein kinase (PKA) pathway is involved in the There is also a body of evidence suggesting that Zn2 is
generation of mossy ber LTP. First, bath application of PKA involved (Weiss et al., 1989). For example, a variety of Zn2
inhibitors blocks the induction of mossy ber LTP (Huang chelators can inhibit the induction of mossy ber LTP (Budde
and Kandel, 1994; Weisskopf et al., 1994) (Fig. 1015C). et al., 1997; Lu et al., 2000b; Li et al., 2001), as can chronic de-
Second, bath application of membrane-permeable cAMP ana- ciency of dietary zinc (Lu et al., 2000b). Conversely, exogenous
logues or of forskolin, an activator of adenylyl cyclase, induces application of Zn2 can induce long-lasting potentiation of
a form of LTP that occludes with tetanus-induced LTP (Huang mossy ber transmission (Li et al., 2001). Mossy ber LTP
and Kandel, 1994; Weisskopf et al., 1994). Third, mutant mice seems to require the entry of Zn2 from the extracellular space
with a targeted deletion of type I adenylyl cyclase show into the postsynaptic neuron, entering via Ca2-permeable
impaired mossy ber LTP (Wu Z-L et al., 1995). These obser- AMPA or kainate receptors or via voltage-gated ion channels
vations do not directly address the question of pre- versus (Li et al., 2001; Jia et al., 2002). Its target in the postsynaptic
postsynaptic locus of induction or expression. Forskolin- cell is not known.
induced LTP is probably expressed presynaptically for the fol-
lowing reasons: First, there is a reduction in paired-pulse LTP at Connections Between Mossy
facilitation (Huang and Kandel, 1994; Weisskopf et al., 1994); Fibers and Interneurons
second, there is no change in the response evoked by ion-
tophoretic glutamate (Weisskopf et al., 1994); and third, there In addition to the large excitatory synapses made by mossy
is an increase in the coefficient of variation of synaptic bers on CA3 pyramidal cells, at least two kinds of connec-
responses, indicative of a presynaptic change (Weisskopf et al., tions to inhibitory interneurons in the stratum lucidum have
1994) (see Box 10.1). However, postsynaptic involvement of been described (see Chapter 4). LTP at these synapses is con-
the PKA signal cascade in mossy ber LTP has also been sug- sidered in Section 10.8.1.
gested by reports that injection of a cAMP analogue into the
postsynaptic cell can facilitate induction of LTP (Hopkins and 10.5.4 Expression of Mossy
Johnston, 1988) and, conversely, that postsynaptic injection of Fiber LTP Is Presynaptic
a PKA peptide inhibitor can impair induction (Yeckel et al.,
1999). There is general agreement that the locus of expression of
mossy ber LTP is presynaptic, in contrast to the long-running
Role of Opioid Peptides in Mossy Fiber Plasticity dispute concerning its locus of induction. Evidence for a presy-
naptic locus of expression was rst presented by Zalutsky and
Mossy ber terminals, like the terminals of the lateral per- Nicoll (1990), who found that paired-pulse facilitation (PPF)
forant path, co-release the opioid peptides enkephalin and of responses evoked by mossy ber stimulation was reduced
dynorphin in a frequency-dependent manner. The involve- after induction of LTP. However interpretation of this result as
402 The Hippocampus Book

evidence of a presynaptic change is itself based on the assump-

tion that PPF is wholly presynaptic, a claim for which there is
no incontrovertible evidence in the mossy ber pathway and
that cannot be taken for granted (see Box 101). However,
related evidence for a presynaptic locus has come from the
demonstration that presynaptic kainate receptors mediate a
component of synaptic facilitation that occludes with mossy
ber LTP (Lauri et al., 2001).
Further support for a presynaptic locus of expression has
emerged from a comparison of the decay constant of
NMDAR-mediated currents in the presence of the activity-
dependent blocker MK801 in potentiated and nonpotentiated
synapses. Potentiated mossy ber synapses show a more rapid
block than nonpotentiated mossy synapses, in contrast to A/C
synapses where the decay rate is unaffected by potentiation
(Weisskopf and Nicoll, 1995). This result implies an increase in
mean Pr following the induction of LTP (see Section 10.4.4),
which could be achieved by an increase in Pr at preexisting
release sites or by activation of silent synapses with a mean Pr
Figure 1016. Mechanisms involved in the expression of mossy
higher than that of the pretetanus population. Activation of
ber LTP. Synaptically released L-glutamate can feed back onto
silent synapses could occur by the insertion of receptor clusters
presynaptic receptors. Under certain circumstances, at least, this
or by the awakening of previously silent release sites. In cul-
may involve kainate receptors, group I mGlu receptors, and Ca2-
tured granule cells making autaptic synapses onto themselves, induced Ca2 release. A possible scheme to link these components
there is evidence that the latter is more likely to be the case is illustrated, which involves convergence of Ca2 and IP3 at the
(Tong et al., 1996). The existence of NMDAR-independent ryanodine receptor. Ca2 entry via voltage-gated Ca2 channels can
LTP at granule cell autapses establishes that the CA3 pyrami- provide an alternative source of Ca2. Increases in Ca2 stimulate
dal cell is not an obligatory component of the LTP machinery adenylyl cyclase (AC) to activate PKA, and this modies L-glutamate
at granule cell terminal boutons. As discussed is section 10.4.4, release. Other mechanisms, including a postsynaptic component,
Kawamura et al. (2003), using an optical method to detect glu- could also be involved in mossy ber LTP.
tamate-induced glial depolarization, also reached the conclu-
sion that LTP at mossy ber synapses, but not at A/C synapses,
was associated with an increase in glutamate release. transmission induced by depolarization with K (Lauri et al,
Quantal analysis has been applied to mossy ber LTP in 2001). This suggests that mossy ber LTP may alter an ion
slices (Xiang et al., 1994) and in a culture system (Lopez- channel to affect the resting membrane potential of mossy
Garcia et al., 1996); in both studies there was an increase in the fiber boutons. One candidate is the hyperpolarization-
coefficient of variation and a reduction in failures at initially activated cation channel Ih (Mellor et al., 2002, but see
nonsilent connections, indicating a presynaptic locus for the Chevaleyre and Castillo, 2002). Thus, a wide range of evidence
expression of LTP (but see Yamamoto et al., 1992). Confocal indicates a presynaptic locus of expression for mossy ber LTP
imaging of excitatory postsynaptic calcium transients (Fig. 1016).
(EPSCaTs) at single complex spines is also consistent with a
presynaptic locus of expression. LTP is accompanied by an 10.5.5 E-LTP and L-LTP at Mossy Fiber
increase in the number of active release sites and/or an Synapses Can Be Distinguished by
increase in Pr at already active sites (Reid et al., 2004). the Effects of Protein Synthesis Inhibitors
Some insight has been gained into the mechanisms under-
lying expression of mossy ber LTP. Thus, mossy ber LTP is Repeated long trains of high-frequency stimulation result in
absent in mice lacking the synaptic vesicle protein Rab3a L-LTP lasting undiminished for 6 hours or more at mossy
(Castillo et al, 1997a) (Fig. 1015D) and the GTP-dependent ber synapses in vitro (Urban and Barrionuevo, 1996). In the
Rab3a binding partner RIM1 (Castillo et al, 2002). RIM1 is presence of transcriptional and translational inhibitors,
located at the active zone and is a substrate for PKA. Therefore potentiation is initially similar, but responses decline to base-
LTP could involve a PKA-dependent interaction between line levels over the following 3 to 4 hours (Huang YY et al.,
Rab3a and RIM1 to enhance transmitter release. Surpris- 1994). As is the case elsewhere in the hippocampus, mossy
ingly, however, forskolin-induced potentiation, which occludes ber LTP can therefore be subdivided into L-LTP, requiring
with mossy ber LTP, is not affected in either knockout. gene expression and protein synthesis, and E-LTP, which is
Therefore, the relation between PKA-dependent signaling and independent of protein synthesis. The cAMP-dependent pro-
the role of these two proteins in mossy ber LTP is unclear. tein kinase cascade is implicated in both early and late mossy
Another possible mechanism was suggested by the nding ber LTP. PKA inhibitors block both phases, and bath applica-
that mossy ber LTP occludes the facilitation of synaptic tion of forskolin, an activator of adenylyl cyclase, induces both
 Synaptic Plasticity in the Hippocampus 403

early protein synthesis-independent and late protein synthe- transmitters and neuromodulators that can affect the thresh-
sis-dependent potentiation. Moreover, early tetanus-induced old for induction or the magnitude and/or duration of LTP.
LTP occludes forskolin-induced LTP and vice versa (Huang We also consider cyclical inuences on LTP, including sleep
YY et al., 1994; Weisskopf et al., 1994). Early LTP is associated and circadian rythms. In this section also we discuss the evi-
with a conspicuous reduction in paired-pulse facilitation dence that newborn granule cells in the dentate gyrus con-
(Zalutsky and Nicoll, 1990); interestingly, this effect attenuates tribute to synaptic plasticity.
over time, suggesting a switch to mechanisms of potentiation
that do not depend on an increase in Pr at individual release 10.6.1 Modulation by Other
sites but, instead, require either postsynaptic changes or an Neurotransmitters and Neuromodulators
increase in the number of synapses or active zones (Huang
and Kandel, 1994). Acetylcholine

10.5.6 Summary Both the early literature (reviewed in Segal and Auerbach,
1997) and more recent studies support the conclusion that the
Despite its independence of the NMDA receptor, it is evident septal cholinergic input can facilitate the induction of hip-
that LTP at mossy ber synapses requires an increase in intra- pocampal LTP. Evidence includes (1) inhibition of LTP by
cellular Ca2; but whether the critical compartment is the broad-spectrum muscarinic antagonists and (2) reduction in
presynaptic terminal or the postsynaptic spine remains the threshold for LTP by muscarinic agonists or by stimulation
unresolved and may depend on the pattern of the plasticity- of the septum. A revealing study in the freely moving rat
inducing stimulation. What is clear is that the increase in Ca2 demonstrates that these effects depend on the behavioral state
is not achieved by activation of NMDA receptors. Instead, in- of the animal (Leung et al., 2003). Basal LTP in the immobile
creases in Ca2 are achieved by a variety of alternative routes, or sleeping rat is not affected by muscarinic antagonists; but if
including the opening of voltage-dependent Ca2 channels tetanic stimulation is delivered to the Schaffer-commissural
following depolarization (caused by the presynaptic action projection when the animal is generating cholinergically
potential in the case of the presynaptic terminal and by bind- driven theta activity as it moves around its environment, an
ing of glutamate to postsynaptic AMPA receptors in the case enhanced level of LTP is induced in area CA1. Pretreatment
of the postsynaptic cell) and the binding of glutamate to type with the muscarinic antagonist scopolamine reduced the level
I mGluRs, leading to release of Ca2 from intracellular stores. of LTP to near-baseline levels, as did injection into the medial
The nding that kainate receptor antagonists, and philantho- septum of the neurotoxin immunoglobulin G (IgG)-saporin,
toxin which blocks Ca2 permeation, can block mossy ber directed against cholinergic neurons. These experiments
LTP under similar experimental conditions raises the possibil- demonstrate that LTP can be modulated by an endogenous,
ity that Ca2 entry through these receptors may also regulate behaviorally driven cholinergic input from the septum. The
plasticity at mossy ber synapses. effect is probably mediated by high-affinity postsynaptic M2
Finally, little is known about the role of mossy ber LTP in receptors (Segal and Auerbach, 1997). It is notable that affer-
the behaving animal. The long (1 second) trains at 100 Hz that ent cholinergic bers and M2 receptors are both present at
are required to produce robust mossy ber LTP are not likely higher density in the stratum oriens than in the stratum radia-
to be mimicked by granule cells in vivo, and it remains to be tum of area CA1. Activation of muscarinic receptors in the
established that naturally occurring patterns of granule cell hippocampus leads to G protein-mediated block of a number
activity lead to LTP at mossy ber synapses. of K channels, including IAHP, IM, and IL, as well as enhanced
NMDA receptor function (Segal and Auerbach, 1997).
Interestingly, direct potentiation of the NMDA receptor con-
ductance occurs at lower concentrations of ACh than that usu-
10.6 LTP Can Be Modulated by Other ally required to block K channels (Harvey et al., 1993). Any
Neurotransmitters, Neuromodulators, or all of these effects may contribute to the positive modula-
and Effectors and by Endogenous and tion of LTP induction by ACh acting on muscarinic receptors.
Circadian Rhythms
Monoamines
The degree and duration of LTP produced by a standard
induction protocol is inuenced by a multitude of factors in Norepinephrine (Noradrenaline). Noradrenergic innervation
addition to the frequency and duration of the tetanus. We of the hippocampus originates primarily from neurons in the
have already discussed the effects of prior activity (metaplas- locus coeruleus (see Chapter 8, Section 8.1.2). In the rst
ticity) and the critical role of inhibitory circuits in setting the study of the role of monamines in LTP, Bliss et al. (1983)
threshold for LTP. A good example of the importance of inhi- investigated the effects of the depletion of norepinephrine and
bition is seen in the dentate gyrus of the hippocampal slice, serotonin. Depletion of norepinephrine was achieved by
where LTP usually cannnot be elicited at all unless the slice is injecting the neurotoxin 6-hydroxydopamine (6-OHDA) into
perfused with a GABAA antagonist. We now examine the role the locus coeruleus, leading to a reduction in the magnitude of
of some of the more important of the many other neuro- LTP of the fEPSP in the dentate gyrus (Bliss et al., 1983).
404 The Hippocampus Book

Curiously, LTP of the population spike was not affected, a 1997). Bath-applied dopaminergic antagonists similarly block
result mirrored in a study showing that -adrenergic receptor L-LTP when given during, but not after, a strong tetanus
antagonists block LTP of the fEPSP but not of the population applied to induce L-LTP (Frey et al., 1990). Moreover, bath-
spike in the dentate gyrus (Munro et al., 2001, but see Stanton applied phosphodiesterase inhibitors, which prolong the
and Sarvey, 1985). There is also evidence that norepinephrine action of cAMP and so enhance the action of the principal
can promote late LTP: Stimulation of adrenergic neurons in downstream effector of dopamine, protein kinase A, can con-
the locus coeruleus by injection of glutamate produces a vert E-LTP into L-LTP (Navakkode et al., 2004). Stimulation of
delayed protein synthesis-dependent potentiation of per- the dopamine receptor-mediated signaling cascade with
forant path-evoked responses in the dentate gyrus (Walling D1/D5 dopamine receptor agonists or with the cAMP activa-
and Harley, 2004). A similar pattern of results is seen in area tor forskolin induces protein sythesis-dependent late potentia-
CA1. Depletion of norepinephrine or bath application of - tion that occludes with L-LTP in area CA1 in vitro (Frey et al.,
adrenergic receptor antagonists blocks LTP, and bath applica- 1993; Huang and Kandel, 1995; but see Mockett et al., 2004).
tion of norepinephrine lowers the threshold for the induction Thus, in area CA1, dopamine receptor activation appears to be
of LTP (Izumi and Zorumski, 1999). Enhancement or reduc- both necessary and sufficient for the induction of L-LTP. This
tion of LTP by drugs that respectively stimulate or block - is not, of course, to say that the release of dopamine or of any
adenergic receptors has also been reported in the mossy ber other neuromodulator is the primary cause of NMDAR-
projection (Hopkins and Johnston, 1988). Because -adrener- dependent LTP. D1/D5 agonists can lower the threshold for
gic receptors are positively linked to adenlyate cyclase, the input-specic tetanus-induced LTP in vitro (Otmakhova and
mechanism of action of norepinephrine is likely to be similar Lisman, 1996) and in vivo (Kusuki et al., 1997). When bath
to that of dopamine, leading (via an increase in cAMP) to acti- applied at high doses, it can under some circumstances pro-
vation of PKA. duce L-LTP nonspecically at many synapses; but this cannot
be what happens in the context of input-specic L-LTP. Here,
Serotonin. The modulatory effects of serotonin (5-hydrox- its action may be, rather, to convert or consolidate E-LTP into
ytryptamine, 5-HT) on LTP have been relatively less studied, L-LTP by facilitating gene expression via enhanced PKA activ-
and there is little consensus among studies. Selective depletion ity. Dopamine exerts its effects through activation of PKA,
of serotonin reduced the magnitude of LTP in the dentate which as we have seen is intimately involved in the dendrite-
gyrus in vivo (Bliss et al., 1983), but this result was not repli- to-nucleus signaling that triggers gene transcription, and in
cated in an in vitro study (Stanton and Sarvey, 1985). At com- local dendritic protein synthesis (see Section 10.4.9). In this
missural-associational synapses in area CA3 in vitro, bath model, then, dopamine, via cAMP and PKA, is directly
application of 5-HT was reported to inhibit LTP (Villani and involved in the synthesis of plasticity-related proteins
Johnston, 1993), whereas depletion of serotonin in area CA1 (Navakkode et al., 2004). An intriguing aspect of dopamine
was without effect (Stanton and Sarvey, 1985). The role of signaling in area CA1 is that it drives the enhancement of LTP
serotonin receptor subclasses in LTP has received some atten- in novel environments (Li et al., 2003; see Lisman and Grace,
tion. Injection of ondansetron, a selective antagonist of 5-HT3 2005 for a model of how the hippocampal-VTA loop controls
receptors, signicantly enhanced the magnitude of LTP in area entry of novel experiences into long-term memory).
CA1 in the freely moving rat, probably by suppressing the Interestingly, noradrenergic rather than dopaminergic signal-
excitability of interneurons mediating feedforward inhibition ing appears to be responsible for novelty-dependent modula-
(Staubli and Xu, 1995). Support for the notion that the pre- tion of LTP in the dentate gyrus (Straube et al., 2003).
dominant effect of 5-HT is to suppress LTP comes from stud-
ies showing that the 5-HT uptake inhibitor uvoxamine Brain-derived Neurotrophic Factor
reduces LTP via activation of 5-HT1A receptors (Kojima et al.,
2003), whereas injection of a 5-HT4 receptor agonist into the There is abundant evidence that the neurotrophic peptide,
lateral ventricle inhibited LTP, an effect that was blocked by a brain-derived neurotrophic factor (BDNF), is released as a
selective 5-HT4 antagonist (Kulla and Manahan-Vaughan, result of synaptic activity at hippocampal synapses.
2002). The cellular distribution of the relevant 5-HT receptors Extracellular BDNF can act on its receptor trkB, a member of
and the downstream signaling pathways by which 5-HT exerts the tropomysin-related family of receptor tyrosine kinases, to
its effects on LTP remain largely unexplored. enhance synaptic efficacy at excitatory synapses and diminish
efficacy at inhibitory synapses (for reviews, see Schinder and
Dopamine. Dopaminergic projections to the hippocampus Poo, 2000 and Ernfors and Bramham, 2003). BDNF can prob-
(primarily to the hilus and stratum radiatum lacunosum) ably be released from both pre- and postsynaptic sites at hip-
originate in the vental tegmental area (VTA) and substantia pocampal synapses and act on populations of pre- and/or
nigra (Scattton et al., 1980; Gasbarri et al., 1997). There is per- postsynaptic trkB receptors whose numbers are regulated by
suasive evidence that dopamine plays an important modula- activity. At Schaffer-commissural synapses in area CA1, BDNF
tory role in the induction of protein synthesis-dependent can act as a potent excitatory transmitter (Katz et al., 1999).
L-LTP in area CA1 (reviewed by Lisman and Grace, 2005). BDNF appears to play a critical role in the induction of
Knockout of the D1 dopamine receptor produces mice that LTP. LTP is impaired in BDNF-null mice (Korte et al., 1995),
lack L-LTP while displaying normal E-LTP (Matthies et al., and the decit can be rescued in vitro by BDNF administered
 Synaptic Plasticity in the Hippocampus 405

directly (Patterson et al., 1996) or by viral transfection (Korte nous levels of IL-1 are elevated in response to stress, infec-
et al., 1996). BDNF can also produce LTP. Bath application of tion, the amyloid peptide A , and aging, and that this may
BDNF to hippocampal slices induces slow-onset, protein syn- partly account for the neurodegeneration and cognitive
thesis-dependent LTP in area CA1 (Kang and Schuman, decits sometimes observed in such circumstances (Murray
1995), whereas in the dentate gyrus repeated pairing of weak and Lynch, 1998; Minogue et al., 2003). The mechanism by
synaptic stimulation with focal dendritic application of BDNF which IL-1 exerts its effects is not fully established but may
leads to rapid-onset LTP that occludes with tetanus-induced involve activation of c-Jun N-terminal kinase (JNK) and
LTP (Kovalchuk et al., 2002). This form of LTP satises p38MAPK as well as caspase-1 and NKB (Vereker et al.,
the criterion for postsynaptic induction, as it is blocked by 2000a,b; Curran et al., 2003; Kelly et al., 2003). Il-1 also
postsynaptic injection of calcium chelators. A late-onset, seems to stimulate superoxide dismutase to increase reactive
transcription-dependent potentiation of synaptic transmis- oxygen species (Vereker et al., 2001). In addition to a role
sion has also been described in the dentate gyrus of the anes- in pathology, however, studies with an IL-1 receptor antago-
thetized rat following transient perfusion with BDNF nist and an IL-1 receptor knockout have suggested that IL-1
(Messaoudi et al., 2002) (Fig. 102D). The synaptic response may be required for the induction of LTP under physiological
climbs gradually after perfusion and persists for many hours conditions (Avital et al., 2003; Ross et al., 2003; Maher et al.,
thereafter. BDNF-induced late LTP occludes with late tetanus- 2005).
induced LTP; but in contrast to the rapid-onset LTP observed Other interleukins, such as IL-2, IL-6, IL-8, and IL-18, also
in the dentate gyrus in vitro, it is not NMDAR-dependent. inhibit the induction of LTP (Tancredi et al., 1990; Bellinger et
However, because it leads to CREB phosphorylation via acti- al., 1995; Li et al., 1997; Curran and OConnor, 2001; Xiong et
vation of ERK1/2 and upregulation of immediate early genes al., 2003). However, certain interleukins have complex inter-
in granule cells, bath application of BDNF activates, at least in actions in their regulation of LTP. For example, IL-10 blocks
part, the normal late LTP-inducing machinery in the postsy- the inhibitory effect of IL-1 on LTP (Kelly et al., 2001). Other
naptic cell downstream from the NMDA receptor (Ying et al., cytokines that have been reported to inhibit LTP include the
2002). Changes in CREB phosphorylation in the entorhinal interferons (DArcangelo et al., 1991; Mendoza-Fernandez et
cortex, where the perforant path projection to the dentate al., 2000) and tumor necrosis factor alpha (TNF) (Tancredi
gyrus originates, suggests there may also be presynaptic et al., 1992). TNF inhibits early LTP, but not late LTP, via acti-
involvement (Gooney et al., 2004). vation of a p38MAPK cascade (Butler et al., 2004) and may
Little information is available on whether the expression of mediate its actions via an effect on the trafficking of AMPA
BDNF-induced LTP is pre- and/or postsynaptically mediated receptors (Beattie et al., 2002)
in the hippocampus. In hippocampal cultures, neurotrophin-
induced potentiation lasting 10 to 15 minutes carries the hall- Corticosteroids
marks of a presynaptic mechanism: reduced paired-pulse
facilitation, increased mini-EPSP frequency without an The effects of stress on hippocampal plasticity were rst
increase in amplitude, a reduction in the coefficient of varia- explored by Foy et al. (1987). Rats were subjected to mild
tion, and comparable enhancement of both AMPAR- and inescapable stress by conning them in a restraining tube for
NMDA-mediated responses (Schinder and Poo, 2000). 30 minutes. Hippocampal slices cut from the brains of animals
Whether endogenous BDNF acts as a retrograde messenger killed immediately afterward displayed signicant impairment
released from the postsynaptic cell to act on presynaptic TrkB in LTP relative to control animals. It later became apparent
receptors remains to be established. A presynaptic component that an important factor is whether the stress is avoidable;
of LTP induced at Schaffer-commissural synapses on CA1 slices from animals that were given shocks but had control
pyramidal cells by theta-burst stimulation requires BDNF. over the termination of the shock showed greater LTP than an
However, analysis of two mutant linesone expressing a unshocked group, whereas those that had no control over ter-
mutant form of BDNF in both CA3 and CA1 cells and the mination showed less LTP (Shors et al., 1989). Hippocampus-
other only in CA1 cellsdisclosed an impairment of theta- dependent learning is also modulated by stress (see Chapter
burst LTP only in the former mutant, suggesting that the 15). Later it is was found that mild stress can facilitate the
source of BDNF is in presynaptic neurons (Zakharenko et al., induction of LTD (Xu et al., 1997; Manahan-Vaughan, 2000).
2003). In other systems, however, such as the neuromuscular In rodents, stress is associated with an increased plasma
junction, there is clear evidence that BDNF acts as a retro- concentration of the adrenal hormone corticosterone, The
grade messenger (Schinder and Poo, 2000). hippocampus is richly endowed with the two types of corti-
costeroid receptor: the high-affinity mineralocorticoid (or
Cytokines type 1) receptor and the low-affinity glucocorticoid (type II)
receptor. Activation of mineralocorticoid receptors promote
A variety of cytokines have been found to interact with LTP. LTP, whereas high corticosterone levels activate the glucocor-
For example, interleukin-1 (IL-1 ) reduces mossy ber LTP ticoid receptor, which is the primary mediator of stress-
(Katsuki et al., 1990) and NMDA receptor-dependent LTP in induced impairment of LTP and spatial learning. Thus, there
area CA1 and the dentate gyrus (Bellinger et al., 1993; is an inverted-U relationship between the plasma concentra-
Cunningham et al., 1996). It has been suggested that endoge- tion of corticosterone and its effects on LTP and learning. At
406 The Hippocampus Book

low concentrations of corticosterone (as occurs following (Warren et al., 1995; Good et al., 1999). Subsequent experi-
adrenalectomy) the occupancy of mineralocorticoid receptors ments have largely been conducted on ovariectomized animals
is low, whereas at high levels (as during severe stress) the glu- where the total absence of estrogen has not always led to the
cocorticoid receptors are fully occupied; both conditions are reduction in LTP that the experiments in normally cycling ani-
associated with a reduction of LTP, promotion of LTD, and mals might have predicted. For example, Day and Good
impaired learning. Consequently, LTP and learning are maxi- (2005) saw no change in LTP following ovariectomy but did
mal at intermediate levels of circulating corticosterone observe impaired LTD that was reversed by chronic estrogen
(Diamond et al., 1992; Kim and Diamond, 2002). It is an over- replacement. Whether these estrogen-induced effects on
simplication, however, to assume that corticosterone is the synaptic plasticity are an indirect result of changes in spine
only mediator of the effects of stress on LTP and behavior; in numbers or reect an action of the hormone on signaling
fact, corticosterone is neither necessary nor sufficient for these pathways or protein synthesis remains to be seen. Puberty in
effects to occur. For example, stress can still suppress LTP in male rats is associated with a bias toward LTD (Hebbard et al.,
rats deprived of corticosterone by adrenalectomy, and lesions 2003). The effect of pregnancy on LTP does not appear to
of the amygdala block the stress-induced impairment of LTP have been specically examined, although a greater exibility
even though corticosterone levels remained high in lesioned in behavior in the watermaze has been noted (Bodensteiner
animals (Kim and Diamond, 2002; Kim et al., 2005c). Priming et al., 2006).
stimuli delivered to the basolateral amygdala has a biphasic
effect on LTP in the dentate gyrus in vivo, enhancing LTP at Sleep, Circadian Rhythms, and LTP
short priming-tetanus intervals and suppressing LTP if given
an hour before the tetanus to the perforant path. The mecha- Sleep is necessary for normal cognitive processing. In particu-
nisms by which activity in the amgygdala is able to block LTP lar, it seems likely that during sleep the brain sorts newly
in the hippocampus have yet to be worked out, but these nd- acquired information to determine what needs to be commit-
ings reect the intimate manner in which hippocampal plas- ted to long-term storage and then assimilates it with previ-
ticity is modulated by the amygdalas view of a stress-lled ously learned experiences. The brain is therefore performing
world (Kim et al., 2005c). complex cognitive processing during sleep that is probably
There is evidence that stress engages some of the same sig- crucial for the consolidation of memories. Several groups have
naling pathways that are involved in synaptic plasticity. For wondered if LTP is modulated during sleep, and in particular
example, phosphorylation of the MAP kinase signaling mole- during dreaming in REM sleep. Our memory for dreams is
cules ERK1/2 is markedly increased by stress, and the surprisingly tenuous, given the vividness of the dreaming
inhibitory effects of stress on learning and LTP can be blocked experience. Can this be linked to suppression of LTP during
by inhibitors of ERK1/2 phosphorylation (Yang et al., 2004). REM sleep? It seems not. In the dentate gyrus, the magnitude
This raises the possibility that the reduction in tetanus- of LTP induced during REM sleep is similar to that obtained
induced LTP during stress reects potentiation caused by during the alert state (Bramham and Srebro, 1989), although
stress itself, pushing the affected synapses toward saturation. there is a signicant reduction in magnitude during slow-
However, neither applied corticosterone nor stress has been wave sleep (Leonard et al., 1987).
reported to enhance baseline levels of hippocampal evoked Other studies have found that sleep deprivation (1272
responses. Because NMDA receptor antagonists given at the hours) can impair the ability to induce LTP (Campbell et al.
time of the stressful experience can block the subsequent 2002; Davis et al. 2003; McDermott et al. 2003; Kim et al.,
reduction in LTP, it is likely that the impairment of LTP is due 2005b), possibly because of the stress that is associated with
to a metaplastic effect mediated by NMDA receptor activation sleep deprivation. However, even when precautions are taken
rather than to saturation of LTP (Abraham, 2004). to minimize stress, short periods (4 hours) of REM sleep dep-
Hippocampal regulation of the hypothalamic-pituitary- rivation result in selective impairment in the maintenance of
adrenal axis is discussed in Chapter 15. L-LTP in the perforant path (Romcy-Pereira and Pavlides,
2004).
10.6.2 Cyclical Inuences A clue to what sleep might be doing has been obtained by
Modulate Induction of LTP mapping plasticity using the expression of the immediate early
gere zif268 (Ribeiro et al. 2002). The induction of LTP in the
Sex Hormones and LTP perforant path, which is associated with a rapid increase in the
expression of zif268, also leads to upregulation of zif268
During proestrus in female rats, when levels of estrogens reach expression during subsequent episodes of REM sleep in
a peak, the number of spines on CA1 pyramdal cells is as extrahippocampal regions, predominantly in the amygdala,
much as 30% higher than during estrus (Woolley et al., 1990) entorhinal, and auditory cortices during the rst REM sleep
(see Chapter 9). Are there corresponding change in the mag- episode and the somatosensory and motor cerebral cortices as
nitude of LTP and/or LTD during the estrus cycle? There have REM sleep recurs. These observations support the hypothesis
been two reports that LTP in area CA1 of female rats with that sleep enables the spread of memory traces from hip-
chronically implanted electrodes is enhanced during proestrus pocampus to the cerebral cortex.
 Synaptic Plasticity in the Hippocampus 407

Exercise promotes LTP at CA1 synapses: LTP is greater induction of LTP (Schmidt-Hieber et al., 2004). However,
when induced in walking rats than in awake but immobile there is clearly some way to go before we can conclude that
animals or during slow-wave or REM sleep (Leung et al., neurogenesis contributes to the cellular basis of learning.
2003). The enhanced synaptic plasticity in walking animals is Among the basic questions that await answers are (1) what is
promoted by the septal cholinergic input to the hippocampus. the lifetime of newborn cells that mature into fully connected
The pyramidal and granule cell subelds respond differ- members of the adult granule cell network, (2) how does the
ently to nightday variations. In the dentate gyrus, LTP is sig- hippocampus maintain a stable representation of stored
nicantly greater when induced during the dark, active stage events in the face of a continuing turnover of granule cells,
of the rats diurnal cycle (Harris and Teyler, 1983; Dana and and (3) why is adult neurogenesis conned to the dentate
Martinez, 1984), consistent with the reduction in LTP gyrus and olfactory bulb. As Schinder and Gage (2004) have
observed during slow-wave sleep and perhaps reecting the pointed out, the tool needed to establish a causal role for neu-
greater baseline fEPSPs obtained in the dark (Barnes et al., rogenesis in learning is a genetic model in which neurogene-
1977). In CA1, in contrast, LTP is greater during the light, sis can be turned on and off at will.
behaviorally inactive stage. The reasons for this difference and
its functional signicance are unknown.

Hibernation 10.7 Long-term Depression and
Depotentiation: Properties and Mechanisms
The temperature of hibernating animals can fall to a value at
which synaptic transmission fails. As temperature rises on 10.7.1 Overview
arousal from hibernation, the magnitude of LTP is greater
than in control animals maintained at the same temperature If activity-dependent plasticity operated only to enhance
(Spangenberger et al., 1995). This no doubt provides a wel- synaptic weights, saturation of synaptic efficacy would even-
come boost to its cognitive faculties at a generally trying time tually ensue. A neural net composed of Hebbian synapses all
for the hamster. of whose synaptic weights were maximal would be incapable
of learning. Although many, if not all, potentiated synapses
10.6.3 Neurogenesis and LTP may eventually decay to baseline through the passage of time,
an activity-driven mechanism to allow erasure, or depotentia-
Neurogenesis proceeds at a truly astonishing rate in the den- tion, of LTP would enhance the computational exibility of
tate gyrus of adult animals (see Chapter 9). In the rat, approx- the network. An additional mechanism permitting activity-
imately 9000 new granule cells are born each day. (Box 91), dependent LTD from baseline values of synaptic efficacy, and
approaching 1% of the total number of granule cells in the independent of LTP, so-called de novo LTD, would further
each dentate gyrus (Section 3.4.1). Performance in some (but enhance the exibility of the system and increase its storage
not all) hippocampus-dependent tasks (e.g., trace eyeblink capacity (Dayan and Willshaw, 1991). Another important
conditioning) is blocked by administration of methyl- consideration is the balance of excitation and inhibition that
azoxymethanol, a toxin that kills dividing cells (Shors et al., determines the overall stability of the network. Unbridled
2001). One explanation for this observation is that newborn potentiation would carry the danger of disrupting this balance
granule cells are necessary for encoding the learned task. What and driving the system to instability and seizure. The homeo-
is the physiological evidence that would support such a con- static checks and balances that the hippocampus deploys at
clusion? First, although most newborn cells die, those that both the cell and network levels to guard against runaway
survive are incorporated into the neural network of the den- excitation are described in Section 10.3.14; in this light, both
tate gyrus, sending axons toward the stratum pyramidale of the homosynaptic properties of LTD and depotentiation, and
area CA3 (Hastings and Gould, 1999; Markakis and Gage, the heterosynaptic depression that in some cases accompanies
1999) and extending dendrites into the molecular layer where the induction of LTP, may be considered additional weapons
they make synaptic contact with axons of the perforant path in the networks homeostatic armory.
(Song et al., 2002). Second, neurogenesis and LTP are A word on nomenclature: In what follows, by LTD we
enhanced by running (van Praag et al., 1999). Third, imma- mean de novo LTD. The depression of a potentiated response
ture granule cells have a lower threshold for LTP than do adult we refer to as depotentiation, and the repotentiation of a
cells (Schmidt-Hieber et al., 2004). The reason for this seems depressed response as de-depression. It is important to keep
to be the much higher input resistance of immature cells these terms distinct because, as we shall see, they embody dif-
(leading to greater depolarization for a given synaptic current) ferent mechanisms. Depotentiation, for example, is true rever-
and the expression by immature cells of T-type Ca2 channels sal or erasure of LTP, not the superposition of depression on a
that support Ca2 spikes following relatively small depolariza- still potentiated background.
tions. These characteristics lead, on the one hand, to greater Heterosynaptic forms of LTD and depotentiation provided
relief of the Mg2 block of the NMDA receptor and, on the the rst indication that, in addition to LTP, hippocampal
other, to enhanced Ca2 entry, both of which promote the synapses could support long-term, activity-dependent reduc-
408 The Hippocampus Book

tions in synaptic efficacy (Lynch et al., 1977; Levy and The molecular mechanisms responsible for the expression
Steward, 1979) (Fig. 1017A,B). Later work also established of NMDAR-dependent LTD have been extensively studied;
that during the rst few minutes after its induction (but not at induction is postsynaptic, and postsynaptic components of
later times) LTP could be destabilized (depotentiated) by expression include dephosphorylation of GluR1 subunits and
low-frequency stimulation (Barrionuevo et al., 1980; Staubli internalization of GluR2-containing AMPA receptors.
and Lynch, 1990; Fujii et al., 1991) (Fig. 1017B). Further Expression of mGluR-LTD appears to involve both presynap-
progress was slow, largely because of the lack of a reliable pro- tic and postsynaptic components.
tocol for inducing LTD. The discovery by Dudek and Bear In this section, we deal rst with homosynaptic forms of
(1992) that prolonged trains of low-frequency stimulation synaptic depression and then go on to examine heterosynap-
(LFS), typically 900 stimuli at 1 Hz, induced both homosy- tic LTD and depotentiation. Different hippocampal subelds
naptic LTD and homosynaptic depotentiation in hippocam- display substantial differences in the types of homosynaptic
pal slices from juvenile rats provided a major impetus to and heterosynaptic LTD and depotentiation they can sustain.
research (Fig. 1017D, Fig. 1019A). Progress in studying LTD Although the number of in vivo studies on LTD is growing,
in the adult animal, however, whether in vivo or in vitro, is still most of the experiments have been conducted on hippocam-
hampered by the lack of effective stimulus protocol(s) that pal slices from young animals in area CA1, and the text neces-
produce LTD most of the time in most laboratories. This has sarily reects this bias. Separate sections are devoted to LTD
led to speculation that de novo LTD is a form of synaptic plas- and depotentiation in the mossy ber projection (see Section
ticity whose major role lies in the sculpting and pruning of 10.7.8) and LTD at connections to and from interneurons (see
synaptic connections during development of the nervous sys- Section 10.8). For a wide-ranging review of LTD and depo-
tem. Nevertheless, to prevent network saturation and to pre- tentiation see Kemp and Bashir (2001); mGluR-dependent
serve an overall mean level of excitability in the network, the LTD has been reviewed by Bashir (2003).
conclusion seems unavoidable that mechanisms to decrease
synaptic weights must also be available to the fully developed 10.7.2 NMDAR-dependent LTD:
hippocampus. Properties and Characteristics
Initial uncertainties about whether the induction of LTD in
area CA1 was NMDAR-dependent (Dudek and Bear, 1992) or In Vitro Studies
mGluR-dependent (Bashir et al., 1993) were resolved when
Oliet et al. (1997) showed that it was possible to obtain either The rst report that LTD could be achieved with low-fre-
result by manipulating the induction protocol, thus conrm- quency stimulation (LFS) (e.g., 100 stimuli at 1 Hz) was pub-
ing the existence of two independent forms of homosynaptic lished by Dunwiddie and Lynch (1978). However, it was not
LTD. In both area CA1 and the dentate gyrus, mGluR-depen- until several years later when Dudek and Bear (1992) (Fig.
dent LTD is reliably induced at all ages by exposure to the 1017D) found that prolonged trains of LFS reliably produced
group 1 agonist DHPG (Palmer et al., 1997) ( Fig. 10-17E) or LTD in young animals that experimental interest in LTD
by low-frequency trains of pairs of pulses at an appropriate began to accelerate. LTD is routinely obtained with the Bear-
inter-pulse interval (Kemp et al., 2000) (see Section 10.7.5). A Dudek protocol (typically 900 stimuli at 1Hz) only when
third form of LTD has been reported at mossy ber synapses slices are prepared from juvenile rats (Dudek and Bear, 1993;
early in postnatal development that requires neither NMDA ODell and Kandel, 1994; Wang and Alger, 1995). LTD induced
nor mGlu receptors (Domenici et al., 1998). with this protocol is long-lasting, input-specic, NMDAR-
The two major forms of LTD have different developmental dependent, apparently saturable, and reversible by tetanic
proles: In vitro, both NMDAR- and mGlu receptor-depend- stimulation (dedepression) (Dudek and Bear, 1992).
ent LTD occur in area CA1 in the neonate, giving way to pre- Dudek and Bear went on to develop the notion of graded
dominantly mGluR-LTD in the adult (Kemp et al., 2000). bidirectional synaptic modiability (Dudek and Bear, 1993)
Pharmacological blockade of the uptake of L-glutamate facili- (Fig. 1019A). In this model, with appropriate high- or low-
tates induction of NMDAR-dependent LTD in perirhinal cor- frequency stimulation, a synapse could be potentiated, depo-
tex (Massey et al., 2004) and in area CA1 (Yang et al., 2005). tentiated, depressed, or de-depressed, thus allowing its efficacy
This suggests that mechanisms exist for NMDAR-dependent to be set anywhere within a range of approximately  50% of
LTD but that normal uptake processes limit their deployment. its base level. The difficulty often encountered of depotentiat-
Because uptake mechanisms are strongly temperature- ing a pathway once LTP has been stabilized in the adult animal
dependent, they are likely to be more effective in vivo than in implies that there are temporal constraints on bidirectionality.
vitro, where experiments are rarely performed at physiological A proposed molecular basis for bidirectionality is discussed
temperatures. This might be one reason why LTD is often not below. [An alternative hypothesis to graded bidirectional
observed in vivo. However, NMDAR-dependent LTD can be changes is that a modiable synapse can switch only between
induced in area CA1 of the adult rat in vivo by low-frequency three digital levels: resting, potentiated, and depressed.
trains of paired-pulse stimuli with a brief 25 msec interpulse Electrophysiological evidence in favor of a binary all-or-none
interval (Thiels et al., 1994; Doyre et al., 1996; Thiels et al., model of LTP at single synapses was presented by Petersen et
2002). al. (1998), but conrmatory studies have not been published.
 A B
160 LFS

fEPSP slope (% change)

30
120
20
fEPSP (% change)

S2
80 10
R

40 0
S1

0 -10
HFS HFS
-50 S1+S 2
Min 0 5 15 60 0 60

0 10 20 50 60 70 80 Day 1 2 3 4 5

Time (min)
C D
20
MPP LPP

fEPSP slope (% change)

30 min
fEPSP slope (% change)

60 0

40
-20
20
0 -40
-20
-40 -60

-60 1 Hz
-80
1 2 3 6 8 -30 -10 0 15 30 45 60
Days
Time (min)
E F

50 DHPG 150
ZIP
fEPSP slope (% change)

fEPSP slope (% change)

100
0

50

-50
0

-100 -50
0 10 20 30 40 50 60 70 -60 0 60 120 180 240 300
Time (min) Time (min)
Figure 1017. Types of hippocampal long-term homosynaptic and (Source: Staubli and Lynch, 1990.) C. Heterosynaptic depotentiation
heterosynaptic depression. A. Heterosynaptic depotentiation of of the medial perforant path in the freely moving rat after tetanic
crossed perforant path responses (S1R) in the anesthetized rat. stimulation of the lateral perforant path. The response remains
The graph plots the amplitude of fEPSPs evoked in the dentate depotentiated over the following days. (Source: Doyre et al., 1997.)
gyrus of an anesthetized rat by test stimuli (S1) to the contralateral D. Prolonged low-frequency stimulation (I Hz for 15 min) induces
perforant path. LTP of the contralateral input is induced only by LTD in area CA1 in vitro. This is the protocol now most commonly
associative tetanization of both ipsilateral and contralateral inputs used to induce LTD. It is most effective in young animals. (Source:
(double arrow, S1S2). The potentiated contralateral response was Dudek and Bear, 1992.) E. LTD induced in area CA1 in vitro by
then heterosynaptically depotentiated by a tetanus to the ipsilateral bath application of the type 1 mGluR agonist DHPG. The perfusion
pathway (S2). A subsequent combined tetanus restored the potenti- uid contained picrotoxin and AP5; hence, under these conditions
ated response (not shown). (Source: Levy and Steward, 1979.) B. the effect is NMDA receptor-independent. Synaptic responses were
Persistent homosynaptic depotentiation in a freely moving rat. recorded using a grease gap technique (Source: Palmer et al., 1997).
Low-frequency trains (1 Hz for 100 to 250 seconds) given 10 to F. ZIP, a membrane permeable substrate inhibitor of PKM ,
15 minutes after the induction of LTP in area CA1 caused a rapid reverses LTP in area CAI in vitro when applied 60 min after induc-
return to baseline. A second tetanus delivered 24 hours later, when tion (lled circles) without affecting the nontetanized control
the response was still at baseline, led to the reinstatement of LTP. pathway (open circles) (Source: Ling et al., 2002).
409
410 The Hippocampus Book

Optical studies of transmission at single visualized synapses followed for more than 24 hours (Kulla et al., 1999). Heterosy-
probably offer the best approach to the issue.] naptic depression can also last for days (Abraham et al., 1994;
When LTP is saturated and then depotentiated by LFS, it Doyre et al., 1997) (Fig. 1017C). A late, protein synthesis-
can subsequently be reinstated from the depotentiated level by dependent component of LTD has been described for
high-frequency stimulation. This experiment demonstrates NMDAR-dependent LTD in area CA1, induced by low-
that depotentiation is not simply a superposition of LTD on a frequency stimulation both in vivo (Manahan-Vaughan et al.
continuing but now hidden LTP because if that were the case m 2000) and in hippocampal slice cultures (Kauderer and
LTP would remain saturated and thus be incapable of Kandel, 2000). An early protein synthesis-dependent phase of
enhancement. Rather, depotentiation is genuine erasure of mGluR-dependent LTD is seen in acute slices (Hber et al.,
LTP, returning the pathway to a state in which LTP can 2000). An experimentally useful equivalence has been estab-
again be elicited (Dudek and Bear, 1992, 1993). However, LTD lished between LTD induced by the mGluR group 1 agonist
and depotentiation are not the same phenomenon (in the DHPG and by a low-frequency paired-pulse protocol; these
sense that they are generated by a common mechanism) two stimulus protocols both produce long-lasting, protein
because different phosphatases are involved in the two cases synthesis-dependent forms of LTD that when saturated mutu-
(Lee et al., 2000) (see Fig. 1019B); and in cortex at least, ally occlude each other, strongly suggesting a shared mecha-
induction is controlled by different NMDA receptor subunits nism (Hber et al., 2001).
(NR2B for LTD and NR2A for depotentiation) (Massey et
al., 2004). Moreover, in knockout mice lacking the A cat- 10.7.3 NMDAR-dependent LTD:
alytic subunit of calcineurin, LTD can be elicited but depo- Induction Mechanisms
tentiation cannot (Zhuo et al., 1999). The limited time
window for depotentiation following the induction of LTP Long-term depression is dened as NMDAR-dependent if it is
also suggests that it is mechanistically different from LTD, blocked by D-AP5 or other NMDA antagonists. There is con-
which can be induced, in principle, at any time in nave vincing evidence that the induction of NMDAR-dependent
synapses. LTD requires a rise in postsynaptic Ca2 and that the polarity
of the change in synaptic efficacy is determined by the ampli-
In Vivo Studies tude and time course of the Ca2 transient. First, the induc-
tion of LTD can be prevented by dialyzing CA1 pyramidal cells
In general, in vivo experiments conrm that it is considerably with a Ca2 chelator (Mulkey and Malenka, 1992). Second,
more difficult to obtain LTD and depotentiation in the intact lowering external Ca2 transforms a protocol that normally
adult rat than in younger animals. Nevertheless, both LTD and induces LTP into one that produces LTD (Mulkey and
depotentiation can be induced with low-frequency trains in Malenka, 1992). Third, Cummings et al. (1996) manipulated
area CA1 of the intact animal, although modications of the Ca2 entry into the postsynaptic cell in a number of ways,
standard protocol are often necessary. There is also evidence including the application of low doses of AP5 to limit Ca2
that susceptibility to LTD is strain-specic, with Wistar and entry during the tetanus, which again converted LTP to LTD.
Sprague-Dawley rats showing a lower threshold for LTD than These observations indicate that the level of the increase in
hooded Listers (Manahan-Vaughan, 2000). The standard 900 intracellular Ca2 is a critical parameter in determining the
pulses/1Hz protocol has been successfully used by some direction of change in synaptic strength. They provide sup-
(Heynen et al., 1996; Manahan-Vaughan, 1997), whereas port for an inuential scheme proposed by Lisman (1989)
others have found it necessary to use trains at higher frequen- that exploits the fact that certain protein phosphatases have a
cies (25 Hz) (Doyle et al., 1997) or paired-pulse trains higher affinity for Ca2 than most protein kinases. This sug-
consisting of pairs of stimuli at 0.5 Hz, with a 25-ms inter- gests a potential mechanism whereby low elevations of Ca2
pulse interval chosen to maximize paired-pulse depression activate protein phosphatases to yield LTD, whereas higher
(Thiels et al., 1994; Doyre et al., 1996). The incidence of LTD concentrations activate protein kinases to yield LTP. In
is greater in freely moving animals if they are subjected to Lismans scheme, activated protein kinases, in particular
mild stress (Xu et al., 1997). LTD has been particularly elusive autophosphorylated CaMKII, enhance synaptic efficacy by
in the dentate gyrus (Errington et al., 1995; Abraham et al., phosphorylating receptor subunits, while this process is
1996) where in the freely moving rat it has been elicited only reversed by activation of phosphatases. Phosphorylation of
under conditions of stress or following perfusion with mGluR AMPA receptors is known to increase AMPA receptor-medi-
agonists (reviewed by Braunewell and Manahan-Vaughan, ated currents, which could account at least in part for the
2001). It is, of course, possible that the optimal patterns of increase in synaptic efficacy in LTP (see Section 10.4).
activity for inducing LTD in the dentate gyrus have yet to be Phosphorylated CaMKII is a substrate for a protein phos-
identied. Depotentiation in the intact animal is discussed phatase (PP1) but PP1 is not itself Ca2-sensitive. To account
below. for LTD, Lisman therefore invoked a phosphatase cascade that
How long is the long in long-term depression? Homosy- begins with activation of the Ca2-sensitive protein phos-
naptic LTD can persist for days in chronically implanted ani- phatase calcineurin (also known as protein phosphatase 2B, or
mals (Doyre et al., 1996), and depotentiation has been PP2B). Activated calcineurin then dephosphorylates a phos-
 Synaptic Plasticity in the Hippocampus 411

phoprotein called inhibitor 1 (I1), which in its activated state cient to induce persistent LTD (Cummings et al., 1996). This
inhibits PP1. Activation of calcineurin thus leads to the acti- approach was taken further by Yang et al. (1999), who used
vation of PP1 by disinhibition, allowing CaMKII to be graduated photolysis of a caged Ca2 compound, DM-EGTA,
dephosphorylated. A critical feature of this scheme is that cal- to control the level and rate of rise of cytosolic calcium. These
cineurin has a higher sensitivity for Ca2 than CaMKII, con- studies made the important point that the rate of rise and the
sistent with the hypothesis that small increases in intracellular magnitude of the Ca2 transient are jointly involved in deter-
Ca2 lead to LTD and larger increases to LTP. Note also that mining the polarity of the change in synaptic efficacy. Large,
the scheme provides for a feedback, or gating, mechanism in rapid rises in Ca2 invariably produced persistent potentia-
which the PKA/inhibitor 1 pathway counteracts the effects of tion, whereas small, prolonged increases always induced per-
the calcineurin pathway when Ca2 levels rise to levels that sistent depression. In both cases calcium-induced plasticity
lead to autophosphorylation of CaMKII and LTP (Lisman, occluded activity-induced LTP or LTD, suggesting a conver-
1989; Blitzer et al., 1998). gence of cellular mechanisms. The importance of the pro-
As is the case for LTP, there is evidence that specic longed duration of the Ca2 transient for the induction of
NMDAR subtypes are involved in LTD. A requirement for the LTD in area CA1 has been further underlined in experiments
NR2B-containing NMDARs was suggested by a knockout of by Mizuno et al. (2001); pharmacological manipulations were
this subunit (Kutsuwada et al., 1996). These animals lacked used to control the magnitude of the Ca2 transient produced
LTD but had major developmental problems and died young. by 1 Hz stimulation, but under no circumstances were trains
Several groups have subsequently used NR2B subtype-specic of less than 200 stimuli successful in inducing LTD. A model
NMDAR antagonists, such as ifenprodil and its more selective of the probable signaling cascades involved in NMDAR-
derivatives Ro-25-6981 and CP-101,606, to investigate the role dependent LTD is shown in Figure 1018A.
of this subunit more directly, but no clear picture has yet
emerged. Thus, at one extreme LTD is reported to be Spike Timing-dependent Induction of LTD
enhanced by NR2B inhibitors (Hendricson et al., 2002) and at
the other blocked by these compounds (Liu et al., 2004a) As we saw in Section 10.3, when plasticity is induced by con-
Other pharmacological evidence has instead provided evi- joint presynaptic stimulation and a back-propagating spike in
dence for a role of NR2C and/or NR2D receptors in LTD the postsynaptic neuron, it is the interval between the onset of
(Hrabetova et al., 2000). There is also evidence that NR2D the two events that determines the sign of the persistent
receptors can mediate a very slow component of the synaptic change. In hippocampal slices, potentiation results when the
response at hippocampal synapses (Lozovaya et al., 2004). The EPSP precedes the back-propagating action potential (AP) in
very slow kinetics of this channel coupled with its low sensi- CA1 pyramidal neurons by 0 to 10 ms (Magee and Johnston,
tivity to Mg2 means that the NR2D component of the 1997). This result was conrmed in experiments on pairs of
NMDAR-mediated EPSC will summate effectively at the low connected hippocampal neurons in organotypic culture
frequencies of stimulation that are typically used to induce (Debanne et al., 1998) (Fig. 1019C) or in dissociated culture
LTD. Presumably, the relative roles of the different NR2 sub- (Bi and Poo, 1998) (Fig. 1019D). Moreover, LTD was induced
units in LTD is determined by a variety of factors. if the order was reversed, with the back-propagating AP pre-
Interestingly, in this context, Pro-BDNF has been shown to ceding the EPSP, extending to the hippocampus a pattern rst
facilitate LTD via activation of p75 leading to up-regulation of established in neocortical neurons (Markram et al., 1997). The
NR2B receptors (Woo et al., 2005). pattern of excitation and inhibition is more complex in hip-
Support for the Lisman hypothesis has come from experi- pocampal slices (Nishiyama et al., 2000). Potentiation results
ments by Malenka and colleagues, who found that LTD was when the onsest of the EPSP and back-propagating AP occur
blocked by injecting specic inhibitors of PP1 and calcineurin within 10 ms of each other, irrespective of order. Symmetri-
into CA1 neurons (Mulkey et al., 1993, 1994). On the other cally placed on either side of the window of potentiation are
hand, a study of transgenic mice expressing an inducible cal- two 20-ms intervals, between 10 and 30 ms (AP preceding
cineurin inhibitor failed to conrm the importance of cal- EPSP) and 10 and 30 ms (EPSP preceding AP), within which
cineurin for LTD; although LTP was enhanced, LTD was not paired activity results in homosynaptic LTD. The latter is
affected (Malleret et al., 2001), suggesting either that the probably explained by the recruitment of feedforward inhibi-
inhibitors were acting on another phosphatase or that another tion (Bi and Rubin, 2005). Furthermore, in contrast to the
high-affinity phosphatase can take over the role normally usual absence of heterosynaptic depression in area CA1 fol-
played by calcineurin. Overall, there is good evidence that the lowing tetanus-induced LTP, heterosynaptic depression occu-
amplitude and kinetics of the synaptically evoked Ca2 tran- pied the same intervals and displayed the same magnitude as
sient is critical for setting the sign of synaptic change. As pre- homosynaptic depression; notably absent, however, was any
dicted by the model, lower Ca2 charge transfer through the indication of heterosynaptic LTP.
NMDA receptor channel is associated with LTD and higher Spike timing-dependent LTD is also NMDAR-dependent,
transfer with LTP (Cho et al., 2001; Wu et al., 2001). but it is far from clear how it relates to conventional LTD
Depolarizing pulses delivered to the postsynaptic cell to induced by prolonged low-frequency stimulation (Lisman
draw Ca2 into the cell via L-type Ca2 channels is itself suffi- and Spruston, 2005). The rules of induction are very different,
412 The Hippocampus Book

A LTD, which was accompanied by a decrease in quantal size,

is probably mediated by a postsynaptic mechanism; mGluR-
dependent LTD, in contrast, is more likely to be presynapti-
AMPA AMPA cally expressed (see below). Subsequently, dephosphorylation
of AMPA receptor subunits and internalization of both
AMPA and NMDA receptors have emerged as mechanisms
AMPA contributing to the early expression of NMDAR-dependent
PP1 LTD. There is also evidence for a presynaptic component of
L-GLU P I1 I1 expression.

NMDA PP2B
Presynaptic Expression Mechanisms

Ca 2+ NMDAR-mediated and AMPAR-mediated currents are

equally downregulated in LTD (Xiao et al., 1995), a nding
that is compatible with a reduction in transmitter release or
with internalization of both receptor types. The evidence for
B downregulation of AMPA receptors is strong and, together
with other evidence for postsynaptic changes discussed above,
makes an overwhelming case for a postsynaptic contribution
AMPA
to NMDAR-dependent LTP. However, this does not exclude a
presynaptic contribution. Evidence for presynaptic involve-
ment in the expression of NMDAR-dependent LTD has
AMPA
MAPK emerged from imaging studies of vesicular turnover using the
PI3K
uorescent styryl dye FM1-43 (Stanton et al., 2001, 2003). The
L-GLU
mTOR rate of release from a rapidly releasing pool of vesicles was sig-
mGlu PTP
nicantly less after the induction of LTD than it was before
induction. The effect was NMDAR-dependent and required
NO release and activation of presynaptic cGMP, raising the
possibility that NO is acting as a retrograde messenger (see
Section 10.4.7).

Figure 1018. Mechanisms involved in the induction of LTD. A.

Postsynaptic Expression Mechanisms
NMDA receptor-dependent LTD. The entry of Ca2 via NMDA
receptors leads to activation of calcineurin (PP2B). This inhibits
A unitary mechanism for bidirectional changes in synaptic
inhibitor-1 (I1), which in turn results in activation of protein phos-
phatase 1 (PP1). This leads to the internalization of AMPARs, prob-
efficacy has obvious attractions (Dudek and Bear, 1993).
ably after they have moved laterally away from the synapse. B. However, an illuminating study on the phosphorylation of
mGluR-dependent LTD. Activation of mGlu receptors, in particular GluR1 residues in LTP and LTD, although conrming the view
mGlu5, also leads to removal of AMPARs from synapses. Here the that depotentiation is a reversal of LTP, suggests that LTP and
signaling mechanisms involve protein tyrosine phosphatases (PTPs), LTD involve phosphorylation and dephosphorylation at dif-
mTOR, P13K and MAPKs. How these enzymes interact is not ferent residues on GluR1 (Lee et al., 2000). In this scheme
known. See text for further details. (Fig. 1019B), AMPA receptors in the baseline state are phos-
phorylated at serine residue 845. High-frequency stimulation
leads to activation of CaMKII and phosphorylation of ser831,
and it is possible that their expression mechanisms are differ- a state associated with increased single-channel conductance
ent too. Do they occlude? Is spike timing-dependent LTD of the receptor (Derkach et al., 1999). Depotentiation by low-
blocked by the same sub type selective NMDA receptor antag- frequency stimulation reverses this process by activation of
onists as conventional LTD? What is the age dependence of protein phosphatases PP1 and PP2A. Low-frequency stimula-
spike timing-dependent LTD? What are the relative roles of tion in the nave slice activates the same okadaic acid-sensitive
conventional LTD and spike timing-dependent LTD in the protein phosphatases, but in this condition it is ser845 that is
behaving animal? These are questions for future research. dephosphorylated, leading to a subunit that is phosphorylated
at neither serine residue. High-frequency stimulation in this
10.7.4 NMDAR-dependent LTD: state activates PKA to rephosphorylate the receptor on ser845
Expression Mechanisms and restore it to its baseline state. Continued high-frequency
stimulation would then progress to activation of CaMKII via
In their characterization of two forms of LTD referred to phosphorylation of S831 and hence to LTP. Note that this
above, Oliet et al. (1997) concluded that NMDAR-dependent scheme cannot explain the relative immunity to depotentia-
 Synaptic Plasticity in the Hippocampus 413

A B
fEPSP slope (% change)

HFS
100
LFS

50 GluR1

PKA CaMKII
0
PP1/2A PP1/2A PS831
PS845
PS845
-50 LTD Naive LTP
0 30 60 90 120 150 180
Time (min)

C D
post before pre pre before post post before pre pre before post

100
100
EPSP (% change)

EPSC (% change)
50
50

0
0

-50

-50

-200 -175 -150 -125 -100 -75 -50 -25 0 25 -80 -40 0 40 80

Spike timing (msec) Spike timing (msec)

Figure 1019. Bidirectional plasticity. A. Bidirectional plasticiy in state via PKA-mediated phosphorylation of serine845. (Source:
area CA1 in vivo. LTD induced by low-frequency stimulation (LFS) Lee et al., 2000.) C, D. Two examples of bidirectional spike timing-
(1 Hz for 15 minutes) (bar) followed by induction of LTP by high- dependent synaptic plasticity (STDP). C. STDP elicited by activat-
frequency stimulation (HFS) (100 Hz for 1 second) (arrow), partial ing pairs of interconnected CA3 cells in organotypic cultures. LTP
depotentiation by LFS, and repotentiation by HFS. (Source: Heynen was induced if an action potential in the presynaptic cell was
et al., 1996.) B. Model of changes in the phosphorylation state of repeatedly evoked 15 ms before an action potential in the postsy-
the GluR1 subunit of the AMPA receptor that underlie NMDA naptic cell; at zero delay or when the postsynaptic spike preceded
receptor-dependent early LTP, depotentiation, early LTD and dede- the presynaptic spike by 15 to 70 ms, LTD was induced. (Source:
pression. In the nave state, the receptor is phosphorylated at serine After Debanne et al., 1998.) D. STDP in pairs of interconnected
845 (S845). E-LTP is expressed via CaMKII-mediated phosphory- pyramidal cells in dissociated hippocampal cultures. Again, LTP is
lation of serine 831. Depotentiation involves the dephosphorylation induced if the presynaptic cell res shortly before the postsynaptic
of S831 via protein phosphatases PP1 and/or PP2A. Early LTD cell, and LTD ensues if the timing is reversed. Each circle represents
involves dephosphorylation of S845, again by PP1/PP2A, and a single pairing experiment. (Source: Bi and Poo, 1998.)
dedepression after LTD brings the receptor back to the nave

tion once LTP has become stabilized, perhaps by protein syn- A number of techniques have been used to demonstrate
thesis-dependent mechanisms. Subsequent studies have downregulation of both AMPA and NMDA receptors in
shown that results from the intact animal are compatible with NMDAR-dependent LTD. They include photolytic uncaging
the model. Similar post-translational changes in GluR1 phos- of caged glutamate (Kandler et al., 1998; Rammes et al., 2003),
phorylation levels on ser845 occur in visual cortex following iontophoresis of NMDA (Montgomery and Madison, 2002),
monocular deprivation (Bear, 2003) and transgenic animals and immunocytochemical estimates of both AMPA and
with point mutations on serine residues 831 and 845 (render- NMDA populations in tissue from animals in which LTD was
ing them nonphosphorylatable) exhibit impaired spatial induced in vivo (Heynen et al., 2000). The use of antibodies
memory, deficits in LTP, and complete suppression of that recognize extracellular epitopes of AMPA receptors on
NMDAR-dependent LTD (Lee et al., 2000). The latter result living neurons (Collingridge et al., 2004) has allowed direct
must be reconciled with the evidence for a protein synthesis- demonstration of AMPA receptor internalization during LTD
dependent component of this form of LTD, discussed below in cultured hippocampal neurons (Beattie et al., 2000). Sub-
(Kauderer and Kandel, 2000). sequent studies using pHluorin-GluR2 have suggested that
414 The Hippocampus Book

internalization occurs at extrasynaptic sites and that this pre- and Henley, 2005). Furthermore, a Ca2-insensitive PICK1
cedes removal of AMPA receptors from the synapse (Ashby et mutant blocks NMDA-induced internalization of AMPA
al., 2004). The simplest model therefore is one in which endo- receptors, implicating this interaction in LTD. However,
cytosis at an extrasynaptic site is followed by lateral diffusion another idea is that GRIP/ABP binds the GluR2 subunit to
of AMPA receptors from synaptic to extrasynaptic locations. maintain a store of AMPA receptors intracellularly (Daw et al.,
Many of the proteins that control the cycling of GluR2 2000). In this scheme, when PICK1 phosphorylates GluR2 it
between cytoplasm and synaptic membrane have been identi- untethers AMPA receptors and so provides a pool near the
ed, and details of their modes of action are beginning to synapse, ready for membrane insertion. It is possible that
emerge (Collingridge et al., 2004). The concept of subunit- these diametrically opposed roles of PICK1, via its interaction
dependent regulation of synaptic plasticity arose from the with GRIP/ABP, can both operate under different conditions.
nding that NSF binds selectively to the GluR2 subunit of Considerable work still needs to be done to understand the
AMPA receptors (Nishimune et al., 1998; Osten et al., 1998; relation between the various proteins involved in AMPA
Song et al., 1998). Of note were the observations that blockade receptor trafficking and LTD.
of the GluR2 interaction with NSF leads to a run-down in Although the role of the GluR2 subunit in LTD is well
AMPAR-mediated synaptic transmission (Nishimune et al., established, it is not the only mechanism that supports
1998; Song et al., 1998), a reduction in the surface expression NMDAR-dependent LTD in the hippocampus. This is shown
of AMPA receptors (Lscher et al., 1999; Noel et al; 1999), and by studies using knockouts, where NMDAR-dependent LTD
occlusion with NMDAR-dependent LTD (Lscher et al., 1999; has been observed in both the GluR2 knockout and the
Lthi et al., 1999). It was subsequently discovered that AP2 GluR2/GluR3 double knockout (Jia et al., 1996; Meng et al.,
binds to a region of GluR2 that overlaps with the NSF site (Lee 2003). Moreover, as we have seen, the dephosphorylation of
et al., 2002). Selective block of the GluR2 interaction with NSF ser845 of GluR1 can also result in LTD (Kameyama et al.,
caused a run-down in synaptic transmission, and selective 1998), suggesting that there are multiple mechanisms for the
block of the GluR2 interaction with AP2 resulted in suppres- induction of LTD. Further studies are required to understand
sion LTD. This suggests a mechanism whereby NSF is nor- the relative contributions of these various mechanisms to LTD
mally bound to GluR2 subunits to stabilize AMPA receptors at in wild-type animals in vivo.
the synapse. However, in response to an LTD-inducing stimu-
lus, AP2 exchanges with NSF, and this initiates clathrin- A Protein Synthesis-dependent Phase
dependent endocytosis (Collingridge et al., 2004). This scheme of NMDAR-dependent LTD
requires a mechanism to initiate the exchange of NSF for
AP2. One possibility is that Ca2 entry during the induction Three investigations published in 2000 established that
of LTD interacts with the neuronal Ca2 sensor hippocalcin NMDAR-dependent LTD can sustain a late, protein-depend-
(Burgoyne, 2004). On binding Ca2, hippocalcin undergoes a ent phase. Different preparations were used in each of these
conformational change to expose a myristoylated region that studies: organotypic cultures (Kauderer and Kandel, 2000),
can target the molecule to the plasma membrane. It also binds acute slices (Hber et al., 2000), and freely moving rats
both AP2 and GluR2 subunits in a Ca2-dependent manner (Manahan-Vaughan et al., 2000). In all three preparations,
(Palmer et al, 2005). Thus, hippocalcin is well placed to late LTD lasting several hours could be induced by prolonged
actively promote the exchange of NSF for AP2. Consistent low-frequency stimulation, and in each case it was blocked by
with this idea, a dominant negative form of hippocalcin blocks protein synthesis inhibitors. However, only in organotypic
induction of NMDAR-dependent LTD (Palmer et al, 2005). culture was late LTD blocked by transcriptional inhibitors.
Thus, hippocalcin is a strong contender for the role of Ca2 Local protein synthesis may therefore provide the message
sensor in LTD. It is likely, however, that other Ca2-sensing captured by the tags set by LTD-inducing stimulation
proteins are also involved in the induction of LTD, and the full (Kauderer and Kandel, 2000; Sajikumar and Frey, 2004b) (see
mechanistic details have yet to be elucidated. Section 10.4.9).
Scaffolding proteins such as GRIP/ABP (Dong et al., 1997;
Srivastava and Ziff, 1999) and the PKC-targeting protein 10.7.5 mGluR-dependent LTD
PICK (Dev et al., 1999; Xia et al., 1999) also bind to the C ter-
minal tail of the GluR2 subunit, where they may tether AMPA mGluR-dependent LTD Induced by Synaptic Activity
receptors at the plasma membrane or at intracellular sites.
There is evidence that PICK1 is directly involved in LTD (Kim Homosynaptic LTD using low-frequency stimulation is
et al., 2001). The proposal is that GRIP/ABP normally anchors blocked by NMDA receptor antagonists (Dudek and Bear,
AMPA receptors at the synapse. However, LTD leads to activa- 1992; Mulkey and Malenka, 1992). Before long, however, it
tion of PICK1, which phosphorylates GluR2 subunits, causing was shown that certain forms of LTD and depotentiation were
them to dissociate from GRIP/ABP so they can be removed prevented by mGluR antagonists rather than by NMDA
from the synapse. Indeed, it has been shown that PICK1 binds antagonists (Bashir et al, 1993; Bashir and Collingridge, 1994;
Ca2, facilitating its interaction with AMPA receptors (Hanley Bolshakov and Siegelbaum, 1994; Yang et al., 1994), thereby
 Synaptic Plasticity in the Hippocampus 415

demonstrating two mechanistically distinct forms of synaptic nists depress the late component of LTD in vivo (Manahan-
depression. Both types were pathway-specic The mechanisti- Vaughan, 1997).
cally distinct nature of these two forms of LTD was conrmed
by the lack of occlusion between the two (Oliet et al., 1997; mGluR-dependent LTD Induced
Palmer et al, 1997). Oliet et al. (1997) also observed a decrease by Exposure to mGluR Agonists
in the quantal size of miniature EPSCs without a change in
miniature frequency in NMDAR-dependent LTD, suggesting a Brief exposure to the group I mGluR agonist dihydroxy-
postsynaptic expression mechanism. In contrast, in mGluR- phenylglycine (DHPG) induces a reassuringly robust LTD in
LTD there was a reduction in miniature frequency, with no adult slice preparations in both area CA1 (Palmer et al., 1997)
change in quantal size, suggesting a presynaptic expression (Fig 10.17E) and the dentate gyrus (Camodeca et al., 1999).
mechanism. Finally, induction of mGluR-dependent LTD did This form of LTD, probably mediated by mGluR5 receptors, is
not erase LTP, as was the case with depotentiation produced blocked by group I or broad-spectrum mGluR antagonists but
by the stimulus protocol that produced NMDAR-dependent not by NMDA receptor antagonists. Experiments to probe the
LTD but, rather, was superimposed on LTP. These experiments induction site of mGluR-LTD point toward a postsynaptic
conrmed the existence of two distinct types of LTD at locus of induction. Intracellular injection of a G protein
Schaffer-commissuralCA1 pyramidal synapses with distinct inhibitor into CA1 pyramidal cells suppressed the ability of
induction and expression mechanisms. DHPG to induce LTD (Watabe et al., 2002). Intracellular
Low-frequency trains to Schaffer-commissural bers in injection of BAPTA, on the other hand, had no such effect
slices from neonatal rats induce a form of mGluR-dependent (Fitzjohn et al., 2001b), suggesting a Ca2-insensitive induc-
LTD that is blocked by postsynaptic injection of calcium tion process. The experiments of Hber et al. (2000) led to the
chelators but is, at least in part, expressed as an increase in conclusion that mGluR-LTP depends on rapid activation of
transmitter release (Bolshakov and Siegelbaum, 1994)a local postsynaptic protein synthesis. In the presence of the
combination that immediately suggests the need for a retro- protein synthesis inhibitor anisomycin, the depression
grade messenger (see Section 10.4.7). Strong, if incomplete, induced by DHPG quickly attenuates, and responses recover
evidence that a lipoxygenase metabolite of arachidonic acid, to baseline within 60 minutes (Hber et al., 2000). LTD
12(S)HPETE, may be the retrograde messenger in this case induced by mGluR agonists thus quickly becomes dependent
has been presented by Feinmark et al. (2003). on protein synthesis. Consistent with the rapid action of ani-
A study by Kemp et al. (2000) illustrates the complexity of somycin, mGluR-induced LTD survived in an isolated in vitro
LTD induction mechanisms. The effects of two protocols in preparation, from which both pre- and postsynaptic cell bod-
area CA1 were compared: trains of single stimuli at 1 Hz and ies had been removed by appropriate scalpel cuts. This result
trains of paired pulses also at 1 Hz. In slices from young ani- implies that local protein synthesis at synaptic sites is required
mals, LTD produced by either protocol was blocked by AP5. early in the expression of mGluR-induced LTD. Given that
Adult animals showed a shift toward mGluR-dependent LTD protein synthetic machinery has not been described at presy-
and in general required longer trains, but it remained possible, naptic boutons, it seems safe to conclude that the induction of
by varying the stimulus parameters, to obtain either mGluR- this early form of protein synthesis-dependent LTD is located
dependent or NMDAR-dependent LTD. Again, the two forms in dendrites. Synaptically driven mGluR-dependent LTD
of LTD did not occlude, indicating the independence of the induced by low-frequency trains of paired pulses was also
associated expression mechanisms. The mechanisms behind blocked by anisomycin (Hber et al., 2000).
the age- and activity-related shift from NMDAR-mediated to There is less agreement concerning the locus of expression
mGluR-mediated LTD are unknown but may reect the devel- of this form of LTD. On the one hand there is evidence that it
opmental regulation of receptor subtypes. Whatever the rea- is presynaptically mediated because (1) there is no postsynap-
sons, these observations make it clear that the precise details tic change in response to ash photolysis of caged glutamate
of activity-induced changes in synaptic efficacy depend on the (Rammes et al., 2003), (2) NMDA and AMPA receptor com-
ne temporal structure of the patterns of neural activity itself. ponents of synaptic responses are equally depressed (Watabe
Although the paired-pulse protocol introduced by Kemp et et al., 2002), (3) paired-pulse facilitation is increased (Fitzjohn
al. (2000) (900 pairs at 1 Hz with an interpulse interval of 50 et al., 2001), and (4) DHPG produces a persistent decrease in
msec) generates mGluR-dependent LTD in area CA1 in vitro the frequency of miniature EPSCs with no change in ampli-
(see also Hber et al., 2000), it has not been shown to be effec- tude in tetrodotoxin-treated cultured hippocampal neurons
tive in vivo. The subtly different paired-pulse protocol formu- (Fitzjohn et al., 2001b). However, there is also evidence that
lated by Thiels et al. (1994), in which the interpulse interval is DHPG-induced LTD involves removal of both AMPA and
set at 25 ms to maximize inhibitory feedback, produces a NMDA receptors from synapses (Snyder et al., 2001). These
depression that is NMDAR-dependent rather than mGluR- disparate observations may be reconciled by the nding that
dependent (Thiels et al, 2002). Furthermore, although group I mGluR-dependent LTD, induced by DHPG, is blocked by
mGluR antagonists are generally more effective at blocking postsynaptically applied inhibitors of protein tyrosine phos-
mGluR-dependent LTD in area CA1 in vitro, group II antago- phatases (PTPs) and of the actin cytoskeleton (Moult et al,
416 The Hippocampus Book

2006).These postsynaptic manipulations blocked the changes (Hber et al., 2001), and, as we have seen, both are blocked by
in paired-pulse facilitation and miniature frequency, suggest- protein synthesis inhibitors (Hber et al., 2000). However,
ing a retrograde signaling process that is dependent on PTPs unlike DHPG-induced LTD, synaptically-induced mGluR-
and the actin cytoskeleton. An alternative explanation is that dependent LTD appears also to be sensitive to PKC inhibitors
AMPA receptors are removed to create silent synapses. If this (Oliet et al, 1997). Whether this reects a heterogeneity in the
occurs preferentially at high-Pr synapses owing to the require- synaptic forms is unknown.
ment for activity to drive the movement, the changes in How the various signalling components that have been
paired-pulse facilitation would simply reect this postsynap- identied in mGluR-dependent LTD (e.g., PI3K-Akt-mTOR,
tic alteration (see Box 101). The lack of change in the MAPKs, PTPs) interrelate is not known. What is clear, how-
response to uncaged glutamate, which act on both spines and ever, is that the signalling mechanisms involved in mGluR-
dendrites, can be reconciled with a postsynaptic model if, induced LTD are very different from those involved in
instead of being internalized, AMPA receptors diffuse laterally NMDAR-dependent LTD (see schemas in Fig. 1018A and B).
to extrasynaptic sites (Moult et al., 2006).
10.7.6 Homosynaptic Depotentiation
Signaling Pathways in mGluR-dependent LTD
Long-term potentiation in vitro can be reversed by low-
DHPG-induced LTD has been widely used as a model to frequency stimulation when it is given within a few minutes
explore the signalling mechanisms involved in mGluR- of the tetanus used to induce LTP (Barrionuevo et al., 1980;
dependent LTD. By denition, mGluRs are coupled to G- Staubli and Lynch, 1990; Fujii et al., 1991), a process some-
proteins, and as expected DHPG-induced LTD requires acti- times referred to as destabilization of LTP. Destabilization is
vation of G-proteins (Watabe et al., 2002; Huang CC et al., a specic example of the more general phenomenon of
2004). Since group I mGluRs are positively linked to PLC, and depotentiation, which refers to the reversal of LTP at any post-
thereby the generation of diacylglycerol and IP3, it might have tetanus interval. Whereas destabilization is a robust phenom-
been anticipated that inhibitors of PKC and/or release of Ca2 enon observed in all hippocampal pathways and at all ages,
from stores would block mGluR-dependent LTD. However, a depotentiation at longer intervals after induction is less easily
variety of potent PKC inhibitors and blockers of Ca2 stores, induced in adult animals.
applied alone or in combination, failed to affect DHPG- In all hippocampal subelds and in both adult and juvenile
induced LTD (Camodeca et al., 1999; Schnabel et al., 2001). animals, LTP can be depotentiated by a prolonged low-fre-
Furthermore, DHPG-induced LTD is insensitive to inhibitors quency train (e.g., 900 stimuli at 1 Hz) beginning immediately
of PKA (Camodeca et al., 1999; Schnabel et al., 2001), while after induction. The time window for induction and the input
inhibitors of CaMKII cause a modest enhancement of the specicity of depotentiation have been documented. The effi-
effect (Schnabel et al., 1999). It has been suggested that ciency of the depotentiating train quickly declines after the
DHPG-induced LTD is blocked by PTK inhibitors in the den- LTP-inducing event; and many, though not all, in vitro studies
tate gyrus (Camodeca et al., 1999) but this nding was not suggest that depotentiation can be induced only if low-
replicated in area CA1 (Moult et al., 2002), suggesting the pos- frequency stimulation is applied within minutes of LTP
sibility of regional differences. Whilst there is general agree- induction. A similar time window is seen in the dentate gyrus
ment for a role of MAPK cascades, there is some contention as of the intact animal (Martin, 1998; Kulla et al., 1999; Straube
to whether the mechanism involves the Ras-activated ERKs and Frey, 2003). In area CA1 of the awake animal, however,
(Gallagher et al., 2004) or the Rap-activated p38 MAPK (Rush Doyle et al. (1997), using 900 stimuli at 5 to 10 Hz, succeeded
et al., 2002; Huang CC et al., 2004). Finally, so far as kinases in depotentiating LTP at a range of intervals from 10 minutes
are concerned, there is evidence for a role of the PI3K-Akt- to 24 hours after induction in area CA1. Lower stimulus fre-
mTOR signalling pathway in DHPG-induced LTD (Hou and quencies were ineffective, and Doyle et al. were unable to
Klann, 2004). With respect to phosphatases, inhibitors of the induce LTD at any frequency. Another question is how long
ser/thr phosphatases that are involved in NMDAR-dependent depotentiation lasts. In the dentate gyrus of the unanes-
LTD (PP1/PP2A and PP2B) do not inhibit DHPG-induced thetized rat, depotentiation has been reported to persist for 24
LTD; rather, PP1/PP2A inhibitors cause a small facilitation of hours (Kulla et al., 1999). This raises the question of the sta-
the effect (Schnabel et al., 2001). However, inhibitors of tyro- bility of the depotentiated response: Do depotentiated
sine phosphatases completely eliminate DHPG-induced LTD synapses, like their potentiated counterparts, retain the poten-
(Moult et al., 2002; Huang CC and Hsu, 2006), a result that tial for reversed plasticity only during a limited time window
identies a major signalling pathway for this robust form of after induction? In other words, is de-depression, like depo-
LTD. tentiation, time-limited? The essence of bidirectional plastic-
Less is known about the signalling pathways involved in ity is that synaptic efficacy can be reset up and down at will,
synaptically-induced mGluR-dependent LTD. The report that but this may be the case only within specic time windows.
the DHPG-induced and synaptically-induced forms mutually There is some evidence to suggest that, at least in area CA1,
occlude each other suggests a common underlying pathway depotentiation appears to be input-specic (Bortolotto et al,
 Synaptic Plasticity in the Hippocampus 417

1994; Huang CC et al., 2001; but see Muller et al., 1995). mounted protection is provided through local protein synthe-
However, in the dentate gyrus in vivo and at mossy ber sis. The protein synthesis blockers emetine and anisomycin
synapses, heterosynaptic effects become apparent (see below). block the protection and do so in slices in which a cut is made
According to the model of early LTP and LTD expression between the cell body layer and the recording and stimulating
proposed by Lee et al. (2000), potentiation occurs as a result of sites in the stratum radiatum of area CA1. The protection is
the phosphorylation of S831 on GluR1, leading to an increase input-specic, but within 40 minutes heterosynaptic protec-
in channel conductance (Derkach et al., 1999). Depotentia- tion against depotentiation is extended to synapses through-
tion results when this residue is dephosphorylated (Lee et al., out the cell by a transcriptional-dependent process. These
2000; Huang CC et al., 2001). Consistent with this idea, it has results are reminiscent of the nding of Huber et al. (2000)
been shown that depotentiation can be, but is not invariably, that mGluR-dependent LTD involves rapid local protein syn-
accompanied by a decrease in single-channel conductance thesis as well as the tagging experiments of Frey and Morris
(Lthi et al, 2004). Indeed, there is a precise relation between (1997) and Kauderer and Kandel (2000). In none of these
the type of LTP and depotentiation observed; when LTP is the cases have the effector molecules responsible or their mode of
result of an increase in single-channel conductance, the action yet been identied.
increase is reversed in depotentiation. However, in cases where Depotentiation exhibits an intriguing form of state
there is no change in single-channel conductance during LTP, dependence in newly unsilenced connections between CA3
neither is there any change in this parameter during depoten- pyramidal cells in organotypic cultures. In this preparation,
tiation. With the latter form of bidirectional plasticity, the about 20% of connected cells are originally silent (that is, the
most likely explanation is changes in the number of AMPA postsynaptic cell does not respond to an action potential
receptors expressed at synapses. Both the protein kinase generated in the presynaptic cell) but can be made active by
(CaMKII) and protein phosphatases (PP1/2A) involved in an LTP-inducing pairing protocol. For the rst few minutes
these processes are calcium-dependent; and, adapting the or so after unsilencing, these new functional connections are
Lisman hypothesis, strong, rapid increases should favor phos- immune to re-silencing. After 30 minutes, the newly activated
phorylation and potentiation, whereas prolonged, modest connections enter a state in which they can be depressed by
increases in calcium should favor dephosphorylation and low-frequency stimulation and after LTP induction can be
depotentiation (Yang et al., 1999; Mizuno et al., 2001). There depotentiated in the usual way by low-frequency stimulation
is disagreement about the pharmacology of depotentiation. (Montgomery and Madison, 2002).
Some authors have reported that AP5 blocks depotentiation We end this section with the reminder that a potent chem-
in area CA1 in vitro (Fujii et al., 1991; Selig et al., 1995b; ical activator of depotentiation, which appears to be effective at
Huang CC et al., 2001; Lthi et al., 2004), whereas others have arbitrarily long intervals after induction, exists in the form of a
found it is blocked by Gp 1/II mGluR antagonists, and not by membrane permeable inhibitor of the autonomously active
NMDAR antagonists, in both area CA1 in acute slices (Bashir PKC isoform PKM (Serrano et al., 2005; see Section 10.4.9).
and Collingridge, 1994; Fitzjohn et al., 1998) and between
pairs of CA3 pyramidal cells in organotypic hippocampal cul- 10.7.7 Heterosynaptic LTD and Depotentiation:
tures (Montgomery and Madison, 2002). As with LTD, the Activity in One Input Can Induce LTD in Another
data suggest that two forms of depotentiation exist: one
depending on activation of NMDA receptors, and the other Heterosynaptic LTD and depotentiation were the rst forms
on mGluRs. The extent to which each form is accessed is likely of long-term activity-dependent depression to be discovered
to depend on the experimental conditions employed. in the hippocampus. In two-pathway experiments in area CA1
The extent to which different NR2 subunits are involved in of the hippocampal slice, Lynch et al. (1977) noticed that a
the NMDA receptor-dependent form of depotentiation in the strong tetanus to one pathway was sometimes accompanied
hippocampus is unknown. The possibility that there may be by a depression of the population spike in the second, nontet-
differences between depotentiation and LTD in this regard is anized pathway. However, even when present, heterosynaptic
however suggested by experiments performed in perirhinal depression rarely lasted for more than a few minutes in slices
cortex (Massey et al., 2004). Depotentiation was not affected from adult animals. Two years later, in a study in the dentate
by selective NR2B inhibitors whilst LTD, recorded in the pres- gyrus of the anesthetized rat, Levy and Steward (1979) showed
ence of an uptake inhibitor, was fully blocked by these com- that previously potentiated responses evoked by stimulation
pounds. Conversely, a partially selective NR2A antagonist, of the contralateral perforant path were depressed by tetanic
NVP-AAM007, blocked depotentiation whilst having no effect stimulation of the ipsilateral perforant path (Fig. 1017A).
on LTD. Depression could also be induced in the nave contralateral
An interesting twist to the depotentiation story was uncov- pathway. The crossed perforant path thus exhibits both het-
ered by Woo and Nguyen (2003). Multiple trains of tetanic erosynaptic LTD and heterosynaptic depotentiation.
stimulation not only produce maximal LTP but protect More recently, heterosynaptic LTD in area CA1 has been
against depotentiation by low-frequency stimulation begin- documented in slices from young animals, induced by a
ning 5 minutes after induction. Surprisingly, this rapidly tetanus (Scanziani et al., 1996; Nagase et al., 2003) or by low-
418 The Hippocampus Book

frequency stimulation (Wasling et al., 2002) delivered to there is evidence it can work in both directions. Tetanization
another pathway. Others have been able to elicit heterosynap- of the LPP induces persistent LTD in nave MPP responses
tic depression only if the heterosynaptic pathway was rst and persistent and reversible depotentiation of potentiated
potentiated (heterosynaptic depotentiation) (Muller et al., MPP responses (Doyre et al., 1997) (Fig. 1017C). Moreover,
1995). Evidence that cytoplasmic Ca2 is a trigger for het- in contrast to homosynaptic depotentiation, there is no
erosynaptic depression has been provided by a detailed study restricted time window for the induction of heterosynaptic
of spike timing-dependent plasticity in slices from 1-month- depotentiation; in the freely moving animal, a tetanus to the
old rats (Nishiyama et al., 2000). When an LTD-inducing pro- LPP can depotentiate MPP responses several days after induc-
tocol was delivered to the rst pathway (that is, when the tion. Both heterosynaptic LTD and depotentiation can persist
afferent input was repeatedly paired with a back-propagating for days, though the depression is in general less persistent
action potential, the latter preceding the former by ~20 ms), than homosynaptic LTP in the same pathways (Doyre et al.,
LTD was also induced in a second, unpaired pathway. The 1997). Doyre et al. were not able to produce heterosynaptic
effect was blocked by injection of an antibody to the IP3 LTD in the LPP; but in Abrahams laboratory, tetanic stimula-
receptor, and it was not seen in IP3 receptor knockout mice tion of the MPP led to LTD that lasted for days in the LPP
(see also Nagase et al., 2003). Nishiyama et al. (2000) attrib- (Abraham et al., 1994).
uted the spread of LTD to the propagation of IP3 receptor- Heterosynaptic depression and depotentiation in the den-
mediated Ca2 waves propagating via the endoplasmic tate gyrus in vivo are largely conned to the fEPSP; Doyre et
reticulum to neighboring synaptic sites. These experiments, al. reported small but signicant heterosynaptic depression of
together with the work of Daw et al. (2002), suggest that the population spike but little if any depotentation. A possible
input specicity in synaptic modication is not an intrin- explanation for this seemingly anomalous result is that het-
sic property but a dynamic variable linked to the pattern erosynaptic depression is accompanied by heterosynaptic E-S
of long-range Ca2 signaling in the postsynaptic dendrite potentiation (Abraham et al., 1985).
(Nishiyama et al., 2000). In complementary work, Daw et al.
(2002) found that the input specicity of LTD was lost in the Heterosynaptic Associative LTD: A Rarely
presence of an inhibitor of IP3, leading to heterosynaptic Observed Form of Synaptic Plasticity
depression. This result is in line with the conclusions of
Nishiyama et al. (2000), since a likely consequence of PI3 Because the nontetanized pathway is notor need not be
kinase inhibition is an increase in IP3 concentration. active at the time of induction, heterosynaptic LTD represents
A novel, unsuspected form of heterosynaptic depression a mechanism for generalized synaptic depression. With asso-
has emerged as a result of the growing interest in the effects of ciative heterosynaptic LTD, in contrast, it is only the synapses
cannabinoids and their receptors in the hippocampus. Tetanic active at the time of induction that are depressed. Convincing
stimulation of the Schaffer-commissural pathway in hip- examples of this form of LTD are rare, and it remains to be
pocampal slices induces a powerful and persistent depression established if associative LTD exists in the hippocampus of the
of neighboring GABAergic terminals projecting to the same intact animal. A report by Stanton and Sejnowski (1989) that
target CA1 pyramidal cells (Chevaleyre and Castillo, 2003). heterosynaptic associative LTD could be induced in area CA1
The effect relies on release of the cannabinoid diacylglycerol in vitro by pairing single stimuli to one pathway with out-of-
(2-AG) from postsynaptic cells; 2-AG binds to cannabinoid phase brief high-frequency trains to a second pathway could
CB1 receptors on the terminals of GABAergic interneurons, not be replicated by Kerr and Abraham, (1993); or Paulsen et
leading by an unknown route to a reduction in transmitter al., (1993). In a variant of this protocol, Christie and Abraham
release. The resulting disinhibition provides a potential (1992) reported that the lateral perforant path in the dentate
mechanism for the phenomenon of E-S potentiation (see gyrus of the anesthetized rat can exhibit associative LTD if it
Section 10.3.12).This and other examples of cannabinoid- has previously been primed by stimulation at theta fre-
mediated synaptic plasticity are reviewed by Chevaleyre et al. quency (Abraham and Goddard, 1983; Christie and Abraham,
(2006). 1992). However, no evidence of priming was seen in area CA1
in the in vitro experiments of Paulsen et al. (1993), suggesting
Heterosynaptic LTD and Depotentiation in the Medial that the effect may be conned to the dentate gyrus or to the
and Lateral Perforant Path of the Intact Rat intact animal. A homosynaptic variant of the Stanton and
Sejnowski protocol has been studied in organotypic hip-
The strongest body of in vivo evidence relating to heterosy- pocampal cultures (Debanne et al., 1994). Repeated pairing of
naptic LTD and depotentiation has come from studies of the a depolarizing pulse to an impaled CA1 neuron with a single
medial and lateral components of the ipsilateral perforant stimulus delivered several hundred milliseconds later to
path (MPP and LPP, respectively) in the intact rat (Abraham Schaffer-commissural bers resulted in prolonged depression
and Goddard, 1983; Abraham et al., 1985, 1994l; Doyre et al., in the stimulated pathway. The time window during which
1997) (Fig. 1017B). Though accounts differ as to the relative associative LTD could be obtained was fairly broad: Maximal
ease of obtaining heterosynaptic effects in the two pathways, depression was obtained with a pairing interval of 800 ms, but
 Synaptic Plasticity in the Hippocampus 419

1600 ms was almost as effective. The effect is associative rmed by the fact that it does not occlude with LTD induced
because it required both afferent activity and strong depolar- by low-frequency stimulation of mossy bers.
izing pulses to re the postsynaptic cell, and it is input-
specic because it was not observed in unstimulated inputs. Depotentiation at Mossy Fiber Synapses
Associative LTD in organotypic cultures is NMDAR-depend-
ent and is blocked in cells dialyzed with Ca2 chelators. Low-frequency stimulation applied immediately, but not 30 to
60 minutes, after a tetanus reverses mossy fiber LTP
10.7.8 LTD and Depotentiation at Mossy (Tzounopoulos et al., 1998; Chen et al., 2001a). Thus, as in area
FiberCA3 Pyramidal Cell Synapses CA1 and the dentate gyrus, LTP in time becomes immune to
depotentiation. Depotentiation is blocked by group II mGluR
LTD at Mossy Fiber Synapses antagonists, and group II agonists can substitute for low-
frequency stimulation in producing time-dependent depoten-
Homosynaptic LTD can be induced at mossy ber synapses in tiation (Chen et al., 2001a). Low-frequency stimulation can
vitro by continuous low-frequency stimulation (1 Hz for 15 induce heterosynaptic depotentiation in a neighboring mossy
minutes), at least in young animals (Battistin and Cherubini, ber pathway, perhaps as a result of glutamate spillover (Chen
1994; Kobayashi et al., 1996; Domenici et al., 1998; Tzouno- et al., 2001). In contrast to LTD, depotentiation can be readily
poulos et al., 1998). However, homosynaptic LTD in mossy elicited in adult animals (Huang CC and Hsu, 2001).
bers appears to be more readily induced by low-intensity, Chen et al. (2001a) have made the interesting observation
high-frequency stimulation than by prolonged low-frequency that depotentiation in the mossy ber pathway is not input-
stimulation, possibly reecting a requirement for the fre- specic. A low-frequency train to a recently potentiated mossy
quency-dependent co-release of dynorphin from mossy ber ber input leads to depotentiation of that input and of a sec-
terminals. Release of dynorphin can also lead to transient ond mossy ber input, not exposed to low-frequency stimula-
heterosynaptic depression (Weisskopf et al., 1993). In the tion, that was potentiated at the same time as the rst. This is
anesthetized rat, there is evidence that dynorphin acts postsy- in contrast to the input specicity of depotentiation that the
naptically to modulate both homosynaptic and heterosynap- same group found in area CA1 (Huang CC and Hsu, 2001).
tic LTD (Derrick and Martinez, 1996). Another study (Huang et al., 2002) provided evidence of the
As with mossy ber LTP, NMDA receptor antagonists do involvement of presynaptic mGluR2 receptors in heterosy-
not block mossy ber LTD, though group II mGluR antago- naptic depotentiation, activated perhaps by spillover from
nists are effective (Kobayashi et al., 1996; Tzounopoulos et al., activated bers. Although these experiments demonstrate the
1998; Chen et al., 2001a). Its induction is not affected by potential for erasing LTP in a nonspecic way in the mossy
clamping the postsynaptic cell at a hyperpolarized membrane ber pathway, it is important to emphasize once again that
potential or by loading the cell with BAPTA (Kobayashi et al., there is as yet no evidence that this occurs in the intact, behav-
1996; Domenici et al., 1998). Given the controversy surround- ing animal.
ing similar experiments addressing the mechanism of mossy
ber LTP, it may be premature to conclude that these results Mossy Fiber LTD Is Developmentally Regulated
provide conclusive evidence for a presynaptic induction
mechanism. Tzounopoulos et al. (1998) have presented evi- In slices from young rats (less than 14 days after birth),
dence for a presynaptic model of mossy ber LTD in which a another form of mossy ber LTD has been observed that can
decrease in the activity of PKA, triggered by a group II be induced by high-frequency stimulation (100 Hz for 1 sec-
mGluR-mediated decrease in adenyl cyclase activity, leads to a ond) and is independent of both NMDA and metabotropic
persistent decrease in release probability at mossy ber termi- glutamate receptors (Battistin and Cherubini 1994). The
nals. Two lines of knockout mice that lack presynaptic pro- induction of this form of LTD is blocked by postsynaptic
teins are further pointers to a presynaptic expression injection of BAPTA, but its expression remains presynaptic
mechanism for mossy ber LTD: One lacks the predominantly (Domenici et al., 1998). Intriguingly, mGluR-dependent LTD
presynaptic receptor mGluR2 (Yokoi et al., 1996) and shows induced by low-frequency stimulation and high-frequency-
impairment in LTD; the other, which shows enhanced LTD, induced LTD do not occlude, suggesting that both induction
lacks RIM, an active zone protein that binds to Rab3a and is and expression mechanisms are different in the two forms of
also a PKA substrate (Castillo et al., 2002). LTD (see Section 10.9.1).
A pairing-induced postsynaptic form of LTD has been
uncovered in the mossy ber projection to CA3 pyramidal LTD: Roles and Rules in the Intact Animal
cells. Postsynaptic depolarization combined with low-fre-
quency (0.33 Hz) stimulation of mossy bers causes persistent In the freely moving animal, mild stress promotes LTD and
depression that depends on Ca2 entry through L-type Ca2 inhibits LTP, though the effect is short-lasting (Xu et al., 1997).
channels in proximal dendrites of CA3 pyramidal cells (Lei et Exploration of a novel environment, on the other hand, can
al., 2003). That this is a novel form of mossy ber LTD is con- promote partial or total depotentiation for an extended
420 The Hippocampus Book

period. In one study, the effect was restricted to a time window one or other side of the synaptic partnership (Bischofberger
of more than an hour but less than 24 hours after induction and Jonas, 2002; Lawrence and McBain, 2003). In this section
(Xu et al., 1998). In another study in which a tetanus protocol we examine LTP and LTD at excitatory synapses onto
was adopted that produced a very long-lasting form of LTP, interneurons and at inhibitory synapses made by interneurons
depotentiation could be induced several weeks (but not onto excitatory and inhibitory neurons.
months) after induction by leaving rats for a few nights in a
rodent Disney World with endless possibilities for exploration 10.8.1 LTP and LTD at Glutamatergic
(Abraham et al., 2002). Synapses on Interneurons
What is the function of LTD and its partner depotentia-
tion? Theoretical analysis of distributed neural nets suggests The subunit composition of AMPA and NMDA receptors on
that a network whose synaptic elements are endowed with the interneurons differ in detail from that found on pyramidal
property of bidirectional plasticity, including heterosynaptic cells (see Chapter 5 and Chapter 6, Section 6.3). The domi-
plasticity, are more stable and have a larger storage capacity nant NMDA subunit is NR2B, the presence of which prolongs
than networks where only potentiation (or depression) is pos- NMDAR-mediated currents (McBain et al., 1999). The AMPA
sible (Willshaw and Dayan, 1990). It is usually presumed that receptor subunit GluR2, which in its edited form limits the
if the encoding of memories and semantic knowledge is calcium permeability of the AMPA receptor ion channel, is
achieved by changes in synaptic weights these changes must generally in low abundance at synapses on inhibitory
be stable, but models where this is not the case have been pro- interneurons, and AMPAR-mediated calcium uxes are corre-
posed (Abraham and Robins, 2005). Such models are even spondingly higher (Lawrence and McBain, 2003). The distinct
more dependent on mechanisms allowing easy reversibility of assemblage of glutamate receptor subtypes at interneuron
changes in synaptic weights. What then is the role of LTD? synaptic membranes is reected in the types of plasticity
Clearly depotentiation has the potential to disrupt memory observed; high-frequency trains can, depending on the proto-
circuits and cause active forgetting. What is the evidence that col or cell type, induce either no change, LTP (Maccaferri and
activity-dependent erasure is necessary for forgetting? To put McBain, 1996; Perez et al., 2001; Lapointe et al., 2003), or LTD
it another way, what sorts of forgetting, if any, could not be (McMahon and Kauer, 1997). Induction of LTP at excitatory
explained by the gradual decay of LTP? After all, forgetting on synapses on stratum oriens/alveus (OA) interneurons in stra-
demand is next to impossible. tum oriens is mGluR1-dependent and requires postsynaptic
One consequence of heterosynaptic LTD is to maintain rel- Ca2 entry; expression appears to be presynaptic (Lapointe et
atively constant the net excitatory drive onto target cells. The al., 2003). There are no reports of long-term changes pro-
observation that there is no increase in the mean ring rate of duced by low-frequency trains. However, LTD at feedforward
place cells following the induction of LTP is consistent with excitatory synapses on CA1 interneurons can be induced
this ideas (McNaughton et al., 1984; Dragoi et al., 2003). by high-frequency stimulation, as rst described at Schaffer-
In the laboratory there are two very different ways of elic- commissural inputs to interneurons in stratum radiatum
iting LTD: (1) with long low-frequency trains tuned to mGluR of area CA1 (McMahon and Kauer, 1997). This form of LTD
group I receptors or to NMDA receptors; and (2) by appro- appears not to be input-specic. Functional suppression,
priate pairings of presynaptic and back-propagating postsy- via LTD, of excitatory feedforward synapses onto interneurons
naptic spikes. There are several fundamental issues to be in stratum radiatum provides a potential mechanism for the
resolved here. How is the tuning realized mechanistically? phenomenon of E-S potentiation (see Section 10.3.12).
Do the endogenous patterns of activity provide anything A pairing strategy (depolarizing pulses combined with
like 1 Hz/900 pulse regimens that are required to obtain sus- theta-burst afferent stimulation) has been successfully
tained LTD? Does LTD or any of the other forms of persistent employed to induce LTP of excitatory afferents to interneu-
synaptic depression described in this section occur at all in the rons of stratum oriens in area CA1, but the same protocol is
hippocampus of the freely moving, behaving animal? The ineffective in the stratum radiatum (Perez et al., 2001). Thus,
nearest this question has come to receiving a positive answer the magnitude and polarity of activity-induced plasticity at
comes from the experiments in behaving animals discussed excitatory synapses on CA1 interneurons depends on the
above. stimulus protocol, the location of the interneuron, and pre-
sumably the distribution of its receptor subunits.
Mossy bers provide an interesting example of terminal-
specic plasticity. As we have seen, both LTP and LTD are
10.8 Synaptic Plasticity expressed presynaptically at mossy berCA3 pyramidal cell
and Inhibitory Pathways synapses. Filopodial extensions of mossy ber giant boutons
and mossy bers with en passant boutons make excitatory
In addition to plasticity at glutamatergic synapses on gluta- synapses on dendrites of CA3 interneurons. In stratum radia-
matergic principal cells, there is a growing body of data on tum of area CA3, tetanic stimulation of mossy bers induces
plasticity at synapses where a GABAergic neuron comprises LTD in interneurons expressing calcium-permeable AMPA
 Synaptic Plasticity in the Hippocampus 421

A B C
P<0.0005
100 LTP 100
50
0
n=33
-50 50
500 pyr
-15 0 15 30
300 mGluR-A 100 No LTP 0
int 50
100
0 0
n=42
-50 -50
0 10 20 30 40 -15 0 15 30 -100 -50 0 50 100
Time (min) Time (min) Spike timing (msec)
Figure 1020. Synaptic plasticity of glutamatergic inputs to hip- non-potentating interneurons are pollted in the lower panel.
pocampal interneurons and at GABAergic inputs from interneu- (Source: Lamsa et al., 2005.) C. Potentiation of GABAergic postsy-
rons. A. Tetanic stimulation of mossy bers produces LTP at CA3 naptic currents (GPSCs) recorded in dissociated hippocampal cul-
pyramidal cells (pyr) (lled circles) but a mean depression in a tures following repeated pairing of presynaptic and postsynaptic
population of interneurons (int) (open circles). Note that responses neurons. LTP of GABAergic transmission is input-specic; and
were blocked by an mGluR agonist (mGluR-A), consistent with unlike pairing-induced plasticity at glutamatergic synapses, it occurs
activation of mossy bers. (Source: Maccaferri et al., 1998.) B. With 20 ms either side of coincident presynaptic and postsynaptic activ-
perforated patch recording, input-specic LTP can be induced by ity. The effect is caused by an L-type calcium channel-dependent
pairing in a subpopulation of small feedforward interneurons in the local suppression of Cl
transporter activity. The resulting decrease
stratum radiatum of area CA1. EPSPs from Interneurons showing in the reversal potential for Cl
results in an increase in the GPSC
LTP are plotted in the upper panel (lled circles are responses from when the cell is clamped at a hyperpolarized potential. (Source:
paired input, open circles from unpaired input); responses from Woodin et al., 2003.)

receptors; and presynaptic mGluRs but not NMDARs are Two other examples of plasticity at excitatory synapses
required for its induction (Maccaferri et al., 1998; Laezza et al., have been identied at hippocampal interneurons. Lamsa et
1999) (Fig. 1020A). In further work from McBains group, al. (2005) observed LTP at feedforward excitatory synapses on
tetanus-induced LTP was observed in stratum lucidum a subset of interneurons (about half their sample of aspiny
interneurons expressing Ca2-impermeable GluR2-containing interneurons) in area CA1 using perforated patch recordings
AMPA receptors. The mechanism underlying LTD depends on to preserve the cytoplasmic integrity of the cell. The authors
whether there are calcium-permeable or calcium-imperme- suggest that LTP at this synapse is important for preserving
able AMPA receptors at that synapse; for both types of the temporal delity of pyramidal cell ring (Fig. 1020B).
synapse, the locus of induction is postsynaptic, although the McBain and his colleagues have identied a metabotropic
evidence points to presynaptic expression in the former case switch mediating bidirectional plasticity at synapses made by
and postsynaptic expression in the latter (Lawrence and mossy bers onto stratum lucidum interneurons (SLINs)
McBain, 2003). At the mossy berbasket cell connection in (Pelkey et al., 2005). The switch is operated by mGluR7 recep-
the dentate gyrus, LTP can be induced by high-frequency stim- tors, which in the resting state are present on presynaptic ter-
ulation paired with depolarization of the basket cell; LTD is the minals of mossy ber-SLIN synapses. Intense activity (such as
outcome of tetanic stimulation without depolarisation. a high-frequency train) activates the receptors and triggers a
Analysis of the number of failures, coefficient of variation, and sequence of events that leads to LTD, expressed as a reduction
paired-pulse modulation indicate a presynaptic locus for LTD in transmitter release from mossy berSLIN terminals. The
(Alle et al., 2001). A general conclusion from these studies is resulting decrease in feedforward inhibition promotes the r-
that the target cell controls the type and extent of plasticity at ing of CA3 pyramidal cells. Another consequence of the
mossy ber terminals. intense activation during the train is that mGluR7 receptors
An intriguing form of E-S potentiation (see Section become internalized; now, without the feedback suppression
10.3.12) at interneurons in the molecular layer of the dentate of transmitter release that they control, the mossy ber termi-
gyrus has been reported by Ross and Soltesz (2001). Tetanic nals onto SLINs display LTP when strongly activated, and
stimulation of the perforant path leads to a persistent decrease feedforward inhibition is restored. By this remarkable meta-
in membrane potential, which increases the probability of the plastic mechanism the balance of excitation and inhibition is
interneuron ring to subsequent stimuli without affecting the maintained in the CA3 network. These two sets of results
amplitude of the EPSP. The effect requires activation of Ca2- demonstrate a greater potential for plasticity in the inhibitory
permeable AMPA receptors and reects downregulation in circuits of the hippocampus than has been generally assumed
the activity of the electrogenic Na/K ATPase pump. (McBain et al., 1999).
422 The Hippocampus Book

10.8.2 LTP and LTD at GABAergic Synapses dependent. Remarkably, the same induction mechanisms that
lead to LTP of excitatory transmission also mediate potentia-
It might be expected that GABAergic synapses would not dis- tion of slow GABAergic responses in the same spines. Huang
play plasticity because they lack the equivalent of the voltage- et al. speculate that the temporal window for coincidence
dependent NMDA receptor that for most excitatory synapses detection is sharpened in spines equipped with the GABAB/
is an essential component of the inductive machinery. GIRK system because subsequent excitatory responses are
However, both LTP and LTD have been observed following held in check by the slow IPSC.
tetanic stimulation of GABAergic afferents to pyramidal cells To conclude this section, we note that to date there have
in area CA1, under conditions in which all ionotropic gluta- been no studies on plasticity at inhibitory synapses on granule
mate channels are blocked (Caillard et al., 1999; Shew et al., cells or at inhibitory synapses on interneurons.
2000). There is disagreement about whether the postsynaptic
cell is (Shew et al., 2000) or is not (Caillard et al., 1999)
involved in the induction of LTP, but the evidence points to a
presynaptic locus of expression. These experiments were car- 10.9 LTP and LTD in Development
ried out on animals 2 weeks old or less. In a study on hip- and Aging and in Animal Models
pocampal slices from 4- to 5-week-old rats, tetanic of Cognitive Dysfunction
stimulation resulted in LTD of GABAergic responses in CA1
pyramidal cells (Lu et al., 2000a). Here the effect appears to Long-term potentiation is often considered as if it were a uni-
have been induced heterosynaptically by co-activation of con- tary process. However, as documented elsewhere in the chap-
vergent excitatory bers, as induction was blocked by bath ter, various factors can greatly inuence the properties of LTP,
application of AP5 or by injecting a calcineurin inhibitor into such as the pathway under investigation, the phase of LTP
the postsynaptic cell. A novel form of heterosynaptic LTD at being studied, and the methods used for its induction.
inhibitory synapses in area CA1 has been discovered by Another critical factor is the age of the animal. It is likely that
Chevaleyre and Castillo (2003). Persistent reduction of evoked the predominant role for synaptic plasticity early in develop-
IPSCs was induced by brief high-frequency stimulation of ment is in the formation and renement of synaptic connec-
Schaffer-commissural bers, again under conditions in which tions. As animals develop, it can be assumed that LTP and
ionotropic glutamate receptors were blocked. The effect was other manifestations of plasticity are involved in modifying
abolished by antagonists of group 1 mGluRs and cannabinoid synaptic strength at existing synapses, but here too there could
CB1 receptors. Chevaleyre and Castillo (2003) concluded that be age-dependent alterations in the underlying mechanisms.
the sequence of events underlying this form of plasticity Juvenile forms of LTP may be less evident but may persist into
begins with mGluR-mediated release of the endogenous ago- adulthood in some circumstances, for example during reactive
nist 2-AG from stimulated pyramidal cells followed by bind- synaptogenesis in response to brain injury and in the dentate
ing of 2-AG to CB1 receptors on the terminals of inhibitory gyrus when newborn granule cells are incorporated into the
boutons. For further discussion of the results of Lu et al. network of mature granule cells (Schmidt-Hieber et al., 2004).
(2000a) and Chevaleyre and Castillo (2003) in the context of
E-S potentiation, see Section 10.3.12. 10.9.1 Hippocampal Synaptic
Woodin et al. (2003) uncovered an unusual form of spike Plasticity During Development
timing-dependent plasticity (STDP) (see Section 10.3.8) that
is independent of the order of the pre- and postsynaptic activ- LTP and the Postnatal Hippocampus
ity at inhibitory synapses onto hippocampal pyramidal cells.
Repeated pairings between the presynaptic action potential Because they have been the most extensively investigated,
and the back-propagating dendritic spike at intervals up to 20 more is known about age-dependent changes in LTP at CA1
ms on either side of coincidence result in long-term reduction synapses than elsewhere in the brain. Silent synapsesdened
in the GABAergic IPSC. The effect is triggered by Ca2 entry as synapses that have an NMDAR-mediated component of
through L-type Ca2 channels, leading to persistent down- synaptic transmission but no detectable AMPAR-mediated
regulation of the K/Cl
co-transporter, with a consequent component (Fig. 1021A)account for approximately 80%
buildup in the concentration of intracellular Cl
and a depo- of CA3CA1connections during the rst 2 days of postnatal
larizing shift in the Cl
equilibrium potential (Fig. 1020C). life in the rat, decreasing dramatically over the next week or so
An unsuspected mechanism for potentiation of the (Durand et al., 1996) (Fig. 1021B). By P5 to P6 they account
GABAergic slow IPSC has been revealed in dendritic spines of for half this number and continue to decline steeply thereafter.
CA1 pyramidal cells (Huang CS et al., 2005). The slow IPSC is The mechanism of LTP at these synapses involves the rapid
mediated by GABAB receptors and the G protein-activated unsilencing of silent synapses through the NMDAR-depend-
inwardly rectifying K (GIRK) channel, both of which are ent appearance of AMPAR-mediated responses. The simplest
located on spines as well as dendrites of CA1 neurons. Pairing mechanism to explain this is the rapid recruitment to the
of single stimuli with depolarization induces LTP of the synapse of AMPA receptors, though presynaptic mechanisms
slow IPSC that is both NMDAR-dependent and CaMKII- have also been proposed (Kullmann, 2003). A reappraisal of
 Synaptic Plasticity in the Hippocampus 423

the role of silent synapses may be needed following the obser- described at cortical synapses (Massey et al., 2004). This sug-
vation that synapses are not initially silent; rather, the gests that the mechanisms for LTD exist in adults but are less
AMPAR-mediated component disappears as a result of the readily accessible. These differences in the ability to induce
synaptic stimulation used to obtain baseline responses (Xiao LTD presumably reect age-related changes in induction
et al., 2004). This activity-dependent silencing is developmen- mechanisms. For example, in very young animals it seems that
tally regulated; it is evident at P3 to P12 but not at P29 to P32, mGluRs and voltage-gated Ca2 channels are important trig-
which could account for the developmental prole of silent gers (Bolshakov and Siegelbaum, 1994). In juveniles and
synapses. Whatever its origin and mechanisms, the unsilenc- adults, both NMDAR- and mGluR-dependent forms of LTD
ing of silent synapses is best regarded as a juvenile form of may coexist, though the relative contributions of each form
LTP; the extent to which it is involved, if at all, during LTP in may change with age (Kemp et al., 2000).
more mature animals or during reactive synaptogenesis is cur- Developmental shifts in the generation of LTD also occur
rently unknown. at mossy ber synapses, as noted in Section 10.7.8. Between
NMDA receptors are likely to be the primary trigger for the rst and second week of life it is possible to induce two
LTP induction throughout life but signaling molecules acti- forms of LTD (Domenici et al, 1998). One form is induced by
vated downstream of these receptors may change as the ani- a high-frequency train and is independent of both NMDARs
mal matures. We have already alluded in section 10.4.3 to the and mGluRs. The second form is induced by prolonged low-
developmental switch in the kinase dependence of E-LTP that frequency stimulation and depends on activation of mGluRs.
occurs between P7 and P8 when LTP is primarily PKA- The high-frequency form of LTD, however, is replaced by LTP
dependent, and P27 when LTP acquires its adult dependence in rats more than 2 weeks of age (Battistin and Cherubini,
on CaMKII (Yasuda et al., 2003) (Fig. 1021C). 1994).
Another study compared LTP mechanisms at CA1
synapses at two stages of development under otherwise iden- 10.9.2 Synaptic Plasticity
tical conditions (Palmer et al., 2004). It was found that LTP and the Aging Hippocampus
involved two mechanisms at P6. The more prevalent form was
exhibited by synapses that had an initially low Pr (and showed Changes in the molecular mechanisms of LTP are unlikely to
pronounced PPF); at these synapses LTP was expressed as an change radically during norml aging: LTP induction, for exam-
increase in Pr (and an associated decrease in PPF) (Fig. ple, still requires the activation of NMDA receptors (Barnes et
1021D). The second form displayed no signicant change in al., 1996). However, during aging there are alterations in the
Pr or PPF but was associated with a decrease in the mean uni- LTP process that could contribute to a decline in cognitive
tary conductance of AMPA receptors. As noted earlier this can function. In general, it can be concluded that aged rats have
most simply be explained by the insertion of additional recep- decits in both the induction and maintenance of LTP; how-
tors but with lower single-channel conductance than preexist- ever, these decits are complex and depend on the pathway
ing receptors. By P12, neither of these mechanisms was under investigation and the experimental protocol (reviewed
present. Instead, LTP was associated with either an increase or by Burke and Barnes, 2006). For example, in aged rats there is
no change in the mean unitary conductance, consistent with an increased rate of decay of LTP at perforant pathgranule
an increase in the single-channel conductance of preexisting cell synapses that correlates with the rate of forgetting (Barnes,
AMPA receptors (or, alternatively, an exchange of high for low 1979). A similar decit in the maintenance of LTP in the per-
conductance receptors) and the insertion of additional recep- forant path input to CA3 neurons has been reported (Dieguez
tors of the same unitary conductance as the preexisting recep- and Barea-Rodriguez, 2004). There is also an increase in the
tors, respectively. Regardless of the precise underlying induction threshold of LTP in aged rats at perforant
mechanisms, this study illustrates the diversity of LTP mecha- pathgranule cell synapses that is detectable with submaximal
nisms that can exist at a single type of synapse. The coexis- induction protocols (Barnes et al., 2000). At CA1 synapses, the
tence of more than one mechanism in the same preparation primary decit with aging seems to be a reduction in the mag-
adds further complexity and may reect the heterogeneity of nitude of LTP, which may be explained by less depolarization
the developmental state of synapses at any one age. during induction and presumably therefore less activation of
NMDA receptors (Deupree et al., 1991; Moore et al., 1993;
LTD and the Postnatal Hippocampus Rosenzweig et al., 1997; Tombaugh et al., 2002). Thus, it
appears that a major factor in age-related changes in LTP is
In addition to marked alterations in the properties and mech- alterations in conditions affecting NMDA receptor function
anisms of LTP during development, there are also signicant and its associated Ca2 signaling. In turn, alterations in NMDA
changes in LTD. At Schaffer-commissural synapses, LTD is receptor-dependent plasticity seem to account for many of the
often induced by a prolonged period of low-frequency stimu- alterations in the dynamics of neuronal assemblies that occur
lation (e.g., 900 pulses at 1 Hz). As noted in Section 10.7, LTD with aging. This is indicated by the ndings that blockade of
is more readily induced in juvenile animals than in adults. NMDA receptors in young rats results in ensemble dynamics
However, NMDAR-dependent LTD can be induced in adults that resemble those of old rats (Kentros et al., 1998; Ekstrom et
when glutamate uptake is blocked (Yang et al, 2005), as earlier al., 2001).
424 The Hippocampus Book

Figure 1021. Developmental changes in LTP. A. Early in develop- of the animal. Note the dramatic decrease in the incidence of silent
ment there are silent synapses that display NMDA but not AMPA synapses with age. (Source: Durand et al., 1996.) C. Sensitivity to the
receptor-mediated EPSCs. These can be rapidly unsilenced by an CaMKII inhibitor KN-93 increases with development. The upper
LTP induction protocol. The traces (left-hand side) show only graph shows that KN-93 has no effect on LTP at CA1 synapses in
failures at a negative membrane potential but NMDA receptor- slices obtained from P7 to P8 rats. However, KN-93 blocked the
mediated synaptic responses at a positive holding potential. The induction of E-LTP in slices obtained from rats at P27 and later.
traces (right-hand side) show only failures before but some AMPA (Source: Yasuda et al., 2003.) D. The expression mechanisms of LTP,
receptor-mediated EPSCs after pairing-induced LTP. The graph assessed as changes in paired-pulse facilitation, alter during devel-
plots the amplitude of EPSCs versus time. (Source: Liao et al., 1995.) opment. The upper graph plots the EPSC amplitude (upper) and
B. The incidence of silent synapses dramatically decreases with paired-pulse ratio (lower) before and following pairing induced LTP
development. The traces are EPSCs at negative and positive mem- (at the time depicted by open square) for a subclass of P6 neurons-
brane potentials at postnatal day 2 (P2) and P6. At P2 the synapse is termed P6(a). The lower graph plots the change in the paired-pulse
silent because only NMDA receptor-mediated responses are ratio following LTP for this group of synapses and for another sub-
observed. At P6 the synapse is no longer silent because both AMPA class of P6 neurons [P6(b)] and P12 neurons. Note the dramatic
and NMDA receptor-mediated responses are observed. The graph decrease in paired pulse facilitation that is specic for P6(a).
plots the incidence of silent synapses versus the developmental age (Source: Palmer et al., 2004.)
 Synaptic Plasticity in the Hippocampus 425

In aged animals it once again appears to be easier to induce even in the aged animal (Valastro et al, 2001), though a decit
LTD and depotentiation (Norris et al., 1996). in LTP has also been reported (Trommer et al., 2004).
Interestingly, a knock-in mouse expressing human ApoE4
10.9.3 Animal Models of Cognitive Decline showed enhanced LTP when young (Kitamura et al., 2004);
LTP dropped back to wild-type levels during adulthood and
Alzheimers Disease has not yet been examined in the aged knock-in mice.
The other major approach to investigating the role of A in
Memory loss is a prominent aspect of Alzheimers disease AD is to apply A exogenously. In most such studies, A has
(AD), and given the strong presumptive links between synap- been found to impair LTP. For example, A blocked both E-
tic plasticity and memory, it is not surprising that several LTP and L-LTP without affecting baseline transmission in
groups have asked whether synaptic plasticity is affected in both the dentate gyrus and area CA1 in vitro (Lambert et al.,
rodent models of the disease (reviewed by Rowan et al., 2003). 1998; Chen et al., 2000; Wang et al., 2002). Not all studies,
The diagnostic, postmortem feature of Alzheimers disease however, have reported decits in LTP following A applica-
in humans is the deposition of amyloid plaque in structures of tion. For example, LTP was enhanced in the perforant path
the temporal lobe (see Chapter 16). Whether plaque deposi- input to dentate gyrus (Wu et al., 1995a). There is also gener-
tion is the cause or a consequence of the disease has not been ally a block of LTP in vivo, though L-LTP appears to be more
resolved, but many believe that excessive levels of amyloid sensitive than E-LTP (Cullen et al., 1997; Freir et al., 2001). At
(A ) is the primary cause of the disease. A is a peptide frag- longer intervals after A injection impairment of baseline
ment that is cleaved from the amyloid precursor protein transmission also becomes apparent (Cullen et al., 1996;
(APP) by the action of secretases. In various familial forms of Stephan et al., 2001), resembling the situation in several of the
early-onset AD, there are mutations in APP that result in its transgenic mouse studies.
misprocessing, leading to deposition of A . Several mouse Interestingly, studies with A fragments suggest that the
lines have been created that overexpress these mutant forms of active forms responsible for the inhibition of LTP are soluble
APP. These mice display age-related cognitive decits and A oligomers (Walsh et al., 2002). Other studies have
amyloid plaque deposition. With respect to alterations in addressed the mechanisms responsible for the A impairment
synaptic function, two main effects have been noted. In some, of LTP and have uncovered interactions of A with the
but not all, studies there is a decrease in LTP (Chapman et al., cholinergic system (Auld et al., 2002), in particular nicotinic
1999; Moechars et al., 1999). In other studies the most dra- receptors and the ERK2 MAPK cascade (Dineley et al., 2001),
matic effect is a decrease in basal synaptic transmission (Hsia the GABA system (Sun and Alkon, 2002), NMDA receptors
et al., 1999; Larson et al., 1999; Fitzjohn et al., 2001a), which (Wu et al., 1995b), and CREB (Tong et al., 2001).
could indirectly affect LTP through the property of coopera- The rst, partially effective, drugs used to treat AD were
tivity (see Section 10.3.1). A decrease in LTP was also observed cholinesterase inhibitors, which exert their effect by potentiat-
in a mouse that overexpressed an amyloidogenic fragment of ing cholinergic function in regions such as the hippocampus.
the C-terminal of APP (Nalbantoglu et al., 1997). An alterna- Cholinergic systems facilitate NMDAR-dependent LTP, possi-
tive approach to investigating the role of APP in AD is to gen- bly in part by direct potentiation of the NMDA receptor con-
erate a mouse with null expression of APP. APP-null mice also ductance (see Section 10.6.1). It is feasible, therefore, that their
display LTP decits under some, but not all, conditions cognitive enhancing effect at the early stages of AD is due to
(Seabrook et al., 1999). In particular, no decit was seen when the facilitation of LTP. The next compound to be introduced
GABA inhibition was blocked, suggesting that the effect on into the clinic for the treatment of AD was memantine, which
LTP is mediated via a relative increase in synaptic inhibition is a weak NMDA receptor antagonist (Parsons et al., 1999). It
(Fitzjohn et al., 2001a). This raises the possibility that APP is may seem paradoxical that an NMDA receptor antagonist can
required for normal functioning of GABA-mediated synaptic have cognitive enhancing properties. However, a plausible
transmission. mechanism of action has been advanced based on studies on
A larger number of cases of familial AD have mutations in hippocampal LTP (Coan et al., 1989). It was found that
PS1, a component of the machinery involved in APP process- bathing hippocampal slices in a Mg2-free medium resulted
ing. Mice expressing PS mutations generally have enhanced in the inhibition of LTP. However, if a concentration of AP5
LTP (Parent et al., 1999; Barrow et al., 2000; Zaman et al., was added to the slices that would ordinarily block LTP, LTP
2000) possibly because Ca2 homeostasis is affected could be obtained. If, however, a higher concentration of AP5
(Schneider et al., 2001). In a conditional forebrain-specic was used, LTP was inhibited. These surprising ndings were
PS1 knockout mice there were no detectable changes in basal interpreted as follows. To induce LTP requires either no acti-
transmission or LTP (Yu et al., 2001), whereas in a mouse vation or perhaps a very low level of NMDA receptor activa-
underexpressing PS1 there was a decit in LTP evoked with tion during baseline stimulation but then a high degree of
multiple tetani (Morton et al., 2002). In humans, the ApoE4 NMDA receptor activity briey during a tetanus. When Mg2
isoform of the lipid-shuttling protein apolipoprotein E was omitted from the bathing medium, the baseline level of
(ApoE) is a risk factor for AD. A mouse with the ApoE gene NMDA receptor activation both before and following the
knocked out showed near-normal LTP (Anderson et al., 1998) tetanus was greatly enhanced; this inhibited the LTP process,
426 The Hippocampus Book

probably by preinduction and partial saturation of LTP. The stages. However, before the onset of classical symptoms,
addition of a standard concentration of AP5 substituted for patients often exhibit cognitive decits. Studies in various
Mg2and suppressed NMDA receptor activation. During the mouse models of Huntingtons disease have reported decits
tetanus, however, the concentration was insufficient to block in LTP at Schaffercommissural synapses (Usdin et al., 1999;
the induction of LTP given the absence of Mg2. That a higher Hodgson et al., 1999; Murphy et al., 2000). Strikingly, and in
concentration was effective demonstrates that it is NMDAR- contrast to control animals, slices from adult mutant mice
dependent LTP that is recovered by AP5 treatment. What is readily displayed NMDA receptor-dependent LTD and depo-
the relevance of this to disease? In various pathological states, tentiation at Schaffercommissural synapses (Murphy et al.,
such as AD, the general dysfunction could lead to depolariza- 2000). Expression of the full-length mutated huntingtin in cell
tion of neurons, for example due to problems with energy lev- lines results in augmentation of NR2B containing NMDA
els affecting glutamate uptake. Indeed, blocking glutamate receptors (Chen et al., 1999), which might explain the
uptake produces a similar NMDA receptor-dependent inhibi- enhancement of LTD (see Section 10.7.3). Mossy ber LTP is
tion of LTP (Katagiri et al., 2001). Memantine is a weak, high also reduced in a mouse model of Huntingtons disease
voltage-dependent NMDA receptor antagonist (Parsons et al., (Gibson et al., 2005). In this study a similar LTP decit was
1993) that can also restore the ability of slices bathed in Mg2- observed in mice lacking complexin II, a protein that is
free medium to undergo LTP (Frankiewiez and Parson), reduced in Huntingtons disease patients.
1999). So its probable mechanism of action is to inhibit spu-
rious activation of NMDA receptors in modestly depolarized Schizophrenia
cells; but during strong depolarization (e.g., during a tetanus)
it is rapidly displaced from NMDA receptors. In this way it can Schizophrenia is another developmentally regulated disorder
facilitate LTP. In addition, its ability to inhibit tonic activation that involves cognitive impairment as well as psychosis. There
of NMDA receptors could endow it with neuroprotective are good reasons to believe that schizophrenia involves patho-
properties (Danysz et al., 2000). logical alterations in synaptic plasticity, in particular LTP-like
processes. Thus, one of the best animal models of schizophre-
Downs Syndrome nia involves injection of phencyclidine (PCP). It is a curious
coincidence that in the same year that the NMDA receptor
The segmental trisomy mouse (Ts65Dn) has an extra chromo- was established as the trigger for LTP at Schaffer-commissural
some 16 (corresponding to human chromosome 21) and has synapses (Collingridge et al., 1983), it was also shown that
accordingly found use as a model for Downs syndrome. The PCP and other dissociative anesthetics such as ketamine are
Ts65Dn mouse shows learning impairments and a reduction potent blockers of NMDA receptors (Anis et al., 1983) and
in LTP at CA1 synapses in vitro (Siarey et al, 1997). In a subse- that PCP blocks induction of LTP at these synapses (Stringer
quent study, a decit was found in theta-burst LTP but not in et al, 1983). Since then, the glutamate hypothesis of schizo-
tetanus-induced LTP; the decit was rescued by blocking phrenia has gained in popularity and is now the dominant
GABA inhibition (Costa and Grybko, 2005). A reduction in hypothesis in the eld. The general notion is that schizophre-
LTP in the dentate gyrus of the Ts65Dn mouse that was res- nia may involve impairment of NMDAR-dependent LTP in
cued by blockade of GABA inhibition has also been observed the hippocampus and other brain regions affected by the dis-
(Kleschevnikov et al, 2004). There is also a decit in LTP in the ease, thereby leading to impairment of LTP, cognitive dysfunc-
Ts1Cje mouse model of Downs syndrome (Siarey et al, 2005). tion, and thought disturbances. Evidence to support the
In this mouse both the cognitive decits and the LTP decits hypothesis that LTP may be affected in schizophrenia is start-
are less severe. Whereas LTP is reduced in mouse models of ing to accumulate. For example, in the post-weaning social
Downs syndrome, LTD is increased (Siarey et al., 1999, 2005). isolation model of schizophrenia there is a reduction in
A mouse expressing most of human chromosome 21, in addi- NMDAR-dependent LTP in the CA1 projection to the subicu-
tion to both copies of the homologous mouse chromosome lum (Roberts and Greene, 2003). There are many ways in
16, shows learning decits and a reduction in LTP in the den- which NMDA receptor-dependent LTP could be directly or
tate gyrus in vivo (ODoherty et al., 2005). Thus, all mouse indirectly affected, and it seems likely that genetic and envi-
models of Downs syndrome so far studied have displayed ronmental inuences can modify synaptic plasticity in various
alterations in LTP or LTD, consistent with the hypothesis that ways. For example, the schizophrenia-susceptibility gene
decits in hippocampal synaptic plasticity may play a causative neuregulin-1 stimulates the internalization of GluR1-contain-
role in the cognitive decits associated with the condition. ing AMPA receptors and leads to depotentiation of NMDAR-
dependent LTP (Kwon et al., 2005).
Huntingtons Disease
Diabetes and Other Diseases Associated
Huntingtons disease is a progressive neurodegenerative disor- with Potential Cognitive Decits
der associated with multiple CAG repeats in the gene encod-
ing the protein huntingtin; it usually starts around the third to Cognitive decits associated with type I diabetes may also
fth decade of life. Neuronal loss is primarily in the striatum involve alterations in hippocampal synaptic plasticity. For
and cerebral cortex but can involve the hippocampus at later example, there is an impairment of LTP at Schaffercommis-
 Synaptic Plasticity in the Hippocampus 427

sural synapses in nonobese diabetic rats and upregulation of plasticity, in particular changes in NMDAR-dependent LTP
NMDA receptors (Valastro et al., 2002). Similarly, rats treated and/or LTD. It seems likely, therefore, that synaptic plasticity
with streptozotocin, a pancreatic beta cell toxin, have in the brain of the type represented by LTP and LTD is
impaired LTP and enhanced LTD at Schaffercommissural disrupted in these diseases and contributes to the etiology of
synapses (Artola et al., 2005). In contrast, LTP appeared nor- cognitive impairment. By extension, it seems likely that alter-
mal at these synapses in Zucker diabetic fatty rats, a model of ations in LTP and LTD in the hippocampus and other brain
type II diabetes (Belanger et al., 2004). regions may be involved in other neurological and psychiatric
disorders, including drug addiction (Hyman and Malenka,
Epilepsy 2001) anxiety, depression, bipolar disorders, motoneuron dis-
ease, and Parkinsons disease.
A link between epilepsy and LTP has long been suspected, but
the evidence to date is not compelling. In two animal models
of human temporal lobe epilepsykindling and pilocarpine-
induced seizuresLTP is reduced, probably because synapses 10.10 Functional Implications
become potentiated during the seizures (Schubert et al., of Hippocampal Synaptic Plasticity
2005). LTP is also impaired in slices prepared from human
epileptic dentate gyrus relative to nonepileptic tissue (Beck et What is the function of LTP? Is it a physiological phenomenon
al., 2000). Kindling is induced in temporal lobe structures, that engages the same neural mechanisms that are responsible
most readily in the amygdala, by daily repetitions of brief for certain forms of learning and memory? Or is it a labora-
high-frequency sinusoidal stimulation, typically at 60 Hz for 1 tory curiosity of no functional signicance? A version of these
second. Initially, stimulation produces no behavioral effects, questions was posed at the end of the rst article on LTP (Bliss
but eventually, after several daily repetitions, the same proto- and Lomo, 1973), and the debate has been with us ever since.
col leads to generalized convulsions (Goddard, 1967). Is there In this concluding section of the chapter, we turn to the func-
an LTP component to kindling? Cain (1989) noted several dif- tional signicance of activity-dependent synaptic plasticity in
ferences between LTP and the pathological processes elicited relation to behavior.
by kindling stimulation. Thus, the development of kindling
may be delayed but is not blocked by NMDA receptor antag- 10.10.1 Synaptic Plasticity
onists. During later stages of kindling, evoked responses and Memory Hypothesis
are often diminished rather than enhanced. Moreover,
repeated tetanic stimulation of afferent pathways of the kind Most functional studies of LTP and LTD address their possible
that induces LTP does not elicit kindling. The kinetics of onset roles in learning, with the majority of studies focusing on
and persistence are also very different in the two cases; kin- NMDAR-dependent forms of plasticity. The aim of such
dling requires several daily sessions to reach full expression research has been caricatured by Stevens (1998) in the ques-
and is then effectively permanent. LTP is induced within tion Does LTP equal memory? There are numerous quali-
seconds or minutes but in most cases returns to baseline cations to such a question. What type of LTP? What type of
within days or weeks. Thus, it seems reasonable to conclude learning? What role for LTD and depotentiation? Is the per-
that the sorts of stimulation that lead to LTP are not in sistence of LTP correlated with and responsible for the per-
themselves epileptogenic, and that LTP lies at the benign end sistence of memory? These questions indicate that the debate
of the spectrum of activity-dependent changes in synaptic has moved on from merely recognizing certain analogies
efficacy. between LTP and learning to specic ideas about how the
mechanisms of induction and expression of synaptic plastic-
Synaptic Plasticity and the Enhancement ity relate to the multiple types and processes of memory that
or Suppression of Memory we now know to exist (see Chapter 13). These ideas share a
common generic theme that has been called the synaptic
Increasing condence in the existence of a causal link between plasticity and memory (SPM) hypothesis (Martin et al.,
LTP and memory has encouraged the development of phar- 2000). The SPM hypothesis asserts that synaptic plasticity (1)
maceutical approaches to the treatment of memory dysfunc- occurs during ordinary brain activity, and (2) is responsible
tion based on targets such as glutamate receptors and for memory formation, as follows (Martin et al., 2000).
transcription factors known to be important in LTP and/or
Activity-dependent synaptic plasticity is induced at
LTD. Though still at a relatively early stage of development,
appropriate synapses during memory formation, and
these approaches hold great promise for the future (see
is both necessary and sufficient for the information
reviews by Cooke and Bliss, 2005 and Lynch, 2006).
storage underlying the type of memory mediated by
the brain area in which that plasticity is observed.
Summary
With respect to the hippocampus, this implies that some of
A common emerging theme is that a number of diseases the neural mechanisms of synaptic plasticity discussed above,
involving cognitive decits have altered hippocampal synaptic such as alterations in AMPA receptor trafficking, should occur
428 The Hippocampus Book

during hippocampus-dependent forms of learning and be ing various forms of associative conditioning. This has some-
necessary for that learning to occur. times been mentioned as a weakness of the SPM hypothesis
Earlier in the chapter, it was pointed out that synaptic plas- (Diamond and Rose, 1994). Associative conditioning depends
ticity displays physiological properties that are highly sugges- on the ability of the CS to predict the US (contingency); sim-
tive of an information storage device, as many have asserted ply pairingCS and US (contiguity)is not sufficient if, for
the hippocampus to be (McNaughton, 1983; Lynch and example, many additional unpaired CS and US presentations
Baudry, 1984; Goelet et al., 1986; Morris et al., 1990; Bliss and are also introduced. The studies of spike timing plasticity (see
Collingridge, 1993; Barnes, 1995; Jeffery, 1997; Shors and Section 10.3) illustrate the temporal contiguity requirements
Matzel, 1997). Such classical properties include associativity, for LTP, but to date they say nothing about contingency.
input-specicity, and persistence. These may be relevant to However, an investigation of amygdalar plasticity reveals that
associative features of learning and memory, as follows. First, LTP and fear conditioning are indeed both sensitive to CS-US
associative induction implies the capacity to relate a pattern of contingency (Bauer et al., 2001). A train of tetanic stimulation
presynaptic neural activity with a pattern of postsynaptic of thalamic afferents to the lateral amygdala in vitro was
activity, as in the association of ideas. Second, they are relevant paired with a series of depolarizing current pulses to the post-
to storage capacity because a synapse-specic mechanism synaptic neuron. This procedure was repeated 15 times at 20-
endows the network with greater storage capacity than would second intervals, resulting in robust LTP. Intriguingly, the
changes in cell excitability. Third, to store information in the addition of unpaired depolarization 10 seconds after each
brain, some persistent change in the nervous system has to pairinganalogous to unpaired US presentationsresulted
occur. Whether these changes necessarily have to last as long in almost no LTP despite the preservation of contiguity. The
as memory is debatable. A qualication is that hippocampal- mechanisms underlying this contingency phenomenon
neocortical consolidation mechanisms may become engaged clearly operate over a time scale of at least 10 seconds. These
that would obviate the need for long-lasting persistence in the ndings illustrate that LTP, at least in the lateral amygdala,
hippocampus, notwithstanding the fact that LTP in the den- reects a property of learning that might have been expected
tate gyrus has been shown to last for more than a year in to arise only at the circuit level.
rodents (Abraham, 2003; Abraham and Robins, 2005). Synaptic plasticity does, of course, have distinct functions
The pioneering studies implicating LTP in memory were in different parts of the nervous system. In the amygdala,
correlational. Barnes (1979) and Barnes and McNaughton it has been implicated in conditioned fear (LeDoux, 2000),
(1985) observed, in the course of work on aging, that the per- thereby reecting a feature of the generic SPM hypothesis
sistence of LTP over time was statistically correlated with the (type of memory mediated by the brain area). However, in
rate of learning and/or the degree of retention of spatial mem- the spinal cord, LTP-like changes have been implicated in
ory in a circular arena task (see Chapter 13 for a description hyperalgesia to painful stimuli (Fitzgerald, 2005). This illus-
of this and other behavioral tasks). Similar correlations have trates one of the themes of this bookhow principles that
been observed many times, another example arising in studies were rst developed for the hippocampus have had a wider
of the overexpression of human mutant amyloid precursor signicance in neuroscience. LTP and LTD may also have
protein (hAPP) in a murine model of Alzheimers disease other functions in the hippocampus. Proposals include
(Hsiao et al., 1996). This is associated with an age-related the plasticity/pathology continuum hypothesis (McEachern
decline in performance in a delayed spatial alternation task and Shaw, 1996) and the notion that synaptic changes
that is correlated with a corresponding decline in LTP, assessed play a role in attentional rather than memory processes (Shors
in vivo and in vitro (Chapman et al., 1999). Lynch (2004) and Matzel, 1997). Distinguishing between these and alter-
reviewed a number of other cases where similar correlations native hypotheses about the functions of synaptic plasti-
are observed. city is not always easy, and exacting analytical studies are
However, it is far from clear that the physiological proper- required.
ties of synaptic plasticity are, in general, likely to be exactly Martin et al. (2000) proposed four criteria for a rigorous
homologous to, and thus correlated with, parameters of learn- test of the SPM hypothesis: detectability, anterograde alter-
ing at the behavioral level. Some properties should be directly ation, retrograde alteration, and mimicry. Broadly speaking,
related. One example is that the mechanisms of induction of they correspond to correlation, necessity, and sufficiency. The
LTP and the encoding of memory should overlap. Other rst of these four criteria, detectability, states that, in associa-
properties are less likely to be correlated because the overt tion with the formation of memory lasting any length of time,
manifestations of memory are not due solely to synaptic prop- LTP or LTD must occur at certain synapses in one or more
ertiesthey also depend on the properties of the network in brain areas and should, in principle, be detectable. The limited
which that plasticity is embedded (see Chapter 14). An exam- number of synapses that change with any one learning expe-
ple of mismatch is that the temporal contiguity requirements rience might make this criterion difficult to meet experimen-
for presynaptic glutamate release and postsynaptic depolar- tally. The possibility of heterosynaptic LTD in conjunction
ization in the induction of NMDAR-dependent LTP (see with learning-induced potentiation poses a similar problem.
Section 10.3) are temporally much tighter than those govern- The second criterion, anterograde alteration, asserts that
 Synaptic Plasticity in the Hippocampus 429

blocking the mechanisms that induce or express changes in radiant heat was sufficient to induce changes in fEPSPs. If
synaptic weights should have the anterograde effect of impair- during exploration the animals brain was temperature-
ing new learning. The intervention may be achieved in various clamped by intermittent infrared heating, exploratory motor
ways: physiologically, pharmacologically, or through molecu- activity was no longer associated with changes in the fEPSP.
lar-genetic manipulation. The NMDA receptor has been a Moser et al. (1993b, 1994) went on to work out the calibration
particular target, but there is a growing body of work examin- functions relating brain temperature to fEPSP magnitude
ing interventions that affect the transition from protein syn- before placing animals in an environment containing six land-
thesis-independent to protein synthesis-dependent forms of marks. A small temperature-independent component of the
LTP and of memory. A separate aspect of necessityour third increase in fEPSP was observed that was correlated with
criterion, retrograde alterationis that altering the pattern of exploration (Fig. 1022B). This increased rapidly at the start
synaptic weights after learning has taken place should affect of exploration and declined gradually to baseline over a short
an animals memory of past experience. This arises because period (15 minutes). Further studies have shown exploratory
altering the spatial distributon of synaptic weights within a behavior to be associated with the activation of immediate
distributed associative matrix such as the hippocampus early genes (IEGs) such as c-fos (Gall et al., 1998) and Arc
should alter the associations retrieved from such a network. (Guzowski et al., 1999). Temperature controls have not, to our
Finally, if changes in synaptic weight are the neural basis of knowledge, been conducted in these studies, but these pat-
trace storage, their articial induction in a specic spatial pat- terns of IEG activation are unlikely to be entirely due to activ-
tern of weights should give rise to an apparent memory ity-associated changes in brain temperature, as the spatial
(mimicry) for a stimulus or event that did not, in practice, distribution of Arc expression in area CA1 over two explo-
actually occur. Although this criterion may seem fanciful, par- ration sessions is exquisitely sensitive to the spatial similarity
ticularly in the hippocampus, it is the ultimate engineering of the two enclosures in which testing takes place (Guzowski
criterion that is most likely to convince skeptics of the func- et al., 1999).
tional importance of LTP. For reasons we shall come to, it may Complementing these studies of fEPSPs in behaving ani-
be easier to meet this criterion in brain areas such as the mals have been studies comparing brain tissue (in vitro) from
amygdala than in the hippocampus. animals that have been exposed to a learning situation with
tissue from animals that have been left unattended. For exam-
10.10.2 Detectability: Is Learning ple, Green and Greenough (1986) saw enhanced fEPSPs in the
Associated with the Induction of LTP? dentate gyrus in response to perforant path stimulation in
slices taken from adult animals that had been reared in com-
The SPM hypothesis requires that synaptic changes must plex environments. This study was particularly well controlled
occur during learning. The experimental design is ostensibly in showing no changes in antidromic potentials or presynap-
straightforward: Synaptic efficacy is compared before and tic ber volleys. However, it is possible that experience can
after a variety of learning experiences. The prediction is that a alter neurogenesis in the dentate gyrus (see Chapter 9). Such
persistent increase in synaptic efficacy should occur at appro- changes could contribute to or account for the ndings of
priate synapses in association with certain types of learning. Green and Greenough (1986); enhanced fEPSPs could reect
Types of learning that engage the hippocampal formation the creation of new neurons and synapses.
should be associated with synaptic potentiation, whereas An increase in glutamate release has also been reported fol-
other types of learning should not. What is less clear is lowing both LTP (see Section 10.3) and watermaze learning
whether such synaptic changes would be readily detectable in (Richter-Levin et al., 1995; McGahon et al., 1996). Richter-
global measures such as fEPSPs. Levin et al. (1997) trained rats in a spatial watermaze task for
varying numbers of trials and then induced LTP in vivo on
Exploration-induced and Spatial Learning- one side of the brain; nally, they examined veratridine-
induced Changes in Hippocampal Field induced glutamate release in synaptosomes prepared from the
Potentials and Transmitter Release hippocampus of the trained animals. Learning was associated
with increased glutamate release. Tissue prepared from the
Considerable excitement surrounded the discovery of what hemisphere in which LTP had been induced showed a greater
appeared to be striking short-term modulation of perforant- increase in glutamate release when taken from rats at an early
path evoked EPSPs in the dentate gyrus during spatial explo- stage of training than at a later stage. Thus, not only were both
ration (Sharp et al., 1989; Green et al., 1990). Exploratory LTP and learning associated with an increase in glutamate
activity was accompanied by an increase in the dentate fEPSP release, but the learning-associated increase occluded the
and a decrease in both amplitude and latency of the popula- increase normally seen after LTP. This striking result would be
tion spike. However, Moser et al. (1993a) discovered that this in accord with the SPM hypothesis were it not for the further
unusual pattern of electrophysiological change during explo- nding that the amount of perforant path-induced LTP in
ration is largely due to changes in brain temperature caused vivo was the same in both the undertrained and extensively
by associated muscular activity (Fig. 1022A). Application of trained groups. This is puzzling. That the magnitude of elec-
430 The Hippocampus Book

trophysiologically induced LTP is unaffected by prior spatial unpaired conditioning. Use of a plasticity-block vector
learning is consistent with the idea that an individual learning (Malinow and Malenka, 2002) to interfere with AMPA recep-
experience may enhance only a small proportion of synapses tor trafficking prevented the learning-associated rectication
in the dentate gyrus. However, if this is the case, why was the of AMPA-mediated transmission and, in separate experi-
LTP seen after extensive learning not associated with an ments, learning itself. Intriguingly, plasticity had to be dis-
increase in glutamate release? Perhaps learning is associated turbed in only a small percentage of neurons (approximately
with a shift in the relative expression of presynaptically medi- 27%) for an effect to be observed, raising the possibility that
ated and postsynaptically mediated LTP. combinatorial effects may be responsible for the learning of
One puzzle about detectability concerns the difficulty in different CS-US pairings.
oberving learning-associated synaptic changes in the hip- A phenomenon resembling LTP also occurs in the primary
pocampus. Storage capacity could be one factor, particularly in motor cortex (M1) following acquisition of a motor skill in
a distributed-associative system such as the hippocampus; rats. Rioult-Pedotti et al. (1998) trained rats to reach through
another could be that heterosynaptic depression (LTD) may a hole in a small plastic box with their preferred paw to
occur at other synapses during learning, rendering eld poten- retrieve food pellets. After a few days of training, recordings
tials an inappropriate way to investigate the issue. Heterosyna- from brain slices revealed that EPSPs from layer II/III of the
ptic LTD might serve a normalizing or scaling function in a contralateral M1 of trained rats were substantially larger than
distributed associative memory system by ensuring that the those recorded in the ipilateral untrained hemisphere or in
sum of the synaptic weights on any given neuron remains untrained animals (Fig. 1022D). Similar LTP-like changes
roughly constant (Willshaw and Dayan, 1990) see Section have since been observed in M1 following motor learning in
10.3.14; fEPSP amplitude would then remain unchanged. vivo (Monls and Teskey, 2004). In a follow-up study,
Given this, the issue might be better pursued in other brain repeated trains of either high- or low-frequency stimulation
structures or with other techniques. were delivered to saturate LTP or LTD, respectively (Rioult-
Pedotti et al., 2000). Strikingly, the capacity for further LTP
Learning-related fEPSP Changes in Other Brain Areas was reduced (and LTD enhanced) following motor learning
despite the synaptic potentials being largerreecting no
Vindicating this suspicion, the detectability criterion has been change in the synaptic plasticity range of the synapses affected
spectacularly upheld in studies of amygdala-dependent fear by or involved in motor learning. This partial occlusion sug-
conditioning. Rogan et al. (1997) monitored evoked potentials gests that LTP and skill learning may engage similar neuronal
elicited by an auditory CS in the lateral amygdala in vivo mechanisms. However, a puzzle about these ndings is that
before and after auditory fear conditioning. Paired presenta- the increase in EPSPs is typically large (about 50%), and the
tions of the auditory CS and foot shock resulted in increased spatial distribution of the weight changes and their informa-
freezing behavior and a parallel potentiation of the CS-evoked tion content are currently unknown. We argued above that, in
potential. Enhancement of electrically evoked responses was the hippocampus at least, learning-related changes in EPSPs
also found in brain slices taken from fear-conditioned rats are likely to be extremely small. This may not be true in the
(McKernan and Shinnick-Gallagher, 1997), and an increase in motor cortex, but the result is surprising nonetheless. It is not
auditory evoked responses following tone fear conditioning clear whether such changes actually encode the learned skill
has since been reported in freely moving mice (Tang et al., thus constituting a motor engramor serve some other, less
2001, 2003). Learning-induced changes in evoked responses specic role in the information processing that accompanies
occlude electrically induced LTP, implying that the two motor learning. Would similar fEPSP increases be seen for
processes engage common mechanisms (Tsvetkov et al., 2002; each new skill that is learned? Alternatively, might such a large
Schroeder and Shinnick-Gallagher, 2005); such changes can change occur only when the rodent motor cortex is engaged in
also be persistent, lasting at least 10 days in the latter case. A skill learning for the rst time?
recent study offers further insights into the mechanisms of Learning-induced changes in evoked responses have also
learning-related plasticity, providing more direct evidence been observed in several other brain areas. For example,
that the learning-related changes are mechanistically similar increased fEPSPs have been observed in the piriform cortex in
to LTP (Rumpel et al., 2005). Using the rectication of AMPA conjunction with olfactory learning (Roman et al., 2004;
receptor-mediated transmission as an index of mutated Sevelinges et al., 2004); and enhancement of LTP accompa-
GluR1 subunit incorporation, as in the studies of LTP expres- nied by facilitation of LTD has also been reported (Lebel et al.,
sion described above (see Section 10.4), tone-fear condition- 2001). Lasting potentiation of transmission at the parallel
ing was shown to be associated with an increase in EPSCs berPurkinje cell synapse of the cerebellum has been
associated with AMPA receptor trafficking in the pathway observed following auditory fear conditioning (Sacchetti et
from the auditory thalamus to the amygdala in vitro (Fig. al., 2004), and induction of an LTP-like effect during learning
1022C). The change was greater when paired conditioning has been observed in the connections of the mbria to the lat-
trials were used in which the tone preceded the fear-evoking eral septum (Jaffard et al., 1996). Synaptic changes are also
shock by an optimum CSUS interval but much weaker after implicated in the experience-dependent reorganization of
 Synaptic Plasticity in the Hippocampus 431

A Exploration (STEM) Heating only B Exploration, with heating subtracted

0.05

Brain temperature (C)

Temp
Brain temperature (C)
0

(C)
3
EPSP slope (mV/ms)

0.05
1.10

Max fEPSP slope

2
1.05
1
1.00

(ratio)
0 0.95

C AMPA receptor trafficking induced by learning D Induction of LTP by motor learning

Paired Trained hemisphere Untrained hemisphere
conditioning 3hrs
36hrs record record
In Vivo 2-6hrs In vitro stimulate stimulate
infection Test 3hrs recording
fEPSP
: fEPSP:
Unpaired
conditioning
M1 M1
0.5 mV

* *
Paired 2.3
*
wm wm
46 10 ms
Rectification index

2.1
Trained hemisphere Untrained hemisphere
42
1.9
synaptic strength

33 + saturated LTP + saturated LTP

31
1.7
50pA 25pA
baseline baseline
25ms 25ms
1.5
GluR1 Non. GluR1 Non.
Paired Unpaired saturated LTD saturated LTD

Figure 1022. Detectability: induction of LTP-like changes in asso- the amygdala is associated with enhanced AMPA receptor traffick-
ciation with learning? A, B. Early studies that explored whether ing, measured using rectication of AMPA receptor-mediated trans-
learning is associated with increased fEPSPs revealed some con- mission as an index of mutated GluR1 subunit incorporation. See
straints on this approach. For example, novelty exploration is asso- text for discussion. (Source: After Rumpel et al., 2005.) D. Motor
ciated with an increase in both fEPSPs and brain temperature. learning restricted to one limb in rodents (food-pellet collecting) is
Control studies established that in most cases the temperature associated with an increase in the size of fEPSPs in layer III of the
change could explain some but not all of the change in fEPSPs that contralateral motor cortex, and this increase occludes the extent of
were at rst thought to be due to learning-associated LTP induc- subsequent LTP induction relative to the opposite hemisphere.
tion. A short-lasting temperature-independent component can be (Source: After Rioult-Pedotti et al., 1998, 2000.)
detected. (Source: Moser et al., 1993a, 1994.) C. Fear conditioning in

cortical representations (e.g., Weinberger, 2004), but a trace eyeblink conditioning (Gruart et al., 2006). In a follow-
detailed treatment of this topic is beyond the scope of this up to their original study, Sacchetti et al. (2002) found that, as
book. in M1 and piriform cortex, learning results in partial occlusion
of subsequent LTP only when assessed up to 1 day after train-
Associative Conditioning and Hippocampal fEPSPs ing; the increase in fEPSPs was previously found to last for 7
days. Perhaps synapse formation occurs in tandem with the
Despite the difficulty of detecting hippocampal LTP-like phe- consolidation of existing memory trace, reestablishing the
nomena after spatial learning, lasting changes in population capacity for normal LTP after several days while preserving
measures of hippocampal activity have been observed after established changes in synaptic strengthalthough metaplas-
certain forms of conditioning (see Weisz et al., 1984). For tic chances might provide an alternative explanation
example, an increase in the slope of the CA1 fEPSP has been (Abraham and Tate, 1997; cf. Moser, et al., 1993a; Lebel et al.,
reported in rat hippocampal slices prepared after contextual 2001). Consistent with other early work of Moser et al. (1993b,
fear conditioning (Sacchetti et al., 2001) and in freely moving 1994), mere exploration of a novel context without shock
mice after auditory fear conditioning (Tang et al., 2003) and resulted in an increase in the evoked response and partial
432 The Hippocampus Book

occlusion of LTP in slices prepared 10 minutes after explo- novo LTD, provide the counterbalance. But perhaps the most
ration but not at later time points. These ndings conrm that provocative question to arise from this study is this: Given that
exploration alone is insufficient to cause lasting potentiation some electrodes detect potentiation, why dont they all? The
of hippocampal fEPSPs (Sacchetti et al., 2001). nding implies that synaptic weight changes are not uni-
The less than compelling literature on detectability in the formly distributed within the network, and suggests instead a
hippocampus considered so far received a major boost from currant bun model of memory, in which representations are
two papers published in 2006. In the rst study, Gruart et al. stored in clusters of potentiated cells that are relatively far
(2006) implanted stimulating and recording electrodes in area apart compared to the inter-electrode distance (250 m) (see
CA1 of the mouse hippocampus, and monitored Schaffer- commentary by Bliss et al., 2006).
commissural responses while animals were trained in a trace
eyeblink conditioning task, using an airpuff as the uncondi- 10.10.3 Anterograde Alteration:
tioned stimulus, and a tone as the conditioned stimulus. Trace Do Manipulations That Block the
conditioning is a hippocampus-dependent form of classical Induction or Expression of
conditioning in which a delay is imposed between the end of Synaptic Plasticity Impair Learning?
the CS and the onset of the US. Gruart et al found that the
amplitude of the evoked fEPSP signicantly increased as The SPM hypothesis requires that any manipulations that
training progressed, and then declined as the conditioned limit or block the induction or expression of synaptic plastic-
response was extinguished. Both learning and the potentia- ity in the hippocampus should have corresponding effects on
tion of the fEPSP were NMDAR-dependent, indicating that hippocampus-dependent learning. Meeting this second crite-
the potentiation was LTP rather than some other facilitatory rion has proved a particularly popular approach, with numer-
process. In the second study, from Mark Bears laboratory, ous pharmacological, molecular-genetic, and occlusion
Whitlock et al. (2006) used a hippocampus-dependent one- studies being conducted. Enhancement of LTP might also
trial learning procedure to look for changes in fEPSPs using an enhance learning, but this prediction is less straightforward
array of 8 recording electrodes implanted in area CA1. In a because, as the BCM theory of metaplasticity implies (see
rst recording session baseline responses were collected. The Section 10.3.13), LTP and LTD exist in dynamic equilibrium.
animals were then trained in an inhibitory avoidance condi- Disrupting this equilibrium might cause dysfunction, even if
tioning task, in which the rat forms an immediate aversion to the agent of disruption is enhanced LTP. Moreover, perturba-
the dark compartment of a two-compartment box after tions that enhance LTP could arise for several reasons, only
receiving a footshock there. Animals were then returned to the some of which would be expected to enhance learning.
recording chamber. Across all trained animals, a signicant
potentiation of the fEPSP was seen in 27% of electrodes, Pharmacological Blockade of the
detectable as soon as recording was begun 15 min after the NMDA Receptor Impairs Learning
footshock, and continuing at a stable level for 4 hours. A
tetanus was then given to test for occlusion. LTP was signi- Morris et al. (1986) showed that chronic intraventricular infu-
cantly greater at non-potentiated than potentiated recording sion of the NMDA antagonist D,L-AP5 blocked spatial but not
sites; this partial occlusion is powerful evidence that the visual discrimination learning in a watermaze (Fig. 1023A).
potentiation observed was LTP. Whitlock et al. wondered if Later work established that this effect was due to the active
their failure to see LTP at more recording sites was a reection D-isomer of AP5 (Davis et al., 1992) and that it occurs with
that learning induces both LTP and LTD. They examined this acute intrahippocampal infusions of D-AP5 (Morris et al.,
possibility using an ingenious biochemical approach. LTP and 1989) and over a range of intrahippocampal drug concentra-
LTD are associated with phosphorylation of AMPA receptors tions comparable to those that impair hippocampal LTP in
at different residues (see Section 10.7.4); using phospho- vivo and in vitro (Davis et al., 1992). Numerous studies have
antibodies specic to the two sites they were able to obtain a since found that competitive NMDA antagonists impair hip-
phosphorylation index of the degree of LTP and LTD in hip- pocampus-dependent learning. Learning paradigms used
pocampal tissue from trained and control animals. The sur- include spatial learning, T-maze alternation, certain types of
prising nding was that training was associated with an olfactory learning, contextual fear-conditioning, social trans-
increase in phosphorylation at ser831, the LTP-specific mission of food preference, avor-place paired associate learn-
residue, but no change at ser845, the LTD specic residue. A ing (delayed reinforcement of low rates of responseDRL),
question for future work is whether this result can be recon- and other operant tasks (Danysz et al., 1988; Tonkiss et al.,
ciled with the widely held assumption that stability in the hip- 1988; Staubli et al., 1989; Tonkiss and Rawlins, 1991; Bolhuis
pocampal network requires LTP at some synapses to be and Reid, 1992; Cole et al., 1993; Lyford et al., 1993; Carama-
balanced by LTD at others. One possibility is that other forms nos and Shapiro, 1994; Fanselow et al., 1994; Li et al., 1997;
of persistent depression, such as heterosynaptic depotentia- Steele and Morris, 1999; Lee and Kesner, 2002; Roberts and
tion and depression, which may have different biochemical Shapiro, 2002; Day et al., 2003; Bast et al., 2005; for reviews see
signatures and which are easier to obtain in adults than de Danysz et al., 1995 and Lynch, 2004). These data strongly sup-
 Synaptic Plasticity in the Hippocampus 433

port the SPM hypothesis, but there are problems with the ing the second task. Sensorimotor disturbances were equally
interpretation of these ndings associated with sensorimotor reduced by both procedures. Bannerman et al. (1995) sug-
side effects of NMDA antagonists and effects of these drugs on gested that blocking NMDA receptors dissociated different
neuronal processes other than LTP. components of spatial learningimpairing an animals ability
In humans, administration of ketamine results in halluci- to learn the required strategy rather than the map of land-
nations. We cannot know what rats are seeing or smelling marks in the room in which the watermaze is situated. This
under the inuence of an NMDA antagonist. However, senso- explanation was apparently refuted by Hoh et al. (1999), who
rimotor disturbances are sometimes observed during water- reported that watermaze strategy learning is unaffected by
maze training with diffuse NMDA receptor blockade, intraperitoneal administration of the NMDA antagonist
including falling off the platform during a wet-dog shake, CGS19755 during nonspatial pretraining at a dose that
thigmotaxis, and failure to climb onto the platform (Cain successfully blocks LTP in freely moving animals in both CA1
et al., 1996; Saucier et al., 1996). Motor disturbances induced and the dentate gyrus. The drug-treated rats learned nons-
by NMDA antagonists are also seen during other tasks. Such patial strategies adequately and showed equivalent perfor-
abnormalities could be due to diffusion of drug to the thala- mance in subsequent spatial learning and spatial reversal
mus disrupting the normal transmission of somatosensory relative to vehicle-treated control animals. Hoh et al. (1999)
and visual information (Sillito, 1985; Salt, 1986; Salt and suggested that relative task difficulty may explain the different
Eaton, 1989) or to the striatum causing motor disturbances outcome of their study and that of Bannerman et al. (1995)
such as accidity (Turski et al., 1990). Clearly, learning cannot the latter study involved a larger pool and a smaller escape
proceed when animals can neither sense nor move properly, platform.
and equally obviously it would be fallacious to assert that the Using single-unit recording to study the effects of the
drug is having a direct effect on learning mechanisms per se. NMDA antagonist 3-(2-carboxypiperazin-4-yl)propyl-1-
Many laboratories have noted that animals treated with non- phosphonic acid (CPP) on hippocampal place elds (see
competitive NMDA antagonists (such as MK-801) show sen- Chapter 11), Kentros et al. (1998) showed that previously
sory inattention and motor stereotypies (Koek et al., 1988; established ring elds are unchanged, and new place elds
Tricklebank et al., 1989; Keith and Rudy, 1990; Tiedtke et al., are acquired normally when rats are placed in a new environ-
1990; Mondadori and Weiskrantz, 1991; Danysz et al., 1995; ment. These ndings are inconsistent with a sensorimotor dis-
Cain et al., 1996). At high doses, AP5-treated animals also dis- turbance hypothesis. Kentros et al. went on to show that the
play stereotypies, but these doses are substantially higher than new place elds are unstable over time, implicating LTP
those necessary to block LTP in vivo following regionally induction in the learning of a new spatial context. Temporal
restricted infusion. These problems are much less apparent instability of place elds was also observed in transgenic mice
with acute intrahippocampal infusion. with altered LTP (Rotenberg et al., 1996). This temporal insta-
Using peripheral injections of an NMDA antagonist, Cain bility could account for Bannerman et al.s (1995) nding of
et al. (1996) showed that the impairment of spatial learning in poor learning by naive animals trained with one trial per day
a watermaze is correlated with the degree of sensorimotor but not the observation that spatially pretrained AP5-treated
impairment. They asserted that the sensorimotor decit is pri- animals learn normally in a new environment.
mary and the learning decit merely secondary. Saucier and Poor spatial memory over time in the presence of NMDA
Cain (1995) also observed that the impairment in watermaze receptor blockade is unlikely to be specic to novel environ-
learning that normally occurs following intraperitoneal ments but task-specic in other ways. Steele and Morris
administration of a competitive NMDA antagonist disappears (1999) trained rats in a delayed matching-to-place (DMP)
if the animals are given sufficient pretraining to prevent the task in the watermaze. In this variant of the task, the platform
drug-induced sensorimotor disturbances seen in experimen- is hidden in a different location each day and stays there for
tally naive animals. Their pretrained animals showed a clear four trials. Normal rats show quite long escape latencies on
block of LTP after drug treatment, no sensorimotor impair- trial 1 (when they do not know where the platform is hidden)
ment, and normal rates of spatial learning. but much shorter latencies on subsequent trials (when they
Other data indicate, however, that a decit in spatial learn- do). Most of the savings in escape latency occur between tri-
ing can still be observed, relative to appropriate control als 1 and 2, indicative of one-trial learning. Infusion of APV
groups, following nonspatial pretraining (Morris, 1989; had no effect on performance at a short memory delay (15-
Bannerman et al., 1995). Bannerman et al. (1995) discovered second intertrial interval) but caused pronounced impair-
that the usual AP5-induced learning decit all but disap- ment at 20 minutes and 2 hours. This delay-dependent decit
peared in animals trained rst as normal animals in one occurred irrespective of whether the animals stayed in the
watermaze (downstairs) before later being trained in a sec- training context throughout the memory delay or were
ond watermaze (upstairs) under the inuence of the drug. returned to the room with their home cages and irrespective
However, if training in the downstairs watermaze was non- of whether the drug was infused chronically into the ventricle
spatial in character, with sight of extramaze cues occluded, the or acutely into the hippocampus. If the AP5-induced impair-
normal AP5-induced decit in spatial learning was seen dur- ment of the DMP task were sensorimotor or attentional in
434 The Hippocampus Book

A NMDA receptor blockade impairs spatial learning B NR1 deletion restricted to CA1
Control D,L-AP5 50 wt fNR1 T29-1 CA1-NR1KO

Quadrants Time (%)

40
30
20
10
0
70
Quadrants Time (%)

60 C zif268 deletion
50
50 +/+ +/- -/-

Quadrants Time (%)

40
30 40

20 30
10 20
0 10
Adj/R
Adj/L
Train

Opp

Adj/R
Adj/L
Train

Opp
D One-shot spatial memory (DMP) 2h TI E CA3-NR1KO on DMP task
I
IT
I
s
15

80 aCSF 80
Day T1 T2 T3 T4 D-AP5
Latency (sec)

Latency (sec)
ITI 60 60
N
short
40 40

N+x 20 20
long 2h
0 0
T1 T2 T3 T4 T1 T2 T3 T4 T1 T2 T3 T4
Trials within a day, averaged across days Trials within a day

Figure 1023. Anterograde alteration of LTP and memory. A. zif-268 impairs LTP and long-term memory in the watermaze.
Chronic intraventricular infusion of the NMDA receptor antagonist Other experiments established that short-term memory was intact.
D,L-AP5 blocks LTP induction in the dentate gyrus (not shown) (Source: After Jones et al., 2001.) D. One-shot learning in the water-
and impairs spatial memory in a watermaze. Swim paths show maze can be studied using the DMP paradigm (see Figure 1323 for
focused searching in the target quadrant only by aCSF controls. explanation). Intrahippocampal D-AP5 infusion blocks memory of
Other experiments established that the drug spared simple visual a one-time experience after 2 hours but leaves short-term (15 sec-
discrimination learning. (Source: After Morris et al., 1986.) B. Cell- onds) memory unaffected. (Source: After Steele and Morris, 1999.)
type and region-specic knockout of the NR1 subunit of the E. Region-specic knockout of NR1 in area CA3 leaves spatial refer-
NMDA receptor in area CA1 of the hippocampus blocks LTP induc- ence memory intact but impairs one-shot memory. (Source: After
tion (not shown) and impairs spatial reference memory. (Source: Nakazawa et al., 2003.)
After Tsien et al., 1996.) C. Knockout of the immediate early gene

nature, a decit would be expected at all delays. The delay- incremental learning of a single location over multiple trials.
dependent memory impairment is inconsistent with Kentros Consequently, the reference memory task may require little
et al.s (1998) proposal that temporal instability occurs only in hippocampal synaptic plasticity in experienced animals,
a novel context. whereas substantial NMDAR-dependent plasticity might
In addition to addressing the potential sensorimotor side remain essential for one-trial encoding, even after extensive
effects of AP5, the study by Steele and Morris (1999) indicated training. Consistent with this, selective impairment for rapid
that rapid spatial learning remains sensitive to NMDA recep- learning has been observed after several genetic interventions.
tor blockade, even after pretraining. (The animals tested in Mice lacking the GluRI (GluR-A) subunit of the AMPA recep-
this study received several days of pretraining on the DMP tor, for example, have profound decits in CA1 LTP, but
task prior to surgery and drug infusion). It may be signicant acquisition of a watermaze reference memory task is normal;
that the DMP task requires the trial-unique encoding of novel severe impairments are found in spatial working memory,
locations, whereas the reference memory task involves the however, such as T-maze nonmatching to place (see Chapter 7
 Synaptic Plasticity in the Hippocampus 435

for a detailed discussion). Similar results have also been training by either daily injections of CPP in rats (Villarreal et
obtained after targeting an entirely different component of the al., 2002) or intraventricular minipump infusions of D-AP5 in
hippocampal circuitrythe recurrent collateral system of both rats and mice (Day and Langston, 2006) does not impair
CA3. Mice with a CA3-specic knockout of the gene coding retention. In fact, Villarreal et al. (2002) reported enhance-
for the NR1 subunit of the NMDA receptor lack associative ment of spatial memory after posttraining NMDA receptor
LTP in this subregion (although the direct entorhinal input antagonism, consistent with their additional nding that
to CA3 is probably also compromised), but the acquisition chronic NMDA receptor blockade prevents the decay of LTP
of a standard watermaze reference memory task remains that usually occurs over a period of several days. Modest
unaffected, at least when all cues are present during retrieval enhancement of retention was also found by Day and
(Nakazawa et al., 2002). Performance in a DMP version of the Langston (2006). Perhaps the blockade of synaptic plasticity
task, in contrast, is markedly impaired (Nakazawa et al., 2003). after training reduces retroactive interference from ongoing
Molecular-genetic approaches are discussed below. experience that might otherwise degrade the original memory
The NMDA receptor is implicated in the induction but not (see also Rosenzweig et al., 2002).
the expression of LTP (see Section 10.3). By analogy, one may Several studies have indicated that low doses of NMDA
predict that NMDA antagonists should impair the encoding receptor antagonists can, paradoxically, enhance the learning
of new memory traces (induction) but have no effect on of certain tasks, such as step-down inhibitory avoidance
memory retrieval (expression). Consistent with this idea, (Mondadori et al., 1989) and social learning (Lederer et al.,
Staubli et al. (1989) found that administering AP5 to animals 1993). These ndings do not challenge the SPM hypothesis
after they had been trained in odor discrimination learning because the effects are observed at doses far too low to block
had no effect on retention, although the drug did impair new LTP in vivo, and different mechanisms are likely to be involved
learning. Entorhinal cortex lesions, on the other hand, cause in antagonist-induced facilitation of learning. For instance,
rapid forgetting of olfactory information (Staubli et al., 1984). the facilitation of inhibitory avoidance by low doses of NMDA
Likewise, Morris (1989) and Morris et al. (1990) found that antagonists is sensitive to pretreatment by steroids such as
APV had no effect on retention of a previously trained water- aldosterone or corticosterone, whereas the impaired inhibi-
maze task, whereas lesions of the hippocampal formation tory avoidance caused by high doses is steroid-insensitive
were disruptive when given shortly after the end of training. (Mondadori and Weiskrantz, 1993; Mondadori et al., 1996).
Similar decits in encoding but not retrieval were noted in the Work with the noncompetitive NMDA receptor antagonist
DMP task of Steele and Morris (1999). Using a one-trial memantine has also led to counterintuitive ndings by virtue
paired-associate learning paradigm, Day et al. (2003) showed of its rapid on-and-off kinetics of channel blocking. At thera-
that intrahippocampal infusion of D-AP5 prior to the sam- peutic doses, memantine does not impair learning or LTP, but
ple trial when two novel stimuli were paired for the rst time it does limit neurotoxicity and so prevent impairments in cog-
caused subsequent choice performance in the memory nitive function or their accumulation over time (Parsons et al.,
retrieval trial to fall to chance. Delaying the infusion of 1999).
D-AP5 by 20 minutes until after the sample trial, such that Studies targeting the NMDA receptor are now comple-
memory retrieval was also conducted with dorsal hippocam- mented by many pharmacological manipulations directed at
pal NMDA receptors blocked, had no effect on memory. Both the diversity of receptors and intracellular signaling pathways
effects were seen at doses sufficient to block LTP in vivo. that play a role in synaptic plasticity. To take just one example,
Similar recents were recently obtained in a purely spatial ver- there is considerable evidence that blockade or knockout of
sion of the task (Bast et al., 2005). group I mGluRs, a class of receptor implicated in both LTP
Long-term memory retention can, however, be impaired and LTD (see Section 10.3), can impair a variety of forms of
by the blockade of hippocampal NMDA receptors soon after, memory, including spatial learning, contextual fear condition-
as well as during, learning (Izquierdo et al., 1993, 1998; ing, and inhibitory avoidance (for reviews see Holscher et al.,
Packard and Teather, 1997; McDonald et al., 2005). These 1999; Riedel et al., 2003; Simonyi et al., 2005). For a fuller
ndings suggest that the early stabilization of memory account of pharmacological interventions targeting sites other
requires additional offline episodes of NMDA receptor acti- than the NMDA receptor, see Lynch (2004).
vation after initial encoding. The involvement of NMDA
receptors in systems consolidation and long-term information Targeted Disruption of the NMDA Receptor and
is less well understood. Shimizu et al. (2000) used the tTA and Downstream Signal-transduction Pathways Impairs
Cre/loxP systems to develop a mouse with an apparently CA1- Hippocampus-dependent Learning
specic (but see Fukaya et al., 2003) inducible deletion of the
NMDA NR1 subunit. Deletion of this subunit by application Some of the earliest uses of targeted molecular engineering
of doxycycline 1 to 7 days after training on a watermaze refer- in neuroscience were in relation to synaptic plasticity and
ence memory task caused subsequent impairment in reten- learning in studies of downstream signal-transduction path-
tion of the platform location. However, chronic blockade of ways (Silva et al., 1992a,b; Grant et al., 1995). These early
hippocampal NMDA receptors during the days after spatial studies revealed intriguing correlations between LTP and
436 The Hippocampus Book

hippocampus-dependent learning. They set the pattern for a small quantities (to avoid seizure activity), the now disinhib-
host of subsequent studies in which homologous recombina- ited area of the dentate showed normal LTP. Errington et al.
tion was used to delete genes in all cells of the body for the (1997) also found that LTP in freely moving thy-1 mutants
lifetime of the animal (Capecchi, 1989). The approach is often was compromised but not totally abolished, with a wide range
characterized by cell biological evidence that the deletion has of variability among individual animals. An important general
been successful, followed by electrophysiological and behav- message of the thy-1 study for the eld is that electrophysio-
ioral analyses of the phenotype of the mutant progeny. The logical results in brain slices are not infallible predictors of
array of generally positive ndings obtained from well over what might happen to synaptic plasticity in the whole animal.
100 relevant knockouts developed since 1992 lends general The study also highlights the value of assessing both learning
support to the SPM hypothesis (see Table 5 in Lynch (2004). and LTP in the same animals.
Numerous problems, however, were soon identied. For A third problem when using mice to assess the SPM
certain critically important proteins, there can be a cata- hypothesis has been the inappropriate interpretation of
strophic outcome such as embryonic or perinatal lethality. behavioral studies. Some studies using the watermaze with
This occurred with an attempted deletion of the NR1 subtype transgenic mice have used an apparatus that is too small to
of the NMDA receptor (Chen and Tonegawa, 1997), although study spatial learning effectively ( 1.2 m) or failed to use
deletion of the NR2A (1) subunit using conventional homol- posttraining probe tests. In a careful factor analysis of hun-
ogous recombination could be successfully carried forward to dreds of experiments, Wolfer et al. (1998) statistically dissoci-
adults, and these mice showed a decit in both hippocampal ated spatial learning in the watermaze from thigmotaxis and
LTP and spatial learning (Sakimura et al., 1995). Other other species-specic swimming strategies. Emphasizing also
mutants display the opposite problem, a null phenotype, the value of looking at LTP in vivo in association with behav-
though it is hard to believe that key brain enzymes have no ioral studies, their work has played an important part in
function. Either inappropriate behavioral assays have been improving the analytical quality of modern studies.
used, or there has been compensation by other closely related Regionally specic mutants can now been made in which
genes (Grant et al., 1995). Another problem is interanimal CRE recombinase is expressed downstream of specic pro-
variability; for example, it is puzzling that some slices from the moters in one line of mice that, after crossing, target genes
CaMKII mutants studied by Silva et al. (1992a) show normal anked with loxP sites in another. The resulting progeny
LTP, whereas most show none unless parallel biochemical enable the development of mutant strains in which a target
pathways can be activated in a probabilistic fashion. Hinds et gene is deleted (or mutated) in only one area of the brain. This
al. (1998) reported a yet higher proportion of CaMKII approach was used to knock out the NR1 subunit of the
mutants showing normal LTP and suggested that the NMDA receptor in area CA1 of the hippocampus only (Tsien
enzyme may have been upregulated. Hinds et al.s (1998) study et al., 1996; Mayford et al., 1997). The mice showed no LTP in
involved animals cross-bred with C57/BL6 mice. The issue of area CA1, normal LTP in the dentate gyrus and neocortex, and
genetic background is important: The embryonic stem cells learning impairment in the watermaze (Fig. 1023B). Using
most widely used to make mutants are derived from a specic multiple single-unit recording, McHugh et al. (1996) discov-
strain of micethe Sv129 strain; and when these mice are ered that these mice had abnormal place elds and a reduction
cross-bred with strains such as C57/BL6, a number of ank- in the correlated ring of cells with overlapping place elds
ing genes derived from the 129 strain are still expressed (see Chapter 11). Such regionally specic manipulations are
alongside the mutated gene. Aspects of the resulting pheno- beginning to allow the targeting of specic components of the
type may reect these anking genes (Gerlai, 1996). Residual hippocampal network. For example, Nakazawa et al. (2002)
Sv129 genes have unpredictable effects and are undesirable created a mouse in which the NR1 gene was deleted only in
because the Sv129 strain is notoriously poor at learning (Lipp CA3, a manipulation that blocks NMDAR-dependent LTP at
and Wolfer, 1998). Recommendations about the desirability of CA3 recurrent synapses but not NMDAR-independent poten-
back-crosses into more suitable strains (such as C57/BL6) tiation in the mossy ber input. These mice learned a water-
were subsequently discussed in the gene-targeting community maze reference memory task as well as the control animals.
and a set of guidelines published (Anonymous, 1997). Retention was likewise unaffected when all extramaze cues
Another issue is that many studies examine LTP only using were present, but performance deteriorated substantially
in vitro brain slices. This is insufficient and a separate consid- when only partial cues were present (see Chapter 13, Fig.
eration concerns the value of looking at LTP in vivo. Nosten- 1321). These ndings are consistent with a long-held view of
Bertrand et al. (1996) found that thy-1 mutants had normal CA3 as an autoassociative network capable of pattern comple-
LTP in area CA1 but no LTP in the dentate gyrus when meas- tionreconstruction of a complete output pattern from par-
ured in anesthetized animals in vivo. Further studies revealed, tial or degraded input (cf. Marr, 1971). According to this view,
however, that normal LTP could be obtained in dentate slices synaptic plasticity at recurrent synapses is necessary for the
when bicuculline was added to reduce inhibition, implying formation of a memory trace that can later be recalled in the
that the machinery for inducing LTP must still be present in presence of partial retrieval cues. In a follow-up study, the
thy-1 mutants. When bicuculline was infused locally in vivo in same mice were found to be impaired in one-trial place learn-
 Synaptic Plasticity in the Hippocampus 437

ing (see above), again supporting the notion of CA3 as a rapid LTD rather than LTPa shift of in the BCM function seen
autoassociative storage device (Nakazawa et al., 2003). in Box Fig. 102A. Hippocampus-dependent learning was
Consistent with this nding, place cell recordings in area CA1 impaired: The mice showed impaired spatial learning using a
of these mice revealed a reduction in spatial tuning during Barnes maze (Bach et al., 1995). Mayford et al. (1996) repli-
early exploration of novel environments, implicating CA3 cated the learning impairments and decits in LTP found by
synaptic plasticity in the renement of the spatial representa- Bach et al. (1995) and Mayford et al. (1995), with suppression
tion supported by the direct entorhinal input to CA1 (see of the transgene by administration of doxycycline relieving
Brun et al., 2002). The Cre-loxP system has now been used to the impairment of both learning and synaptic plasticity. Work
target many components of the plasticity machinery in addi- by Giese et al. (1998) complemented these studies. Instead of
tion to the NMDA receptor (for reviews see Matynia et al., using transgenic techniques, they introduced a point muta-
2002; Lynch, 2004). For example, mice lacking the AMPA tion (threonine to alanine) into the gene encoding CaMKII
receptor GluRI (GluR-A) subunit show impairments in CA1 to block autophosphorylation at residue 286, thereby prevent-
LTP and spatial working memory, although the acquisition of ing the transition of this kinase into a CaM-independent state
spatial reference memory is unaffected (Zamanillo et al., 1999; without disrupting its CaM-dependent activity. LTP could not
Reisel et al., 2002) (see Chapter 7). The impairments in LTP be elicited in area CA1 in the mutant mice across a range of
and memory can be rescued by adding a forebrain-specic stimulation frequencies, a slightly different prole to that
GFP-tagged GluRI expression system to these mice, conrm- shown by Mayfords transgenic mouse. Giese et al.s mice also
ing the specicity of the impairment (Schmitt et al., 2005). had profound decits in spatial learning in the watermaze
Region- and cell type-specic interventions are powerful, and showed altered dependence on extramaze versus intra-
as they carry the potential to investigate gene function in spe- maze cues in single-unit recording studies of place cells (Cho
cic cells, to manipulate pre- or postsynaptic sites of synapses et al., 1998). Although this mutation is not inducible, tempo-
independently, and to intervene in biochemical pathways for ral control can be achieved by combining pharmacological
which there is no known ligand. The CaMKII promoter has and molecular genetic techniques. Heterozygous Thr286
been useful in such experiments because it is not activated mutant mice have no decit in contextual fear conditioning or
until around day 20, obviating certain developmental prob- LTP, but injection of a low dose of CPP that in wild type mice
lems. The approach, however, is expensive as numerous lines has no effect results in impairment of both learning and plas-
have to be developed of which only some show useful expres- ticity, suggesting that the NMDAR-dependent activation of
sion patterns. Moreover, although achieving precise regional CaMKII is necessary for both processes (Ohno et al., 2001,
and cell-type specicity is advantageous over pharmacological 2002).
manipulations, it does not yet allow the same precise tempo- Other early studies to use the tTA technique were those of
ral control as drugs. In the previous section, experiments were Mansuy et al. (1998a,b) and Winder et al. (1998). They
described in which animals were normal in one phase of an reported that transgenic mice overexpressing a truncated
experiment but treated in another. This enabled dissocia- form of the phosphatase calcineurin display normal early-LTP
tions between encoding and retrieval processes of memory and short-term memory but defective late-LTP and long-term
to be addressed using highly selective anatagonists such as memory. However, evidence that the latter decit was second-
D-AP5. The corresponding way forward, using genetic inter- ary to some other problem came from behavioral work show-
ventions, is via conditional mutants in which inducible pro- ing that a change in training protocol could rescue the
moters are engineered to put gene activation (or inactivation) impaired long-term memory. This suggested that the decit in
under experimental control. Use of the tetracycline transacti- these animals was more likely in the transition from short- to
vator systems, rtTA and tTA (see Box 7.1), provides one exam- long-term memory than in the mechanisms underlying either
ple of this approach (see Furth et al., 1994; Kistner et al., on its own. Regulation of calcineurin overexpression using the
1996). Inducible genetic interventions also nesse the compli- rtTA technique was examined in animals tested in the water-
cations of altered neuronal development that can occur with maze (Mansuy et al., 1998b). In the latter study, animals com-
nonconditional knockouts (Lathe and Morris, 1994; Mayford pleted training and were rst tested with the transgene off.
et al., 1997) and can otherwise only be addressed using res- They learned the platform location as indexed by good per-
cue experiments. formance in a posttraining probe test. When the transgene
Using inducible techniques, Mayford et al. (1996) created a was then turned on, performance in a second probe test fell
transgenic mouse in which overexpression of a constitutive to chance; and when it was later turned off again, perform-
CaMKII was under the control of tetracycline. This study built ance recovered. These results suggest that calcineurin con-
upon previous work using standard transgenic mice that strains memory retrieval. The rtTA system has also been used
overexpress the autophosphorylated form of CaMKII through to enhance the inhibition of calcineurin. Malleret et al. (2001)
a point mutation of Thr286 (Mayford et al., 1995). These ani- created a mouse with inducible overexpression of a C-termi-
mals had exhibited normal CA1 LTP in response to high-fre- nal autoinhibitory domain that reduces calcineurin activity.
quency stimulation at 100 Hz, but stimulation in the 5- to Expression of the transgene facilitated LTP both in vivo and in
10-Hz range (encompassing theta) preferentially resulted in vitro but did not affect LTD. Parallel enhancement of hip-
438 The Hippocampus Book

pocampus-dependent memory was observed. Following an tion, but not the expression, of odor-context paired associate
analogous strategy, a regulatory subunit of calcinuerin, CNB1, memory.
was deleted in a forebrain-specic manner, resulting in Good temporal control can also be achieved using anti-
reduced calcineurin activity (Zeng et al., 2001). Although the sense techniques. These methods lie somewhere between the
mutation was not inducible, it did not become detectable until genetic and pharmacological approaches, having some of
5 weeks of age, obviating to some extent the possibility of advantages and disadvantages of both. Examples of their use
developmental consequences. No enhancement of LTP was include studies in which the expression of mRNA for two dif-
evident following a standard high-frequency tetanus, but an ferent potassium channels was reduced by repeated intracere-
overall leftward shift in the function relating LTP/ LTD to broventricular injections of oligodeoxyribonucleotides to
tetanus frequency was observed together with impaired LTD. reveal a dissociation between learning and plasticity. Antisense
Perhaps for the latter reason, these mice did not show disruption of the presynaptic A-type potassium channel, Kv
enhanced learning and memory. Performance in a watermaze 1.4, eliminated both early and late phases of CA1 LTP without
reference memory task was normal, but acquisition of a task affecting LTP in the dentate gyrus in rats (Meiri et al., 1998).
requiring the learning of multiple novel locations was However, this antisense knockdown of plasticity had no
impaired, as was performance in a radial-arm maze working effect on spatial learning. Given that threshold changes in the
memory task. These ndings highlight the potential impor- intensity of the tetanus required to induce LTP have been seen
tance of decreases, as well as increases, in synaptic strength. with some genetic manipulations (Kiyama et al., 1998), the
Perhaps LTD-like mechanisms are particularly important in fact that Meiri et al. (1998) used different intensities to induce
tasks that require exible learning of new information in con- LTP signicantly strengthens their conclusion that CA1 LTP
junction with rapid suppression of information that is no was successfully blocked in the antisense group. However, this
longer relevant (cf. Manahan-Vaughan and Braunewell, 1999; study did not include in vivo observations, a range of tetaniza-
Kemp and Manahan-Vaughan, 2004). tion frequencies (Mayford et al., 1995; Migaud et al., 1998), or
Despite its power, the tTA system still requires several days information about the regional spread of the antisense oligo.
for complete induction and repression, and advances in tem- The latter point is important, as LTP may have been moni-
poral control are eagerly awaited. A novel inducible system has tored in an area along the longitudinal axis of the hippocam-
since been developed based on fusion of a target protein with pus that was affected by the oligo, whereas learning may have
the ligand-binding domain of the estrogen receptor modied utilized neurons along the full length of this axis. It should be
to bind tamoxifen instead of estrogen (Kida et al., 2002). recognized, however, that antisense disruption of Kv 1.1, a dif-
Using this system, expression can be turned on and off within ferent potassium channel that is highly localized in dendrites
hours, but the range of potential target proteins is limited of CA3 neurons, had no effect on LTP in either the CA1 sub-
compared to the tTA approach. Another development is the eld or the dentate gyrus but did cause profound decits in
increasing use of viral vector-mediated gene transfer. This spatial learning (Meiri et al., 1997). Clearly, the longitudinal
approach enables the targeted and temporally controlled over- axis objection cannot apply here, and the signicance of both
expression of transgenes following local transfection within a these studies should be recognized. The antisense approach
specic brain region. For example, the virus-mediated expres- has also generated discrepancies between LTP and memory in
sion of calcium-permeable AMPA receptors in CA1 resulted the opposite direction (i.e., enhanced LTP associated with
in enhanced CA1 LTP in vitro and facilitated place learning impaired memory) (Pineda et al., 2004), although such nd-
and memory (Okada et al., 2003). However, dentate transfec- ings are potentially less troublesome for the synaptic plasticity
tion impaired watermaze performance despite similar and memory hypothesis, as discussed below. Other studies
enhancememt of LTP, perhaps reecting functional differ- have reported parallel effects on both LTP and memory
ences between these two subregions. The behavioral conse- (Guzowski et al., 2000; Hou et al., 2004). In the former study,
quences of enhancing LTP are considered in more detail in the for example, antisense inhibition of the expression of hip-
following section. In another study, intrahippocampal infu- pocampal Arc protein (see Section 10.9) resulted in a decit in
sion of a viral vector conveying an inactive mutant form of long-term spatial memory in a watermaze reference memory
CREB was found to block long-term, but not short-term, task and impairment in the late maintenance of dentate LTP in
memory for socially transmitted food preference (Brightwell vivo (Guzowski et al., 2000), consistent with a role for Arc in
et al., 2005). Conversely, overexpression of CREB induced the long-term synaptic modications postulated to underlie
by intraamygdalar infusion of a viral vector results in memory storage.
enhancement of amygdala-dependent memory (Josselyn et The development of methods for gene silencing does not
al., 2001; Jasnow et al., 2005). Alternatively, the virus-medi- end with the antisense approach. A more effective alternative
ated delivery of Cre can be used to trigger deletion of a target to antisense knockdown is the RNA interference technique
gene anked by lox-P sites (e.g., Scammell et al., 2003). Using (RNAi). In a renement of this technique, small interfering
this technique, Rajji et al. (2006) were able to target deletion RNA (siRNA) sequences are targeted to a specic brain region
of the NR1 gene to area CA3 of the dorsal hippocampus, a by local elecroporation or by viral transfection, thereby offer-
manipulation that resulted in an impairment in the acquisi- ing excellent regional and temporal control (Akaneya et al.,
 Synaptic Plasticity in the Hippocampus 439

2005). Although highly promising, this approach has not yet position for bidirectional plasticity. Perhaps it is this abnor-
been applied to studies of hippocampal synaptic plasticity and mality or simply a deciency in LTD that underlies the learn-
memory. ing problems evident in these animals.

Do Treatments That Enhance LTP Also Enhance Memory? Parallels Between Different Phases of LTP
Expression and Memory Consolidation
As we discussed earlier, the circuit-level consequences of
enhancing synaptic plasticity may be far from straightforward. There is considerable evidence that both hippocampus-
Synaptic plasticity might normally be optimally tuned for the dependent memories and increases in synaptic strength each
efficient encoding of information, such that any disturbance undergo a period of consolidation in the minutes to hours fol-
even enhancementhas disruptive effects on memory. lowing their formation (McGaugh, 2000). A variety of inter-
Nonetheless, many pharmacological and molecular-genetic ventions can interfere with both memory consolidation and
manipulations do enhance both LTP and memory. Perhaps the stabilization of LTP, including (1) disrupting the activity of a
best-known example of the pharmacological approach range of protein kinases (Izquierdo and Medina, 1997;
involves treatment with ampakines, which decrease the rate Micheau and Riedel, 1999) such as CaMKII (Lisman et al.,
of AMPA receptor desensitization and slow the deactivation 2002; Colbran and Brown, 2004), MAPK/ERK (Kelleher et al.,
of receptor currents after agonist application (Arai et al., 1994, 2004a; Sweatt, 2004; Thomas and Huganir, 2004), tyrosine
1996). Consequently, they facilitate the induction of hip- kinases (Purcell and Carew, 2003), PKA (Nguyen and Woo,
pocampal LTP (Staubli et al., 1994); and there is now a con- 2003), and PKC (Sun and Alkon, 2005); (2) disrupting the
siderable body of evidence that they can enhance the encoding expression of immediate early genes such as Zif268 (Jones et
of memory in a variety of tasks (Lynch, 2006). A number of al., 2000) (Fig. 1023C) or arc (Guzowski et al., 2000); (3)
other compounds have also been reported to enhance both temporary neuronal inactivation (e.g., Ambrogi Lorenzini et
learning and hippocampal LTP (Cooke and Bliss, 2005). al., 1999; Florian and Roullet, 2004; Daumas et al., 2005); and
Similarly, Manabe et al. (1998) reported that mice lacking (4) inhibition of macromolecular synthesis (for review see
the nociceptin/orphaninFQ receptor show enhanced LTP in Davis and Squire, 1984; Frey and Morris, 1998a; Dudai and
area CA1 (possibly also due to a change in K channel func- Morris, 2000). Growing evidence also supports a role for cell
tion) and both a modest but signicant decrease in escape adhesion molecules in consolidation processes (Murase and
latency in the watermaze and enhanced memory consolida- Schuman, 1999; Benson et al., 2000; Gall and Lynch, 2004).
tion in step-through avoidance learning. In a widely discussed Studies such as these offer broad support for the idea that last-
study, Tang et al. (1999) showed that overexpression of the ing changes in synaptic strength underlie the formation of
juvenile 2B subunit of the NMDA receptor both facilitated long-term memory.
LTP across a range of induction frequencies and enhanced For instance, late-LTP (i.e., LTP lasting around 4 hours or
memory in novel object recognition, cue and context fear- more), like cellular consolidation, requires synthesis of new
conditioning, and the earliest stages of learning a watermaze. proteins (for review see Frey and Morris, 1998a). As discussed
This pattern of results has since been observed following in Section 10.4.9, application of protein synthesis inhibitors
many other genetic interventions (e.g., Futatsugi et al., 1999; such as anisomycin prevents late-LTP. Importantly, however,
Madani et al., 1999; Routtenberg et al., 2000; Malleret et al., the events leading to protein synthesis do not have to occur at
2001; Nakamura et al., 2001; Jeon et al., 2003). exactly the same time as the stimulation that induces synaptic
In contrast, Migaud et al. (1998) reported that a PSD-95 potentiation. Tetanization in the presence of a protein synthe-
mutant that shows enhanced hippocampal LTP and decreased sis inhibitor still results in late-LTP if a strong tetanus is
LTD across a range of induction frequencies displays pro- applied to a separate pathway up to 2 hours previously (Frey
found impairment of watermaze performance. The nding of and Morris, 1997). Together with other ndings, this result
enhanced LTP and impaired spatial memory has since been has led to the notion that a high-frequency tetanus sets
reported in several other strains of mutant mice (Uetani et al., synaptic tags that can capture plasticity proteins as they
2000; Zeng et al., 2001; Kaksonen et al., 2002; Cox et al., 2003; become available (Frey and Morris, 1997, 1998b). The same
Vaillend et al., 2004), whereas other interventions result in may be true of memory formation. Activation of neuromod-
normal (or near-normal) spatial memory despite enhance- ulatory systems by novelty, reward, or punishment (Frey et al.,
ment of LTP (Jun et al., 1998; Gu et al., 2002; Cox et al., 2003; 2001; Richter-Levin and Akirav, 2003; Lisman and Grace,
Pineda et al., 2004). However, impairment of other forms of 2005) may cause widespead upregulation of gene expression
memory were reported in some of these studies. In some and protein synthesis via activation of cAMP-, PKA-, and
cases, impaired behavioral performance might result from the CREB-dependent pathways. This would enable capture of
dysfunction of extrahippocampal systems (Gerlai et al., 1998), proteins by synapses that have recently been potentiated and
but other factors may also be relevant. For instance, the PSD- tagged or are about to be potentiated as a result of normal
95 mutant mice created by Migaud et al. (1998) exhibited a experience (Frey and Morris, 1998a). However, MAPK-medi-
shift in M of the BCM function well to the left of its optimal ated upregulation of translationperhaps in dendrites
440 The Hippocampus Book

might also be critical under some circumstances (Martin et reached an impasse in 1993 when a series of articles (Cain et
al., 2000a; Steward and Schuman, 2003; Kelleher et al., 2004b) al., 1993; Jeffery and Morris, 1993; Korol et al., 1993;
(see Section 10.4 for a discussion of dendritic protein synthe- Sutherland et al., 1993) reported an inability to replicate ear-
sis). The interaction of specic synaptic signals with wide- lier ndings indicating that saturation induced reversible
spread transcriptional or translational upregulation may occlusion of subsequent spatial learning (McNaughton et al.,
provide an explanation for the facilitated retention of mem- 1986; Castro et al., 1989). Jeffery and Morris (1993) and Korol
ory for episodes occurring shortly before or after events of et al. (1993) conducted exact replications of part of Castro et
motivational importance, episodes that would otherwise be al.s (1989) experiment. In neither study was any learning
rapidly forgotten (cf. Seidenbecher et al., 1995). decit observed. Reid and Stewart (1997) succeeded in repli-
Consistent with this scenario, a number of studies have cating Castro et al.s ndings, including decay of the effect over
suggested that activation of CREB is necessary for both the time, but they used electroconvulsive seizures, which among
induction of hippocampal late-LTP and the formation of long- other effects can cause indiscriminate induction of LTP rather
term memory (i.e., cellular consolidation) in mice and rats than explicit saturation on a single pathway.
(Bourtchuladze et al., 1994; Bernabeu et al., 1997; Guzowski Bliss and Richter-Levin (1993) suggested several reasons the
and McGaugh, 1997; Kogan et al., 2000; Pittenger et al., 2002; negative results might have been obtained: (1) cumulative LTP
Brightwell et al., 2005; Wood et al., 2005). Barco et al. (2002) of perforant path terminals may not have reached a true state
recently created a mouse in which a constitutively active form of saturation; (2) perforant path terminals may have been suf-
of CREB, VP16-CREB, is expressed in a forebrain-restricted ciently saturated, but not those of other extrinsic or intrinsic
and temporally regulated manner. In these animals, the thresh- hippocampal pathways that are also critical for learning (e.g.,
old for late-LTP induction in the Schaffer-commissural to CA1 CA3-CA1 terminals); (3) appropriate saturation of the full
pyramidal cell pathway was reduced, such that a weak tetanus septotemporal axis of the hippocampus may not have been
capable of inducing only a decremental early-LTP in normal achieved with stimulation at a single site in the angular bundle.
mice induced late-LTP in VP16-CREB animals. These ndings Evidence for this third idea was presented by Barnes et al.
suggest that elevated CREB-mediated transcription in these (1994), who found upregulation of the immediate early gene
animals leads to synthesis of a pool of proteins or mRNAs that zif-268 restricted to the dorsal hippocampus after stimulation
can subsequently be captured by synapses tagged during weak at one site in the angular bundle. They also noted differences
tetanization, without the need for new transcription. A follow- in the sensitivity of various learning tasks to LTP saturation.
up study implicates the overexpression of BDNF in this phe- Moser et al. (1998) designed a study with these issues in
nomenon (Barco et al., 2002). However, it is not yet known mind. There were three key features: (1) use of an array of
whether VP16-CREB mice exhibit enhanced long-term mem- cross-bundle stimulation electrodes designed to maximally
ory for weakly encoded information, consistent with their LTP activate the perforant path, with the cathode switched fre-
phenotype. quently between active electrodes; (2) use of a separate probe
stimulating electrode to test whether the asymptotic LTP
Does Saturation of LTP or LTD Occlude induced by the electrode array was true saturation of LTP on
the Encoding of New Memory Traces? that pathway; (3) use of animals given unilateral hippocampal
lesions to reduce the amount of tissue to be potentiated.
In studies of expression mechanisms for LTP, occlusion was Subsequent to multiple high-frequency (HF) trains or control
often deployed as a tool to identify that LTP had been engaged. low-frequency (LF) stimulation, the rats were trained in a
The same approach can be deployed in behavioral studies. It is standard watermaze task. Controls learned normally. The HF
important, however, to distinguish between the cumulative group showed a bimodal distribution, with some animals
induction of LTP (or LTD) and the saturation of either pro- learning where the platform was located and others failing to
cess. Successive episodes of LTP may have a cumulative effect learn. When all animals were subsequently tested for induction
on synaptic strength, at least at a population level (cf. Petersen of LTP from the probe site in the perforant path, the HF ani-
et al., 1998), but not saturate the plasticity available on the mals in which it was impossible to induce further LTP (i.e., the
pathway being stimulated (Jeffery, 1997). Cumulative LTP saturated subgroup) were the ones that had failed to learn
may enhance neural throughput and so improve learning. the watermaze, whereas those in whom LTP could still be
Berger (1984) obtained just such a result in a study of nicti- induced had learned a little about where the platform was
tating membrane conditioning in rabbits, the original species located. Thus, true saturation of LTP in the perforant path
in which LTP was discovered. In contrast, true saturation of does impair spatial learning. These ndings vindicate the ear-
LTP prior to behavioral training should prevent new learning lier claims of McNaughton et al. (1986) and Castro et al.
because no further LTP would then be possible. Similar con- (1989).
siderations would apply to saturation of LTD, which can be Despite these positive ndings, there remains skepticism
achieved with three successive trains of stimulus pairs (Thiels about the analytical potential of the occlusion approach. One
et al., 1998). concern is that repeated tetanization may result in acute
Research on the behavioral effects of LTP saturation pathological phenomena, such as seizure-like after-discharges,
 Synaptic Plasticity in the Hippocampus 441

that would cause learning decits (McEachern and Shaw, Lynch, 1990; Bashir and Collingridge, 1994). Stubli and
1996). However, Moser et al. (1998) found no after-discharges Chun (1996) found that a few minutes of 5-Hz stimulation
during tetanization. Also, learning was impaired only in the can depotentiate recently induced LTP in area CA1 in vitro.
animals with saturated LTP, despite all rats having received the The efficacy of depotentiation declines rapidly as the interval
same course of tetanic stimulation. A second point concerns between the tetanus and 5 Hz is increased, with little effect
homeostatic compensatory changes, such as altered inhibitory being obtained 30 minutes after LTP induction. Dentate LTP
transmission, synapse formation, and reduced postsynaptic in vivo can also be reversed by 5-Hz stimulation when deliv-
sensitivity. These compensatory changes were considered in a ered up to a few minutes after tetanization, but later stimula-
review article by Moser and Moser (1999) but primarily as tion is ineffective (Martin, 1998; Kulla, 1999). Unfortunately,
factors contributing to the difficulty often experienced in sat- none of the protocols for inducing depotentiation has yet
urating LTP. Even when LTP saturation is successful, it might been tested for its ability to cause forgetting in behaving ani-
still be argued that such changes, rather than saturation itself, mals.
are responsible for the learning impairment. A third area of Xu et al. (1998) reported that exposing freely moving ani-
disquiet is that LTP saturation does, of course, increase synap- mals to a novel but nonstressful recording chamber can
tic weight; a global increase in the efficacy of synaptic trans- reverse recently induced LTP without affecting a control path-
mission might, on its own, disrupt normal hippocampal way. They speculated that exposure to novelty has the effect of
information processing. However, the number of studies erasing hitherto unconsolidated information (see also
reporting normal learning despite the induction of substantial Manahan-Vaughan and Braunewell, 1999). Support for this
LTP suggests that a cumulative increase in synaptic weight interpretation is offered in a report (Izquierdo et al., 1999) in
does not in itself disrupt the encoding of new information. In which exposure to novelty limited the ability of an animal to
fact, Moser et al. (1998) found no correlation between the remember a one-trial inhibitory avoidance task carried out up
magnitude of LTP induced by cross-bundle tetanization and to 1 hour previously. Exploration of novelty shortly before or
subsequent learning. long after the training trial was without effect. The effect
Moser et al.s (1998) study is unlikely to be the last word. appears to be NMDAR- and CaMKII-dependent.
First, saturation itself is not well understood physiologically. Instead of erasing LTP, the induction of further LTP should
Doing so could offer new insights into the maximum amounts alter the spatial distribution of synaptic weights in the distrib-
of potentiation the hippocampus could sustain. Second, satu- uted network of areas CA3 and CA1 and so prevent successful
ration might be achieved in other ways. It might also be real- retrieval of information stored earlier. In an early study of this
ized by bilateral stimulation of the ventral hippocampal kind, McNaughton et al. (1986) trained rats to remember the
commissure to potentiate the commissural/associational location of an escape tunnel in a Barnes arena and then
pathway in CA3 and CA1 (Bliss and Richter-Levin, 1993). induced LTP in the dentate gyrus via chronically implanted
Another study has suggested that as little as a single LTP- electrodes in the angular bundle. Recently acquired reference
inducing tetanus delivered to the Schaffer-commissural path- memory was disrupted by LTP induction, whereas well estab-
way can impair trace eyeblink conditioning (Gruart et al., lished spatial memory was unaffected. Results consistent with
2006). A pharmacological, rather than an electrophysiological, these have been reported by Brun et al. (2001), who trained
approach should also be considered using drugs such as rats in a spatial reference memory task in the watermaze for 5
BDNF (Bramham and Messaoudi, 2005) or agonists of adeny- days, before also inducing LTP via stimulating electrodes
late cyclase, protein kinase A, or mitogen-activated protein straddling the angular bundle of the perforant path. In con-
kinase (MAPK), to induce slow-onset but asymptotic synaptic trast to nonstimulated and low-frequency control groups, rats
potentiation. Multiple approaches are required to rigorously that had been subjected to high-frequency tetanization were
test this component of the SPM hypothesis. completely unable to remember the platform location in a
subsequent probe trial (Fig. 1024B). Nevertheless, the same
10.10.4 Retrograde Alteration: Does Further animals were able to learn a new platform location in a differ-
Induction or Reversal of LTP Cause Forgetting? ent watermaze environment as well as the controls, a result
consistent with the nding reported by Otnaess et al. (1999)
If memory traces related to a recent learning experience are that pretraining eliminates the LTP saturation-induced decit
temporarily stored in the hippocampus, procedures that in new spatial reference learning (cf. Bannerman et al., 1995;
induce further LTP in the same network, or successfully erase Saucier and Cain, 1995).
LTP, should cause forgetting. Erasure might be achieved (1) In the study of Brun et al. (2001), a group of rats was
using trains of suitable depotentiating (e.g., low frequency) tetanized in the presence of the NMDA receptor antagonist
stimulation or (2) applying drugs or enzyme inhibitors that CPP. These animals remembered the platform location as well
interrupt expression of LTP when given shortly after its induc- as the controls in a subsequent retention test (carried out in
tion (e.g., kinase inhibitors). the absence of CPP). In conjunction with the intact ability of
Depotentiation can be induced using continuous trains of animals with asymptotic or near-asymptotic LTP to learn a
single pulses at 1 to 5 Hz (Barrionuevo et al., 1980; Stubli and new watermaze task, these results suggest that neither high-
442 The Hippocampus Book

frequency tetanization nor LTP induction per se causes signif- the recall of trace eyeblink conditioning (Gruart et al., 2006),
icant hippocampal dysfunction, and they add further cre- although more widespread alteration of synaptic strengths is
dence to the notion that memories are stored as patterns of sometimes necessary to disrupt memory (Leung and Shen,
changes in synaptic strength, and that these patterns can be 2006).
disrupted by the addition of tetanus-induced LTP that consti- Finally, Pastalkova et al (2006) have shown that application
tutes meaningless noise. This notion is further supported by of ZIP, a membrane permeable inhibitor of PKM that is
the nding that LTP induction at CA3CA1 synapses prevents thought to mediate the persistence of hippocampal LTP over

Figure 1024. Retrograde alteration of LTP and memory. A. afferents of the perforant path that ordinarily induces LTP in the
Depotentiation of LTP should erase memory. One way in this might dentate gyrus, they were later unable to remember the location of
be achieved is through experiences such as novelty exploration. Rats the hidden platform, as shown in post-training probe tests (see Fig.
in which LTP had been recently induced showed persistent potenti- 13(23 for explanation of behavioural task). Giving such stimulation
ation when placed in a familiar environment (top), but LTP was in the presence of an NMDA antagonist had no effect on memory.
erased when they explored a novel room. The novelty acts in an (Source: After Brun et al., 1998). D. Application of ZIP (see text)
analogous manner to distraction. (Source: After Xu et al., 1998.) B. reverses established LTP to baseline. E. Rats trained to avoid a zone
Alternatively, if memory traces involve a particular spatial distribu- on a rotating arena where they receive a footshock (place avoidance)
tion of synaptic potentiation, articial induction of further LTP take longer to enter this region as training progresses. Application of
(black symbols) should make it difficult to retrieve the appropriate ZIP at the end of training abolishes this memory. (Source:
memory (1,2,3) later. C. When rats, soon after learning a spatial Pastalkova et al, 2006).
watermaze task, were given high-frequency stimulation to multiple

A Novel room exploration depotentiates LTP B Retrograde alteration of synaptic weights

C Tetanisation after learning impairs memory

D ZIP induced reversal of LTP E ZIP induced loss of place avoidance memory
 Synaptic Plasticity in the Hippocampus 443

time (see Section 10.4.9), causes both a reversal of LTP estab- that in hippocampal LTP we have found a window into the
lished 22 hrs earlier and a loss of place-avoidance memory synaptic basis of a central aspect of cognitionepisodic
(Fig 10-24D,E). With this striking new result, it seems that memory. At the same time, the study of synaptic plasticity has
predictions of the retrograde alteration criterion are being led to a heightened appreciation of the richness and complex-
upheld. ity of synaptic function. Synapses are simply more interesting
when they are endowed with memory. The question remains,
10.10.5 Mimicry however, as to whether this avalanche of new knowledge about
the physiology, pharmacology, and molecular and develop-
A key test of any hypothesis concerning memory encoding mental biology of synapses has taught us anything fundamen-
and storage must be a mimicry experiment, in which apparent tal about how the brain stores memories.
memory is generated articially without the usual require- Many of the issues encountered in the last section are a
ment for a learning experience. This would constitute a criti- product of the painful realization that although we now know
cal test that changes in synaptic efficacy are sufficient for a great deal about plasticity at the level of the individual neu-
memory, rather than merely necessary. ron, and even the individual synapse, we know relatively little
Pairing of electrical stimulation in specic CS and US about how individual synapses integrate to inuence the
pathways to the amygdala should result in potentiation of behavior of the single neuron. We know even less about how
interconnecting pathways. Subsequent behavioral testing networks of neurons interact to encode and store information
might reveal that this LTP constitutes the engineering of an and about the processes of retrieval that allow memory to be
emotional memory. Such an experiment may be difficult in expressed as changes in behavior. The problem of detectabil-
practice, and no such reports have yet appeared; nor have any ity discussed in Section 10.10.2 reects our ignorance of hip-
reports of the artical engineering of hippocampal memory pocampal function at the network level. If LTP is so
been reported. However, several detectability studies have pervasiveestimates cited in Section 10.3.2 suggest that most
employed a methodology that might be regarded as a halfway synapses made by interconnected hippocampal pyramidal
house between detectability and mimicry. Studies in which cells support LTPwhy has it been so difficult to detect in the
sensory stimulation is replaced with electrical stimulation of a intact, learning animal? Does it reect the sparseness of
particular neural pathway, as a discriminative cue (Mouly et encoding, or is it that every instance of LTP is accompanied by
al., 2001; Roman et al., 2004) or a conditioned stimulus LTD at other synapses, thus compromising the experimenters
(Matthies et al., 1986; Laroche et al., 1989; Doyere and ability to detect changes in eld potential responses? A strik-
Laroche, 1992) fall into this category. In such experiments, ing increase in the number of CA1 neurons expressing mRNA
posttraining LTP-like changes in evoked potentials are often transcripts actiivity related gene arc/arg3.1 has been detected
found in response to stimulation of the same pathway that while animals explore a new environment, and it has been
was stimulated during learning. The difficulty lies in ensuring possible to show that different but overlapping networks of
that the electrical stimulation adequately mimics natural sen- arc-expressing cells are engaged when different environments
sory stimulation, such that any learning-related changes in are explored (Guzowski et al., 1999) (see also discussion in
synaptic strength might plausibly be expected to occur Chapter 13). This is a promising approach, but it does not
albeit perhaps in a more diffuse and sparse fashionduring allow us to make the unequivocal claim that LTP has been
normal learning. In this respect, studies involving stimula- detected, as arc and other potential markers such as zif268 and
tion of sensory pathways are arguably easier to interpret than homer are not exclusively activated by LTP. In this respect,
studies in which, for instance, perforant path tetanization is phospho specic antibodies directed to protein kinases that
used as a CS (Laroche et al., 1989; Doyere and Laroche, 1992). are phosphorylated on different residues in LTP and LTD (see
A range of other relevant studies examining the articial section 10.4.6) should offer a powerful tool in the analysis of
induction of receptive eld plasticity in the neocortex has plasticity at the network level.
been reported, but they lie outside the scope of this book (e.g., Estimates of correlated activity between populations of
Talwar and Gerstein, 2001; Weinberger, 2004). Nonetheless, neurons is another promising approach to the question: Does
we are unaware of any study to date that adequately meets the LTP happen in real life? A suggestive example was offered by
mimicry criterion. Mehta et al. (2000), who examined changes in the statistical
coupling of cells as animals repeatedly moved through their
10.10.6 Synaptic Plasticity, Learning, adjacent place elds in a toroidal running maze. Each time the
and Memory: The Story So Far animal traversed the circular maze, entering the CA3 place
eld just before the CA1 place eld, the CA3 place cell repeat-
In this chapter we have covered the more important advances edly red immediately before the CA1 cell. After several cir-
in the eld of synaptic plasticity that have emerged from neu- cuits, the place eld of the CA3 cell moved toward that of the
roscience laboratories around the world over the last three CA1 cell. This is the result that would be predicted if the two
decades as a result of the enormous interest generated by the cells were coupled and the strength of the connection between
discovery of LTP. That interest is driven, at heart, by the belief them had become strengthened. Is it LTP? Does this change in
444 The Hippocampus Book

coupling reect some gain in the animals knowledge of the dence that we now have the conceptual and instrumental tools
maze? Experiments on remapping of place cells when animals to tackle these obstacles. It is this sense of optimism that gives
are put in a new environment suggest that NMDAR-mediated the eld its vitality and excitement. If the eld of conscious-
processes are required for the stability and ne tuning of the ness has its easy and hard problems, so has LTPwith this dif-
new mapping. These are promising but perhaps not com- ference: that its hard problem, the link between plasticity and
pelling examples, an obvious difficulty being that changes in memory, is solvable.
cell ring are an indirect measure of presumed changes in
synaptic connectivity. Nevertheless, techniques for listening
into the conversations between large ensembles of neurons ACKNOWLEDGMENTS
offer one hope of establishing the role of synaptic potentiation
T.V.P.B. thanks those who over the years gave him refuge to write:
in network behavior.
Susan and Robert Alain at Lac Laroche, Quebec and on the north
It is already possible to implant several multiunit recording Norfolk coast Julia Peyton-Jones at Southrepps and Mike and Eithne
assemblies into each of the hippocampal subelds and to Doy at Cley. Wai Han Yau skillfully turned sketches and scans into g-
record the activity of many tens of neurons in each subeld. ures. G.L.C. is grateful to Andy Doherty for help with preparing g-
Some of these neurons will be connected. Cross-correlational ures. R.G.M.M. thanks Stephen Martin for assistance with the text
analysis of unit ring can pick out those that are and allow and Simon Rempel for providing material for one of the gures. The
estimates to be made of the strength of the connection work of T.V.P.B., G.L.C., and R.G.M.M. has been supported for many
between them. The timing of monosynaptic connections years in whole or in part by the MRC. We acknowledge support also
between CA3 and CA1 neurons reduces the possibility that from the Human Frontiers Science Program and the EU.
both cells are being driven by another cell. A protocol to detect
LTP might be the following. Record simultaneously from REFERENCES
many such pairs while the animal is learning a CA3/CA1-
dependent task, and changes in the weights that connect at Abbott LF, Nelson SB (2000) Synaptic plasticity: taming the beast.
least some of the pairs should become apparent. Now depo- Nat Neurosci 3(Suppl):11781183.
tentiate the potentiated pairs by stimulating the presynaptic Abeliovich A, Chen C, Goda Y, Silva AJ, Stevens CF, Tonegawa S
neuron, via the appropriate electrode in the recording assem- (1993) Modied hippocampal long-term potentiation in PKC
bly, soon after the change is detected. Do that to enough pairs gamma-mutant mice. Cell 75:12531262.
Abraham WC (2003) How long will long-term potentiation last?
whose altered weights marks them as participating in the rep-
Philos Trans R Soc Lond B Biol Sci 358:735744.
resentation of the memory, and what has been learned,
Abraham WC (2004) Stress-related phenomena. Hippocampus
according to the SPM hypothesis, should be attenuated or lost. 14:675676.
Now restore the synaptic weights to the pairs that have been Abraham WC, Bear MF (1996) Metaplasticity: the plasticity of synap-
depotentiated by tetanic stimulation, pair on pair, again using tic plasticity. Trends Neurosci 19:126130.
the presynaptic recording electrode as a stimulating electrode, Abraham WC, Mason-Parker SE, Bear MF, Webb S, Tate WP (2001)
and the memory should be recovereda primitive form of Heterosynaptic metaplasticity in the hippocampus in vivo: a
mimicry. This is not, admittedly, the formation of a new BCM-like modiable threshold for LTP. Proc Natl Acad Sci USA
memory but recovery of a lost memory by restoring synaptic 98:1092410929.
strengths specically to a subpopulation of the synapses that Abraham WC, Goddard GV (1983) Asymmetric relationships
encoded the original represention of that memory. Achieving between homosynaptic long-term potentiation and heterosy-
naptic long-term depression. Nature 305:717719.
the stimulation specicity to achieve this project may be diffi-
Abraham WC, Robins A (2005) Memory retention: the synaptic sta-
cult, but thought experiments of this kind challenge us to con-
bility versus plasticity dilemma. Trends Neurosci 28:7378.
sider what would be needed to test the SPM hypothesis with Abraham WC, Tate WP (1997) Metaplasticity: a new vista across the
the appropriate rigor. eld of synaptic plasticity. Prog Neurobiol 52:303323.
Although this and other experiments are for the future, Abraham WC, Bliss TVP, Goddard GV (1985) Heterosynaptic changes
they are within the bounds of what is technically feasible and accompany long-term but not short-term potentiation of the
make at least the point that the SPM hypothesis is potentially perforant path in the anaesthetized rat. J Physiol 363:335349.
falsiable. Before the role of the hippocampus in episodic Abraham WC, Gustafsson B, Wigstrm H (1986) Single high-
memory was revealed in H.M. and before the Hebb synapse strength afferent volleys can produce long-term potentiation in
was made manifest in LTP, neuroscientists who thought about the hippocampus in vitro. Neurosci Lett 70:217222.
the neural basis of learning were faced with what must have Abraham WC, Gustafsson B, Wigstrm H (1987) Hippocampal long-
term potentiation involves an increase in synaptic excitation rel-
seemed a series of unbridgeable chasms between the experi-
ative to inhibition. Int J Neurosci 35:128.
ence to be remembered and its representation and storage in
Abraham WC, Christie BR, Logan B, Lawlor P, Dragunow M (1994)
the neural network, between the network and its constituent Immediate early gene expression associated with the persistence
neurons, and between the neuron and its synapses, the ulti- of heterosynaptic long-term depression in the hippocampus.
mate units of information storage. The enormous advances in Proc Natl Acad Sci USA 91:1004910053.
our knowledge of the mechanisms and functions of synaptic Abraham WC, Mason-Parker SE, Logan B (1996) Low-frequency
plasticity summarized in this chapter can give us some con- stimulation does not readily cause long-term depression or
 Synaptic Plasticity in the Hippocampus 445

depotentiation in the dentate gyrus of awake rats. Brain Res Anwyl R (1999) Metabotropic glutamate receptors: electrophysiolog-
722:217221. ical properties and role in plasticity. Brain Res Rev 29:83120.
Abraham WC, Logan B, Greenwood JM, Dragunow M (2002) Arai A, Kessler M, Xiao P, Ambros-Ingerson J, Rogers G, Lynch G
Induction and experience-dependent consolidation of stable (1994) A centrally active drug that modulates AMPA receptor
long-term potentiation lasting months in the hippocampus. J gated currents. Brain Res 638:343346.
Neurosci 22:96269634. Arai A, Kessler M, Ambros-Ingerson J, Quan A, Yigiter E, Rogers G,
Adams JP, Dudek SM (2005) Late-phase long-term potentiation: get- Lynch G (1996) Effects of a centrally active benzoylpyrrolidine
ting to the nucleus. Nat Rev Neurosci 6:737743. drug on AMPA receptor kinetics. Neuroscience 75:573585.
Aiba A, Chen C, Herrup K, Rosenmund C, Stevens CF, Tonegawa S Artola A, Singer W (1987) Long-term potentiation and NMDA
(1994) Reduced hippocampal long-term potentiation and con- receptors in rat visual cortex. Nature 330:649652
text-specic decit in associative learning in mGluR1 mutant Artola A, Singer W (1993) Long-term depression of excitatory synap-
mice. Cell 79:365375. tic transmission and its relationship to long-term potentiation.
Akaneya Y, Jiang B, Tsumoto T (2005) RNAi-induced gene silencing Trends Neurosci 16:480487.
by local electroporation in targeting brain region. J Neurophysiol Artola A, Kamal A, Ramakers GM, Biessels GJ, Gispen WH (2005)
93:594602. Diabetes mellitus concomitantly facilitates the induction of
Alberini CM (2005) Mechanisms of memory stabilization: are con- long-term depression and inhibits that of long-term potentia-
solidation and reconsolidation similar or distinct processes? tion in hippocampus. Eur J Neurosci 22:169178.
Trends Neurosci 28:5156. Ascher P, Nowak L (1988) The role of divalent cations in the N-
Alford S, Frenguelli BG, Schoeld JG, Collingridge GL (1993) methyl-D-aspartate responses of mouse central neurones in cul-
Characterization of Ca2 signals induced in hippocampal CA1 ture. J Physiol (Lond) 399:247266.
neurones by the synaptic activation of NMDA receptors. J Ashby MC, De La Rue SA, Ralph GS, Uney J, Collingridge GL, Henley
Physiol 469:693716. JM (2004) Removal of AMPA receptors (AMPARs) from
Alle H, Jonas P, Geiger JRP (2001) PTP and LTP at a hippocampal synapses is preceded by transient endocytosis of extrasynaptic
mossy berinterneuron synapse. Proc Natl Acad Sci USA AMPARs. J Neurosci 24:51725176.
98:1470814713. Assaf SY, Chung SH (1984) Release of endogenous Zn2 from brain
Allen PB, Hvalby O, Jensen V, Errington ML, Ramsay M, Chaudhry tissue during activity. Nature 308:734746.
FA, Bliss TV, Storm-Mathisen J, Morris RG, Andersen P, Auger C, Marty A (2000) Quantal currents at single-site central
Greengard P (2000) Protein phosphatase-1 regulation in the synapses. J Physiol (Lond) 526:311
induction of long-term potentiation: heterogeneous molecular Auld DS, Kornecook TJ, Bastianetto S, Quirion R (2002) Alzheimers
mechanisms. J Neurosci 20:35373543. disease and the basal forebrain cholinergic system: relations to
Ambrogi Lorenzini CG, Baldi E, Bucherelli C, Sacchetti B, Tassoni G beta-amyloid peptides, cognition, and treatment strategies. Prog
(1999) Neural topography and chronology of memory consoli- Neurobiol 68:209245.
dation: a review of functional inactivation ndings. Neurobiol Ault B, Evans RH, Francis AA, Oakes DJ, Watkins JC (1980) Selective
Learn Mem 71:118. depression of excitatory amino acid induced depolarizations by
Andersen P (1960) Interhippocampal impulses. II. Apical dendritic magnesium ions in isolated spinal cord preparations. J Physiol
activation of CA1 neurons. Acta Physiol Scand 48:178208. (Lond) 307:413428.
Andersen P, Sundberg SH, Sveen O, Wigstrm H (1977) Specic Avital A, Goshen I, Kamsler A, Segal M, Iverfeldt K, Richter-Levin G,
long-lasting potentiation of synaptic transmission in hip- Yirmiya R (2003) Impaired interleukin-1 signaling is associated
pocampal slices. Nature 266:736737. with decits in hippocampal memory processes and neural
Andersen P, Sundberg SH, Sveen O, Swann JW, Wigstrm H (1980) plasticity. Hippocampus 13:826834.
Possible mechanisms for long-lasting potentiation of synaptic Bach ME, Hawkins RD, Osman M, Kandel ER, Mayford M (1995)
transmission in hippocampal slices from guinea-pigs. J Physiol Impairment of spatial but not contextual memory in CaMKII
302:463482. mutant mice with a selective loss of hippocampal LTP in the
Anderson R, Barnes JC, Bliss TV, Cain DP, Cambon K, Davies HA, range of the theta frequency. Cell 81:905915.
Errington ML, Fellows LA, Gray RA, Hoh T, Stewart M, Large Bagal AA, Kao JP, Tang CM, Thompson SM (2005) Long-term poten-
CH, Higgins GA (1998) Behavioural, physiological and mor- tiation of exogenous glutamate responses at single dendritic
phological analysis of a line of apolipoprotein E knockout spines. Proc Natl Acad Sci USA 102:1443414439.
mouse. Neuroscience 85:93110. Balschun D, Wolfer D, Bertocchini F, Barone V, Conti A, Zuschratter
Aniksztejn L, Roisin MP, Amsellem R, Ben-Ari Y (1989) Long-term W, Missiaen L, Lipp H-P, Frey JU, Sorrentino V (1999) Dele-
potentiation is not associated with a sustained enhanced release tion of the ryanodine receptor type 3 (RyR3) impairs forms
of endogenous excitatory amino acids. Neuroscience 28:387392. of synaptic plasticity and spatial learning. EMBO J 18:
Anis NA, Berry SC, Burton NR, Lodge D (1983) The dissociative 52645273.
anaesthetics, ketamine and phencyclidine, selectively reduce Balschun D, Wolfer DP, Gass P, Mantamadiotis T, Welzl H, Schutz G,
excitation of central mammalian neurones by N-methyl-aspar- Frey JU, Lipp H P (2003) Does cAMP response element-binding
tate. Br J Pharmacol 79:565575. protein have a pivotal role in hippocampal synaptic plasticity and
Anonymous (1997) Mutant mice and neuroscience: recommenda- hippocampus-dependent memory? J. Neurosci 23:63046314.
tions concerning genetic background. Neuron 19:755759. Bannerman DM, Good MA, Butcher SP, Ramsay M, Morris RG
Antonova I, Arancio O, Trillat AC, Wang HG, Zablow L, Udo H, (1995) Distinct components of spatial learning revealed by prior
Kandel ER, Hawkins RD (2001) Rapid increase in clusters of training and NMDA receptor blockade. Nature 378:182186.
presynaptic proteins at onset of long-lasting potentiation. Bannerman DM, Chapman PF, Kelly PAT, Butcher SP, Morris RGM
Science 294:15471450. (1994) Inhibition of nitric-oxide synthase does not prevent the
446 The Hippocampus Book

induction of long-term potentiation in-vivo. J Neurosci Long-term potentiation of NMDA receptor-mediated synaptic
14:74157425. transmission in the hippocampus. Nature 349:156158.
Barbour B, Szatkowski M, Ingledew N, Attwell D (1989) Arachidonic Bashir ZI, Bortolotto ZA, Davies CH, Berretta N, Irving AJ, Seal AJ,
acid induces a prolonged inhibition of glutamate uptake into Henley JM, Jane DE, Watkins JC, Collingridge GL (1993)
glial cells. Nature 342:918920. Induction of LTP in the hippocampus needs synaptic activation
Barco A, Alancon JM, Kandel ER (2002) Expression of constitu- of glutamate metabotropic receptors. Nature 363:347350.
tively active CREB protein facilitates the late phase of long- Bast T, da Silva BM, Morris RG (2005) Distinct contributions of hip-
term potentiation by enhancing synaptic capture. Cell 108: pocampal NMDA and AMPA receptors to encoding and
689703. retrieval of one-trial place memory. J Neurosci 25:58455856.
Barnes CA (1979) Memory decits associated with senescence: a neu- Battistin T, Cherubini E (1994) Developmental shift from long-term
rophysiological and behavioral study in the rat. J Comp Physiol depression to long-term potentiation at mossy ber synapses in
Psychol 93:74104. the rat hippocampus. Eur J Neurosci 6:17501755.
Barnes CA (1995) Involvement of LTP in memory: are we searching Bauer EP, LeDoux JE, Nader K (2001) Fear conditioning and LTP in
under the street light? Neuron 15:751754. the lateral amygdala are sensitive to the same stimulus contin-
Barnes CA, McNaughton BL, Goddard GV, Douglas RM, Adamec R gencies. Nat Neurosci 4:687688.
(1977) Circadian rhythm of synaptic excitability in rat and Bear MF (1996) A synaptic basis for memory storage in the cerebral
monkey central nervous system. Science 197:9192. cortex. Proc Natl Acad Sci USA 93:1345313459.
Barnes CA, McNaughton BL (1985) An age comparison of the rates Bear MF, Linden DJ (2001) The mechanisms and meaning of long-
of acquisition and forgetting of spatial information in relation term synaptic depression in the mammalian brain. In: Synapses
to long-term enhancement of hippocampal synapses. Behav (Cowan WM, Sdhof TC, Stevens CF, eds), pp 455518.
Neurosci 99:10401048. Baltimore: Johns Hopkins University Press.
Barnes CA, Jung MW, McNaughton BL, Korol DL, Andreasson K, Bear MF, Cooper LN, Ebner FF (1987) A physiological basis for a the-
Worley PF (1994) LTP saturation and spatial learning disrup- ory of synapse modication. Science 237:4248.
tion: effects of task variables and saturation levels. J Neurosci Beattie EC, Carroll RC, Yu X, Morishita W, Yasuda H, von Zastrow M,
14:57935806. Malenka RC (2000) Regulation of AMPA receptor endocytosis
Barnes CA, Rao G, McNaughton BL (1996) Functional integrity of by a signaling mechanism shared with LTD. Nat Neurosci
NMDA-dependent LTP induction mechanisms across the life- 3:12911300.
span of F-344 rats. Learn Mem 3:124137. Beattie EC, Stellwagen D, Morishita W, Bresnahan JC, Ha BK, Von
Barnes CA, Rao G, Houston FP (2000) LTP induction threshold Zastrow M, Beattie MS, Malenka RC (2002) Control of synaptic
change in old rats at the perforant pathgranule cell synapse. strength by glial TNFalpha. Science 295:22822285.
Neurobiol Aging 21:613620. Beck H, Goussakov IV, Lie A, Helmstaedter C, Elger CE (2000)
Barria A, Derkach V, Soderling T (1997a) Identication of the Synaptic plasticity in the human dentate gyrus. J Neurosci
Ca2/calmodulin-dependent protein kinase II regulatory phos- 20:70807086.
phorylation site in the alpha-amino-3-hydroxyl-5-methyl-4- Belanger A, Lavoie N, Trudeau F, Massicotte G, Gagnon S (2004)
isoxazole-propionate-type glutamate receptor. J Biol Chem Preserved LTP and watermaze learning in hyperglycaemic-
272:3272732730. hyperinsulinemic ZDF rats. Physiol Behav 83:483494.
Barria A, Malinow R (2005) NMDA receptor subunit composition Bellinger FP, Madamba S, Siggins GR (1993) Interleukin 1 beta
controls synaptic plasticity by regulating binding to CaMKII. inhibits synaptic strength and long-term potentiation in the rat
Neuron 48:289301. CA1 hippocampus. Brain Res 628:227.
Barria A, Muller D, Derkach V, Griffith LC, Soderling TR (1997b) Bellinger FP, Madamba SG, Campbell IL, Siggins GR (1995) Reduced
Regulatory phosphorylation of AMPA-type glutamate receptors long-term potentiation in the dentate gyrus of transgenic
by CaM-KII during long-term potentiation. Science 276: mice with cerebral overexpression of interleukin-6. Neurosci
20422045. Lett198:9598.
Barrionuevo G, Brown T (1983) Associative long-term potentiation Benke TA, Luthi A, Isaac JT, Collingridge GL (1998) Modulation of
in hippocampal slices. Proc Natl Acad Sci USA 70:73477351. AMPA receptor unitary conductance by synaptic activity.
Barrionuevo G, Shottler F, Lynch G (1980) The effects of repetitive Nature 393:793797.
low-frequency stimulation on control and potentiated synap- Benson DL, Schnapp LM, Shapiro ML, Huntley GW (2000) Making
tic responses in the hippocampus. Life Sci 27:23852391. memories stick: cell-adhesion molecules in synaptic plasticity.
Barrow PA, Empson RM, Gladwell SJ, Anderson CM, Killick R, Yu X, Trends Cell Biol 10:473482.
Jefferys JG, Duff K (2000) Functional phenotype in transgenic Berberich S, Punnakkl P, Jensen V, Pawlak V, Seeburg PH, Hvalby O,
mice expressing mutant human presenilin-1. Neurobiol Dis Khr G (2005) Lack of NMDA Receptor Subtype Selectivity for
7:119126. Hippocampal Long-Term Potentiation. J Neurosci 25:69076910.
Bashir ZI (2003) On long-term depression induced by activation of Berger TW (1984) Long-term potentiation of hippocampal synaptic
G-protein coupled receptors. Neurosci Res 45:363367. transmission affects rate of behavioral learning. Science
Bashir ZI, Collingridge GL (1994) An investigation of depotentiation 224:627630.
of long-term potentiation in the CA1 region of the hippocam- Bernabeu R, Bevilaqua L, Ardenghi P, Bromberg E, Schmitz P,
pus. Exp Brain Res 100:437443. Bianchin M, Izquierdo I, Medina JH (1997) Involvement of hip-
Bashir ZI, Tam B, Collingridge GL (1990) Activation of the glycine pocampal cAMP/cAMP-dependent protein kinase signaling
site in the NMDA receptor is necessary for the induction of LTP. pathways in a late memory consolidation phase of aver-sively
Neurosci Lett 108:2616. motivated learning in rats. Proc Natl Acad Sci USA 94:
Bashir ZI, Alford S, Davies SN, Randall AD, Collingridge GL (1991) 70417046.
 Synaptic Plasticity in the Hippocampus 447

Berretta N, Berton F, Bianchi R, Brunelli M, Capogna M, Francesconi Bodensteiner KJ, Cain P, Ray AS, Hamula LA (2006) Effects of preg-
W (1991) Long-term potentiation of NMDA receptor-mediated nancy on spatial cognition in female Hooded Long-Evan rats.
EPSP in guinea-pig hippocampal slices. Eur J Neurosci Horm Behav 49:303314.
3:850854. Boehm J, Kang MG, Johnson RC, Esteban J, Huganir RL, Malinow R
Bi G-Q, Poo M-M (1998) Synaptic modications in cultured hip- (2006) Synaptic incorporation of AMPA receptors during LTP is
pocampal neurons: dependence on spike timing, synaptic controlled by a PKC phosphorylation site on GluR1. Neuron
strength, and postsynaptic cell type. J Neurosci 18:1046410472. 51:21325.
Bi G-Q, Rubin J (2005) Timing in synaptic plasticity: from detection Bohme GA, Bon C, Stutzmann JM, Doble A, Blanchard JC (1991)
to integration. Trends Neurosci 28:222228. Possible involvement of nitric-oxide in long-term potentiation.
Bi P, Poo M-M (2001) Synaptic modication by correlated activity: Eur J Pharm 199:379381.
Hebbs postulate revisited. Annu Rev Neurosci 24:139166. Bolhuis JJ, Reid IC (1992) Effects of intraventricular infusion of the
Bienenstock EL, Cooper LN, Munro PW (1982) Theory for the devel- N-methyl-D-aspartate (NMDDA receptor antgonist AP5 on
opment of neuron selectivity: orientation specicity and binoc- spatial memory of rats in a radial arm maze. Behav Brain Res
ular interaction in visual cortex. J Neurosci 2:3248. 47:151157.
Bischofberger J, Jonas P (2002) TwoB or not twoB: differential trans- Bolshakov V, Siegelbaum S (1994) Postsynaptic induction and presy-
mission at glutamatergic mossy berinterneuron synapses in naptic expression of hippocampal long-term depression. Science
the hippocampus. Trends Neurosci 25:600603. 264:11481151.
Bliss T, Errington M, Fransen E, Godfraind J-M, Kauer J, Kooy RF, Bolshakov VY, Siegelbaum SA (1995) Regulation of hippocampal
Maness PF, Furley AJW (2000) Long-term potentiation in mice transmitter release during development and long-term potenti-
lacking the neural cell adhesion molecule L1. Curr Biol ation. Science 269:17301734.
10:16071610. Bon CL, Garthwaite J (2003) On the role of nitric oxide in hip-
Bliss TVP, Collingridge GL (1993) A synaptic model of memory: pocampal long-term potentiation. J Neurosci 23:19411948.
long-term potentiation in the hippocampus. Nature 361:3139. Bortolotto ZA, Collingridge GL (1993) Characterisation of LTP
Bliss TVP, Dolphin AC (1984) Where is the locus of long-term poten- induced by the activation of glutamate metabotropic receptors
tiation in the hippocampus? In: The neurobiology of learning and in area CA1 of the hippocampus. Neuropharmacology 32:1.
memory (McGaugh JL, Lynch G, eds), pp 451458. New York: Bortolotto ZA, Collingridge GL (2000) A role for protein kinase C in
Guilford Press. a form of metaplasticity that regulates the induction of long-
Bliss TVP, Gardner-Medwin AR (1973) Long-lasting potentiation of term potentiation at CA1 synapses of the adult rat hippocam-
synaptic transmission in the dentate area of the unanaesthetized pus. Eur J Neurosci 12:40554062.
rabbit following stimulation of the perforant path. J Physiol Bortolotto ZA, Bashir ZI, Davies CH, Collingridge GL (1994) A
(Lond) 232:357374. molecular switch activated by metabotropic glutamate recep-
Bliss TVP, Lmo T (1973) Long-lasting potentiation of synaptic trans- tors regulates induction of long-term potentiation. Nature
mission in the dentate area of the anaesthetized rabbit following 368:740743.
stimulation of the perforant path. J Physiol (Lond) 232:331356. Bortolotto ZA, Collingridge GL (1995) On the mechanism of long-
Bliss TVP, Lynch MA (1988) Long-term potentiation of synaptic term potentiation induced by (1S,3R)-1-aminocyclopentane-
transmission in the hippocampus: properties and mechanisms. 1,3-dicarboxylic acid (ACPD) in rat hippocampal slices.
In: Long-term potentiation in the hippocampus: from biophysics to Neuropharmacology 34:10031014.
behavior (Landeld PW, Deadwyler SA, eds), pp 372. New Bortolotto ZA, Collingridge GL (1998) Involvement of calcium/
York: Alan R Liss. calmodulin-dependent protein kinases in the setting of a molec-
Bliss TVP, Richter-Levin G (1993) Spatial-learning and the saturation ular switch involved in hippocampal LTP. Neuropharmacology
of long-term potentiation. Hippocampus 3:123126. 37:535544.
Bliss TVP, Gardner-Medwin AR, Lmo T (1973) Synaptic plasticity in Bortolotto ZA, Clarke VR, Delany CM, Parry MC, Smolders I, Vignes
the hippocampal formation. In: Macromolecules and behavior M, Ho KH, Miu P, Brinton BT, Fantaske R, Ogden A, Gates M,
(Ansell GB, Bradley PB, eds), pp 193203. Baltimore: University Orstein PL, Lodge D, Bleakman D, Collingridge GL (1999)
Park Press. Kainate receptors are involved in synaptic plasticity. Nature 402:
Bliss TVP, Goddard GV, Riives M (1983) Reduction of long-term 297301.
potentiation in the dentate gyrus of the rat following selective Bortolotto ZA, Anderson WW, Isaac JT, Collingridge GL (2002)
depletion of monoamines. J Physiol 334:475491. Synaptic plasticity in the hippocampal slice preparation. In:
Bliss TVP, Errington AL, Lynch A, Williams JH (1990) Presynaptic Current protocols in neuroscience (Crawley JN, Gerfen CR, McKay
mechanisms in hippocampal long-term potentiation. Cold R, Rogawski MA, Sibley DR, Skolnick P, eds). New York: Wiley.
Spring Harbor Symp Quant Biol 55:119129. Bourtchuladze R, Frenguelli B, Blendy J, Cioffi D, Schutz G, Silva AJ
Bliss TVP, Collingridge GL, Morris RGM, (eds) (2003) LTP: enhanc- (1994) Decient long-term memory in mice with a targeted
ing neuroscience for 30 years. Philos Trans R Soc B 358:607842. mutation of the cAMP-responsive element-binding protein. Cell
Bliss TVP, Collingridge Gl, Laroche S (2006) From Zap to Zip: a story 79:5968.
to forget. Science 313:10581059. Bozdagi O, Shan W, Tanaka H, Benson DL, Huntley GW (2000)
Blitzer RD, Wong T, Nouranifar R, Iyengar R, Landau EM (1995) Increasing numbers of synaptic puncta during late-phase LTP:
Postsynaptic cAMP pathway gates early LTP in hippocampal N-cadherin is synthesized, recruited to synaptic sites, and
CA1 region. Neuron 15:14031414. required for potentiation. Neuron 28:245259.
Blitzer RD, Connor JH, Brown GP, Wong T, Shenolikar S, Iyengar G, Bradshaw KD, Emptage NJ, Bliss TVP (2003) A role for dendritic
Landau EM (1998) Gating of CaMKII by cAMP-regulated pro- protein synthesis is hippocampal late LTP. Eur J Neurosci
tein phosphatase activity during LTP. Science 280:19401943. 18:31503152.
448 The Hippocampus Book

Bramham CR, Messaoudi E (2005) BDNF function in adult synaptic impairs long-term potentiation in rat hippocampal slices. J
plasticity: the synaptic consolidation hypothesis. Prog Neurobiol Neurophysiol 88:10731076.
76:99125. Camodeca N, Breakwell NA, Rowan MJ, Anwyl R (1999) Induction of
Bramham CR, Srebro B (1989) Synaptic plasticity in the hippocam- LTD by activation of group I mGluR in the dentate gyrus in
pus is modulated by behavioral state. Brain Res 493:7486. vitro. Neuropharmacology 38:15971606.
Braunewell K-H, Manahan-Vaughan D (2001) Long-term depres- Capecchi MR (1989) Altering the genome by homologous recombi-
sion: a cellular basis for learning? Rev Neurosci 12:121140. nation. Science 244:12881292.
Bredt DS, Nicoll RA (2003) AMPA receptor trafficking at excitatory Caramanos Z, Shapiro ML (1994) Spatial memory and N-methyl-D-
synapses. Neuron 40:361379. aspartate receptor antagonists AP5 and MK-801: memory
Brightwell JJ, Smith CA, Countryman RA, Neve RL, Colombo PJ impairments depend on familiarity with the environment, drug
(2005) Hippocampal overexpression of mutant CREB blocks dose, and training duration. Behav Neurosci 108:3043.
long-term but not short-term memory for a socially transmit- Castillo PE, Janz R, Sudhof TC, Tzounopoulos T, Malenka RC, Nicoll
ted food preference. Learn Mem 12:1217. RA (1997a) Rab3A is essential for mossy ber long-term poten-
Brown TH, McAfee DA (1982) Long-term potentiation in the supe- tiation in the hippocampus. Nature 388:590593.
rior cervical ganglion. Science 215:14111413. Castillo PE, Malenka RC, Nicoll RA (1997b) Kainate receptors medi-
Brun VH, Ytterbo K, Morris RG, Moser MB, Moser EI (2001) Retro- ate a slow postsynaptic current in hippocampal CA3 neurons.
grade amnesia for spatial memory induced by NMDA receptor- Nature 388:182186.
mediated long-term potentiation. J Neurosci 21: 356362. Castillo PE, Schoch S, Schmitz F, Sudhof TC, Malenka RC (2002)
Brun VH, Otnass MK, Molden S, Steffenach HA, Witter MP, Moser RIM1alpha is required for presynaptic long-term potentiation.
MB, Moser EI (2002) Place cells and place recognition main- Nature 425:327330.
tained by direct entorhinal-hippocampal circuitry. Science Castro CA, Silbert LH, McNaughton BL, Barnes CA (1989) Recovery
296:22432246. of spatial learning decits after decay of electrically induced
Buchs PA, Muller D (1996) Induction of long-term potentiation is synaptic enhancement in the hippocampus. Nature 342:
associated with major ultrastructural changes of activated 545548.
synapses. Proc Natl Acad Sci USA 93:80408045. Cavus I, Teyler T (1996) Two forms of long-term potentiation in area
Budde T, Minta A, White JA, Kay AR (1997) Imaging free zinc in CA1 activate different signal transduction cascades. J Neuro-
synaptic terminals in live hippocampal slices. Neuroscience physiol 76:30383047.
79:347358. Chan CS, Weeber EJ, Kurup S, Sweatt JD, Davis RL (2003) Integrin
Burgoyne RD (2004) The neuronal calcium-sensor proteins. Biochim requirement for hippocampal synaptic plasticity and spatial
Biophys Acta 1742:5968. memory. J Neurosci 23:71077116.
Burke SN, Barnes CA (2006) Neural plasticity in the ageing brain. Chapman PF, Kairiss EW, Keenan CL, Brown TH (1990) Long-term
Nat Rev Neurosci 7:3040. potentiation in the amygdala. Synapse 6:271278.
Burrone J, OByrne M, Murthy VN (2002) Multiple forms of synap- Chapman PF, White GL, Jones MW, Cooper-Blacketer D, Marshall VJ,
tic plasticity triggered by selective suppression of activity in Irizarry M, Younkin L, Good MA, Bliss TV, Hyman BT, Younkin
individual neurons. Nature 420:414418. SG, Hsiao KK. (1999) Impaired synaptic plasticity and learning
Bushell TJ, Sansig G, Collett VJ, van der Putten H, Collingridge GL in aged amyloid precursor protein transgenic mice. Nat Neurosci
(2002) Altered short-term synaptic plasticity in mice lacking the 2:271276.
metabotropic glutamate receptor mGlu7. Sci World J 2: Chavez-Noriega LE, Bliss TVP, Halliwell JV (1989) The EPSP-spike
730737. (E-S) component of long-term potentiation in the rat hip-
Butler MP, OConnor JJ, Moynagh PN (2004) Dissection of tumor- pocampal slice is modulated by GABAergic but not cholinergic
necrosis factor-alpha inhibition of long-term potentiation mechanisms. Neurosci Lett 104:5864.
(LTP) reveals a p38 mitogen-activated protein kinase-depend- Chen C, Tonegawa S (1997) Molecular genetic analysis of synaptic
ent mechanism which maps to early- but not late-phase LTP. plasticity, activity-dependent neural development, learning, and
Neuroscience 124:319326. memory in the mammalian brain. Annu Rev Neurosci 20:
Buzsaki G, Haas HL, Anderson EG (1987) Long-term potentiation 157184.
induced by physiologically relevant stimulus patterns. Brain Res Chen L, Chetkovich DM, Petralia RS, Sweeney NT, Kawasaki Y,
435:331333. Wenthold RJ, Bredt DS, Nicoll RA (2000) Stargazin regulates
Caillard O, Ben-Ari Y, Gaiarsa JL (1999) Long-term potentiation of synaptic targeting of AMPA receptors by two distinct mecha-
GABAergic synaptic transmission in neonatal rat hippocampus. nisms. Nature 408:936943.
J Physiol (Lond) 518:109119. Chen N, Luo T, Wellington C, Metzler M, McCutcheon K, Hayden
Cain DP (1989) Long-term potentiation and kindling: how similar MR, Raymond LA (1999) Subtype-specic enhancement of
are the mechanisms? Trends Neurosci 12:610. NMDA receptor currents by mutant huntingtin. J Neurochem
Cain DP, Hargreaves EL, Boon F, Dennison Z (1993) An examination 72:18901898.
of the relations between hippocampal long-term potentiation, Chen QS, Kagan BL, Hirakura Y, Xie CW (2000) Impairment of hip-
kindling, afterdischarge, and place learning in the watermaze. pocampal long-term potentiation by Alzheimer amyloid beta-
Hippocampus 3:153163. peptides. J Neurosci Res 60:6572.
Cain DP, Saucier D, Hall J, Hargreaves EL, Boon F (1996) Detailed Chen Y-L, Huang C-C, Hsu K-S (2001a) Time-dependent reversal of
behavioral analysis of watermaze acquisition under AP5 or long-term potentiation by low-frequency stimulation at the
CNQX: contribution of sensorimotor disturbances to drug- mossy ber-CA3 synapses. J Neurosci 21:27052714.
induced acquisition decits. Behav Neurosci 110:86102. Chen HX, Otmakhov N, Strack S, Colbran RJ, Lisman JE (2001b) Is
Campbell IG, Guinan MJ, Horowitz JM (2002) Sleep deprivation persistent activity of calcium/calmodulin-dependent kinase
 Synaptic Plasticity in the Hippocampus 449

required for the maintenance of LTP? J Neurophysiol 85: Colicos MA, Collins BE, Sailor MJ, Goda Y (2001) Remodeling of
13681376. synaptic actin induced by photoconductive stimulation. Cell
Chetkovich DM, Gray R, Johnston D, Sweatt JD (1991) N-Methyl-D- 107:605616.
aspartate receptor activation increases cAMP levels and voltage- Colledge M, Snyder EM, Crozier RA, Soderling JA, Jin Y, Langeberg
gated Ca2 channel activity in area CA1 of hippocampus. Proc LK, Lu H, Bear MF, Scott JD (2003) Ubiquitination regulates
Natl Acad Sci USA 88:64676471. PSD-95 degradation and AMPA receptor surface expression.
Chetkovich DM, Sweatt JD (1993) NMDA receptor activation Neuron 40:595607.
increases cyclic AMP in area CA1 of the hippocampus via cal- Collingridge GL (1985) Long term potentiation in the hippocampus:
cium/calmodulin stimulation of adenylyl cyclase. J Neurochem mechanisms of initiation and modulation by neurotransmit-
61:19331942. ters. Trends Pharmacol Sci 6:407411.
Chetkovich DM, Chen L, Stocker TJ, Nicoll RA, Bredt DS (2002) Collingridge GL, Kehl SJ, McLennan H (1983) Excitatory amino acids
Phosphorylation of the postsynaptic density-95 (PSD-95)/discs in synaptic transmission in the Schaffer collateralcommissural
large/zona occludens-1 binding site of stargazin regulates bind- pathway of the rat hippocampus. J Physiol 334:3346.
ing to PSD-95 and synaptic targeting of AMPA receptors. J Collingridge GL, Herron CE, Lester RA (1988) Synaptic activation of
Neurosci 22:57915796. N-methyl-D-aspartate receptors in the Schaffer collateral-com-
Chevaleyre V, Castillo PE (2002) Assessing the role of Ih channels in missural pathway of rat hippocampus. J Physiol 399: 283300.
synaptic transmission and mossy ber LTP. Proc Natl Acad Sci Collingridge GL, Isaac JT, Wang YT (2004) Receptor trafficking and
USA 99:95389543. synaptic plasticity. Nat Rev Neurosci 5:952962.
Chevaleyre V, Castillo PE (2003) Heterosynaptic LTD of hippocam- Conquet F, Bashir ZI, Davies CH, Daniel H, Ferraguti F, Bordi F,
pal GABAergic synapses: a novel role of endocannabinoids in Franz-Bacon K, Reggiani A, Matarese V, Conde F, et al (1994)
regulating excitability. Neuron 38:461472. Motor decit and impairment of synaptic plasticity in mice
Chevaleyre V, Castillo PE (2004) Endocannabinoid-mediated meta- lacking mGluR1. Nature 372:237243.
plasticity in the hippocampus. Neuron 43:871881. Contractor A, Swanson G, Heinemann SF (2001) Kainate receptors
Chevaleyre V, Takahashi KA, Castillo PE (2006) Endocannabinoid- are involved in short- and long-term plasticity at mossy ber
mediated synaptic plasticity in the CNS. Annu Rev Neurosci synapses in the hippocampus. Neuron 29:209216.
29:3776. Contractor A, Rogers C, Maron C, Henkemeyer M, Swanson GT,
Cho K, Aggleton JP, Brown MW, Bashir ZI (2001) An experimental Heinemann SF (2002) Trans-synaptic Eph-receptor-ephrin sig-
test of the role of postsynaptic calcium levels in determining nalling in hippocampal mossy ber LTP. Science 296:18641869.
synaptic strength using perirhinal cortex of rat. J Physiol 532: Cooke SF, Bliss TV (2005) Long-term potentiation and cognitive
459466. drug discovery. Curr Opin Invest Drugs 6:2534.
Cho YH, Giese KP, Tanila H, Silva AJ, Eichenbaum H (1998) Abnor- Cooke SF, Bliss TV (2006) Plasticity in the human central nervous
mal hippocampal spatial representations in alphaCaMKIIT286A system. Brain 129:16591673.
and CREBalphaDelta- mice. Science 279:867869. Cooke SF, Wu J, Plattner F, Errington M, Rowan M, Peterson, Hirano
Choi S, Klingauf J, Tsien RW (2000) Postfusional regulation of cleft A, Bradshaw KD, Anwyl R, Bliss TVP, Giese KP (2006)
glutamate concentration during LTP at silent synapses. Nat Autophosphorylation of CaMKII is not a general requirement
Neurosci 3:330. for NMPA receptor-dependent LTP in the adult mouse. J Physiol
Christie BR, Abraham WC (1992) Priming of associative long-term (in press).
depression in the dentate gyrus by theta frequency synaptic Costa AC, Grybko MJ (2005) Decits in hippocampal CA1 LTP
activity. Neuron 9:7984. induced by TBS but not HFS in the Ts65Dn mouse: a model of
Clark KA, Collingridge GL (1995) Synaptic potentiation of dual- Down syndrome. Neurosci Lett 382:317322.
component excitatory postsynaptic currents in the rat hip- Cox PR, Fowler V, Xu B, Sweatt JD, Paylor R, Zoghbi HY (2003) Mice
pocampus. J Physiol 482:3952. lacking tropomodulin-2 show enhanced long-term potentia-
Coan EJ, Saywood W, Collingridge GL (1987) Mk-801 blocks NMDA tion, hyperactivity, and decits in learning and memory. Mol
receptor-mediated synaptic transmission and long-term poten- Cell Neurosci 23:112.
tiation in rat hippocampal slices. Neurosci Lett 80:111114. Cremer H, Chazal G, Carleton A, Goridis C, Vincent JD, Lledo P
Coan EJ, Irving AJ, Collingridge GL (1989) Low-frequency activation (1998) Long-term but not short-term plasticity at mossy ber
of the NMDA receptor system can prevent the induction of LTP. synapses is impaired in neural cell adhesion molecule-decient
Neurosci Lett 105:205210. mice. Proc Natl Acad Sci USA 95:1324213247.
Cohen AS, Abraham WC (1996) Facilitation of long-term potentia- Crpel F, Jaillard D (1991) Pairing of pre- and postsynaptic activities
tion by prior activation of metabotropic glutamate receptors. J in cerebellar Purkinje cells induces long-term changes in synap-
Neurophysiol 76:953962. tiic efficacy in vitro. J Physiol 432:123141.
Colbert CM, Levy WB (1993) Long-term potentiation of perforant Crick F (1984) Memory and molecular turnover. Nature 312:101.
path synapses in hippocampal CA1 in vitro. Brain Res 606:8791. Cull-Candy SG, Usowicz MM (1987) Multiple-conductance channels
Colbran RJ, Brown AM (2004) Calcium/calmodulin-dependent pro- activated by excitatory amino acids in cerebellar neurons.
tein kinase II and synaptic plasticity. Curr Opin Neurobiol Nature 325:525528.
14:318327. Cullen WK, Wu J, Anwyl R, Rowan MJ (1996) Beta-amyloid produces
Cole BJ, Klewer M, Jones GH, Stephens DN (1993) Contrasting a delayed NMDA receptor-dependent reduction in synaptic
effects of the competitive NMDA antagonist CPP and non- transmission in rat hippocampus. Neuroreport 8:8792.
competitive NMDA antagonist MK801 on performance of an Cullen WK, Suh YH, Anwyl R, Rowan MJ (1997) Block of LTP in rat
operant delayed matching to position task in rats. Pyscho- hippocampus in vivo by beta-amyloid precursor protein frag-
pharmacology (Berl) 111:465471. ments. Neuroreport 8:32133217.
450 The Hippocampus Book

Cummings JA, Nicola SM, Malenka RC (1994) Induction in the rat Davies J, Francis AA, Jones AW, Watkins JC (1981) 2-Amino-5-phos-
hippocampus of long-term potentiation (LTP) and long-term phonovalerate (2AP5), a potent and selective antagonist of
depression (LTD) in the presence of a nitric oxide synthase amino acid-induced and synaptic excitation. Neurosci Lett
inhibitor Neurosci Lett 1994 18:110114 21:7781.
Cummings JA, Mulkey RM, Nicoll RA, Malenka RC (1996) Ca2 sig- Davies SN, Lester RA, Reymann KG, Collingridge GL (1989)
nalling requirements for long-term depression in the hip- Temporally distinct pre- and post-synaptic mechanisms main-
pocampus. Neuron 16:825833. tain long-term potentiation. Nature 338:500503.
Cunningham AJ, Murray CA, ONeill LA, Lynch MA, OConnor JJ Davis C, Harding JW, Wright JW (2003). REM sleep deprivation-
(1996) Interleukin-1 beta (IL-1 beta) and tumour necrosis fac- induced decits in the latency-to-peak induction and mainte-
tor (TNF) inhibit long-term potentiation in the rat dentate nance of long-term potentiation within the CA1 region of the
gyrus in vitro. Neurosci Lett 203:1720. hippocampus. Brain Res 973:293297.
Curran B, OConnor JJ (2001) The pro-inammatory cytokine inter- Davis HP, Squire LR (1984) Protein synthesis and memory: a review.
leukin-18 impairs long-term potentiation and NMDA receptor- Psychol Bull 96:518559.
mediated transmission in the rat hippocampus in vitro. Davis S, Butcher SP, Morris RG (1992) The NMDA receptor antago-
Neuroscience 108:8390. nist D-2-amino-5-phosphonopentanoate (D-AP5) impairs spa-
Curran BP, Murray HJ, OConnor JJ (2003) A role for c-Jun N-termi- tial learning and LTP in vivo at intracerebral concentrations
nal kinase in the inhibition of long-term potentiation by inter- comparable to those that block LTP in vitro. J Neurosci 12:2134.
leukin-1beta and long-term depression in the rat dentate gyrus Davis S, Vanhoutte P, Pages C, Caboche J, Laroche S (2000) The
in vitro. Neuroscience 118:347357. MAPK/ERK cascade targets both Elk-1 and cAMP response ele-
Dan Y, Poo M-M (2004) Spike timing-dependent plasticity of neural ment-binding protein to control long-term potentiation-
circuits. Neuron 44:2230. dependent gene expression in the dentate gyrus in vivo. J
Dana RC, Martinez JL (1984) Effect of adrenalectomy on the circa- Neurosci 20:45634572.
dian rhythm of LTP. Brain Res 308:392395. Daw MI, Chittajallu R, Bortolotto ZA, Dev KK, Duprat F, Henley JM,
Danscher G, Fjerdingstad EJ, Fjerdingstad E, Fredens K (1976) Heavy Collingridge GL, Isaac JT (2000) PDZ proteins interacting with
metal content in subdivisions of the rat hippocampus (zinc, C-terminal GluR2/3 are involved in a PKC-dependent regula-
lead, copper). Brain Res 112:442446. tion of AMPA receptors at hippocampal synapses. Neuron
Danysz W, Parsons CG, Mobius HJ, Stoffler A, Quack G (2000) 28:873876.
Neuroprotective and symptomatological action of memantine Daw MI, Bortolotto ZA, Saulle E, Zaman S, Collingridge GL, Isaac JT
relevant for Alzheimers disease - a unied glutamatergic (2002) Phosphatidylinositol 3 kinase regulats synapse specicity
hypothesis on the mechanism of action. Neurotox Res 2:8597. of hippocampal long-term depression. Nat Neurosci 5:835836.
Danysz W, Wroblewski JT, Costa E (1988) Learning impairment in Day M, Good M (2005) Ovariectomy-induced disruption of long-
rats by N-methyl-D-aspartate receptor antagonist. Neurophar- term synaptic depression in the hippocampal CA1 region in
macology 27:653656. vivo is attenuated with chronic estrogen replacement. Neurobiol
Danysz W, Zajaczkowski W, Parsons CG (1995) Modulation of learn- Learn Mem 83:1321.
ing processes by ionotropic glutamate receptor ligands. Behav Day M, Langston RF (2006) Post-training N-methyl-D-aspartate
Pharmacol 6:455474. receptor blockade offers protection from retrograde interfer-
Daoudal G, Hanada Y, Debanne D (2002) Bidirectional plasticity of ence but does not affect consolidation of weak or strong mem-
excitatory postsynaptic potential (EPSP)-spike coupling in CA1 ory traces in the watermaze. Neuroscience 137:1928.
hippocampal pyramidal neurons. Proc Natl Acad Sci USA Day M, Langston R, Morris RG (2003) Glutamate-receptor-mediated
99:1451214517. encoding and retrieval of paired-associate learning. Nature
DArcangelo G, Grassi F, Ragozzino D, Santoni A, Tancredi V, Eusebi 424:205209.
F (1991) Interferon inhibits synaptic potentiation in rat hip- Dayan P, Willshaw DJ (1991) Optimising synaptic learning rules in
pocampus. Brain Res 564:245248. linear associative memories. Biol Cybern 65:253265.
Daumas S, Halley H, Frances B, Lassalle JM (2005) Encoding, con- Debanne D, Gahwiler BH, Thompson SM (1994) Asynchronous
solidation, and retrieval of contextual memory: differential presynaptic and postsynaptic activity induces associative long-
involvement of dorsal CA3 and CA1 hippocampal subregions. term depression in area CA1 of the rat hippocampus in vitro.
Learn Mem 12:375382. Proc Natl Acad Sci USA 91:11481152.
Davies CH, Collingridge GL (1993) The physiological regulation of Debanne D, Gahwiler BH, Thompson SM (1998) Long-term synap-
synaptic inhibition by GABAB autoreceptors in rat hippocam- tic plasticity between pairs of individual CA3 pyramidal cells in
pus. J Physiol 472:245265. rat hippocampal slice cultures. J Physiol (Lond) 507:237247.
Davies CH, Collingridge GL (1996) Regulation of EPSPs by the Debanne D, Gahwiler BH, Thompson SM (1999) Heterogeneity of
synaptic activation of GABAB autoreceptors in rat hippocam- synaptic plasticity at unitary CA3CA1 and CA3CA3 connec-
pus. J Physiol 496:451470. tions in rat hippocamal slice cultures. J Neurosci 19:
Davies SN, Collingridge GL (1989) Role of excitatory amino acid 1066410671.
receptors in synaptic transmission in area CA1 of rat hip- Deisseroth K, Bito H, Schulman H, Tsien RW (1995) A molecular
pocampus. Proc R Soc B 236:373384. mechanism of metaplasticity. Curr Biol 5:13341338.
Davies CH, Davies SN, Collingridge GL (1990) Paired-pulse depres- Deisseroth K, Bito H, Tsien RW (1996) Signaling from synapse to
sion of monosynaptic GABA-mediated inhibitory postsynaptic nucleus: postsynaptic CREB phosphorylation during multiple
responses in rat hippocampus. J Physiol 424:513531. forms of hippocampal synaptic plasticity. Neuron 16:89101.
Davies CH, Starkey SJ, Pozza MF, Collingridge GL (1991) GABA-B Deisseroth K, Heist EK, Tsien RW (1998) Translocation of calmod-
autoreceptors regulate the induction of LTP. Nature 349: ulin to the nucleus supports CREB phosphorylation in hip-
609611. pocampal neurons. Nature 392:198202.
 Synaptic Plasticity in the Hippocampus 451

Derkach V, Barria A, Soderling TR (1999) Ca2/calmodulin-kinase II Domenici MR, Berretta N, Cherubini E (1998) Two distinct forms of
enhances channel conductance of alpha-amino-3-hydroxy-5- long-term depression coexist at the mossy berCA3 synapse in
methyl-4-isoxazolepropionate type glutamate receptors. Proc the hippocampus during development. Proc Natl Acad Sci USA
Natl Acad Sci USA 96:32693274. 95:83108315.
Derrick BE, Martinez JL (1994a) Opioid receptor activation is one Dong H, OBrien RJ, Fung ET, Lanahan AA, Worley PF, Huganir RL
factor underlying the frequency dependence of mossy ber LTP. (1997) GRIP: a synaptic PDZ domain-containing protein that
J Neurosci 14:43594367. interacts with AMPA receptors. Nature 386:279284.
Derrick BJ, Martinez JL (1994b) Frequency-dependent associative Douglas RM, Goddard GV (1975) Long-term potentiation of the
long-term potentiation at the hippocampal mossy ber-CA3 perforant pathgranule cell synapse in the rat hippocampus.
synapse. Proc Natl Acad Sci USA 91:1029010294. Brain Res 86:205215.
Derrick BE, Martinez JL (1996) Associative, bidirectional modica- Douglas RM, Goddard GV, Riives M (1982) Inhibitory modulation of
tions at the hippocampal mossy ber-CA3 synapse. Nature long-term potentiation: evidence for a postsynaptic locus of
381:429434. control. Brain Res 240:259272.
Desmond NL, Levy WB (1983) Synaptic correlates of associative Doyre V, Laroche S (1992) Linear relationship between the mainte-
potentiation/depression: an ultrastructural study in the hip- nance of hippocampal long-term potentiation and retention of
pocampus. Brain Res 265:2130. an associative memory. Hippocampus 2:3948.
Desmond NL, Levy WB (1986a) Changes in the postsynaptic density Doyre V, Errington ML, Laroche S, Bliss TVP (1996) Low-frequency
with long-term potentiation in the dentate gyrus. J Comp Neurol trains of paired stimuli induce long-term depression in area
253:476482. CA1 but not in dentate gyrus of the intact rat. Hippocampus
Desmond NL, Levy WB (1986b) Changes in the numerical density of 6:5257.
synaptic contacts with long-term potentiation in the hippocam- Doyre V, Srebro B, Laroche S (1997) Heterosynaptic LTD and depo-
pal dentate gyrus. J Comp Neurol 253:466475. tentiation in the medial perforant path in the freely moving rat.
Desmond NL, Levy WB (1990) Morphological correlates of long- J Neurophysiol 77:571578.
term potentiation imply the modication of existing synapses, Doyle CA, Cullen WK, Rowan MJ, Anwyl R (1997) Low-frequency
not synaptogenesis, in the hippocampal dentate gyrus. Synapse stimulation induces homosynaptic depotentiation but not long-
5:139143. term depression of synaptic transmission in the adult anaes-
DeSouza S, Fu J, States BA, Ziff EB (2002) Differential palmitoylation thetized and awake rat hippocampus in vivo. Neuroscience
directs the AMPA receptor-binding protein ABP to spines or to 77:7585.
intracellular clusters. J Neurosci 22:34933503. Dozmorov M, Li R, Abbas AK, Hellberg F, Farre C, Huang FS, Jilderos
Deupree DL, Turner DA, Watters CL (1991) Spatial performance cor- B, Wigstrom H (2006) Contribution of AMPA and NMDA
relates with in vitro potentiation in young and aged Fischer 344 receptors to early and late phases of LTP in hippocampal slices.
rats. Brain Res 554:19. Neurosci Res 55:182188.
Dev KK, Nishimune A, Henley JM, Nakanishi S (1999) The protein Dragoi G, Harris KD, Buzsaki G (2003) Place representation within
kinase C alpha binding protein PICK1 interacts with short but hippocampal networks is modied by long-term potentiation.
not long form alternative splice variants of AMPA receptor sub- Neuron 39:843853.
units. Neuropharmacology 38:635644. Dragunow M (1996) A role for immediate-early transcription factors
Diamond DM, Rose GM (1994) Does associative LTP underlie classi- in learning and memory. Behav Genet 26:293299.
calal conditioning? Psychobiology 22:263269. Du J-L, Poo M-M (2004) Rapid BDNF-induced retrograde modica-
Diamond DM, Bennett MC, Fleshner M, Rose GM (1992) Inverted- tion in a developing retinotectal system. Nature 429:878883.
U relationship between the level of peripheral corticosterone Dudai Y, Morris RGM (2000) To consolidate or not to consolidate:
and the magnitude of hippocampal primed burst potentiation. what are the questions? In: Brain, perception, memory (Bolhuis
Hippocampus 2:421430. JJ, ed), pp 149162. Oxford: Oxford University Press.
Diamond JS, Bergles DE, Jahr CE (1998) Glutamate release moni- Dudek SM, Bear MF (1992) Homosynaptic long-term depression
tored with astrocyte transporter currents during LTP. Neuron and effects of N-methyl-D-aspartate receptor blockade. Proc
21:425433. Natl Acad Sci USA 89:43634367.
Dieguez D, Jr., Barea-Rodriguez EJ (2004) Aging impairs the late Dudek SM, Bear MF (1993) Bidirectional long-term modication of
phase of long-term potentiation at the medial perforant synaptic effectiveness in the adult and immature hippocampus.
pathCA3 synapse in awake rats. Synapse 52:5361. J Neurosci 13:29102918.
Dineley KT, Westerman M, Bui D, Bell K, Ashe KH, Sweatt JD (2001) Dudek SM, Fields RD (2002) Somatic action potentials are sufficient
Beta-amyloid activates the mitogen-activated protein kinase for late-phase LTP-related signaling. Proc Natl Acad Sci USA
cascade via hippocampal alpha7 nicotinic acetylcholine recep- 99:39623967.
tors: in vitro and in vivo mechanisms related to Alzheimers dis- Duffy SN, Craddock KJ, Abel T, Nguyen PV (2001) Environmental
ease. J Neurosci 21:41254133. enrichment modies the PKA-dependence of hippocampal LTP
Dingledine R, Hynes MA, King GL (1986) Involvement of N-methyl- and improves hippocampus-dependent memory. Learn Mem
D-aspartate receptors in epileptiform bursting in the rat hip- 8:2634.
pocampal slice. J Physiol 380:175189. Duffy JN, Nguyen PV (2003) Postsynaptic application of a peptide
Do VH, Martinez CO, Martinez JL Jr, Derrick BE (2002) Long-term inhibitor of cAMP-dependent protein kinase blocks expression
potentiation in direct perforant path projections to the hip- of long-lasting synaptic potentiation in hippocampal neurons. J
pocampal CA3 region in vivo. J Neurophysiol 87: 669678. Neurosci 23:1142.
Dolphin AC, Errington ML, Bliss TV (1982) Long-term potentiation Dumuis A, Sebben M, Haynes L, Pin JP, Bockaert J (1988) NMDA
of the perforant path in vivo is associated with increased gluta- receptors activate the arachidonic acid cascade system in striatal
mate release. Nature 297:496498. neurons. Nature 336:6870.
452 The Hippocampus Book

Dunwiddie T, Lynch G (1978) Long-term potentiation and depres- trols synaptic trafficking underlying plasticity. Nat Neurosci
sion of synaptic responses in the rat hippocampus: localization 6:136143.
and frequency dependency. J Physiol (Lond) 276:353367. Evans RH, Francis WW, Watkins JC (1977) Selective antagonism by
Durand GM, Kovalchuk Y, Konnerth A (1996) Long-term potentia- Mg2 of amino acid-induced depolarization of spinal neurones.
tion and functional synapse induction in developing hippocam- Experientia 15:489491.
pus. Nature 381:7175. Fan Y, Fricker D, Brager DH, Chen X, Lu HC, Chitwood RA, Johnston
Eaton SA, Jane DE, Jones PL, Porter RH, Pook PC, Sunter DC, D (2005) Activity-dependent decrease of excitability in rat hip-
Udvarhelyi PM, Roberts PJ, Salt TE, Watkins JC (1993) pocampal neurons through increases in I(h). Nat Neurosci
Competitive antagonism at metabotropic glutamate receptors 8:15421551.
by (S)-4-carboxyphenylglycine and (RS)-alpha-methyl-4-car- Fanselow MS, Kim JJ, Yipp J, De Oca B (1994) Differential effects of
boxyphenylglycine. Eur J Pharmacol 244:195197. the N-methyl-D-aspartate antagonist DL-2-amino-5-phospho-
Eberwine J, Miyashiro K, Kacharmina JE, Job C (2001) Local transla- novalerate on acquisition of fear of auditory and contextual
tion of classes of mRNAs that are targeted to neuronal den- cues. Behav Neurosci 108:235240.
drites. Proc Natl Acad Sci USA 98:70807085. Feinmark SJ, Begum R, Tsvetkov E, Goussakov IV, Funk CD,
Ehrlich I, Malinow R (2004) Postsynaptic density 95 controls Siegelbaum SA, Bolshakov VY (2003) 12-Lipoxygenase metabo-
AMPA receptor incorporation during long-term potentiation lites of arachidonic acid mediate metabotropic glutamate recep-
and experience-driven synaptic plasticity. J Neurosci 24: tor-dependent long-term depression at hippocampal CA3CA1
916927. synapses. J Neurosci 23:1142711435.
Ekstrom AD, Meltzer J, McNaughton BL, Barnes CA (2001) NMDA Feng B, Tse HW, Skifter DA, Morley R, Jane DE, Monaghan DT
receptor antagonism blocks experience-dependent expansion of (2004) Structure-activity analysis of a novel NR2C/NR2D-
hippocampal place elds. Neuron 31:631638. preferring NMDA receptor antagonist: 1-(phenanthrene-2-
El Husseini A, Schnell E, Dakoji S, Sweeney N, Zhou Q, Prange O, carbonyl) piperazine-2,3-dicarboxylic acid. Br J Pharmacol
Gauthier-Campbell C, Aguilera-Moreno A, Nicoll RA, Bredt DS 141:508516.
(2002) Synaptic strength regulated by palmitate cycling on PSD- Fifkova E, Van Harreveld A (1977) Long-lasting morphological
95. Cell 108:849863. changes in dendritic spines of dentate granular cells following
Emptage N, Bliss TV, Fine A (1999) Single synaptic events evoke stimulation of the entorhinal area. J Neurocytol 6:211230.
NMDA receptor-mediated release of calcium from internal Fischer M, Kaech S, Knutti D, Matus A (1998) Rapid actin-based
stores in hippocampal dendritic spines. Neuron 22:115124. plasticity in dendritic spines. Neuron 20:847854.
Emptage NJ, Reid CA, Fine A, Bliss TVP (2003) Optical quantal Fitzgerald M (2005) The development of nociceptive circuits. Nat
analysis reveals a presynaptic component of LTP at hippocam- Rev Neurosci 6:507520.
pal Schaffer-associational synapses. Neuron 38:797804. Fitzjohn SM, Bortolotto ZA, Palmer MJ, Doherty AJ, Ornstein PL,
Engert F, Bonhoeffer T (1997) Synapse specicity of long-term poten- Schoepp DD, Kingston AE, Lodge D, Collingridge GL (1998)
tiation breaks down at short distances. Nature 388:279284. The potent mGlu receptor antagonist LY341495 identies roles
Engert F, Bonhoeffer T (1999) Dendritic spine changes associated for both cloned and novel mGlu receptors in hippocampal
with hippocampal long-term synaptic plasticity. Nature synaptic plasticity. Neuropharmacology 37:14451458.
399:6670. Fitzjohn SM, Morton RA, Kuenzi F, Rosahl TW, Shearman M, Lewis
English JD, Sweatt JD (1996) Activation of p42 mitogen-activated H, Smith D, Reynolds DS, Davies CH, Collingridge GL,
protein kinase in hippocampal long term potentiation. J Biol Seabrook GR (2001a) Age-related impairment of synaptic
Chem 271:2432924332. transmission but normal long-term potentiation in transgenic
English JD, Sweatt JD (1997) A requirement for the mitogen-acti- mice that overexpress the human APP695SWE mutant form of
vated protein kinase cascade in hippocampal long-term poten- amyloid precursor protein. J Neurosci 21:4691.
tiation. J Biol Chem 272:1910319106. Fitzjohn SM, Palmer MJ, May JE, Neeson A, Morris SA, Collingridge
Ernfors P, Bramham CR (2003) The coupling of a trkB tyrosine GL (2001b) A characterization of long-term depression induced
residue to LTP. Trends Neurosci 26:171173. by metabotropic glutamate receptor activation in the rat hip-
Errington ML, Lynch MA, Bliss TVP (1987) Long-term potentiation pocampus in vitro. J Physiol 537:421430.
in the dentate gyrus: induction and increased glutamate release Fitzsimonds RM, Song H-J, Poo M-M (1997) Propagation of activity-
are blocked by D(-)aminophosphonovalerate. Neuroscience dependent synaptic depression in simple neural networks.
20:279284. Nature 388:439448.
Errington ML, Bliss TVP, Richter-Levin G, Yenk K, Doyere V, Laroche Florian C, Roullet P (2004) Hippocampal CA3-region is crucial for
S (1995) Stimulation at 1-5 Hz does not produce long-term acquisition and memory consolidation in Morris watermaze
depression or depotentiation in the hippocampus of the adult task in mice. Behav Brain Res 154:365374.
rat in-vivo. J Neurophysiol 74:17931799. Fonseca R, Nageri UV, Morris RGM, Bonhoeffer T (2004) Competing
Errington ML, Bliss TVP, Morris RJ, Laroche S, Davis S (1997) Long- for memory: hippocampal LTP under regimes of reduced pro-
term potentiation in awake mutant mice. Nature 387:666667. tein synthesis. Neuron 44:10111020.
Errington ML, Galley PT, Bliss TVP (2003). Long-term potentiation Foy MR, Stanton ME, Levine S, Thompson RF (1987) Behavioral
in the dentate gyrus of the anaesthetized rat is accompanied by stress impairs long-term potentiation in rodent hippocampus.
an increase in extracellular glutamate: real-time measurements Behav Neural Biol 48:138149.
using a novel dialysis electrode. Philos Trans R Soc B 358: Frankiewicz T, Parsons CG (1999) Memantine restores long term
675687. potentiation impaired by tonic N-methyl-D-aspartate (NMDA)
Esteban JA, Shi SH, Wilson C, Nuriya M, Huganir RL, Malinow R receptor activation following reduction of Mg2 in hippocam-
(2003) PKA phosphorylation of AMPA receptor subunits con- pal slices. Neuropharmacology 38:12531259.
 Synaptic Plasticity in the Hippocampus 453

Frederickson CJ, Danscher G (1990) Zinc-containing neurons in Retention of NMDA receptor NR2 subunits in the lumen of
hippocampus and related CNS structures. Prog Brain Res endoplasmic reticulum in targeted NR1 knockout mice. Proc
83:7184. Natl Acad Sci USA 100:48554860.
Freir DB, Holscher C, Herron CE (2001) Blockade of long-term Fukazawa Y, Saitoh Y, Ozawa F, Ohta Y, Mizuno K, Inokuchi K (2003)
potentiation by beta-amyloid peptides in the CA1 region of the LTP is accompanied by enhanced F-actin content within the
rat hippocampus in vivo. J Neurophysiol 85:708713. dendritic spine that is essential for late LTP maintenance in vivo.
French PJ, OConnor V, Jones MW, Davis S, Errington ML, Voss K, Neuron 38:447460.
Truchet B, Wotjak C, Stean T, Doyere V, Maroun M, Laroche S, Fukunaga K, Muller D, Miyamoto E (1995) Increased phosphoryla-
Bliss TVP (2001) Subeld-specic immediate early gene expres- tion of Ca2/calmodulin-dependent protein kinase II and its
sion associated with hippocampal long-term potentiation in endogenous substrates in the induction of long-term potentia-
vivo. Eur J Neurosci 13:958976. tion. J Biol Chem 270:61196124.
Frey U, Morris RGM (1997) Synaptic tagging and long-term potenti- Furth PA, St Onge L, Boger H, Gruss P, Gossen M, Kistner A, Bujard
ation. Nature 385:533536. H, Hennighausen L (1994) Temporal control of gene expression
Frey U, Morris RGM (1998a) Synaptic tagging: implications for late in transgenic mice by a tetracycline-responsive promoter. Proc
maintenance of hippocampal long-term potentiation. TINS Natl Acad Sci USA 91:93029306.
21:181188. Futatsugi A, Kato K, Ogura H, Li ST, Nagata E, Kuwajima G, Tanaka
Frey U, Morris RGM (1998b) Weak before strong: dissociating synap- K, Itohara S, Mikoshiba K (1999) Facilitation of NMDAR-inde-
tic-tagging and plasticity-factor accounts of late-LTP. pendent LTP and spatial learning in mutant mice lacking ryan-
Neuropharmacology 37:545552. odine receptor type 3. Neuron 24:701713.
Frey U, Krug M, Reymann KG, Matthies H (1988) Anisomycin, an Gall CM, Lynch G (2004) Integrins, synaptic plasticity and epilepto-
inhibitor of protein synthesis, blocks late phases of LTP phe- genesis. Adv Exp Med Biol 548:1233.
nomena in the hippocampal CA1 region in vitro. Brain Res Gall CM, Hess US, Lynch G (1998) Mapping brain networks
452:5765. engaged by, and changed by, learning. Neurobiol Learn Mem
Frey U, Krug M, Brodemann R, Reymann K, Matthies H (1989) 70:1436.
Long-term potentiation induced in dendrites separated from Gallagher SM, Daly CA, Bear MF, Huber KM (2004) Extracellular
rats CA1 pyramidal somata does not establish a late phase. signal-regulated protein kinase activation is required for
Neurosci Lett 97:135-139. metabotropic glutamate receptor-dependent long-term depres-
Frey U, Schroeder H, Matthies H (1990) Dopaminergic antagonists sion in hippocampal area CA1. J Neurosci 24:48594864.
prevent long-term maintenance of posttetanic LTP in the CA1 Garthwaite J, Charles SL, Chess-Williams R (1988) Endothelium-
region of rat hippocampal slices. Brain Res 522:6975. derived relaxing factor release on activation of NMDA receptors
Frey U, Huang YY, Kandel ER (1993) Effects of cAMP simulate a late suggests role as intercellular messenger in the brain. Nature
stage of LTP in hippocampal CA1 neurons. Science 260: 336:385388.
16611664. Gasbarri A, Sulli A, Packard MG (1997) The dopaminergic mesen-
Frey U, Schollmeier K, Reymann KG, Seidenbecher T (1995) cephalic projections to the hippocampal formation in the rat.
Asymptotic hippocampal long-term potentiation in rats does Prog Neuropsychopharmacol Biol Psychiatry 21:122.
not preclude additional potentiation at later phases. Neuro- Ge W-P, Yang X-J, Zhang Z, Wang H-K, Shen W, Deng Q-P, Duan
science 67:799807. S (2006) Long-term potentiation of neuron-glia synapses
Frey U, Frey S, Schollmeier F, Krug M (1996) Inuence of actinomy- mediated by Ca2-permeable AMPA receptors. Science 312:
cin-D, an RNA-synthesis inhibitor, on long-term potentiation in 15331537.
rat hippocampal neurons in vivo and in vitro. J Physiol (Lond) Geinisman Y, de Toledo-Morrell L, Morrell F (1991) Induction of
490:703711. long-term potentiation is associated with an increase in the
Frey S, Bergado-Rosado J, Seidenbecher T, Pape HC, Frey JU (2001) number of axospinous synapses with segmented postsynaptic
Reinforcement of early long-term potentiation (early-LTP) in densities. Brain Res 566:7788.
dentate gyrus by stimulation of the basolateral amygdala: het- Geinisman Y, deToledo-Morrell L, Morrell F, Heller RE, Rossi M,
erosynaptic induction mechanisms of late-LTP. J Neurosci Parshall RF (1993) Structural synaptic correlate of long-term
21:36973703. potentiation: formation of axospinous synapses with multiple,
Frick A, Magee JC, Johnston D (2004) LTP is accompanied by an completely partioned transmission zones. Hippocampus
enhanced local excitability of pyramidal neuron dendrites. Nat 3:435446.
Neurosci 7:126135. Gnin A, French P, Doyere V, Davis M, Errington ML, Maroun M,
Friedlander MJ, Fregnac Y, Burke JP (1993) Temporal covariance Stean T, Truchet B, Webber M, Wills T, Richter-Levin G, Sanger
of postsynaptic membrane potential and synaptic input: G, Hunt SP, Mallet J, Laroche S, Bliss TVP, OConnor V (2003)
role in synaptic efficacy in visual cortex. Prog Brain Res 95: LTP but not seizure is associated with up-regulation of AKAP-
191205. 150. Eur J Neurosci 17:331340.
Froemke RC, Dan Y (2002) Spike-timing-dependent synaptic Gerlai R (1996) Gene-targeting studies of mammalian behaviour: is
modication induced by natural spike trains. Nature 416: it the mutation or the background genotype? Trends Neurosci
433438. 19:177181.
Fujii S, Saito K, Miyakawa H, Ito K, Kato K (1991) Reversal of long- Gerlai R, Henderson JT, Roder JC, Jia Z (1998) Multiple behavioral
term potentiation (depotentiation) induced by tetanus stimula- anomalies in GluR2 mutant mice exhibiting enhanced LTP.
tion of the input to CA1 neurons of guinea pig hippocampal Behav Brain Res 95:3745.
slices. Brain Res 555:112122. Gianotti C, Nunzi M G, Gispen W H, Corradetti R (1992) Phospho-
Fukaya M, Kato A, Lovett C, Tonegawa S, Watanabe M (2003) rylation of the presynaptic protein B-50 (GAP-43) is increased
454 The Hippocampus Book

during electrically induced long-term potentiation. Neuron Gustafsson B, Asztely F, Hanse E, Wigstrm H (1989) Onset charac-
8:843848. teristics of long-term potentiation in the guinea-pig hippocam-
Gibson HE, Reim K, Brose N, Morton AJ, Jones S (2005) A similar pal ca1 region invitro. Eur J Neurosci 1:382394.
impairment in CA3 mossy ber LTP in the R6/2 mouse model Guzowski JF, McGaugh JL (1997) Antisense oligodeoxynucleotide-
of Huntingtons disease and in the complexin II knockout mediated disruption of hippocampal cAMP response element
mouse. Eur J Neurosci 22:17011712. binding protein levels impairs consolidation of memory for
Giese KP, Fedorov NB, Filipkowski RK, Silva AJ (1998) Autopho- watermaze training. Proc Natl Acad Sci USA 94:26932698.
sphorylation at Thr(286) of the alpha calcium-calmodulin Guzowski JF, McNaughton BL, Barnes CA, Worley PF (1999)
kinase II in LTP and learning. Science 279:870873. Environment-specic expression of the immediate-early gene
Goddard GV (1967) Development of epileptic seizures through brain Arc in hippocampal neuronal ensembles. Nat Neurosci
stimulation at low intensity. Nature 214:10201021. 2:11201124.
Goelet P, Castellucci VF, Schacher S, Kandel ER (1986) The long and Guzowski JF, Lyford GL, Stevenson GD, Houston FP, McGaugh JL,
the short of long-term memorya molecular framework. Worley PF, Barnes CA (2000) Inhibition of activity-dependent
Nature 322:419422. arc protein expression in the rat hippocampus impairs the
Golding NL, Staff NP, Spruston N (2003) Dendritic spikes as a mech- maintenance of long-term potentiation and the consolidation
anism for cooperative long-term potentiation. Nature of long-term memory. J Neurosci 20:39934001.
418:326331. Haley JE, Schaible E, Pavlidis P, Murdock A, Madison DV (1996)
Good M, Day M, Muir JL (1999) Cyclical changes in endogenous lev- Basal and apical synapses of CA1 pyramidal cells employ differ-
els of oestrogen modulate the induction of LTD and LTP in the ent LTP induction mechanisms. Learn Mem 3:289295.
hippocampal CA1 region. Eur J Neurosci 11:44764480. Haley JE, Wilcox GL, Chapman PF (1992) The role of nitric oxide in
Gooney M, Messaoudi E, Maher CO, Bramham CR, Lynch MA hippocampal long-term potentiation. Neuron 8:211216.
(2004) BDNF-induced LTP in dentate gyrus is impaired with Hanley JG, Henley JM (2005) PICK1 is a calcium-sensor for NMDA-
age: analysis of changes in cell signaling events. Neurobiol Aging induced AMPA receptor trafficking. EMBO J 24:32663278.
25:13231331. Hanse E, Gustafsson B (1992) Postsynaptic, but not presynaptic,
Grant SG, Karl KA, Kiebler MA, Kandel ER (1995) Focal adhesion activity controls the early time course of long-term potentiation
kinase in the brain: novel subcellular localization and specic in the dentate gyrus. J Neurosci 12:32263240.
regulation by Fyn tyrosine kinase in mutant mice. Genes Dev Hardingham GE, Arnold FJ, Bading H (2001) Nuclear calcium sig-
9:19091921. naling controls CREB-mediated gene expression triggered by
Green EJ, Greenough WT (1986) Altered synaptic transmission in synaptic activity. Nat Neurosci 4:261267.
dentate gyrus of rats reared in complex enviornments: evidence Harris EW, Cotman CW (1986) Long-term potentiation of guinea
from hippocampal slices maintained in vitro. J Neurophysiol pig mossy ber responses is not blocked by N-methyl-D-aspar-
55:739751. tate antagonists. Neurosci Lett 70:132137.
Green EJ, McNaughton BL, Barnes CA (1990) Exploration-depend- Harris EW, Ganong AH, Cotman CW (1984) Long-term potentiation
ent modulation of evoked responses in fascia dentata: dissocia- in the hippocampus involves activation of N-methyl-D-aspar-
tion of motor, EEG, and sensory factors and evidence for a tate receptors. Brain Res 323:132137.
synaptic efficacy change. J Neurosci 10:14551471. Harris KM (1999) Structure, development, and plasticity of dendritic
Grosshans DR, Clayton DA, Coultrap SJ, Browning MD (2002) LTP spines. Curr Opin Neurobiol 9:343348.
leads to rapid surface expression of NMDA but not AMPA Harris KM, Stevens JK (1989) Dendritic spines of CA1 pyramidal
receptors in adult rat CA1. Nat Neurosci 5:2733. cells in the rat hippocampus: serial electron microscopy with
Grover LM, Teyler TJ (1990) Two components of long-term potenti- reference to their biophysical characteristics. J Neurosci
ation induced by different patterns of afferent activation. Nature 9:29822997.
347:477479. Harris KM, Teyler TJ (1983) Age differences in a circadian inuence
Gruart A, Dolores Munoz M, Delgado-Garcia JM (2006) on hippocampal LTP. Brain Res 261:6973.
Involvement of the CA3-CA1 synapse in the acquisition of asso- Harris KM, Fiala JC, Ostroff L (2003) Structural changes at dendritic
ciative learning in behaving mice. J Neurosci 26:10771087. spine synapses during long-term potentiation. Philos Trans R
Grunwald IC, Korte M, Adelmann G, Plueck A, Kullander K, Adams Soc Lond B Biol Sci 358:745748.
RH, Frotscher M, Bonhoeffer T, Klein R (2004) Hippocampal Harrison NL, Gibbons SJ (1994) Zn2: an endogenous modulator of
plasticity requires postsynaptic ephrinBs. Nat Neurosci 7:3340. ligand- and voltage-gated ion channels. Neuropharmacology
Grutzendler J, Kasthuri N, Gan WB (2002) Long-term dendritic 33:935952.
spine stability in the adult cortex. Nature 420:751752. Harvey J, Collingridge GL (1992) Thapsigargin blocks the induction
Gu Y, McIlwain KL, Weeber EJ, Yamagata T, Xu B, Antalffy BA, Reyes of long-term potentiation in rat hippocampal slices. Neurosci
C, Yuva-Paylor L, Armstrong D, Zoghbi H, Sweatt JD, Paylor R, Lett 139:197200.
Nelson DL (2002) Impaired conditioned fear and enhanced Harvey J, Balasubramaniam R, Collingridge GL (1993) Carbachol
long-term potentiation in Fmr2 knock-out mice. J Neurosci can potentiate N-methyl-D-aspartate responses in the rat hip-
22:27532763. pocampus by a staurosporine and thapsigargin-insensitive
Gustafsson B, Wigstrm H (1986) Hippocampal long lasting poten- mechanism. Neurosci Lett 162:165168.
tiation produced by pairing single volleys and brief condition- Hastings NB, Gould E (1999) Rapid extension of axons into the CA3
ing tetani evoked in separate afferents. J Neurosci 6:15751582. region by adult-generated granule cells. J Comp Neurol 413:
Gustafsson B, Wigstrm H, Abraham WC, Huang YY (1987) Long- 146154.
term potentiation in the hippocampus using depolarizing cur- Haug FM (1967) Electron microscopical localization of the zinc in
rent pulses as the conditioning stimulus to single volley synaptic hippocampal mossy ber synapses by a modied sulde silver
potentials. J Neurosci 7:774780. procedure. Histochemie 8:355368.
 Synaptic Plasticity in the Hippocampus 455

Hawkins RD, Son H, Arancio O (1998) Nitric oxide as a retrograde plasmic toxicity, and selective striatal neurodegeneration.
messenger during long-term potentiation in hippocampus. Prog Neuron 23:181192.
Brain Res 118:155172. Holscher C, Gigg J, OMara SM (1999) Metabotropic glutamate
Hayashi Y, Shi SH, Esteban JA, Piccini A, Poncer JC, Malinow R receptor activation and blockade: their role in long-term poten-
(2000) Driving AMPA receptors into synapses by LTP and tiation, learning and neurotoxicity. Neurosci Biobehav Rev
CaMKII: requirement for GluR1 and PDZ domain interaction. 23:399410.
Science 287:22622267. Hoh T, Beiko J, Boon F, Weiss S, Cain DP (1999) Complex behavioral
Hebb DO (1949) The organization of behavior. New York: Wiley. strategy and reversal learning in the watermaze without NMDA
Hebbard PC, King RR, Malsbury CW, Harley CW (2003) Two orga- receptor-dependent long-term potentiation. J Neurosci 19:RC2.
nizational effects of pubertal testosterone in male rats: transient Hopkins WF, Johnston D (1988) Noradrenergic enhancement of
social memory and a shift away from long-term potentiation long-term potentiation at mossy ber synapses in the hip-
following a tetanus in hippocampal CA1. Exp Neurol pocampus. J Neurophysiol 59:667687.
182:470475. Hopper R, Lancaster B, Garthwaite J (2004) On the regulation of
Henderson JT, Georgiou J, Jia Z, Robertson J, Elowe S, Roder JC, NMDA receptors by nitric oxide. Eur J Neurosci 19:16751682.
Pawson T (2001) The receptor tyrosine kinase EphB2 regulates Horwood JM, Dufour F, Laroche S, Davis S (2006) Signalling mech-
NMDA-dependent synaptic function. Neuron 32:10411056. anisms mediated by the phosphoinositide 3-kinase/Akt cascade
Hendricson AW, Miao CL, Lippmann MJ, Morrisett RA (2002) in synaptic plasticity and memory in the rat. Eur J Neurosci
Ifenprodil and ethanol enhance NMDA receptor-dependent 23:33753384.
long-term depression. J Pharmacol Exp Ther 301:938944. Hosokawa T, Rusakov DA, Bliss TVP, Fine A (1995) Repeated confo-
Henze DA, Urban NN, Barrionuevo G (2000) The multifarious hip- cal imaging of individual dendritic spines in the living hip-
pocampal mossy ber pathway: a review. Neuroscience 98: pocampal slice: evidence for changes in length and orientation
407427. associated with chemically-induced LTP. J Neurosci 15:
Hernandez AI, Blace N, Crary JF, Serrano PA, Leitges M, Libien JM, 55605573.
Weinstein G, Tcherapanov A, Sacktor TC (2003) Protein kinase Hou Q, Gao X, Zhang X, Kong L, Wang X, Bian W, Tu Y, Jin M, Zhao
M zeta synthesis from a brain mRNA encoding an indepen- G, Li B, Jing N, Yu L (2004) SNAP-25 in hippocampal CA1
dent protein kinase C zeta catalytic domain. Implications for region is involved in memory consolidation. Eur J Neurosci
the molecular mechanism of memory. J Biol Chem 278: 20:15931603.
4030540316. Hou L, Klann E (2004) Activation of the phosphoinositide 3-kinase-
Hernandez RV, Navarro MM, Rodriquez WA, Martinez JL, LeBaron Akt-mammalian target of rapamycin signaling pathway is
RG (2005) Differences in the magnitude of long-term potentia- required for metabotropic glutamate receptor-dependent long-
tion produced by theta burst and high frequency stimulation term depression. J Neurosci 24:63526361.
protocols matched in stimulus number. Brain Res Protocols Howell GA, Welch MG, Frederickson CJ (1984) Stimulation-induced
15:613. uptake and release of zinc in hippocampal slices. Nature
Herron CE, Williamson R, Collingridge GL (1985) A selective N- 308:736738.
methyl-D-aspartate antagonist depresses epileptiform activity in Hrabetova S, Serrano P, Blace N, Tse HW, Skifter DA, Jane DE,
rat hippocampal slices. Neurosci Lett 61:255260. Monaghan DT, Sacktor TC (2000) Distinct NMDA receptor
Herron CE, Lester RA, Coan EJ, Collingridge GL (1986) Frequency- subpopulations contribute to long-term potentiation and long-
dependent involvement of NMDA receptors in the hippocam- term depression induction. J Neurosci 20:RC81, 8186.
pus: a novel synaptic mechanism. Nature 322:265268. Hsia AY, Salin PA, Castillo PE, Aiba A, Abeliovich A, Tonegawa
Hessler NA, Shirke AM, Malinow R (1993) The probability of trans- S, Nicoll RA (1995) Evidence against a role for metabotropic
mitter release at a mammalian central synapse. Nature 366: glutamate receptors in mossy ber LTP: the use of mutant mice
569572. and pharmacological antagonists. Neuropharmacology 34:
Hevroni D, Rattner A, Bundman M, Lederfein D, Gabarah 15671572.
A, Mangelus M, Silverman MA, Kedar H, Naor C, Kornuc Hsia AY, Masliah E, McConlogue L, Yu GQ, Tatsuno G, Hu K,
M, Hanoch T, Seger R, Theill LE, Nedivi E, Richter-Levin G, Kholodenko D, Malenka RC, Nicoll RA, Mucke L (1999)
Citri Y (1998) Hippocampal plasticity involves extensive gene Plaque-independent disruption of neural circuits in Alzheimers
induction and multiple cellular mechanisms. J Mol Neurosci disease mouse models. Proc Natl Acad Sci USA 96:32283233.
10:7598. Hsiao K, Chapman P, Nilsen S, Eckman C, Harigaya Y, Younkin S,
Heynen AJ, Abraham WC, Bear MF (1996) Bidirectional modica- Yang F, Cole G (1996) Correlative memory decits, Ab eleva-
tion of CA1 synapses in the adult hippocampus in vivo. Nature tion, and amyloid plaques in transgenic mice. Science 274:
381:163166. 99102.
Heynen AJ, Quinlan EM, Bae DC, Bear MF (2000) Bidirectional, Hrabetova S, Serrano P, Blace N, Tse HW, Skifter DA, Jane DE,
activity-dependent regulation of glutamate receptors in the Monaghan DT, Sacktor TC (2000) Distinct NMDA receptor
adult hippocampus in vivo. Neuron 28:527536. subpopulations contribute to long-term potentiation and long-
Hinds HL, Tonegawa S, Malinow R (1998) CA1 long-term potentia- term depression induction. J Neurosci 20:RC81, 8186.
tion is diminished but present in hippocampal slices from Hu GY, Hvalby O, Walaas SI, Albert KA, Skjeo P, Andersen P,
alpha-CaMKII mutant mice. Learn Mem 5:344354. Greengard P (1987) Protein kinase C injection into hippocam-
Hodgson JG, Agopyan N, Gutekunst CA, Leavitt BR, LePiane F, pal pyramidal cells elicits features of long term potentiation.
Singaraja R, Smith DJ, Bissada N, McCutcheon K, Nasir J, Jamot Nature 328:426429.
L, Li XJ, Stevens ME, Rosemond E, Roder JC, Phillips AG, Rubin Huang CC, Hsu K-S (2001) Progress in understanding the factors
EM, Hersch SM, Hayden MR (1999) A YAC mouse model for regulating reversibility of long-term potentiation. Rev Neurosci
Huntingtons disease with full-length mutant huntingtin, cyto- 12:5168.
456 The Hippocampus Book

Huang CC, Hsu KS (2006) Sustained activation of metabotropic glu- Isaac JT, Nicoll RA, Malenka RC (1995) Evidence for silent synapses:
tamate receptor 5 and protein tyrosine phosphatases mediate implications for the expression of LTP. Neuron 15:427434.
the expression of (S)-3,5-dihydroxyphenylglycine-induced Isaac JT, Luthi A, Palmer MJ, Anderson WW, Benke TA, Collingridge
long-term depression in the hippocampal CA1 region. J GL (1998) An investigation of the expression mechanism of LTP
Neurochem 96:179194. of AMPA receptor-mediated synaptic transmission at hip-
Huang CC, Luang YC, Hsu K-S (2001) Characterization of the mech- pocampal CA1 synapses using failures analysis and dendritic
anism underlying the reversal of long term potentiation by low recordings. Neuropharmacology 37:13991410.
frequency stimulation at hippocampal CA1 synapses. J Biol Ito I, Hidaka H, Sugiyama H (1991) Effects of KN-62, a specic
Chem 2276:4810848117. inhibitor of calcium/calmodulin-dependent protein kinase II,
Huang CC, Chen Y-L, Liang Y-C, Hsu K-S (2002) Role for cAMP and on long-term potentiation in the rat hippocampus. Neurosci
protein phosphatase in the presynaptic expression of mouse Lett 121:119121.
hippocampal mossy ber depotentiation. J Physiol (Lond) Izquierdo I, Medina JH (1997) Memory formation: the sequence of
543:767778. biochemical events in the hippocampus and its connection to
Huang CC, You JL, Wu MY, Hsu KS (2004) Rap1-induced p38 mito- activity in other brain structures. Neurobiol Learn Mem
gen-activated protein kinase activation facilitates AMPA recep- 68:285316.
tor trafficking via the GDI.Rab5 complex. Potential role in Izquierdo I, Medina JH, Bianchin M, Walz R, Zanatta MS, Da Silva
(S)-3,5-dihydroxyphenylglycene-induced long term depression. RC, Bueno e Silva M, Ruschel AC, Paczko N (1993) Memory
J Biol Chem 279:1228612292. processing by the limbic system: role of specic neurotransmit-
Huang CS, Song-Hai S, Ule J, Ruggiu M, Barker LA, Darnell RB, Jan ter systems. Behav Brain Res 58:9198.
YN, Jan LY (2005) Common molecular pathways mediate long- Izquierdo I, Izquierdo LA, Barros DM, Mello e Souza T, de Souza
term potentiation of synaptic excitation and slow synaptic inhi- MM, Quevedo J, Rodrigues C, SantAnna MK, Madruga M,
bition. Cell 123:105118. Medina JH (1998) Differential involvement of cortical receptor
Huang Y, Lu W, Ali DW, Pelkey KA, Pitcher GM, Lu YM, Aoto H, mechanisms in working, short-term and long-term memory.
Roder JC, Sasaki T, Salter MW, MacDonald JF (2001b) Behav Pharmacol 9:421427.
CAKbeta/Pyk2 kinase is a signaling link for induction of long- Izquierdo I, Schroder N, Netto CA, Medina JH (1999) Novelty
term potentiation in CA1 hippocampus. Neuron 29:485496. causes time-dependent retrograde amnesia for one-trial avoid-
Huang YY, Colino A, Selig DK, Malenka RC (1992) The inuence of ance in rats through NMDA receptor- and CaMKII-dependent
prior synaptic activity on the induction of long-term potentia- mechanisms in the hippocampus. Eur J Neurosci 11:33233328.
tion. Science 255:730733. Izumi Y, Zorumski CF (1999) Norepinephrine promotes long-term
Huang YY, Kandel ER (1994) Recruitment of long-lasting and pro- potentiation in the adult rat hippocampus in vitro. Synapse
tein kinase A-dependent long-term potentiation in the CA1 31:196202.
region of hippocampus requires repeated tetanization. Learn Jaffard R, Vouimba RM, Marighetto A, Garcia R (1996) Long-
Mem 1:7482. term potentiation and long-term depression in the lateral sep-
Huang YY, Kandel ER (1995) D1/D5 receptor agonists induce a pro- tum in spatial working and reference memory. J Physiol
tein synthesis-dependent late potentiation in the CA1 region of 90:339341.
the hippocampus. Proc Natl Acad Sci USA 92:24462450. Jaffe D, Johnston D (1990) Induction of long-term potentiation at
Huang YY, Li XC, Kandel ER (1994) cAMP contributes to mossy ber hippocampal mossy-ber synapses follows a Hebbian rule. J
LTP by initiating both a covalently mediated early phase and Neurophysiol 64:948961.
macromolecular synthesis-dependent late phase. Cell 79:6979. Jahr CE, Stevens CF (1987) Glutamate activates multiple single
Hber KM, Kayser MS, Bear MF (2000) Role for rapid dendritic pro- channel conductances in hippocampal neurons. Nature 325:
tein synthesis in hippocampal mGluR-dependent long-term 522525.
depression. Science 288:12541257. Jahr CE, Stevens CF (1993) Calcium permeability of the N-methyl-D-
Hber KM, Roder JC, Bear MF (2001) Chemical induction of aspartate receptor channel in hippocampal neurons in culture.
mGluR5-and protein synthesis-dependent long-term depres- Proc Natl Acad Sci USA 90:1157311577.
sion in hippocampal area CA1. J Neurophysiol 86:321325. Jasnow AM, Shi C, Israel JE, Davis M, Huhman KL (2005) Memory
Huntley GW (2002) Dynamic aspects of cadherin-mediated plastic- of social defeat is facilitated by cAMP response element-binding
ity in synapse development and plasticity. Biol Cell 94:335344. protein overexpression in the amygdala. Behav Neurosci
Hvalby O, Lacaille JC, Hu GY, Andersen P (1987) Postsynaptic long- 119:11251130.
term potentiation follows coupling of dendritic glutamate Jeffery KJ (1997) LTP and spatial learningwhere to next?
application and synaptic activation. Experientia 43:599601. Hippocampus 7:95110.
Hyman SE, Malenka RC (2001) Addiction and the brain: the neuro- Jeffery KJ, Morris RGM (1993) Cumulative long-term potentiation in
biology of compulsion and its persistence. Nat Rev Neurosci the rat dentate gyrus correlates with, but does not modify, per-
2:695703. formance in the watermaze. Hippocampus 3:133140.
Ikeda H, Heinke B, Ruscheweyh R, Sandkuhler J (2003) Synaptic Jeon D, Yang YM, Jeong MJ, Philipson KD, Rhim H, Shin HS (2003)
plasticity in spinal lamina I projection neurons that mediate Enhanced learning and memory in mice lacking Na/Ca2
hyperalgesia. Science 299:12371240. exchanger 2. Neuron 38:965976.
Impey S, Obrietan K, Wong ST, Poser S, Yano S, Wayman G, Jester JM, Campbell LW, Sejnowski TJ (1995) Associative EPSP-spike
Deloulme JC, Chan G, Storm DR (1998) Cross talk between potentiation induced by pairing orthodromic and antidromic
ERK and PKA is required for Ca2 stimulation of CREB- stimulation in rat hippocampal slices. J Physiol (Lond)
dependent transcription and ERK nuclear translocation. 484:589705.
Neuron 21:869883. Jia Y, Jeng JM, Sensi SL, Weiss JH (2002) Zn2 currents are mediated
 Synaptic Plasticity in the Hippocampus 457

by calcium-permeable AMPA/kainate channels in cultured (2004) Glutamate release increases during mossy-CA3 LTP but
murine hippocampal neurones. J Physiol 543:3548. not during Schaffer-CA1 LTP. Eur J Neurosci 19:15911600.
Jia Z, Agopyan N, Miu P, Xiong Z, Henderson J, Gerlai R, Taverna FA, Kazanietz MG, Caloca MJ, Eroles P, Fujii T, Garcia-Bermejo ML,
Velumian A, MacDonald J, Carlen P, Abramow-Newerly W, Reilly M, Wang H (2000) Pharmacology of the receptors for the
Roder J (1996) Enhanced LTP in mice decient in the AMPA phorbol ester tumor promoters: multiple receptors with differ-
receptor GluR2. Neuron 17:945956. ent biochemical properties. Biochem Pharmacol 60:14171424.
Jia Z, Lu Y, Henderson J, Taverna F, Romano C, Abramow-Newerly W, Keith JR, Rudy JW (1990) Why NMDA-receptor-dependent long-
Wojtowicz JM, Roder J (1998) Selective abolition of the NMDA term potentiation may not be a mechanism of learning and
component of long-term potentiation in mice lacking mGluR5. memory: reappraisal of the NMDA-receptor blockade strategy.
Learn Mem 5:331343. Psychobiology 18:251257.
Jones MW, Errington ML, French PJ, Fine A, Bliss TVP, Garel S, Kelleher RJI, Govindarajan A, Tonegawa S (2004a) Tranlational regu-
Charnay P, Bozon B, Laroche S, Davis S (2001) A requirement latory mechanisms in persistent forms of synaptic plasticity.
for the immediate early gene Zif268 in the expression of late LTP Neuron 44:5973.
and long-term memories. Nat Neurosci 3:289296. Kelleher RJI, Govindarajan A, Jung H-Y, Kang H, Tonegawa S (2004b)
Josselyn SA, Shi C, Carlezon WA Jr, Neve RL, Nestler EJ, Davis M Translational control by MAPK signaling in long-term synaptic
(2001) Long-term memory is facilitated by cAMP response ele- plasticity and memory. Cell 116:467479.
ment-binding protein overexpression in the amygdala. J Kelly A, Lynch MA (2000) Long-term potentiation in dentate gyrus
Neurosci 21:24042412. of the rat is inhibited by the phosphoinositide 3-kinase
Jun K, Choi G, Yang SG, Choi KY, Kim H, Chan GC, Storm DR, Albert inhibitor, wortmannin. Neuropharmacology 39:643651.
C, Mayr GW, Lee CJ, Shin HS (1998) Enhanced hippocampal Kelly A, Mullany PM, Lynch MA (2000) Protein synthesis in entorhi-
CA1 LTP but normal spatial learning in inositol 1,4,5-trisphos- nal cortex and long-term potentiation in dentate gyrus.
phate 3-kinase(A)-decient mice. Learn Mem 5:317330. Hippocampus 10:431437.
Katz KW, Rose CR, Thoenen H, Konnerth A (1999) Neurotrophin- Kelly A, Lynch A, Vereker E, Nolan Y, Queenan P, Whittaker E, ONeill
evoked rapid excitation through TrkB receptors. Nature LA, Lynch MA (2001) The anti-inammatory cytokine, inter-
401:918921. leukin (IL)-10, blocks the inhibitory effect of IL-1 beta on long
Kaksonen M, Pavlov I, Voikar V, Lauri SE, Hienola A, Riekki R, Lakso term potentiation: a role for JNK. J Biol Chem 276:
M, Taira T, Rauvala H (2002) Syndecan-3-decient mice exhibit 4556445572.
enhanced LTP and impaired hippocampus-dependent memory. Kelly A, Vereker E, Nolan Y, Brady M, Barry C, Loscher CE, Mills KH,
Mol Cell Neurosci 21:158172. Lynch MA (2003) Activation of p38 plays a pivotal role in the
Kameyama K, Lee HK, Bear MF, Huganir RL (1998) Involvement of inhibitory effect of lipopolysaccharide and interleukin-1 beta
a postsynaptic protein kinase A substrate in the expression of on long term potentiation in rat dentate gyrus. J Biol Chem
homosynaptic long-term depression. Neuron 21:11631175. 278:1945319462.
Kandler K, Katz LC, Kauer JA (1998) Focal photolysis of caged gluta- Kelso SR, Brown TH (1986) Differential conditioning of associative
mate produces long-term depression of hippocampal glutamate synaptic enhancement in hippocampal brain slices. Science
receptors. Nat Neurosci 1:119123. 232:8587.
Kang HJ, Schuman EM (1995) Neurotrophin-induced modulation of Kelso SR, Ganong AH, Brown TH (1986) Hebbian synapses in hip-
synaptic transmission in the adult hippocampus. J Physiol Paris pocamus. Proc Natl Acad Sci USA 83:53265330.
89:1122. Kemp N, Bashir ZI (2001) Long-term depression: a cascade of induc-
Kang H, Schuman EM (1996) A requirement for local protein syn- tion and expression mechanisms. Prog Neurobiol 65:339365.
thesis in neurotrophin-induced hippocampal synaptic plastic- Kemp A, Manahan-Vaughan D (2004) Hippocampal long-
ity. Science 273:14021406. term depression and long-term potentiation encode different
Kapur A, Yeckel MF, Gray R, Johnston D (1998) L-Type calcium aspects of novelty acquisition. Proc Natl Acad Sci USA 101:
channels are required for one form of hippocampal mossy ber 81928197.
LTP. J Neurophysiol 79:21812190. Kemp N, McQueen J, Faulkes S, Bashir ZI (2000) Different forms of
Kasai H, Matsuzaki M, Noguchi K (2003) Structure-stability-func- LTD in the CA1 region of the hippocampus: role of age and
tion relationships of dendritic spines. Trends Neurosci stimulus protocol. Eur J Neurosci 12:360366.
26:360368. Kentros C, Hargreaves E, Hawkins RD, Kandel ER, Shapiro M, Muller
Katagiri H, Tanaka K, Manabe T (2001) Requirement of appropriate RV (1998) Abolition of long-term stability of new hippocampal
glutamate concentrations in the synaptic cleft for hippocampal place cell maps by NMDA receptor blockade. Science
LTP induction. Eur J Neurosci 14:547553. 280:21212126.
Katsuki H, Nakai S, Hirai Y, Akaji K, Kiso Y, Satoh M (1990) Kerchner GA, Wilding TJ, Huettner JE, Zhuo M (2002) Kainate
Interleukin-1 beta inhibits long-term potentiation in the CA3 receptor subunits underlying presynaptic regulation of trans-
region of mouse hippocampal slices. Eur J Pharmacol 181: mitter release in the dorsal horn. J Neurosci-22:80108017.
323326. Kerr DS, Abraham WC (1993) Associative stimulation does not facil-
Kauderer BS, Kandel ER (2000) Capture of a protein synthesis- itate long-term depression in region CA1 of the hippocampus.
dependent component of long-term depression. Proc Natl Acad Synapse 14:305313.
Sci USA 97:1334213347. Kida S, Josselyn SA, de Ortiz SP, Kogan JH, Chevere I, Masushige S,
Kauer JA, Malenka RC, Nicoll RA (1988) A persistent postsynaptic Silva AJ (2002) CREB required for the stability of new and reac-
modication mediates long-term potentiation in the hip- tivated fear memories. Nat Neurosci 5:348355.
pocampus. Neuron 1:911917. Kim CH, Chung HJ, Lee HK, Huganir RL (2001) Interaction of the
Kawamura Y, Manita S, Nakamura T, Inoue M, Kudo Y, Miyakawa H AMPA receptor subunit GluR2/3 with PDZ domains regulates
458 The Hippocampus Book

hippocampal long-term depression. Proc Natl Acad Sci USA Konorski J (1948) Conditioned reexes and neuron organization.
98:1172511730. Cambridge, UK: Hefner.
Kim CH, Takamiya K, Petralia RS, Sattler R, Yu S, Zhou W, Kalb R, Korol DL, Abel TW, Church LT, Barnes CA, McNaughton BL (1993)
Wenthold R, Huganir R (2005a) Persistent hippocampal CA1 Hippocampal synaptic enhancement and spatial learning in the
LTP in mice lacking the C-terminal PDZ ligand of GluR1. Nat Morris swim task. Hippocampus 3:127132.
Neurosci 8:985987. Kombian SB, Malenka RC (1994) Simultaneous LTP of non-NMDA-
Kim EY, Mahmoud GS, Grover LM (2005b) REM sleep deprivation and LTD of NMDA-receptor-mediated responses in the nucleus
inhibits LTP in vivo in area CA1 of the rat hippocampus. accumbens. Nature 368:242246.
Neurosci Lett 388:163167. Korte M, Carroll P, Wolf E, Brem G, Thoenen H, Bonhoeffer T (1995)
Kim JJ, Diamond DM (2002) The stressed hippocampus, synaptic Hippocampal long-term potentiation is impaired in mice lack-
plasticity and lost memories. Nat Rev Neurosci 3:453462. ing brain-derived neurotrophic factor. Proc Natl Acad Sci USA
Kim JJ, Koo J-W, Lee H-J, Han J-S (2005c) Amygdalar inactivation 92:88568860.
blocks stress-induced impairments in hippocampal long-term Korte M, Griesbeck O, Gravel C, Carroll P, Staiger V, Thoenen H,
potentiation and spatial memory. J Neurosci 25:15321539. Bonhoeffer T (1996) Virus-mediated gene transfer into hip-
Kirov SA, Sorra KE, Harris KM (1999) Slices have more synapses pocampal CA1 region restores long-term potentiation in brain-
than perfusion-xed hippocampus from both young and derived neurotrophic factor mutant mice. Proc Natl Acad Sci
mature rats. J Neurosci 19:28762886. USA 93:1254712552.
Kistner A, Gossen M, Zimmermann F, Jerecic J, Ullmer C, Lubbert H, Kovalchuk Y, Eilers J, Lisman J, and Konnerth A (2000). NMDA
Bujard H (1996) Doxycycline-mediated quantitative and tissue- receptor-mediated subthreshold Ca(2) signals in spines of
specic control of gene expression in transgenic mice. Proc Natl hippocampal neurons. J Neurosci 20:17911799.
Acad Sci USA 93:1093310938. Kovalchuk Y, Hanse E, Katz KW, Konnerth A (2002) Postsynaptic
Kitamura HW, Hamanaka H, Watanabe M, Wada K, Yamazaki C, induction of BDNF-mediated long-term potentiation. Science
Fujita SC, Manabe T, Nukina N (2004) Age-dependent enhance- 295:17301734.
ment of hippocampal long-term potentiation in knock-in mice Krasteniakov NV, Martina M, Bergeron R (2004) Subthreshold con-
expressing human apolipoprotein E4 instead of mouse tribution of N-methyl-D-aspartate receptors to long-term
apolipoprotein E. Neurosci Lett 369:173178. potentiation induced by low-frequency pairing in rat hip-
Kiyama Y, Manabe T, Sakimura K, Kawakami F, Mori H, Mishina M pocampal CA1 pyramidal cells. Neuroscience 126:8394.
(1998) Increased thresholds for long-term potentiation and Krucker T, Siggins GR, Halpain S (2003) Dynamic actin laments are
contextual learning in mice lacking the NMDA-type glutamate required for stable long-term potentiation (LTP) in area CA1 of
receptor epsilon1 subunit. J Neurosci 18:67046712. the hippocampus. Proc Natl Acad Sci USA 97:68566861.
Kleschevnikov AM, Sokolov MV, Kuhnt U, Dawe GS, Stephenson JD, Krug M, Lossner B, Ott T (1984) Anisomycin blocks the late phase of
Voronin LL (1997) Changes in paired-pulse facilitation corre- long-term potentiation in the dentate gyrus of freely moving
late with induction of long-term potentiation in area CA1 of rat rats. Brain Res Bull 13:3942.
hippocampal slices. Neuroscience 76:829843. Kuhl D, Skehel P (1998) Dendritic localization of mRNAs. Curr Opin
Kleschevnikov AM, Belichenko PV, Villar AJ, Epstein CJ, Malenka RC, Neurobiol 8:600606.
Mobley WC (2004) Hippocampal long-term potentiation sup- Kulla A, Reymann KG, Manahan-Vaughan D (1999) Time-dependent
pressed by increased inhibition in the Ts65Dn mouse, a genetic induction of depotentiation in the dentate gyrus of freely mov-
model of Down syndrome. J Neurosci 24:81538160. ing rats: involvement of group 2 metabotropic glutamate recep-
Kobayashi K, Poo M-M (2004) Spike train timing-dependent asso- tors.Eur J Neurosci 11:38643872.
ciative modication of hippocampal CA3 recurrent synapses by Kulla A, Manahan-Vaughan D (2002) Modulation of serotonin 5-
mossy bers. Neuron 41:445454. HT(4) receptors of long-term potentiation and depotentiation
Kobayashi K, Manabe T, Takahashi T (1996) Presynaptic long-term in the dentate gyrus of freely moving rats. Cereb Cortex
depression at the hippocampal mossy ber-CA3 synapse. 12:150162.
Science 273:648650. Kulla A, Reymann KG, Manahan-Vaughan D (1999) Time-dependent
Koek W, Woods JH, Winger GD (1988) MK-801, a proposed non- induction of depotentiation in the dentate gyrus of freely-mov-
competitive antagonist of excitatory amino acid neurotransmis- ing rats: involvement of group 2 metabotropic receptors. Eur J
sion, produces phencyclidine-like behavioral effects in pigeons, Neurosci 11:38643872.
rats and rhesus monkeys. J Pharmacol Exp Ther 245:969974. Kullmann DM (1994) Amplitude uctuations of dual-component
Kogan JH, Frankland PW, Silva AJ (2000) Long-term memory under- EPSCs in hippocampal pyramidal cells: implications for long-
lying hippocampus-dependent social recognition in mice. term potentiation. Neuron 12:11111120.
Hippocampus 10:4756. Kullmann DM (2003) Silent synapses: what are they telling us about
Kojima N, Wang J, Mansuy IM, Grant SG, Mayford M, Kandel ER long-term potentiation? Philos Trans R Soc Lond B Biol Sci
(1997) Rescuing impairment of long-term potentiation in fyn- 358:727733.
decient mice by introducing Fyn transgene. Proc Natl Acad Sci Kullmann DM, Erdemli G, Asztely F (1996) LTP of AMPA and
USA 94:47614765. NMDA receptor-mediated signals: evidence for presynaptic
Kojima T, Matsumoto M, Togashi H, Tachibana K, Kemmotsu O, expression and extrasynaptic glutamate spill-over. Neuron
Yoshioka M (2003) Fluvoxamine suppresses the long-term po- 17:461474.
tentiation in the hippocampal CA1 eld of anesthetized rats: an Kusuki T, Imahori Y, Ueda S, Inokuchi K (1997) Dopaminergic mod-
effect mediated via 5-HT1A receptors. Brain Res 959: 165168. ulation of LTP induction in the dentate gyrus of the intact brain.
Kolta A, Lynch G, Ambros-Ingerson J (1998) Effects of aniracetam Neuroreport 8:20372040.
after LTP induction are suggestive of interactions on the kinet- Kutsuwada T, Sakimura K, Manabe T, Takayama C, Katakura N,
ics of the AMPA receptor channel. Brain Res 788:269286. Kushiya E, Natsume R, Watanabe M, Inoue Y, Yagi T, Aizawa S,
 Synaptic Plasticity in the Hippocampus 459

Arakawa M, Takahashi T, Nakamura Y, Mori H, Mishina M Lederer R, Radeke E, Mondadori C (1993) Facilitation of social learn-
(1996) Impairment of suckling response, trigeminal neuronal ing by treatment with an NMDA receptor antagonist. Behav
pattern formation, and hippocampal LTD in NMDA receptor Neural Biol 60:220224.
epsilon 2 subunit mutant mice. Neuron 16:333344. LeDoux JE (2000) Emotion circuits in the brain. Annu Rev Neurosci
Kwon OB, Longart M, Vullhorst D, Hoffman DA, Buonanno A (2005) 23:155184.
Neuregulin-1 reverses long-term potentiation at CA1 hip- Lee HK, Barbarosie M, Kameyama K, Bear MF, Huganir RL (2000)
pocampal synapses. J Neurosci 25:93789383. Regulation of distinct AMPA receptor phosphorylation
Laezza F, Doherty JJ, Dingledine R (1999) Long-term depression in sites during bidirectional synaptic plasticity. Nature 405:
hippocampal interneurones: joint requirement for pre- and 955959.
postsynaptic events. Science 285:14111414. Lee I, Kesner RP (2002) Differential contribution of NMDA receptors
Lambert MP, Barlow AK, Chromy BA, Edwards C, Freed R, Liosatos in hippocampal subregions to spatial working memory. Nat
M, Morgan TE, Rozovsky I, Trommer B, Viola KL, Wals P, Zhang Neurosci 5:162168.
C, Finch CE, Krafft GA, Klein WL (1998) Diffusible, nonbril- Lee KS, Schottler F, Oliver M, Lynch G (1980) Brief bursts of high-
lar ligands derived from Abeta1-42 are potent central nervous frequency stimulation produce two types of structural change in
system neurotoxins. Proc Natl Acad Sci USA 95:64486453. rat hippocampus. J Neurophysiol 44:247258.
Lamsa K, Heeroma JH, Kullmann DM (2005) Hebbian LTP in feed- Lee PR, Cohen JE, Becker KG, Fields RD (2005) Gene expression in
forward inhibitory interneurons and the temporal delity of the conversion of early-phase to late-phase long-term potentia-
input discrimination. Nat Neurosci 8:916924. tion. Ann N Y Acad Sci 1048:259271.
Lanahan A, Worley P (1998) Immediate-early genes and synaptic Lee SH, Liu L, Wang YT, Sheng M (2002) Clathrin adaptor AP2 and
function. Neurobiol Learn Mem 70:3743. NSF interact with overlapping sites of GluR2 and play distinct
Langdon RB, Johnson JW, Barrionuevo G (1995) Posttetanic poten- roles in AMPA receptor trafficking and hippocampal LTD.
tiation and presynaptically induced long-term potentiation at Neuron 36:661674.
the mossy ber synapse in rat hippocampus. J Neurobiol Lei S, McBain CJ (2002) Distinct NMDA receptors provide differen-
26:370385. tial modes of transmission at mossy ber-interneuron synapses.
Lapointe V, Morin F, Ratte S, Croce A, Conquet F, Lacaille JC (2003) Neuron 33:921933.
Synapse-specic mGluR1-dependent long-term potentiation in Lei S, Pelkey KA, Topolnik L, Congar P, Lacaille JC, McBain CJ (2003)
interneurons regulates mouse hippocampal inhibition. J Physiol Depolarization-induced long-term depression at hippocampal
(Lond) 555:125135. mossy fiber-CA3 pyramidal neuron synapses. J Neurosci
Larkum ME, Zhu JJ, Sakmann B (1999) A new cellular mechanism 23:97869795.
for coupling inputs arriving at different cortical layers. Nature Lein ES, Zhao X, Gage FH (2004) Dening a molecular atlas of the
398:338341. hippocampus using DNA microarrays and high-throughput in
Laroche S, Doyere V, Bloch V (1989) Linear relation between the situ hybridization. J Neurosci 24:38793889.
magnitude of long-term potentiation in the dentate gyrus and Lemphrect R, Le Doux J (2004) Structural plasticity and memory.
associative learning in the rat: a demonstration using commis- Nat Rev Neurosci 5:4554.
sural inhibition and local infusion of an N-methyl-D-aspartate Lengyel I, Voss K, Cammaroto M, Bradshaw K, Brent V, Murphy KP,
receptor antagonist. Neuroscience 28:375386. Giese KP, Rostas J, Bliss TV (2004) Autonomous activity of
Laroche S, Jay TM, Thierry AM (1990) Long-term potentiation in the CaMKII is only transiently increased after the induction of
prefrontal cortex following stimulation of the hippocampal long-term potentiation in the rat hippocampus. Eur J Neurosci
CA1/subicular region. Neurosci Lett 114:184190. 20:30633072.
Larson J, Lynch G (1986) Induction of synaptic potentiation in hip- Leonard AS, Davare MA, Horne MC, Garner CC, Hell JW (1998)
pocampus by patterned stimulation involves two events. Science SAP97 is associated with the alpha-amino-3-hydroxy-5-
232:985987. methylisoxazole-4-propionic acid receptor GluR1 subunit. J
Larson J, Lynch G, Games D, Seubert P (1999) Alterations in synap- Biol Chem 1998 273:1951819524.
tic transmission and long-term potentiation in hippocampal Leonard BJ, McNaughton BL, Barnes CA (1987) Suppression of hip-
slices from young and aged PDAPP mice. Brain Res 840:2335. pocampal synaptic plasticity during slow-wave sleep. Brain Res
Lathe R, Morris RGM (1994) Analyzing brain-function and dysfunc- 425:174177.
tion in transgenic animals. Neuropathol Appl Neurobiol 20: Leung LS, Shen B (2006) Hippocampal CA1 kindling but not long-
350358. term potentiation disrupts spatial memory performance. Learn
Lauri SE, Bortolotto ZA, Bleakman D, Ornstein PL, Lodge D, Isaac JT, Mem 13:1826.
Collingridge GL (2001) A critical role of a facilitatory presynap- Leung LS, Shen B, Rajakumar N, Ma J (2003) Cholinergic activity
tic kainate receptor in mossy ber LTP. Neuron 32: 697709. enhances hippocampal long-term potentiation in CA1 during
Lauri SE, Bortolotto ZA, Nistico R, Bleakman D, Ornstein PL, Lodge walking in rats. J Neurosci 23:9297-9304.
D, Isaac JT, Collingridge GL (2003) A role for Ca2 stores in Levy WB, Steward O (1979) Synapses as associative memory elements
kainate receptor-dependent synaptic facilitation and LTP at in the hippocampal formation. Brain Res 175:233245.
mossy ber synapses in the hippocampus. Neuron 39:327341. Levy WB, Steward O (1983) Temporal contiguity requirements for
Lawrence JJ, McBain CJ (2003) Containing the detonation: feedfor- long-term associative potentiation/depression in the hippocam-
ward inhibition in the CA3 hippocampus. Trends Neurosci pus. Neuroscience 8:791797.
26:631640. Li AJ, Katafuchi T, Oda S, Hori T, Oomura Y (1997) Interleukin-6
Lebel D, Grossman Y, Barkai E (2001) Olfactory learning modies inhibits long-term potentiation in rat hippocampal slices. Brain
predisposition for long-term potentiation and long-term Res 748:3038.
depression induction in the rat piriform (olfactory) cortex. Li HB, Matsumoto K, Yamamoto M, Watanabe H (1997) NMDA but
Cereb Cortex 11:485489. not AMPA receptor antagonists impair the delay-interposed
460 The Hippocampus Book

radial maze performance of rats. Pharmacol Biochem Behav family transcription factors in the nervous system. Neuron 35:
58:249253. 605623.
Li L, Carter J, Gao X, Whitehead J, Tourtellotte WG (2005) The neu- Lopez-Garcia JC, Arancio O, Kandel ER, Baranes D (1996) A presy-
roplasticity-associated arc gene is a direct transcriptional target naptic locus for long-term potentiation of elementary synaptic
of early growth response (Egr) transcription factors. Mol Cell transmission at mossy ber synapses in culture system. Proc
Biol 25:1028610300. Natl Acad Sci USA 93:47124717.
Li S, Cullen WK, Anwyl R, Rowan MJ (2003) Dopamine-dependent Lovinger DM, Wong KL, Murakami K, Routtenberg A (1987) Protein
facilitation of LTP induction in hippocampal CA1 by exposure kinase C inhibitors eliminate hippocampal long-term potentia-
to spatial novelty. Nat Neurosci 6:526531. tion. Brain Res 436:177183.
Li Y, Hough CJ, Frederickson CJ, Sarvey JM (2001) Induction Lozovaya NA, Grebenyuk SE, Tsintsadze T, Feng B, Monaghan DT,
of mossy ber () CA3 long-term potentiation requires Krishtal OA (2004) Extrasynaptic NR2B and NR2D subunits of
translocation of synaptically released Zn2. J Neurosci 21: NMDA receptors shape superslow afterburst EPSC in rat hip-
80158025. pocampus. J Physiol 558:451463.
Liao D, Hessler NA, Malinow R (1995) Activation of postsynaptically Lu W, Man H, Ju W, Trimble WS, MacDonald JF, Wang YT (2001)
silent synapses during pairing-induced LTP in CA1 region of Activation of synaptic NMDA receptors induces membrane
hippocampal slice. Nature 375:400404. insertion of new AMPA receptors and LTP in cultured hip-
Linden DJ (1999) The return of the spike: postsynaptic action poten- pocampal neurons. Neuron 29:243254.
tials and the induction of LTP and LTD. Neuron 22:661666. Lu YM, Jia Z, Janus C, Henderson JT, Gerlai R, Wojtowicz JM, Roder
Ling DS, Benardo LS, Sacktor TC (2006) Protein kinase Mzeta JC (1997) Mice lacking metabotropic glutamate receptor 5 show
enhances excitatory synaptic transmission by increasing the impaired learning and reduced CA1 long-term potentiation
number of active postsynaptic AMPA receptors. Hippocampus (LTP) but normal CA3 LTP. J Neurosci 17:51965205.
16:443452. Lu YM, Roder JC, Davidow J, Salter MW (1998) Src activation in the
Ling DS, Benardo LS, Serrano PA, Blace N, Kelly MT, Crary JF, induction of long-term potentiation in CA1 hippocampal neu-
Sacktor TC (2002) Protein kinase Mzeta is necessary and suffi- rons. Science 279:13631367.
cient for LTP maintenance. Nat Neurosci 5:295296. Lu YM, Mansuy IM, Kandel ER, Roder J (2000a) Calcineurin-medi-
Lipp HP, Wolfer DP (1998) Genetically modied mice and cognition. ated LTD of GABAergic inhibition underlies the increased
Curr Opin Neurobiol 8:272280. excitability of CA1 neurons associated with LTP. Neuron
Lisman J (1989) A mechanism for the Hebb and anti-Hebb processes 26:197205.
underlying learning and memory. Proc Natl Acad Sci USA Lu YM, Taverna FA, Tu R, Ackerley CA, Wang YT, Roder J (2000b)
86:95749578. Endogenous Zn(2) is required for the induction of long-term
Lisman J, Spruston N (2005) Postsynaptic depolarization requir- potentiation at rat hippocampal mossy ber-CA3 synapses.
ments for LTP and LTD: a critique of spike timing-dependent Synapse 38:187197.
plasticity. Nat Neurosci 8:839-841. Luscher C, Malenka RC, Nicoll RA (1998) Monitoring glutamate
Lisman J, Schulman H, Cline H (2002) The molecular basis of release during LTP with glial transporter currents. Neuron
CaMKII function in synaptic and behavioural memory. Nat Rev 21:435441.
Neurosci 3:175190. Lscher C, Xia H, Beattie EC, Carroll RC, von Zastrow M, Malenka
Lisman JE (1997) Bursts as a unit of neural information: making RC, Nicoll RA (1999) Role of AMPA receptor cycling in synap-
unreliable synapses reliable. Trends Neurosci 20:3843. tic transmission and plasticity. Neuron 24:649658.
Lisman JE, Grace AA (2005) The hippocampal-VTA loop: controlling Lthi A, Laurent JP, Figurov A, Muller D, Schachner M (1994)
the entry of information into long-term memory. Neuron Hippocampal long-term potentiation and neural cell adhesion
46:703713. molecules L1 and NCAM. Nature 372:777779.
Lissin DV, Gomperts SN, Caroll RC, Christine CW, Kalman D, Lthi A, Chittajallu R, Duprat F, Palmer M, Benke TA, Kidd FL,
Kitamura M, Hardy S, Nicoll RA, Malenka RC, von Zastrow M Henley JM, Isaac JTR (1999) Hippocampal LTD expression
(1998) Activity differentially regulates the surface expression of involves a pool of AMPARs regulated by the NSF-GluR2 inter-
synaptic AMPA and NMDA glutamate receptors. Proc Natl Acad action. Neuron 24:389399.
Sci USA 95:70977102. Lthi A, Wikstrm MA, Palmer MJ, Matthews P, Benke TA, Isaac JTR,
Liu L, Wong TP, Pozza MF, Lingenhoehl K, Wang Y, Sheng M, Collingridge GL (2004) Bi-directional modulation of AMPA
Auberson YP, Wang YT (2004a) Role of NMDA receptor sub- receptor unitary conductance by synaptic activity. BMC
types in governing the direction of hippocampal synaptic plas- Neurosci 5:44.
ticity. Science 304:10211024. Lyford GL, Gutnikov SA, Clark AM, Rawlins JN (1993) Determi-
Liu S, Ninan I, Antonova I, Battaglia F, Trinchese F, Narasanna A, nants of non-spatial working memory decits in rats given
Kolodilov N, Dauer W, Hawkins RD, Arancio O (2004b) Alpha- intraventricular infusions of the NMDA antagonist AP5.
synuclein produces a long-lasting increase in neurotransmitter Neuropsychologia 31:10791098.
release. EMBO J 23:45064516. Lynch G (2006) Glutamate-based therapeutic approaches: ampak-
Lledo PM, Zhang X, Sudhof TC, Malenka RC, Nicoll RA (1998) ines. Curr Opin Pharmacol 6:8288.
Postsynaptic membrane fusion and long-term potentiation. Lynch G, Baudry M (1984) The biochemistry of memory: a new and
Science 279:399403. specic hypothesis. Science 224:10571063.
Lmo T (1966) Frequency potentiation of excitatory synaptic activ- Lynch G, Larson J, Kelso S, Barrionuevo G, Schottler F (1983)
ity in the dentate area of the hippocampal formation. Acta Intracellular injections of EGTA block induction of hippocam-
Physiol Scand 68(Suppl 277):128. pal long-term potentiation. Nature 305:719721.
Lonze B, Ginty DD (2002) Function and regulation of CREB Lynch GS, Dunwiddie T, Gribkoff V (1977) Heterosynaptic depres-
 Synaptic Plasticity in the Hippocampus 461

sion: a postsynaptic correlate of long-term potentiation. Nature whole-cell recordings of long-term potentiation in hippocam-
266:737739. pal slices. Nature 346:177180.
Lynch MA (2004) Long-term potentiation and memory. Physiol Rev Malinow R, Schulman H, Tsien RW (1989) Inhibition of postsynap-
84:87136. tic PKC or CaMKII blocks induction but not expression of LTP.
Ma L, Zablow L, Kandel ER, Siegelbaum SA (1999) Cyclic AMP Science 245:862866.
induces functional presynaptic boutons in hippocampal CA3- Malleret G, Haditsch U, Genoux D, Jones MW, Bliss TVP, Vanhoose
CA1 neuronal cultures. Nat Neurosci 2:2430. AM, Weitlauf C, Kandel ER, Winder DG, Mansuy IM (2001)
Maccaferri G, McBain CJ (1996) Long-term potentiation in distinct Inducible and reversible enhancement of learning, memory and
subtypes of hippocampal nonpyramidal neurons. J Neurosci long-term potentiation by genetic inhibitior of calcineurin. Cell
16:53345343. 104:675686.
Maccaferri G, Toth K, McBain CJ (1998) Target-specic expression of Man HY, Wang Q, Lu WY, Ju W, Ahmadian G, Liu L, DSouza S, Wong
presynaptic mossy ber plasticity. Science 279:13681370. TP, Taghibiglou C, Lu J, Becker LE, Pei L, Liu F, Wymann MP,
MacDermott AB, Mayer ML, Westbrook GL, Smith SJ, Barker JL MacDonald JF, Wang YT (2003) Activation of PI3-kinase is
(1986) NMDA-receptor activation increases cytoplasmic cal- required for AMPA receptor insertion during LTP of mEPSCs in
cium concentration in cultured spinal cord neurones. Nature cultured hippocampal neurons. Neuron 38:611624.
321: 51922. Manabe T, Nicoll RA (1994) Long-term potentiation: evidence
MacDonald JF, Wojtowicz JM (1980) Two conductance mechanisms against an increase in transmitter release probability in the CA1
activated by applications of L-glutamic, L-aspartic, DL-homocys- region of the hippocampus. Science 265:18881892.
teic, N-methyl-D-aspartic, and DL-kainic acids to cultured mam- Manabe T, Noda Y, Mamiya T, Katagiri H, Houtani T, Nishi M, Noda
malian central neurones. Can J Physiol Pharmacol 58: T, Takahashi T, Sugimoto T, Nabeshima T, Takeshima H (1998)
13931397. Facilitation of long-term potentiation and memory in mice
Madani R, Hulo S, Toni N, Madani H, Steimer T, Muller D, Vassalli JD lacking nociceptin receptors. Nature 394:577581.
(1999) Enhanced hippocampal long-term potentiation and Manahan-Vaughan D (1997) Group 1 and 2 metabotropic receptors
learning by increased neuronal expression of tissue-type plas- play differential roles in hippocampal long-term depression and
minogen activator in transgenic mice. EMBO J 18: 30073012. long-term potentiation in freely moving rats. J Neurosci
Madison DV, Schuman EM (1994) Locally distributed synaptic 17:33033311.
potentiation in the hippocampus. Science 263:532536. Manahan-Vaughan D (2000) Long-term depression in freely moving
Magee JC, Johnston D (1997) A synaptically controlled, associative rats is dependent upon strain variation, induction protocol and
signal for Hebbian plasticity in hippocampal neurons. Science behavioral state. Cereb Cortex 10:483487.
275:209213. Manahan-Vaughan D, Braunewell KH (1999) Novelty acquisition is
Magleby KL (1979) Facilitation, augmentation and potentiation of associated with induction of hippocampal long-term depres-
transmitter release. Prog Brain Res 49:175182. sion. Proc Natl Acad Sci USA 96:87398744.
Maher FO, Nolan Y, Lynch MA (2005) Downregulation of IL-4- Manahan-Vaughan D, Kulla A, Frey JU (2000) Requirement of trans-
induced signalling in hippocampus contributes to decits in lation but not transcription for the maintenance of long-term
LTP in the aged rat. Neurobiol Aging 26:717728. depression in the CA1 region of freely moving rats. J Neurosci
Mainen ZF, Jia Z, Roder J, Malinow R (1998) Use-dependent AMPA 20:85728576.
receptor block in mice lacking GluR2 suggests postsynaptic site Mansuy IM, Mayford M, Jacob B, Kandel ER, Bach ME (1998a)
for LTP expression. Nat Neurosci 1:579586. Restricted and regulated overexpression reveals calcineurin as a
Malenka RC, Madison DV, Nicoll RA (1986) Potentiation of synaptic key component in the transition from short-term to long-term
transmission in the hippocampus by phorbol esters. Nature memory. Cell 92:3949.
321:175177. Mansuy IM, Winder DG, Moallem TM, Osman M, Mayford M,
Malenka RC, Kauer JA, Perkel DJ, Mauk MD, Kelly PT, Nicoll RA, Hawkins RD, Kandel ER (1998b) Inducible and reversible gene
Waxham MN (1989) An essential role for postsynaptic calmod- expression with the rtTA system for the study of memory.
ulin and protein kinase activity in long-term potentiation. Neuron 21:257265.
Nature 340:554557. Manzoni OJ, Weisskopf MG, Nicoll RA (1994) MCPG antagonizes
Malenka RC, Lancaster B, Zucker RS (1992) Temporal limits on the metabotropic glutamate receptors but not long-term potentia-
rise in postsynaptic calcium required for the induction of long- tion in the hippocampus. Eur J Neurosci 6:10501054.
term potentiation. Neuron 9:121128. Marder CP, Buonomano DV (2003) Differential effects of short- and
Maletic-Savatic M, Malinow R, Svoboda K (1999) Rapid dendritic long-term potentiation on cell ring in the CA1 region of the
morphogenesis in CA1 hippocampal dendrites induced by hippocampus. J Neurosci 23:112121.
synaptic activity. Science 283:19231927. Marder CP, Buonomano DV (2004) Timing and balance of inhibition
Malgaroli A, Ting AE, Wendland B, Bergamaschi A, Villa A, Tsien RW, enhance the effect of long-term potentiation on cell ring. J
Scheller RH (1995) Presynaptic component of long-term poten- Neurosci 24:88738884.
tiation visualized at individual hippocampal synapses. Science Margrie TW, Rostas JA, Sah P (1998) Long-term potentiation of
268:16241628. synaptic transmission in the avian hippocampus. J Neurosci
Malinow R, Malenka RC (2002) AMPA receptor trafficking and 18:12071216.
synaptic plasticity. Annu Rev Neurosci 25:103126. Markakis EA, Gage FH (1999) Adult-generated neurons in the den-
Malinow R, Miller JP (1986) Postsynaptic hyperpolarization during tate gyrus send axonal projections to eld CA3 and are sur-
conditioning reversibly blocks induction of long-term potentia- rounded by synaptic vesicles. J Comp Neurol 406:449460.
tion. Nature 320:529530. Markram H, Tsodyks M (1996) Redistribution of synaptic efficacy
Malinow R, Tsien RW (1990) Presynaptic enhancement shown by between neocortical pyramidal cells. Nature 382:807810.
462 The Hippocampus Book

Markram H, Lubke J, Frotscher M, Sakmann B (1997) Regulation of Mayford M, Wang J, Kandel ER, ODell TJ (1995) CaMKII regulates
synaptic efficacy by coincidence of postsynaptic APs and EPSPs. the frequency-response of hippocampal synapses for the pro-
Science 275:213215. duction of both LTP and LTD. Cell 81:891904.
Marr D (1971) Simple memory: a theory for archicortex. Philos Trans Mayford M, Bach ME, Huang Y-Y, Wang L, Hawkins R, Kandel ER
R Soc Lond B Biol Sci 262:2381. (1996) Control of memory formation through regulated
Martin KC (2004) Local protein synthesis during axon guidance and expression of a CaMKII transgene. Science 274:16781683.
synaptic plasticity. Curr Opin Neurobiol 14:305310. Mayford M, Mansuy IM, Muller RU, Kandel ER (1997) Memory and
Martin KC, Kosik KS (2002) Synaptic taggingwhos it? Nat Neurosci behavior: a second generation of genetically modied mice.
Rev 3:813820. Curr Biol 7:R580(R589.
Martin KC, Barad M, Kandel ER (2000a) Local protein synthesis and McBain CJ, Freund TF, Mody I (1999) Glutamatergic synapses onto
its role in synapse-specic plasticity. Curr Opin Neurobiol hippocampal interneurons: precision timing without lasting
10:587592. plasticity. Trends Neurosci 22:228235.
Martin SJ (1998) Time-dependent reversal of dentate LTP by 5 Hz McDermott CM, LaHoste GJ, Chen C, Musto A, Bazan NG, Magee JC
stimulation. Neuroreport 9:37753781. (2003). Sleep deprivation causes behavioral, synaptic, and mem-
Martin SJ, Morris RGM (1997) (R,S)-Alpha-methyl-4-car- brane excitability alterations in hippocampal neurons. J
boxyphenylglycine (MCPG) fails to block long-term potentia- Neurosci 23:96879695.
tion under urethane anaesthesia in vivo. Neuropharmacology McDonald RJ, Hong NS, Craig LA, Holahan MR, Louis M, Muller RU
10:13391354. (2005) NMDA-receptor blockade by CPP impairs post-training
Martin SJ, Grimwood PD, Morris RG (2000) Synaptic plasticity and consolidation of a rapidly acquired spatial representation in rat
memory: an evaluation of the hypothesis. Annu Rev Neurosci hippocampus. Eur J Neurosci 22:12011213.
23:649711. McEachern JC, Shaw CA (1996) An alternative to the LTP orthodoxy:
Massey PV, Johnson BE, Moult PR, Auberson YP, Brown MW, Molnar a plasticity-pathology continuum model. Brain Res Brain Res
E, Collingridge GL, Bashir ZI (2004) Differential roles of NR2A Rev 22:5192.
and NR2B-containing NMDA receptors in cortical long-term McGahon B, Holscher C, McGlinchey L, Rowan MJ, Lynch MA
potentiation and long-term depression. J Neurosci 24: (1996) Training in the Morris watermaze occludes the syner-
78217828. gism between ACPD and arachidonic acid on glutamate release
Matsuda S, Mikawa S, Hirai H (1999) Phosphorylation of serine-880 in synaptosomes prepared from rat hippocampus. Learn Mem
in GluR2 by protein kinase C prevents its C terminus from bind- 3:296304.
ing with glutamate receptor-interacting protein. J Neurochem McGaugh JL (2000) Memorya century of consolidation. Science
73:17651768. 287:248251.
Matsuo R, Murayama A, Saitoh Y, Sakaki Y, Inokuchi K (2000) McHugh TJ, Blum KI, Tsien JZ, Tonegawa S, Wilson MA (1996)
Identication and cataloging of genes induced by long-lasting Impaired hippocampal representation of space in CA1-specic
long-term potentiation in awake rats. J Neurochem 74: NMDAR1 knockout mice. Cell 87:13391349.
22392249. McKernan MG, Shinnick-Gallagher P (1997) Fear conditioning
Matsuzaki M, Honkura N, Ellis-Davies GCR, Kasai H (2004) induces a lasting potentiation of synaptic currents in vitro.
Structural basis of long-term potentiation in single dendritic Nature 390:607611.
spines. Nature 429:761766. McKinney RA, Capogna M, Durr R, Gahwiler BH, Thompson SM
Matthies H, Reymann KG (1993) Protein kinase A inhibitors prevent (1999) Miniature synaptic events maintain dendritic spines via
the maintenance of hippocampal long-term potentiation. AMPA receptor activation. Nat Neurosci 2:4449.
Neuroreport 4:712714. McMahon LL, Kauer JA (1997) Hippocampal interneurons express a
Matthies H, Ruthrich H, Ott T, Matthies HK, Matthies R (1986) Low novel form of synaptic plasticity. Neuron 18:295305.
frequency perforant path stimulation as a conditioned stimulus McNaughton BL (1980) Evidence for two physiologically distinct
demonstrates correlations between long-term synaptic potenti- perforant pathways to the fascia dentata. Brain Res 199:119.
ation and learning. Physiol Behav 36:811821. McNaughton BL (1982) Long-term synaptic enhancement and
Matthies H, Becker A, Schroeder H, Kraus J, Hollt V, Krug M (1997) short-term potentiation in rat fascia dentata act through differ-
Dopamine D1-decient mutant mice do not express the late ent mechanisms. J Physiol 324:249262.
phase of hippocampal long-term potentiation. Neuroreport McNaughton BL (1983) Activity-dependent modulation of hip-
8:35333535. pocampal synaptic efficacy: some implications for memory
Matus A (2005) Growth of dendritic spines: a continuing story. Curr processes. In: Neurobiology of the hippocampus (Siefert W, ed),
Opin Neurobiol 15:6772. pp 233252. London: Academic Press.
Matynia A, Kushner SA, Silva AJ (2002) Genetic approaches to McNaughton BL, Douglas RM, Goddard GV (1978) Synaptic
molecular and cellular cognition: a focus on LTP and learning enhancement in fascia dentata: cooperativity among coactive
and memory. Annu Rev Genet 36:687720. afferents. Brain Res 157:277293.
Mauceri D, Cattabeni F, Di Luca M, Gardoni F (2004) Calcium/ McNaughton BL, Barnes CA, Rao G (1984) Presynaptic versus post-
calmodulin-dependent protein kinase II phosphorylation drives synaptic control over long-term synaptic enhancement. In:
synapse-associated protein 97 into spines. J Biol Chem Neurobiology of learning and memory (Lynch G, McGaugh JL and
279:2381323821. Weinberger NM, eds), pp. 466-469. New York: Guilford Press.
Mayer ML, Westbrook GL, Guthrie PB (1984) Voltage-dependent McNaughton BL, Barnes CA, Rao G, Baldwin J, Rasmussen M (1986)
block by Mg2 of NMDA responses in spinal cord neurones. Long-term enhancement of hippocampal synaptic transmission
Nature 309:261263. and the acquisition of spatial information. J Neurosci 6:563571.
 Synaptic Plasticity in the Hippocampus 463

Mehta MR, Quirk MC, Wilson MA (2000) Experience-dependent Monaghan DT, Cotman CW (1885) Distribution of N-methyl-D-
asymmetric shape of hippocampal receptive elds. Neuron aspartate-sensitive L-[3H]-binding sites in rat brain. J Neurosci
25:707715. 5:29092919.
Meiri N, Ghelardini C, Tesco G, Galeotti N, Dahl D, Tomsic D, Mondadori C, Weiskrantz L (1991) Memory facilitation by NMDA
Cavallaro S, Quattrone A, Capaccioli S, Bartolini A, Alkon DL receptor blockade. In: Long-term potentiation: a debate of current
(1997) Reversible antisense inhibition of Shaker-like Kv1.1 issues (Baudry M, Davis J, eds), pp 259266. Cambridge, MA:
potassium channel expression impairs associative memory in MIT Press.
mouse and rat. Proc Natl Acad Sci USA 94:44304434. Mondadori C, Weiskrantz L (1993) NMDA receptor blockers facili-
Meiri N, Sun MK, Segal Z, Alkon DL (1998) Memory and long-term tate and impair learning via different mechanisms. Behav Neural
potentiation (LTP) dissociated: normal spatial memory despite Biol 60:205210.
CA1 LTP elimination with Kv1.4 antisense. Proc Natl Acad Sci Mondadori C, Weiskrantz L, Buerki H, Petschke F, Fagg GE (1989)
USA 95:1503715042. NMDA receptor antagonists can enhance or impair learning
Mellor J, Nicoll RA (2001) Hippocampal mossy ber LTP is inde- performance in animals. Exp Brain Res 75:449456.
pendent of postsynaptic calcium. Nat Neurosci 4:125126. Mondadori C, Borkowski J, Gentsch C (1996) The memory-facilitat-
Mellor J, Nicoll RA, Schmitz D (2002) Mediation of hippocampal ing effects of the competitive NMDA-receptor antagonist CGP
mossy ber long-term potentiation by presynaptic Ih channels. 37849 are steroid-sensitive, whereas its memory-impairing
Science 295:143147. effects are not. Psychopharmacology (Berl) 124:380383.
Mendoza-Fernandez V, Andrew RD, Barajas-Lopez C (2000) Moechars D, Dewachter I, Lorent K, Reverse D, Baekelandt V, Naidu
Interferon-alpha inhibits long-term potentiation and unmasks a A, Tesseur I, Spittaels K, Haute CV, Checler F, Godaux E, Cordell
long-term depression in the rat hippocampus. Brain Res B, Van Leuven F (1999) Early phenotypic changes in transgenic
885:1424. mice that overexpress different mutants of amyloid precursor
Meng Y, Zhang Y, Jia Z (2003) Synaptic transmission and plasticity in protein in brain. J Biol Chem 274:64836492.
the absence of AMPA glutamate receptor GluR2 and GluR3. Monls MH, Teskey GC (2004) Skilled-learning-induced potentia-
Neuron 39:163176. tion in rat sensorimotor cortex: a transient form of behavioural
Messaoudi E, Ying SW, Kanhema T, Croll SD, Bramham CR (2002) long-term potentiation. Neuroscience 125:329336.
Brain-derived neurotrophic factor triggers transcription- Montgomery JM, Madison DV (2002) State-dependent heterogeneity
dependent, late phase long-term potentiation in vivo. J Neurosci in synaptic depression between pyramidal cell pairs. Neuron
22:74537461. 33:765777.
Micheau J, Riedel G (1999) Protein kinases: which one is the memory Montgomery JM, Pavlidis P, Madison DV (2001) Pair recordings
molecule? Cell Mol Life Sci 55:534548. reveal all-silent synaptic connections and the postsynaptic
Micheva KD, Buchanan J, Holz RW, Smith SJ (2003) Retrograde reg- expression of long-term potentiation. Neuron 29:691701.
ulation of synaptic vesicle endocytosis and recycling. Nat Moore CI, Browning MD, Rose GM (1993) Hippocampal plasticity
Neurosci 6:925932. induced by primed burst, but not long-term potentiation, stim-
Migaud M, Charlesworth P, Dempster M, Webster LC, Watabe AM, ulation is impaired in area CA1 of aged Fischer 344 rats.
Makhinson M, He Y, Ramsay MF, Morris RG, Morrison JH, Hippocampus 3:5766.
ODell TJ, Grant SG (1998) Enhanced long-term potentiation Moosmang S, Haider N, Klugbauer N, Adelsberger H, Langwieser N,
and impaired learning in mice with mutant postsynaptic den- Muller J, Stiess M, Marais E, Schulla V, Lacinova L, Goebbels S,
sity-95 protein. Nature 396:433439. Naye K-L, Storm DR, Hofmann F, Kleppisch T (2005) Role of
Miller B, Sarantis M, Traynelis SF, Attwell D (1992) Potentiation of hippocampal Cav1.2 Ca2 channels in NMDA receptor-inde-
NMDA receptor currents by arachidonic acid. Nature pendent synaptic plasticity and spatial memory. J Neurosci
235:722725. 25:98839892.
Miller SG, Kennedy MB (1986) Regulation of brain type II Ca2/ More JC, Nistico R, Dolman NP, Clarke VR, Alt AJ, Ogden AM,
calmodulin-dependent protein kinase by autophosphorylation: Buelens FP, Troop HM, Kelland EE, Pilato F, Bleakman D,
a Ca2-triggered molecular switch. Cell 44:861870. Bortolotto ZA, Collingridge GL, Jane DE (2004) Characterisa-
Minogue AM, Schmid AW, Fogarty MP, Moore AC, Campbell VA, tion of UBP296: a novel, potent and selective kainate receptor
Herron CE, Lynch MA (2003) Activation of the c-Jun N-termi- antagonist. Neuropharmacology 47:4664.
nal kinase signaling cascade mediates the effect of amyloid-beta Morris R, Anderson E, Lynch GS, Baudry M (1986) Selective impair-
on long term potentiation and cell death in hippocampus: a role ment of learning and blockade of long-term potentiation by an
for interleukin-1beta? J Biol Chem 278:2797127980. N-methyl-D-aspartate receptor antagonist, AP5. Nature 319:
Mizuno T, Kanazawa I, Sakurai M (2001) Differential induction of 774776.
LTP and LTD is not determined solely by instantaneous calcium Morris RG (1989) Synaptic plasticity and learning: selective impair-
concentration: an essential involvement of a temporal factor. ment of learning rats and blockade of long-term potentiation in
Eur J Neurosci 14:701708. vivo by the N-methyl-D-aspartate receptor antagonist AP5. J
Mockett B, Coussens C, Abraham WC (2002) NMDA receptor-medi- Neurosci 9:30403057.
ated metaplasticity during the induction of long-term depres- Morris RG, Halliwell RF, Bowery N (1989) Synaptic plasticity and
sion by low-frequency stimulation. Eur J Neurosci 15:18191826. learning. II. Do different kinds of plasticity underlie different
Mockett BG, Brooks WM, Tate WP, Abraham WC (2004) Dopamine kinds of learning? Neuropsychologia 27:4159.
D1/D5 receptor activation fails to initiate an activity-independ- Morris RGM, Davis S, Butcher SP (1990) Hippocampal synatic plas-
ent late-phase LTP in rat hippocampus. Brain Res Rev 1021: ticity and NMDA receptors: a role in information storage?
92100. Philos Trans R Soc Lond 329:187204.
464 The Hippocampus Book

Morton RA, Kuenzi FM, Fitzjohn SM, Rosahl TW, Smith D, Zheng H, Murai KK, Pasquale EB (2004) Eph receptors, ephrins, and synaptic
Shearman M, Collingridge GL, Seabrook GR (2002) Impair- function. Neuroscientist 10:304414.
ment in hippocampal long-term potentiation in mice under- Murai KK, Nguyen LN, Irie F, Yamaguchi Y, Pasquale EB (2003)
expressing the Alzheimers disease related gene presenilin-1. Control of hippocampal dendritic spine morphology through
Neurosci Lett 319:3740. ephrin-A3/EphA4 signaling. Nat Neurosci 6: 153160.
Moser E, Moser M-B (1999) Is learning blocked by saturation of Murase S, Schuman EM (1999) The role of cell adhesion molecules
synaptic weights in the hippocampus? Neurosci Biobehav Rev in synaptic plasticity and memory. Curr Opin Cell Biol 11:
23:661672. 549553.
Moser E, Mathiesen I, Andersen P (1993a) Association between brain Murase S, Mosser E, Schuman EM (2002) Depolarization drives beta-
temperature and dentate eld potentials in exploring and swim- catenin into neuronal spines promoting changes in synaptic
ming rats. Science 259:13241326. structure and function. Neuron 35:91105.
Moser E, Moser M-B, Andersen P (1993b) Synaptic potentiation in Murphy KP, Carter RJ, Lione LA, Mangiarini L, Mahal A, Bates GP,
the rat dentate gyrus during exploratory learning. Neuroreport Dunnett SB, Morton AJ (2000) Abnormal synaptic plasticity
5:317320. and impaired spatial cognition in mice transgenic for exon 1 of
Moser MB, Trommald M, Andersen P (1994) An increase in dendritic the human Huntingtons disease mutation. J Neurosci 20:
spine density on hippocampal CA1 pyramidal cells following 51155123.
spatial learning in adult rats suggests the formation of new Murphy KPSJ, Williams JH, Bettache N, Bliss TVP (1994) Pho-
synapses. Proc Natl Acad Sci USA 91:1267312675. tolytic release of nitric-oxide modulates NMDA receptor-
Moser EI, Krobert KA, Moser MB, Morris RG (1998) Impaired spa- mediated transmission but does not induce long-term potenti-
tial learning after saturation of long-term potentiation. Science ation at hippocampal synapses. Neuropharmacology 33:
281:20382042. 13751385.
Moult PR, Schnabel R, Kilpatrick IC, Bashir ZI, Collingridge GL Murray CA, Lynch MA (1998) Evidence that increased hippocampal
(2002) Tyrosine dephosphorylation underlies DHPG-induced expression of the cytokine interleukin-1 beta is a common trig-
LTD. Neuropharmacology 43:175180. ger for age- and stress-induced impairments in long-term
Moult PR, Gladding CM, Sanderson TM, Fitzjohn SM, Bashir ZI, potentiation. J Neurosci 18:29742981.
Molnar E, Collingridge GL (2006) Tyrosine phosphatases regu- Nagappan G, Lu B (2005) Activity-dependent modulation of the
late AMPA receptor trafficking during metabotropic glutamate BDNF receptor TrKB: mechanisms and implications. Trends
receptor-mediated long-term depression. J Neurosci 26: Neurosci 28:464471.
25442554. Nagase T, Ito K-I, Kato K, Kaneko K, Kohda K, Hoshino A,
Mouly AM, Fort A, Ben-Boutayab N, Gervais R (2001) Olfactory Matsumoto M, Inoue T, Fujii S, Kato A, Mikoshiba K (2003)
learning induces differential long-lasting changes in rat central Long-term potentiation and long-term depression in hip-
olfactory pathways. Neuroscience 102:1121. pocampal CA1 neurons of mice lacking the IP3 type 1 receptor.
Moyer JRJ, Thompson LT, Disterhoft JF (1996) Trace eyeblink condi- Neuroscience 117:821830.
tioning increases CA1 excitability in a transient and learning- Nakagawa T, Futai K, Lashuel HA, Lo I, Okamoto K, Walz T, Hiyashi
specic manner. J Neurosci 16:55365546. Y, Sheng M (2004) Quarternary structure, protein dynamics,
Mulkey RM, Malenka RC (1992) Mechanisms underlying induction and synaptic function of SAP97 controlled by L27 domain
of homosynaptic long-term depression in area CA1 of the hip- interactions. Neuron 44:453467.
pocampus. Neuron 9:967975. Nakamura K, Manabe T, Watanabe M, Mamiya T, Ichikawa R,
Mulkey RM, Herron CE, Malenka RC (1993) An essential role for Kiyama Y, Sanbo M, Yagi T, Inoue Y, Nabeshima T, Mori H,
protein phosphatases in hippocampal long-term depression. Mishina M (2001) Enhancement of hippocampal LTP, reference
Science 261:10511055. memory and sensorimotor gating in mutant mice lacking a
Mulkey RM, Endo S, Shenolikar S, Malenka RC (1994) Involvement telencephalon-specic cell adhesion molecule. Eur J Neurosci
of a calcineurin/inhibitor-1 phosphatase cascade in hippocam- 13:179189.
pal long-term depression. Nature 369:486488. Nakanishi K, Saito H, Abe K (2001) The supramammillary nucleus
Muller D, Joly M, Lynch G (1988) Contributions of quisqualate and contributes to associative EPSP-spike potentiation in the rat
NMDA receptors to the induction and expression of LTP. dentate gyrus in vivo. Eur J Neurosci 13:793800.
Science 242:16941697. Nakazawa K, Quirk MC, Chitwood RA, Watanabe M, Yeckel MF, Sun
Muller D, Bittar P, Boddeke H (1992) Induction of stable long-term LD, Kato A, Carr CA, Johnston D, Wilson MA, Tonegawa S
potentiation in the presence of the protein kinase C antagonist (2002) Requirement for hippocampal CA3 NMDA receptors in
staurosporine. Neurosci Lett 135:1822. associative memory recall. Science 297:211218.
Muller D, Hefft S, Figurov A (1995) Heterosynaptic interactions Nakazawa K, Sun LD, Quirk MC, Rondi-Reig L, Wilson MA,
between LTP and LTD in CA1 hippocampal slices. Neuron Tonegawa S (2003) Hippocampal CA3 NMDA receptors are
14:599605. crucial for memory acquisition of one-time experience. Neuron
Muller D, Wang D, Skibo G, Toni N, Cremer H, Calaora V, Rougon G, 38:305315.
Kiss JZ (1996) PSA-NCAM is required for activity-induced Nalbantoglu J, Tirado-Santiago G, Lahsaini A, Poirier J, Goncalves O,
synaptic plasticity. Neuron 17:413422. Verge G, Momoli F, Welner SA, Massicotte G, Julien JP, Shapiro
Munro CAM, Walling SG, Evans JH, Harley CW (2001) -Adrenergic ML (1997) Impaired learning and LTP in mice expressing the
blockade in the dentate gyrus in vivo prevents high frequency- carboxy terminus of the Alzheimer amyloid precursor protein.
induced long-term potentiation of EPSP slope, but not long- Nature 387:500505.
term potentiaion of population spike amplitude. Hippocampus Navakkode S, Sakijumar S, Frey JU (2004) The type IV-specic phos-
11:329336. phodiesterase inhibitor rolipram and its effects on hippocampal
 Synaptic Plasticity in the Hippocampus 465

long-term potentiation and synaptic tagging. J Neurosci 24: ODell TJ, Kandel ER, Grant SG (1991) Long-term potentiation in
77407744. the hippocampus is blocked by tyrosine kinase inhibitors.
Nayak A, Zastrow DJ, Lickteig R, Zahniser NR, Browning MD (1998) Nature 353:558560.
Maintenance of late-phase LTP is accompanied by PKA-depen- ODell TJ, Hawkins RD, Kandel ER, Arancio O (1991b) Tests of the
dent increase in AMPA receptor synthesis. Nature 394:680683. roles of two diffusible substances in long-term potentiation: evi-
Ngezahayo A, Schachner M, Artola A (2000) Synaptic activity modu- dence for nitric oxide as a possible early retrograde messenger.
lates the induction of bidirectional synaptic changes in adult Proc Natl Acad Sci USA 88:1128511289.
mouse hippocampus. J Neurosci 20:24512458. ODoherty A, Ruf S, Mulligan C, Hildreth V, Errington ML, Cooke S,
Nguyen PV, Abel T, Kandel ER (1994) Requirement for a critical Sesay A, Modino S, Vanes L, Hernandez D, Linehan JM, Sharpe
period of transcription for induction of a late phase of LTP. PT, Brandner S, Bliss TV, Henderson DJ, Nizetic D, Tybulewicz
Science 265:11041107. VL, Fisher EM (2005) An aneuploid mouse strain carrying
Nguyen PV, Woo NH (2003) Regulation of hippocampal synaptic human chromosome 21 with Down syndrome phenotypes.
plasticity by cyclic AMP-dependent protein kinases. Prog Science 309:20332037.
Neurobiol 71:401437. Oh MC, Derkach VA, Guire ES, Soderling TR (2006) Extrasynaptic
Nguyen PV, Duffy SN, Young JZ (2000) Differential maintenance membrane trafficking regulated by GluR1 serine 845 phospho-
and frequency-dependent tuning of LTP at hippocampal rylation primes AMPA receptors for long-term potentiation. J
synapses of specic strains of inbred mice. J Neurophysiol Biol Chem 281:752758
84:24842493. Oh MM, Kuo AG, Wu WW, Sametsky EA, Disterhoft JF (2003)
Nicholls JG, Martin AR, Wallace BG, Fuchs PA (2001) From neuron to Watermaze learning enhances excitability of CA1 pyramidal
brain: celluar approach to the function of the nervous system, 4th neurons. J Neurophysiol 90:21712179.
ed. Sunderland, MA: Sinauer. Ohno M, Frankland PW, Chen AP, Costa RM, Silva AJ (2001)
Nicoll RA, Schmitz D (2005) Synaptic plasticity at hippocampal Inducible, pharmacogenetic approaches to the study of learning
mossy ber synapses. Nat Rev Neurosci 6:863876. and memory. Nat Neurosci 4:12381243.
Nimchinsky EA, Sabatini BL, Svoboda K (2002) Structure and func- Ohno M, Frankland PW, Silva AJ (2002) A pharmacogenetic
tion of dendritic spines. Annu Rev Physiol 64:313353. inducible approach to the study of NMDA/alphaCaMKII sig-
Nishimune A, Isaac JT, Molnar E, Noel J, Nash SR, Tagaya M, naling in synaptic plasticity. Curr Biol 12:654656.
Collingridge GL, Nakanishi S, Henley JM (1998) NSF binding to Okada T, Yamada N, Tsuzuki K, Horikawa HP, Tanaka K, Ozawa S
GluR2 regulates synaptic transmission. Neuron 21:8797. (2003) Long-term potentiation in the hippocampal CA1 area
Nishiyama M, Hong K, Mikoshiba K, Poo M-M, Kato K (2000) and dentate gyrus plays different roles in spatial learning. Eur J
Calcium stores regulate the polarity and input specicity of Neurosci 17:341349.
synaptic modication. Nature 408:584588. Okamoto K-I, Nagai T, Miyawaki A, Hayashi Y (2004) Rapid and per-
Noel J, Ralph GS, Pickard L, Williams J, Molnar E, Uney JB, sistent modulation of actin dynamics regulates postsynaptic
Collingridge GL, Henley JM (1999) Surface expression of AMPA reorganisation underlying bidirectional plasticity. Nat Neurosci
receptors in hippocampal neurons is regulated by an NSF- 7:11041112.
dependent mechanism. Neuron 23:365376. Okumoto S, Looger LL, Micheva KD, Reimer RJ, Smith SJ, Frommer
Norris CM, Korol DL, Foster TC (1996) Increased susceptibility to WB (2005) Detection of glutamate release from neurons by
induction of long-term depression and long-term potentiation genetically encoded surface-displayed FRET nanosensors. Proc
reversal during aging. J Neurosci 16:53825392. Natl Acad Sci USA 102:87408745.
Nosten-Bertrand M, Errington ML, Murphy KP, Tokugawa Y, Oliet SHR, Malenka RC, Nicoll RA (1997) Two distinct forms of
Barboni E, Kozlova E, Michalovich D, Morris RG, Silver J, long-term depression co-exist in CA1 hippocampal pyramidal
Stewart CL, Bliss TV, Morris RJ (1996) Normal spatial learning cells. Neuron 18:969982.
despite regional inhibition of LTP in mice lacking Thy-1. Nature OMara SM, Commins S, Anderson M (2000) Synaptic plasticity in
379:826829. the hippocampal area CA1-subiculum projection: implications
Nowak L, Bregestovski P, Ascher P, Herbet A, Prochiantz A (1984) for theories of memory. Hippocampus 10:447456.
Magnesium gates glutamate-activated channels in mouse cen- Opazo P, Watabe AM, Grant SG, ODell TJ (2003) Phosphatidylinos-
tral neurones. Nature 307:462465. itol 3-kinase regulates the induction of long-term potentiation
Obenaus A, Mody I, Baimbridge KG (1989) Dantrolene-Na through extracellular signal-related kinase-independent mecha-
(Dantrium) blocks induction of long-term potentiation in hip- nisms. J Neurosci 3679:36793688.
pocampal slices. Neurosci Lett 98:172178. Osten P, Srivastava S, Inman GJ, Vilim FS, Khatri L, Lee LM, States
OConnor JJ, Rowan MJ, Anwyl R (1994) Long-lasting enhance- BA, Einheber S, Milner TA, Hanson PI, Ziff EB (1998) The
ment of NMDA receptor-mediated synaptic transmission by AMPA receptor GluR2 C terminus can mediate a reversible,
metabotropic glutamate receptor activation. Nature 367: ATP-dependent interaction with NSF and alpha- and beta-
557559. SNAPs. Neuron 21:99110.
OConnor JJ, Rowan MJ, Anwyl R (1995) Tetanically induced LTP Osten P, Khatri L, Perez JL, Kohr G, Giese G, Daly C, Schulz TW,
involves a similar increase in the AMPA and NMDA receptor Wensky A, Lee LM, Ziff EB (2000) Mutagenesis reveals a role for
components of the excitatory postsynaptic current: investiga- ABP/GRIP binding to GluR2 in synaptic surface accumulation
tions of the involvement of mGlu receptors. J Neurosci of the AMPA receptor. Neuron 27:313325.
15:20132020. Ostroff LE, Fiala JC, Allwardt B, Harris KM (2002) Polyribosomes
ODell TJ, Kandel ER (1994) Low-frequency stimulation erases LTP redistribute from dendritic shafts into spines with enlarged
through an NMDA receptor-mediated activation of protein synapses during LTP in developing rat hippocampal slices.
phosphatases. Learn Mem 1:129139. Neuron 35:535545.
466 The Hippocampus Book

Otani S, Abraham WC (1989) Inhibition of protein synthesis in the potentiation dependent on mGluR1a in hippocampal inhibi-
dentate gyrus, but not the entorhinal cortex, blocks mainte- tory interneurones. Proc Natl Acad Sci USA 98:94019406.
nance of long-term potentiation in rats. Neurosci Lett Petersen CCH, Malenka RC, Nicoll RA, Hopeld JJ (1998) All-or-
106:175180. none potentiation at CA3-CA1 synapses. Proc Natl Acad Sci USA
Otani S, Marshall CJ, Tate WP, Goddard GV, Abraham WC (1989) 95:47324737.
Maintenance of long-term potentiation in rat dentate gyrus Pettit DL, Perlman S, Malinow R (1994) Potentiated transmission
requires protein synthesis but not messenger RNA synthesis and prevention of further LTP by increased CaMKII activity
immediately post-tetanization. Neuroscience 28:519526. in postsynaptic hippocampal slice neurons. Science 266:
Otmakhova NA, Lisman JE (1996) D1/D5 dopamine receptor activa- 18811885.
tion increases the magnitude of early long-term potentiation at Philpot BD, Esinosa JS, Bear MF (2003) Evidence for enhanced
CA1 hippocampal synapses. J Neurosci 16:74787486. NMDA receptor function as a basis for metaplasticity in visual
Otmakhova NA, Otmakhov N, Mortenson LH, Lisman JE (2000) cortex. J Neurosci 23:55835588.
Inhibition of the cAMP pathway decreases early long-term Pickard L, Noel J, Duckworth JK, Fitzjohn SM, Henley JM,
potentiation at CA1 hippocampal synapses. J Neurosci Collingridge GL, Molnar E (2001) Transient synaptic activation
20:44464451. of NMDA receptors leads to the insertion of native AMPA
Otnaess MK, Brun VH, Moser MB, Moser EI (1999) Pretraining pre- receptors at hippocampal neuronal plasma membranes.
vents spatial learning impairment after saturation of hippocam- Neuropharmacology 41:700713.
pal long-term potentiation. J Neurosci 19:RC49. Pike FG, Meredith RM, Olding AW, Pauslen O (1999) Postsynaptic
Packard MG, Teather LA (1997) Posttraining injections of MK-801 bursting is essential for Hebbian induction of associative long-
produce a time-dependent impairment of memory in two term potentiation at excitatory synapses in rat hippocampus. J
watermaze tasks. Neurobiol Learn Mem 68:4250. Phyiol (Lond) 518:571576.
Palmer CL, Lim W, Hastie PG, Toward M, Korolchuk VI, Burbidge Pineda VV, Athos JI, Wang H, Celver J, Ippolito D, Boulay G,
SA, Banting G, Collingridge GL, Isaac JT, Henley JM (2005) Birnbaumer L, Storm DR (2004) Removal of G(ialpha1) con-
Hippocalcin functions as a calcium sensor in hippocampal LTD. straints on adenylyl cyclase in the hippocampus enhances LTP
Neuron 47:487494. and impairs memory formation. Neuron 41: 153163.
Palmer MJ, Irving AJ, Seabrook GR, Jane DE, Collingridge GL (1997) Piomelli D, Volterra A, Dale N, Siegelbaum SA, Kandel ER, Schwartz
The group I mGlu receptor agonist DHPG induces a novel form JH, Berladetti F (1987) Lipoxygenase metabolites of arachidonic
of LTD in the CA1 region of the hippocampus. Neuropharmaco- acid are second messengers for presynaptic inhibition of Aplysia
logy 36:1517532. sensory cells. Nature 328:3843.
Palmer MJ, Isaac JT, Collingridge GL (2004) Multiple, developmen- Pittenger C, Huang YY, Paletzki RF, Bourtchouladze R, Scanlin H,
tally regulated expression mechanisms of long-term potentia- Vronskaya S, Kandel ER (2002) Reversible inhibition of
tion at CA1 synapses. J Neurosci 24:49034911. CREB/ATF transcription factors in region CA1 of the dorsal
Parent A, Linden DJ, Sisodia SS, Borchelt DR (1999) Synaptic trans- hippocampus disrupts hippocampus-dependent spatial mem-
mission and hippocampal long-term potentiation in transgenic ory. Neuron 34:447462.
mice expressing FAD-linked presenilin 1. Neurobiol Dis 6:5662. Plant K, Pelkey KA, Bortolotto ZA, Morita D, Terashima A, McBain
Parsons CG, Danysz W, Quack G (1999) Memantine is a clinically CJ, Collingridge GL, Isaac JT (2006) Transient incorporation of
well tolerated N-methyl-D-aspartate (NMDA) receptor antago- native GluR2-lacking AMPA receptors during hippocampal
nista review of preclinical data. Neuropharmacology 38: long-term potentiation. Nat Neurosci 9:6024.
735767. Pokorska A, Vanhoutte P, Arnold FJL, Silvagno F, Hardingham GE,
Parsons CG, Gruner R, Rozental J, Millar J, Lodge D (1993) Patch Bading H (2003) Synaptic activity induces signalling to CREB
clamp studies on the kinetics and selectivity of N-methyl-D- without increasing global levels of cAMP in hippocampal neu-
aspartate receptor antagonism by memantine (1-amino-3,5- rons. J Neurochem 84:447452.
dimethyladamantan). Neuropharmacology 32:13371350. Poncer JC, Esteban JA, Malinow R (2002) Multiple mechanisms for
Passafaro M, Piech V, Sheng M (2001) Subunit-specic temporal and the potentiation of AMPA receptor-mediated transmission by
spatial patterns of AMPA receptor exocytosis in hippocampal alpha-Ca2/calmodulin-dependent protein kinase II. J Neurosci
neurons. Nat Neurosci 4:917926. 22:44064411.
Pastalkova E, Serrano P, Pinkhasova D, Wallace E, Fenton AA, Sacktor Popov VI, Davies HA, Rogachevsky VV, Patrushev IV, Errington ML,
TC (2006) Storage of spatial information by the maintenance Gabbott PLA, Bliss TVP, Stewart MG (2004) Remodelling of
mechanism of LTP. Science, 313:11411144. synaptic morphology but unchanged synaptic density during
Patterson SL, Abel T, Deuel TA, Martin KC, Rose JC, Kandel ER late phase LTP: a serial section electron micrographic study of
(1996) Recombinant BDNF rescues decits in basal synaptic the dentate gyrus in the anesthetized rat. Neuroscience
transmission and hippocampal LTP in BDNF knockout mice. 128:251262.
Neuron 16:11371145. Purcell AL, Carew TJ (2003) Tyrosine kinases, synaptic plasticity and
Paulsen O, Li Y-G, Hvalby O, Andersen P, Bliss TVP (1993) Failure to memory: insights from vertebrates and invertebrates. Trends
induce long-term depression by an anti-correlation procedure Neurosci 26:625630.
in area CA1 of the rat hippocampal slice. Eur J Neurosci Quinta-Ferreira ME, Matias CM (2004) Hippocampal mossy ber
5:12411246. calcium transients are maintained during long-term potentia-
Pelkey KA, Lavezzari G, Racca C, Roche KW, McBain CJ (2005) tion and are inhibited by endogenous zinc. Brain Res 1004:
mGluR7 is a metaplastic switch controlling bidirectional plas- 5260.
ticity of feedforward inhibition. Neuron 46:89102. Racine RJ, Milgram NW, Hafner S (1983) Long-term potentiation
Perez Y, Morin F, Lacaille JC (2001) A Hebbian form of long-term phenomena in the rat limbic forebrain. Brain Res 260: 217231.
 Synaptic Plasticity in the Hippocampus 467

Rae MG, Martin DJ, Collingridge GL, Irving AJ (2000) Role of Ca2 JD (1999) The mitogen-activated protein kinase cascade couples
stores in metabotropic L-glutamate receptor-mediated supra- PKA and PKC to cAMP response element binding protein phos-
linear Ca2 signaling in rat hippocampal neurons. J Neurosci phorylation in area CA1 of hippocampus. J Neurosci
20:86288636. 19:43374348.
Rajji T, Chapman D, Eichenbaum H, Greene R (2006) The role of Roberts L, Greene JR (2003) Post-weaning social isolation of rats
CA3 hippocampal NMDA receptors in paired associate learn- leads to a diminution of LTP in the CA1 to subiculum pathway.
ing. J Neurosci 26:908915. Brain Res 991:271273.
Rammes G, Zeilhofer HU, Collingridge GL, Parsons CG, Swandulla D Roberts M, Shapiro M (2002) NMDA receptor antagonists impair
(1999) Expression of early hippocampal CA1 LTP does not lead memory for nonspatial, socially transmitted food preference.
to changes in AMPA-EPSC kinetics or sensitivity to cycloth- Behav Neurosci 116:10591069.
iazide. Pugers Arch 437:191196. Rogan MT, Staubli UV, LeDoux JE (1997) Fear conditioning induces
Rammes G, Palmer M, Eder M, Dodt H-U, Zieglgansberger W, associative long-term potentiation in the amygdala. Nature
Collingridge GL (2003) Activation of mGlu receptors induces 390:604607.
LTD without affecting postsynaptic sensitivity of CA1 neurons Roman FS, Truchet B, Chaillan FA, Marchetti E, Soumireu-Mourat B
in rat hippocampal slices. J Physiol 546:455460. (2004) Olfactory associative discrimination: a model for study-
Raymond CR, Redman SJ (2006) Spatial segregation of neuronal cal- ing modications of synaptic efficacy in neuronal networks
cium signals encodes different forms of LTP in rat hippocam- supporting long-term memory. Rev Neurosci 15:117.
pus. J Physiol (Lond) 570:97111. Romcy-Pereira R, Pavlides C (2004) Distinct modulatory effects of
Raymond CR, Redman SJ, Crouch MF (2002) The phosphoinositide sleep on the maintenance of hippocampal and medial prefrontal
3-kinase and p70 S6 kinase regulate long-term potentiation in cortex LTP. Eur J Neurosci 20:34533462.
hippocampal neurons. Neuroscience 109:531536. Rose GM, Dunwiddie TV (1986) Induction of hippocampal long-
Reid CA, Dixon D, Takahashi M, Bliss TVP, Fine A (2004) Optical term potentiation using physiologically patterned stimulation.
quantal analysis indicates LTP at single hippocampal mossy Neurosci Lett 69:244248.
ber synapses is expressed through increased release probability, Rosenblum K, Dudai Y, Richter-Levin G (1996) Long-term potentia-
recruitment of new release sites and activation of silent tion increases tyrosine phosphorylation of the N-methyl-D-
synapses. J Neurosci 24:36183626. aspartate receptor subunit 2B in rat dentate gyrus in vivo. Proc
Reid IC, Stewart CA (1997) Seizures, memory and synaptic plasticity. Natl Acad Sci USA 93:1045710460.
Seizure 6:351359. Rosenmund C, Clements JD, Westbrook GL (1993) Nonuniform
Reisel D, Bannerman DM, Schmitt WB, Deacon RM, Flint J, probability of glutamate release at a hippocampal synapse.
Borchardt T, Seeburg PH, Rawlins JN (2002) Spatial memory Science 262:754757.
dissociations in mice lacking GluR1. Nat Neurosci 5:868873. Rosenzweig ES, Rao G, McNaughton BL, Barnes CA (1997) Role of
Ribeiro S, Mello CV, Velho T, Gardner TJ, Jarvis ED, Pavlides C (2002) temporal summation in age-related long-term potentiation-
Induction of hippocampal long-term potentiation during wak- induction decits. Hippocampus 7:549558.
ing leads to increased extrahippocampal zif-268 expression dur- Rosenzweig ES, Barnes CA, McNaughton BL (2002) Making room
ing ensuing rapid-eye-movement sleep. J Neurosci 22: for new memories. Nat Neurosci 5:68.
1091410923. Ross FM, Allan SM, Rothwell NJ, Verkhratsky A (2003) A dual role for
Richter-Levin G, Akirav I (2003) Emotional tagging of memory for- interleukin-1 in LTP in mouse hippocampal slices. J
mationin the search for neural mechanisms. Brain Res Brain Neuroimmunol 144:6167.
Res Rev 43:247256. Ross ST, Soltesz I (2001) Long-term plasticity in interneurons in the
Richter-Levin G, Errington ML, Maegawa H, Bliss TV (1994) dentate gyrus. Proc Natl Acad Sci USA 98:88748879.
Activation of metabotropic glutamate receptors is necessary for Rostas JA, Brent VA, Voss K, Errington ML, Bliss TV, Gurd JW (1996)
long-term potentiation in the dentate gyrus and for spatial Enhanced tyrosine phosphorylation of the 2B subunit of the N-
learning. Neuropharmacology 33:853857. methyl-D-aspartate receptor in long-term potentiation. Proc
Richter-Levin G, Canevari L, Bliss TV (1995) Long-term potentiation Natl Acad Sci USA 93:1045210456.
and glutamate release in the dentate gyrus: links to spatial learn- Rotenberg A, Mayford M, Hawkins RD, Kandel ER, Muller RU (1996)
ing. Behav Brain Res 66:3740. Mice expressing activated CaMKII lack low frequency LTP and
Richter-Levin G, Canevari L, Bliss TVP (1997) Spatial training and do not form stable place cells in the CA1 region of the hip-
high-frequency stimulation engage a common pathway to pocampus. Cell 87:13511361.
enhance glutamate release in the hippocampus. Learn Mem Rouach N, Byrd K, Petralia RS, Elias GM, Adesnik H, Tomita S,
4:445450. Karimzadegan S, Kealey C, Bredt DS, Nicoll RA (2005) TARP
Riedel G, Reymann K (1993) An antagonist of the metabotropic glu- gamma-8 controls hippocampal AMPA receptor number, distri-
tamate receptor prevents LTP in the dentate gyrus of freely mov- bution and synaptic plasticity. Nat Neurosci 8:15251533.
ing rats. Neuropharmacology 32:929931. Routtenberg A, Rekart JL (2005) Post-translational protein modica-
Riedel G, Platt B, Micheau J (2003) Glutamate receptor function in tion as the substrate for long-term memory. Trends Neurosci
learning and memory. Behav Brain Res 140:1-47. 28:1219.
Rioult-Pedotti MS, Friedman D, Hess G, Donoghue JP (1998) Routtenberg A, Lovinger D (1985) Selective increase in phosphoryla-
Strengthening of horizontal cortical connections following skill tion of a 47 kDA protein (F1) directly related to long-term
learning. Nat Neurosci 1:230234. potentiation. Behav Neural Biol 43:311.
Rioult-Pedotti M-S, Friedman D, Donoghue JP (2000) Learning- Routtenberg A, Cantallops I, Zaffuto S, Serrano P, Namgung U (2000)
induced LTP in neocortex. Science 290:533536. Enhanced learning after genetic overexpression of a brain
Roberson ED, English JD, Adams JP, Selcher JC, Kondratick C, Sweatt growth protein. Proc Natl Acad Sci USA 97:76577662.
468 The Hippocampus Book

Rowan MJ, Klyubin I, Cullen WK, Anwyl R (2003) Synaptic plasticity Salt TE (1986) Mediation of thalamic sensory input by both NMDA
in animal models of early Alzheimers disease. Philos Trans R Soc receptors and non-NMDA receptors. Nature 322:263265.
Lond B Biol Sci 358:821828. Salt TE, Eaton SA (1989) Function of non-NMDA receptors and
Rozov A, Burnashev N (1999) Polyamine-dependent facilitation of NMDA receptors in synaptic responses to natural somatosen-
postsynaptic AMPA receptors counteracts paired-pulse depres- sory stimulation in the ventrobasal thalamus. Exp Brain Res
sion. Nature 401:594598. 77:646652.
Rumbaugh G, Sia GM, Garner CC, Huganir RL (2003) Synapse- Sanes JR, Lichtman JW (1999) Can molecules explain long-term
associated protein-97 isoform-specic regulation of surface potentiation? Nat Neurosci 2:597604.
AMPA receptors and synaptic function in cultured neurons. J Sanna PP, Cammalleri M, Berton F, Simpson C, Lutjens R, Bloom FE,
Neurosci 23:45674576. Francesconi W (2002) Phosphatidylinositol 3-kinase is required
Rumelhart DE, Hinton GE, Williams RJ (1986) Learning representa- for the expression but not for the induction or the maintenance
tions by back-propagation errors. Nature 323:533536. of long-term potentiation in the hippocampal CA1 region. J
Rumpel S, LeDoux J, Zador A, Malinow R (2005) Postsynaptic recep- Neurosci 22:33593365.
tor trafficking underlying a form of associative learning. Science Sastry BR, Goh JW, Auyeung A (1986) Associative induction of post-
308:8388. tetanic and long-term potentiation in CA1 neurons of rat hip-
Rush AM, Wu J, Rowan MJ, Anwyl R (2002) Group I metabotropic pocampus. Science 232:988990.
glutamate receptor (mGluR)-dependent long-term depression Saucier D, Cain DP (1995) Spatial learning without NMDA receptor-
mediated via p38 mitogen-activated protein kinase is inhibited dependent long-term potentiation. Nature 378:186189.
by previous high-frequency stimulation and activation of Saucier D, Hargreaves EL, Boon F, Vanderwolf CH, Cain DP (1996)
mGluRs and protein kinase C in the rat dentate gyrus in vitro. J Detailed behavioral analysis of watermaze acquisition under
Neurosci 22:61216128. systemic NMDA or muscarinic antagonism: nonspatial pretrain-
Sabatini BL, Oertner TG, Svoboda K (2002) The life cycle of Ca(2) ing eliminates spatial learning decits. Behav Neurosci 110:
ions in dendritic spines. Neuron 33:43952. 103116.
Sacchetti B, Lorenzini CA, Baldi E, Bucherelli C, Roberto M, Tassoni Scammell TE, Arrigoni E, Thompson MA, Ronan PJ, Saper CB,
G, Brunelli M (2001) Long-lasting hippocampal potentiation Greene RW (2003) Focal deletion of the adenosine A1 receptor
and contextual memory consolidation. Eur J Neurosci 13: in adult mice using an adeno-associated viral vector. J Neurosci
22912298. 23:57625770.
Sacchetti B, Lorenzini CA, Baldi E, Bucherelli C, Roberto M, Tassoni Scanziani M, Malenka RC, Nicoll RA (1996) Role of intercellular
G, Brunelli M (2002) Time-dependent inhibition of hippocam- interactions in heterosynaptic long-term depression. Nature
pal LTP in vitro following contextual fear conditioning in the 380:446450.
rat. Eur J Neurosci 15:143150. Scattton B, Simon H, Le Moal M, Bischoff S (1980) Origin of
Sacchetti B, Scelfo B, Tempia F, Strata P (2004) Long-term synaptic dopaminergic innervation of the rat hippocampal formation.
changes induced in the cerebellar cortex by fear conditioning. Neurosci Lett 18:125131.
Neuron 42:973982. Schinder AF, Gage FH (2004) A hypothesis about the role of adult
Sacktor TC, Osten P, Valsamis H, Jiang X, Naik MU, Sublette E (1993) neurogenesis in hippocampal function. Physiology (Bethesda)
Persistent activation of the zeta isoform of protein kinase C in 19:253261.
the maintenance of long-term potentiation. Proc Natl Acad Sci Schinder AF, Poo M-M (2000) The neurotrophin hypothesis for
U S A 90:83428346. synaptic plasticity. Trends Neurosci 23:639645.
Sajikumar S, Frey JU (2004a) Resetting of synaptic tags is time- and Schmidt-Hieber C, Jonas P, Bischofberger J (2004) Enhanced synap-
activity dependent in rat hippocampal CA1 in vitro. tic plasticity in newly generated granule cells of the adult hip-
Neuroscience 129:503507. pocampus. Nature 429:184187.
Sajikumar S, Frey JU (2004b) Late-associativity, synaptic tagging, and Schmitt WB, Sprengel R, Mack V, Draft RW, Seeburg PH, Deacon
the role of dopamine during LTP and LTD. Neurobiol Learn RM, Rawlins JN, Bannerman DM (2005) Restoration of spatial
Mem 82:1225. working memory by genetic rescue of GluR-A-decient mice.
Sajikumar S, Navakkode S, Sacktor TC, Frey JU (2005) Synaptic tag- Nat Neurosci 8:270272.
ging and cross-tagging: the role of protein kinase Mzeta in Schmitz D, Mellor J, Nicoll RA (2001) Presynaptic kainate receptor
maintaining long-term potentiation but not long-term depres- mediation of frequency facilitation at hippocampal mossy ber
sion. J Neurosci 25:57505756. synapses. Science 291:19721976.
Sakamoto T, Porter LL, Asanuma H (1987) Long-lasting potentiation Schmitz D, Mellor J, Breustedt J, Nicoll RA (2003) Presynaptic kainate
of s ynaptic potentials in the motor cortex produced by stimu- receptors impart an associative property to hippocampal mossy
lation of the sensory cortex in the cat: a basis of motor learning. ber long-term potentiation. Nat Neurosci 6: 10581063.
Brain Res 413:360364. Schnabel R, Kilpatrick IC, Collingridge GL (1999) An investigation
Sakimura K, Kutsuwada T, Ito I, Manabe T, Takayama C, Kushiya into signal transduction mechanisms involved in DHPG-
E, Yagi T, Aizawa S, Inoue Y, Sugiyama H, et al. (1995) Reduced induced LTD in the CA1 region of the hippocampus.
hippocampal LTP and spatial learning in mice lacking NMDA Neuropharmacology 38:15851596.
receptor epsilon 1 subunit. Nature 373:151155. Schnabel R, Kilpatrick IC, Collingridge GL (2001) Protein phos-
Salin PA, Weisskopf MG, Nicoll RA (1995) A comparison of the role phatase inhibitors facilitate DHPG-induced LTD in the CA1
of dynorphin in the hippocampal mossy ber pathway in region of the hippocampus. Br J Pharmacol 132:10951101.
guinea-pig and rat. J Neurosci 15:69396945. Schneider I, Reverse D, Dewachter I, Ris L, Caluwaerts N, Kuiperi C,
Salin PA, Scanziani M, Malenka RC, Nicoll RA (1996) Distinct short- Gilis M, Geerts H, Kretzschmar H, Godaux E, Moechars D, Van
term plasticity at two excitatory synapses in the hippocampus. Leuven F, Herms J (2001) Mutant presenilins disturb neuronal
Proc Natl Acad Sci USA 93:1330413309. calcium homeostasis in the brain of transgenic mice, decreasing
 Synaptic Plasticity in the Hippocampus 469

the threshold for excitotoxicity and facilitating long-term AMPA receptors after synaptic NMDA receptor activation.
potentiation. J Biol Chem 276:1153911544. Science 284:18111816.
Schroeder BW, Shinnick-Gallagher P (2005) Fear learning induces Shimizu E, Tang YP, Rampon C, Tsien JZ (2000) NMDA receptor-
persistent facilitation of amygdala synaptic transmission. Eur J dependent synaptic reinforcement as a crucial process for mem-
Neurosci 22:17751783. ory consolidation. Science 290:11701174.
Schubert M, Siegmund H, Pape HC, Albrecht D (2005) Kindling- Shors TJ, Matzel LD (1997) Long-term potentiation: whats learning
induced changes in plasticity of the rat amygdala and hip- got to do with it? Behav Brain Sciences 20:597655.
pocampus. Learn Mem 12:520526. Shors TJ, Seib TB, Levine S, Thompson RF (1989) Inescapable versus
Schulz PE, Cook EP, Johnston D (1994) Changes in paired-pulse escapable shock modulates long-term potentiation in the rat
facilitation suggest presynaptic involvement in long-term hippocampus. Science 244:224226.
potentiation. J Neurosci 14:53255337. Shors TJ, Miesegaes G, Beylin A, Zhao M, Rydel T, Gould E (2001)
Schuman EM, Madison DV (1991) A requirement of the intercellular Neurogenesis in the adult is involved in the formation of trace
messenger nitric oxide in long-term potentiation. Science memories. Nature 410:372376.
254:15031506. Shors TJ, Townsend DA, Zhao M, Kozorovitskiy Y, Gould E (2002)
Schwartzkroin PA, Wester K (1975) Long-lasting facilitation of a Neurogenesis may relate to some but not all types of hippocam-
synaptic potential following tetanization in the in vitro hip- pal-dependent learning. Hippocampus 12:578584.
pocampal slice. Brain Res 89:107119. Shouval HZ, Bear MF, Cooper LN (2002) A unied model of NMDA
Seabrook GR, Smith DW, Bowery BJ, Easter A, Reynolds T, Fitzjohn receptor-dependent bidirectional synaptic plasticity. Proc Natl
SM, Morton RA, Zheng H, Dawson GR, Sirinathsinghji DJ, Acad Sci USA 99:1083110836.
Davies CH, Collingridge GL, Hill RG (1999) Mechanisms con- Si K, Lindquist S, Kandel ER (2003) A neuronal isoform of the aplysia
tributing to the decits in hippocampal synaptic plasticity in CPEB has prion-like properties. Cell 115:879891.
mice lacking amyloid precursor protein. Neuropharmacology Siarey RJ, Stoll J, Rapoport SI, Galdzicki Z (1997) Altered long-term
38:349359. potentiation in the young and old Ts65Dn mouse, a model for
Segal M, Auerbach JM (1997) Muscarinic receptors involved in hip- Down syndrome. Neuropharmacology 36:15491554.
pocampal plasticity. Life Sci 13/14:10851091. Siarey RJ, Carlson EJ, Epstein CJ, Balbo A, Rapoport SI, Galdzicki Z
Seidenbecher T, Balschun D, Reymann KG (1995) Drinking after (1999) Increased synaptic depression in the Ts65Dn mouse, a
water deprivation prolongs unsaturated LTP in the dentate model for mental retardation in Down syndrome. Neurophar-
gyrus of rats. Physiol Behav 57:10011004. macology 38:19171920.
Sejnowski TJ (1977) Storing covariance with nonlinearly interacting Siarey RJ, Villar AJ, Epstein CJ, Galdzicki Z (2005) Abnormal synap-
neurons. J Math Biol 4:303321. tic plasticity in the Ts1Cje segmental trisomy 16 mouse model
Selig DK, Hjelmstad GO, Herron C, Nicoll RA, Malenka RC (1995a) of Down syndrome. Neuropharmacology 49:122128.
Independent mechanisms for long-term depression of AMPA Sillito AM (1985) Inhibitory circuits and orientation selectivity in the
and NMDA responses. Neuron 15:417426. visual cortex. In: Models of the visual cortex (Dobson DRaVG,
Selig DK, Lee HK, Bear MF, Malenka RC (1995b) Reexamination of ed), pp 396407. New York: Wiley.
the effects of MCPG on hippocampal LTP, LTD, and depotenti- Silva AJ, Paylor R, Wehner JM, Tonegawa S (1992a) Impaired spatial
ation. J Neurophysiol 74:10751082. learning in alpha-calcium calmodulin kinase II mutant mice.
Selig DK, Nicoll RA, Malenka RC (1999) Hippocampal long-term Science 257:206211.
potentiation preserves the delity of postsynaptic responses to Silva AJ, Stevens CF, Tonegawa S, Wang Y (1992b) Decient hip-
presynaptic bursts. J Neurosci 19:12361246. pocampal long-term potentiation in alpha-calcium-calmodulin
Sergueeva OA, Fedorov NB, Reymann KG (1993) An antagonist kinase II mutant mice. Science 257:201206.
of glutamate metabotropic receptors, (RS)-alpha-methyl- Simonyi A, Schachtman TR, Christoffersen GR (2005) The role of
4-arboxyphenylglycine, prevents the LTP-related increase in metabotropic glutamate receptor 5 in learning and memory
postsynaptic AMPA sensitivity in hippocampal slices. processes. Drug News Perspect 18:353361.
Neuropharmacology 32:933935. Singer W (1990) The formation of cooperative cell assemblies in the
Serrano P, Yao Y, Sacktor TC (2005) Persistent phosphorylation by visual cortex. J Exp Biol 153:177197.
protein kinase Mzeta maintains late-phase long-term potentia- Skrede KKR, Westgaard RH (1971) The transverse hippocampal slice:
tion. J Neurosci 25:19791984. a well-dened cortical structure maintained in vitro. Brain Res
Sevelinges Y, Gervais R, Messaoudi B, Granjon L, Mouly AM (2004) 35:589593.
Olfactory fear conditioning induces eld potential potentia- Snyder EM, Philpot BD, Huber KM, Dong X, Fallon JR, Bear MF
tion in rat olfactory cortex and amygdala. Learn Mem 11: (2001) Internalization of ionotropic glutamate receptors in
761769. response to mGluR activation. Nat Neurosci 4: 10791085.
Sharp PE, McNaughton BL, Barnes CA (1989) Exploration-ependent Soderling TR, Derkach VA (2000) Postsynaptic protein phosphoryla-
modulation of evoked response in fascia dentata: fundamental tion and LTP. Trends Neurosci 23:7580.
observations and time-course. Psychobiology 17:257269. Sokolov MV, Rossokhin AV, Kasyanov A, Gasparini S, Berretta N,
Shew T, Yip S, Sastry BR (2000) Mechanisms involved in tetanus- Cherubini E, Voronin L (2003) Associative mossy ber LTP
induced potentiation of fast IPSCs in rat hippocampal CA1 induced by pairing presynaptic stimulation with postsynaptic
neurons. J Neurophysiol 83:33883401. hyperpolarization of CA3 neurons in rat hippocampal slice. Eur
Shi S, Hayashi Y, Esteban JA, Malinow R (2001) Subunit-specic rules J Neurosci 17:14251437.
governing AMPA receptor trafficking to synapses in hippocam- Son H, Hawkins RD, Martin K, Kiebler M, Huang PL, Fishman MC,
pal pyramidal neurons. Cell 105:331343. Kandel ER (1996) Long-term potentiation is reduced in mice
Shi SH, Hayashi Y, Petralia RS, Zaman SH, Wenthold RJ, Svoboda K, that are doubly mutant in endothelial and neuronal nitric oxide
Malinow R (1999) Rapid spine delivery and redistribution of synthase. Cell 87:10151023.
470 The Hippocampus Book

Song H, Stevens CF, Gage FH (2002) Neural stem cells from adult Staubli U, Xu FB (1995) Effects of 5-HT3 receptor antagonism on
hippocampus develop essential properties of functional CNS hippocampal theta rythm, memory and LTP induction in the
neurons. Nat Neurosci 5:438445. freely moving rat. J Neurosci Methods 15:24452452.
Song I, Kamboj S, Xia J, Dong H, Liao D, Huganir RL (1998) Staubli U, Ivy G, Lynch G (1984) Hippocampal denervation causes
Interaction of the N-ethylmaleimide-sensitive factor with rapid forgetting of olfactory information in rats. Proc Natl Acad
AMPA receptors. Neuron 21:393400. Sci USA 81:58855887.
Song S, Miller KD, Abbott LF (2000) Competitive Hebbian learning Staubli U, Thibault O, DiLorenzo M, Lynch G (1989) Antagonism of
through spike-timing-dependent synaptic plasticity. Nat NMDA receptors impairs acquisition but not retention of olfac-
Neurosci 3:919926. tory memory. Behav Neurosci 103:5460.
Sorra KE, Harris KM (1998) Stability in synapse number and size at Staubli U, Perez Y, Xu FB, Rogers G, Ingvar M, Stone Elander S, Lynch
2 hr after long-term potentiation in the hippocampus. J G (1994) Centrally active modulators of glutamate receptors
Neurosci 18:658671. facilitate the induction of long-term potentiation in vivo. Proc
Sourdet V, Russier M, Daoudal G, Ankri N, Debanne D (2003) Long- Nat Acad of Sci USA 91:1115811162.
term enhancement of neuronal excitability and temporal Staubli U, Chun D, Lynch G (1998) Time-dependent reversal of long-
delity mediated by metabotropic glutamate receptor subtype term potentiation by an integrin antagonist. J Neurosci
5. J Neurosci 23:1023810248. 18:34603469.
Spangenberger H, Nikmanesh FG, Igelmund P (1995) Long-term Steele RJ, Morris RG (1999) Delay-dependent impairment of a match-
potentiation at low temperature is stronger in hippocampal ing-to-place task with chronic and intrahippocampal infusion of
slices from hibernating Turkish hamsters compared to warm- the NMDA-antagonist D-AP5. Hippocampus 9:118136.
acclimated hamsters and rats. Neurosci Lett 194:127129. Stein V, House DR, Bredt DS, Nicoll RA (2003) Postsynaptic density-
Sprengel R, Suchanek B, Amico C, Brusa R, Burnashev N, Rozov A, 95 mimics and occludes hippocampal long-term potentiation
Hvalby O, Jensen V, Paulsen O, Andersen P, Kim JJ, Thompson and enhances long-term depression. J Neurosci 23:55035506.
RF, Sun W, Webster LC, Grant SG, Eilers J, Konnerth A, Li J, Stellwagen D, Malenka RC (2006) Synaptic scaling mediated by glial
McNamara JO, Seeburg PH (1998) Importance of the intracel- TNF-alpha. Nature 440:10541059.
lular domain of NR2 subunits for NMDA receptor function in Stent GS (1973) A physiological mechanism for Hebbs postulate of
vivo. Cell 92:279289. learning. Proc Natl Acad Sci USA 70:9971001.
Squire LR, Barondes SH (1972) Variable decay of memory and its Stephan A, Laroche S, Davis S (2001) Generation of aggregated beta-
recovery in cycloheximide-treated mice. Proc Natl Acad Sci USA amyloid in the rat hippocampus impairs synaptic transmission
69:14161420. and plasticity and causes memory decits. J Neurosci 21:5703-
Srivastava S, Ziff EB (1999) ABP: a novel AMPA receptor binding 5714.
protein. Ann NY Acad Sci 868:561564. Stevens CF (1998) A million dollar question: does LTP  memory?
Staff NP, Spruston N (2003) Intracellular correlate of EPSP-spike Neuron 20:12.
potentiation in CA1 pyramidal neurons is controlled by Stevens CF, Wang Y (1993) Reversal of long-term potentiation by
GABAergic modulation. Hippocampus 13:801805. inhibitors of haem oxygenase. Nature 364:147149.
Stanton PK, Sejnowski TJ (1989) Associative long-term depression in Stevens CF, Wang Y (1994) Changes in reliability of synaptic function
the hippocampus induced by Hebbian covariance. Nature as a mechanism for plasticity. Nature 371:704707.
339:215218. Steward O (1983) Alterations in polyribosomes associated with den-
Stanton PK, Sarvey JM (1985) Depletion of norepinephrine, but not dritic spines during the reinnervation of the dentate gyrus of
serotonin, reduces long-term potentiation in the dentate gyrus the adult rat. J Neurosci 3:177188.
of rat hippocampal slices. J Neurosci 5:21692176. Steward O, Schuman E (2001) Protein synthesis at synaptic sites on
Stanton PK, Heinemann U, Muller W (2001) FM1-43 imaging reveals dendrites. Annu Rev Neurosci 24:299235.
cGMP-dependent long-term depression of presynaptic trans- Steward O, Schuman EM (2003) Compartmentalized synthesis and
mitter release. J Neurosci 21:RC167. degradation of proteins in neurons. Neuron 40:347359.
Stanton PK, Winterer J, Bailey CP, Kyrozis A, Raginor I, Laube G, Veh Straube T, Frey JU (2003) Time-dependent depotentiation in the
RW, Nguyen CP, Muller W (2003) Long-term depression of dentate gyrus of freely moving rats by repeated brief 7 Hz stim-
presynaptic release from the readily releasable vesicle pool ulation. Neurosci Lett 339:8284.
induced by NMDA receptor-dependent retrograde nitric oxide. Straube T, Korz V, Balschun D, Frey JU (2003) Requirement of beta-
J Neurosci 23:59355944. adrenergic receptor activation and protein synthesis for LTP-
Stanton PK, Winterer J, Zhang XL, Muller W (2005) Imaging LTP of reinforcement by novelty in rat dentate gyrus. J Phyiol (Lond)
presynaptic release of FM1-43 from the rapidly recycling vesicle 552:953960.
pool of Schaffer collateral-CA1 synapses in rat hippocampal Stringer JL, Greeneld LJ, Hackett JT, Guyenet PG (1983) Blockade of
slices. Eur J Neurosci 22:24512461. long-term potentiation by phencyclidine and sigma opiates in
Staubli U, Chun D (1996) Factors regulating the reversibility of long- the hippocampus in vivo and in vitro. Brain Res 280:127138.
term potentiation. J Neurosci 16:853860. Stuart GJ, Sakmann B (1994) Active propagation of somatic action
Staubli U, Lynch G (1987) Stable hippocampal long-term potentia- potentials into neocortical pyramidal cell dendrites. Nature
tion elicited by theta-pattern stimulation. Brain Res 435: 367:6972.
227234. Sun MK, Alkon DL (2002) Impairment of hippocampal CA1 het-
Staubli U, Lynch G (1990) Stable depression of potentiated synaptic erosynaptic transformation and spatial memory by beta-amy-
responses in the hippocampus with 1-5 Hz stimulation. Brain loid(25-35). J Neurophysiol 87:24412449.
Res 513:113118. Sun MK, Alkon DL (2005) Protein kinase C isozymes: memory ther-
 Synaptic Plasticity in the Hippocampus 471

apeutic potential. Curr Drug Targets CNS Neurol Disord Thiels E, Kanterewicz BI, Norman ED, Trzaskos JM, Klann E (2002)
4:541552. Long-term depression in the adult hippocampus in vivo
Sutherland RJ, Dringenberg HC, Hoesing JM (1993) Induction of involves activation of extracellular signal-regulated kinase and
long-term potentiation at perforant path dentate synapses does phosphorylation of Elk-1. J Neurosci 22:20542062.
not affect place learning or memory. Hippocampus 3:141147. Thomas GM, Huganir RL (2004) MAPK cascade signalling and
Sweatt JD (2001a) Protooncogenes subserve memory formation in synaptic plasticity. Nat Rev Neurosci 5:173183.
the adult CNS. Neuron 31:671674. Thomas KL, Laroche S, Errington ML, Bliss TV, Hunt SP (1994)
Sweatt JD (2001b) The neuronal MAP kinase cascade: a biochemical Spatial and temporal changes in signal transduction pathways
signal integration system subserving synaptic plasticity and during LTP. Neuron 13:737745.
memory. J Neurochem 76:110. Thomas MJ, ODell TJ (1995) The molecular switch hypothesis fails
Sweatt JD (2004) Mitogen-activated protein kinases in synaptic plas- to explain the inconsistent effects of the metabotropic gluta-
ticity and memory. Curr Opin Neurobiol 14:311317. mate receptor antagonist MCPG on long-term potentiation.
Talwar SK, Gerstein GL (2001) Reorganization in awake rat auditory Brain Res 695:4552.
cortex by local microstimulation and its effect on frequency-dis- Thomas MJ, Watabe AM, Moody TD, Makhinson M, ODell TJ
crimination behavior. J Neurophysiol 86:15551572. (1998) Postsynaptic complex spike bursting enables the induc-
Tanaka H, Shan W, Phillips GR, Arndt K, Bozdaghi O, Shapiro L, tion of theta frequency synaptic stimulation. J Neurosci
Huntley GW, Benson DL, Colman DR (2000) Molecular modi- 18:71187126.
cation of N-cadherin in response to synaptic activity. Neuron Tiedtke PI, Bischoff C, Schmidt WJ (1990) MK-801-induced stereo-
25:93107. typy and its antagonism by neuroleptic drugs. J Neural Transm
Tancredi V, Zona C, Velotti F, Eusebi F, Santoni A (1990) Interleukin- Gen Sect 81:173182.
2 suppresses established long-term potentiation and inhibits its Togashi H, Abe K, Mizoguchi A, Takaoka K, Chisaka O, Takeichi M
induction in the rat hippocampus. Brain Res 525:149151. (2002) Cadherin regulates dendritic spine morphogenesis.
Tancredi V, DArcangelo G, Grassi F, Tarroni P, Palmieri G, Santoni A, Neuron 35:7789.
Eusebi F (1992) Tumor necrosis factor alters synaptic transmis- Tombaugh GC, Rowe WB, Chow AR, Michael TH, Rose GM (2002)
sion in rat hippocampal slices. Neurosci Lett 146:176178. Theta-frequency synaptic potentiation in CA1 in vitro distin-
Tang J, Wotjak CT, Wagner S, Williams G, Schachner M, Dityatev A guishes cognitively impaired from unimpaired aged Fischer 344
(2001) Potentiated amygdaloid auditory-evoked potentials and rats. J Neurosci 22:99329940.
freezing behavior after fear conditioning in mice. Brain Res Tomita S, Adesnik H, Sekiguchi M, Zhang W, Wada K, Howe JR,
919:232241. Nicoll RA, Bredt DS (2005) Stargazin modulates AMPA recep-
Tang J, Wagner S, Schachner M, Dityatev A, Wotjak CT (2003) Poten- tor gating and trafficking by distinct domains. Nature
tiation of amygdaloid and hippocampal auditory evoked poten- 435:10521058.
tials in a discriminatory fear-conditioning task in mice as a Tomita S, Fukata M, Nicoll RA, Bredt DS (2004) Dynamic interaction
function of tone pattern and context. Eur J Neurosci 18:639650. of stargazin-like TARPs with cycling AMPA receptors at
Tang L, Hung CP, Schuman EM (1998) A role for the cadherin fam- synapses. Science 303:15081511.
ily of cell adhesion molecules in hippocampal long-term poten- Tong G, Malenka RC, Nicoll RA (1996) Long-term potentiation in
tiation. Neuron 20:11651175. cultures of single hippocampal granule cells: a presynaptic form
Tang YG, Zucker RS (1997) Mitochondrial involvement in post-teta- of plasticity. Neuron 16:11471157.
nic potentiation of synaptic transmission. Neuron 18:483491. Tong L, Thornton PL, Balazs R, Cotman CW (2001) Beta -amyloid-
Tang Y-P, Shimizu E, Dube GR, Rampon C, Kerchner GA, Zhuo M, (1-42) impairs activity-dependent cAMP-response element-
Liu G, Tsien JZ (1999) Genetic enhancement of learning and binding protein signaling in neurons at concentrations in which
memory in mice. Nature 401:6369. cell survival Is not compromised. J Biol Chem 276:1730117306.
Tao H, Zhang LI, Bi G, Poo M-M (2000) Selective presynaptic prop- Toni N, Buchs PA, Nikonenko I, Bron CR, Muller D (1999) LTP pro-
agation of long-term potentiation in dened neural networks. J motes formation of multiple spine synapses between a single
Neurosci 20:32333243. axon terminal and a dendrite. Nature 402:421425.
Tardin C, Cognet L, Bats C, Lounis B, Choquet D (2003) Direct imag- Tonkiss J, Rawlins JNP (1991) The competitive NMDA antagonist
ing of lateral movements of AMPA receptors inside synapses. AP5, but not the noncompetitive antagonist MK801, induces a
EMBO J 22:46564665. delay-related impairment in spatial working memory in rats.
Tashiro A, Yuste R (2003) Structure and molecular organization of Exp Brain Res 85:349358.
dendritic spines. Histol Histopathol 18:617634. Tonkiss J, Morris RGM, Rawlins JNP (1988) Intra-ventricular infu-
Terashima A, Cotton L, Dev KK, Meyer G, Zaman S, Duprat F, Henley sion of the NMDA antagonist AP5 impairs performance on a
JM, Collingridge GL, Isaac JT (2004) Regulation of synaptic non-spatial operant DRL task in the rat. Exp Brain Res 73:
strength and AMPA receptor subunit composition by PICK1. J 181188.
Neurosci 24:53815390. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E,
Thiels E, Barrionuevo G, Berger T (1994) Excitatory stimulation dur- Svoboda K (2002) Long-term in vivo imaging of experience-
ing postsynaptic inhibition induces long-term depression in dependent synaptic plasticity in adult cortex. Nature 420:
hippocampus in vivo. J Neurophysiol 72:30093116. 788794.
Thiels E, Norman ED, Barrionuevo G, Klann E (1998) Transient and Tricklebank MD, Singh L, Oles RJ, Preston C, Iversen SD (1989) The
persistent increases in protein phosphatase activity during long- behavioural effects of MK-801: a comparison with antagonists
term depression in the adult hippocampus in vivo. Neuroscience acting non-competitively and competitively at the NMDA
86:10231029. receptor. Eur J Pharmacol 167:127135.
472 The Hippocampus Book

Trommald M, Hulleberg G, Andersen P (1996) Long-term potentia- Vereker E, Campbell V, Roche E, McEntee E, Lynch MA (2000a)
tion is associated with new excitatory spine synapses on rat den- Lipopolysaccharide inhibits long term potentiation in the rat
tate granule cells. Learn Mem 3:218228. dentate gyrus by activating caspase-1. J Biol Chem 275:
Trommer BL, Shah C, Yun SH, Gamkrelidze G, Pasternak ES, Ye GL, 2625226258.
Sotak M, Sullivan PM, Pasternak JF, LaDu MJ (2004) ApoE iso- Vereker E, ODonnell E, Lynch MA (2000b) The inhibitory effect of
form affects LTP in human targeted replacement mice. interleukin-1beta on long-term potentiation is coupled with
Neuroreport 15:26552658. increased activity of stress-activated protein kinases. J Neurosci
Tsien JZ, Chen DF, Gerber D, Tom C, Mercer EH, Anderson DJ, 20:68116819.
Mayford M, Kandel ER, Tonegawa S (1996a) Subregion-and cell Vereker E, ODonnell E, Lynch A, Kelly A, Nolan Y, Lynch MA (2001)
type-restricted gene knockout in mouse brain. Cell Evidence that interleukin-1beta and reactive oxygen species
87:13171326. production play a pivotal role in stress-induced impairment of
Tsvetkov E, Carlezon WA, Benes FM, Kandel ER, Bolshakov VY LTP in the rat dentate gyrus. Eur J Neurosci 14:18091819.
(2002) Fear conditioning occludes LTP-induced presynaptic Vickers CA, Dickson KS, Wyllie DJ (2005) Induction and mainte-
enhancement of synaptic transmission in the cortical pathway nance of late-phase long-term potentiation in isolated dendrites
to the lateral amygdala. Neuron 34:289300. of rat hippocampal CA1 pyramidal neurones. J Physiol (Lond)
Turrigiano GG, Nelson SB (2004) Homeostatic plasticity in the 568:803813.
developing nervous system. Nat Rev Neurosci 5:97107. Vignes M, Collingridge GL (1997) The synaptic activation of kainate
Turski L, Klockgether T, Turski WA, Schwarz M, Sontag KH (1990) receptors. Nature 388:179182.
Blockade of excitatory neurotransmission in the globus pallidus Villani F, Johnston D (1993) Serotonin inhibits induction of long-
induces rigidity and akinesia in the rat: implications for excita- term potentiation at commissural synapses in hippocampus.
tory neurotransmission in pathogenesis of Parkinsons diseases. Brain Res 606:304308.
Brain Res 512:125131. Villarreal DM, Do V, Haddad E, Derrick BE (2001) NMDA receptor
Tzounopoulos T, Janz R, Sudhof TC, Nicoll RA, Malenka RC (1998) antagonists sustain LTP and spatial memory: active processes
A role for cAMP in long-term depression at hippocampal mossy mediate LTP decay. Nat Neurosci 5:4852.
ber synapses. Neuron 21:837845. Volianskis A, Jensen MS (2003) Transient and sustained types of
Ueno S, Tsukamoto M, Hirano T, Kikuchi K, Yamada MK, Nishiyama long-term potentiation in the CA1 area of the rat hippocampus.
N, Nagano T, Matsuki N, Ikegaya Y (2002) Mossy ber Zn2 J Physiol (Lond) 550:459492.
spillover modulates heterosynaptic N-methyl-D-aspartate recep- Walling SG, Harley CW (2004) Locus ceruleus activation initiates
tor activity in hippocampal CA3 circuits. J Cell Biol 158:215220. delayed synaptic potentiation of perforant path input to the
Uetani N, Kato K, Ogura H, Mizuno K, Kawano K, Mikoshiba K, dentate gyrus in awake rats: a novel beta-adrenergic- and pro-
Yakura H, Asano M, Iwakura Y (2000) Impaired learning with tein synthesis-dependent mammalian plasticity mechanism. J
enhanced hippocampal long-term potentiation in PTPdelta- Neurosci 24:598604.
decient mice. EMBO J 19:27752785. Walsh DM, Klyubin I, Fadeeva JV, Cullen WK, Anwyl R, Wolfe MS,
Urban NN, Barrionuevo G (1996) Induction of Hebbian and non- Rowan MJ, Selkoe DJ (2002) Naturally secreted oligomers of
Hebbian mossy ber long-term potentiation by distinct pat- amyloid beta protein potently inhibit hippocampal long-term
terns of high-frequency stimulation. J Neurosci 16:42934299. potentiation in vivo. Nature 416:535539.
Urban NN, Henze DA, Lewis DA, Barrionuevo G (1996) Properties of Wang HW, Pasternak JF, Kuo H, Ristic H, Lambert MP, Chromy B,
LTP induction in the CA3 region of the primate hippocampus. Viola KL, Klein WL, Stine WB, Krafft GA, Trommer BL (2002)
Learn Mem 3:8695. Soluble oligomers of beta amyloid (1-42) inhibit long-term
Usdin MT, Shelbourne PF, Myers RM, Madison DV (1999) Impaired potentiation but not long-term depression in rat dentate gyrus.
synaptic plasticity in mice carrying the Huntingtons disease Brain Res 924:133140.
mutation. Hum Mol Genet 8:839846. Wang JH, Alger BE (1995) GABAergic and developmental inuences
Vaillend C, Billard JM, Laroche S (2004) Impaired long-term spatial on homosynaptic LTD and depotentiation in rat hippocampus.
and recognition memory and enhanced CA1 hippocampal LTP J Neurosci 15:15771586.
in the dystrophin-decient Dmd(mdx) mouse. Neurobiol Dis Wang Z, Xu N-I, Wu C-P, Duan S, Poo M-M (2003) Bidirectional
17:1020. changes in spatial dendritic integration accompany long-term
Valastro B, Ghribi O, Poirier J, Krzywkowski P, Massicotte G (2001) synaptic modications. Neuron 37:463472.
AMPA receptor regulation and LTP in the hippocampus of Warren SG, Humphreys AG, Juraska JM, Greenough WT (1995) LTP
young and aged apolipoprotein E-decient mice. Neurobiol varies across the estrous cycle: enhanced synaptic plasticity in
Aging 22:915. proestrus rats. Brain Res 703:2630.
Valastro B, Cossette J, Lavoie N, Gagnon S, Trudeau F, Massicotte G Wasling P, Hanse E, Gustafsson B (2002) Long-term depression in the
(2002) Up-regulation of glutamate receptors is associated with developing hippocampus: low induction threshold and synapse
LTP defects in the early stages of diabetes mellitus. Diabetologia nonspecicity. J Neurosci 22:18231830.
45:642650. Watabe AM, Carlisle HJ, ODell TJ (2002) Postsynaptic induction and
Van Harreveld A, Fifkova E (1975) Swelling of dendritic spines in presynaptic expression of group 1 mGlurR-dependent LTD in
the fascia dentata after stimulation of the perforant bers as a the hippocampal CA1 region. J Neurophysiol 87:13951403.
mechanism of post-tetanic potentiation. Exp Neurol 49: Wathey JC, Lytton WW, Jester JM, Sejnowski TJ (1992) Computer
736749. simulations of EPSP-spike (E-s) potentiation in hippocampal
Van Praag H, Christie BR, Sejnowski TJ, Gage FH (1999) Running CA1 pyramidal cells. J Neurosci 12:607618.
enhances neurogenesis, learning and long-term potentiation in Weinberger NM (2004) Specic long-term memory traces in primary
mice. Proc Natl Acad Sci USA 96:1342713431. auditory cortex. Nat Rev Neurosci 5:279290.
 Synaptic Plasticity in the Hippocampus 473

Weiss JH, Koh JY, Christine CW, Choi DW (1989) Zinc and LTP. Willshaw D, Dayan P (1990) Optimal plasticity from matrix memo-
Nature 338:212. ries: what goes up must come down. Neural Comput 2: 8593.
Weisskopf MG, Nicoll RA (1995) Presynaptic changes during mossy Wilson RI, Nicoll RA (2001) Endogenous cannabinoids mediate ret-
ber LTP revealed by NMDA receptor-mediated synaptic rograde signalling at hippocampal synapses. Nature 410:
responses. Nature 376:256259. 588592.
Weisskopf MG, Zalutsky RA, Nicoll RA (1993) The opioid peptide Winder DG, Mansuy IM, Osman M, Moallem TM, Kandel ER (1998)
dynorphin mediates heterosynaptic depression of hippocampal Genetic and pharmacological evidence for a novel, intermediate
mossy ber synapses and modulates long-term potentiation. phase of long-term potentiation suppressed by calcineurin. Cell
Nature 362:423427. 92:2537.
Weisskopf MG, Castillo PE, Zalutsky RA, Nicoll RA (1994) Mediation Wisden W, Errington ML, Williams S, Dunnett SB, Waters C,
of hippocampal mossy ber long-term potentiation by cyclic Hitchcock D, Evan G, Bliss TV, Hunt SP (1990) Differential
AMP. Science 265:18781882. expression of immediate early genes in the hippocampus and
Weisz DJ, Clark GA, Thompson RF (1984) Increased responsivity of spinal cord. Neuron 4:603614.
dentate granule cells during nictitating membrane response Wolfer DP, Stagljar-Bozicevic M, Errington ML, Lipp HP (1998)
conditioning in rabbit. Behav Brain Res 12:145154. Spatial memory and learning in transgenic mice: fact or artifact?
West AE, Griffith EC, Greenberg ME (2002) Regulation of trans- News Physiol Sci 13:118123.
cription factors by neuronal activity. Nat Rev Neurosci 3: Woo NH, Nguyen PV (2003) Protein synthesis is required for synap-
921931. tic immunity to depotentiation. J Neurosci 23:11251132.
Whitlock, J.R., Heynen, A.J., Shuler MG, Bear MF (2006) Learning Woo NH, Teng HK, Siao CJ, Chiaruttini C, Pang PT, Milner TA,
induces LTP in the hippocampus. Science 313:10931097. Hempstead BL, Lu B (2005) Activation of p75NTR by proBDNF
Wibrand K, Messaoudi E, Havik B, Steenslid V, Lovlie R, Steen VM, facilitates hippocampal long-term depression. Nat Neurosci
Bramham CR (2006) Identication of genes co-upregulated 8:10691077.
with Arc during BDNF-induced long-term potentiation in adult Wood MA, Kaplan MP, Park A, Blanchard EJ, Oliveira AM, Lombardi
rat dentate gyrus in vivo. Eur J Neurosci 23:15011511. TL, Abel T (2005) Transgenic mice expressing a truncated
Wigstrm H, Gustafsson B (1983) Facilitated induction of hip- form of CREB-binding protein (CBP) exhibit decits in hip-
pocampal long-lasting potentiation during blockade of inhibi- pocampal synaptic plasticity and memory storage. Learn Mem
tion. Nature 301:603604. 12:111119
Wigstrm H, Gustafsson B (1985) On long-lasting potentiation in Woodin MA, Ganguly K, Poo M-M (2003) Coincident pre- and post-
the hippocampusa proposed mechanism for its dependence synaptic activity modies GABAergic synapses by postsynaptic
on coincident presynaptic and postsynaptic activity. Acta Physiol changes in Cl
transporter. Neuron 39:807820.
Scand 123:519522. Woolley DE, Gould E, Frankfurt M, McEwen BS (1990) Naturally
Wigstrm H, Gustafsson B, Huang C-C, Abraham WC (1986) occurring uctuations in dendritic spine density on adult hip-
Hippocampal long-term potentiation is induced by pairing sin- pocampal pyramidal neurons. J Neurosci 10:40354039.
gle afferent volleys with intracellularly injected depolarizing Wu H, Nash JE, Zamorano P, Garner CC (2002) Interaction of SAP97
current pulses. Acta Physiol Scand 126:317319. with minus-end-directed actin motor myosin. VI. Implica-
Wikstrm MA, Matthews P, Roberts D, Collingridge GL, Bortolotto tions for AMPA receptor trafficking. J Biol Chem 277:
ZA (2003) Parallel kinase cascades are involved in the induction 3092830934.
of LTP at hippocampal CA1 synapses. Neuropharmacology Wu J, Anwyl R, Rowan MJ (1995a) Beta-amyloid-(1-40) increases
45:828836. long-term potentiation in rat hippocampus in vitro. Eur J
Wilkinson DG (2001) Multiple roles of Eph receptors and ephrins Pharmacol 284:R1R3.
inneural development. Nat Rev Neurosci 2:155164. Wu J, Anwyl R, Rowan MJ (1995b) Beta-amyloid selectively aug-
Williams CE, Guevremont D, Kennard JT, Mason-Parker SE, Tate WP, ments NMDA receptor-mediated synaptic transmission in rat
Abraham WC (2003) Long-term regulation of N-methyl-D- hippocampus. Neuroreport 6:24092413.
aspartate receptor subunits and associated synaptic proteins fol- Wu J, Rush A, Rowan MJ, Anwyl R (2001) NMDA receptor- and
lowing hippocampal synaptic plasticity. Neuroscience 118: metabotropic glutamate receptor-dependent synaptic plasticity
10031013. induced by high frequency stimulation in the rat dentate gyrus
Williams JH, Bliss TVP (1988) Induction but not maintenance of cal- in vitro. J Physiol 533:745755.
cium-induced long-term potentiation in dentate gyrus and area Wu L, Saggau P (1994) Presynaptic calcium is increased during nor-
CA1 of the hippocampal slice is blocked by nordihydrogua- mal synaptic transmission and paired-pulse facilitation, but not
iaretic acid. Neurosci Lett 88:8185. in long-term potentiation in area CA1 of hippocampus. J
Williams JH, Errington ML, Lynch MA, Bliss TVP (1989) Arachido- Neurosci 14:645654.
nic acid induces a long-term activity-dependent enhancement Wu J, Rowan MJ, Anwyl R (2006) Long-term potentiation is medi-
of synaptic transmission in the hippocampus. Nature 341: ated by multiple kinase cascades involving CaMKII or either
739742. PKA or p42/44 MAPK in the adult rat dentate gyrus in vitro. J
Williams S, Johnston D (1989) Long-term potentiation of hip- Neurophysiol 95:351927.
pocampal mossy ber synapses is blocked by postsynaptic injec- Wu Z-L, Thomas SA, Villacres EC, Xia Z, Simmons ML, Chavkin C,
tion of calcium chelators. Neuron 3:583588. Palmiter RD, Storm DR (1995) Altered behavior and long-term
Williams JH, Li YG, Nayak A, Errington ML, Murphy KP, Bliss TV potentiation in type I adenyl cyclase mutant mice. Proc Natl
(1993) The suppression of long-term potentiation in rat hip- Acad Sci USA 92:220224.
pocampus by inhibitors of nitric oxide synthase is temperature Wyllie DJ, Nicoll RA (1994) A role for protein kinases and phos-
and age dependent. Neuron 11:877884. phatases in the Ca(2)-induced enhancement of hippocampal
474 The Hippocampus Book

AMPA receptor-mediated synaptic responses. Neuron 13: Ying SW, Futter M, Rosenblum K, Webber MJ, Hunt SP, Bliss TVP,
635643. Bramham CR (2002) Brain-derived neurotrophic factor induces
Xia J, Zhang X, Staudinger J, Huganir RL (1999) Clustering of AMPA long-term potentiation in intact adult hippocampus: require-
receptors by the synaptic PDZ domain-containing protein ment for ERK activation coupled to CREB and upregulation of
PICK1. Neuron 22:179187 Arc synthesis. J Neurosci 22:15321540.
Xiang Z, Greenwood AC, Kairiss EW, Brown TW (1994) Quantal Yokoi M, Kobayashi K, Manabe T, Takahashi T, Sakaguchi I, Katsuura
mechanism of long-term potentiation in hippocampal mossy- G, Shigemoto R, Ohishi H, Nomora S, Nakamura K, Nakao K,
ber synapses. J Neurophysiol 71:25522556. Katsuki M, Nakanishi S (1996) Impairment of hippocam-
Xiao MY, Karpefors M, Gustafsson B, Wigstrm H (1995) On the pal mossy ber LTD in mice lacking mGluR2. Science 273:
linkage between AMPA and NMDA receptor-mediated EPSPs in 645647.
homosynaptic long-term depression in the hippocampal CA1 Yu H, Saura CA, Choi SY, Sun LD, Yang X, Handler M, Kawarabayashi
region of young rats. J Neurosci 15:44964506. T, Younkin L, Fedeles B, Wilson MA, Younkin S, Kandel ER,
Xiao MY, Wasling P, Hanse E, Gustafsson B (2004) Creation of Kirkwood A, Shen J (2001) APP processing and synaptic plas-
AMPA-silent synapses in the neonatal hippocampus. Nat ticity in presenilin-1 conditional knockout mice. Neuron
Neurosci 7:236243. 31:713726
Xie X, Smart TG (1994) Modulation of long-term potentiation in rat Yuan LL, Adams JP, Swank M, Sweatt JD, Johnston D (2002) Protein
hippocampal pyramidal neurons by zinc. Pugers Arch 427: kinase modulation of dendritic K channels in hippocampus
481486. involves a mitogen-activated protein kinase pathway. J Neurosci
Xiong H, Boyle J, Winkelbauer M, Gorantla S, Zheng J, Ghorpade A, 22:48604868.
Persidsky Y, Carlson KA, Gendelman HE (2003) Inhibition of Yuste R, Bonhoeffer T (2001) Morphological changes in dendritic
long-term potentiation by interleukin-8: implications for spines associated with long-term plasticity. Annu Rev Neurosci
human immunodeficiency virus-1-associated dementia. J 24:10711089.
Neurosci Res 71:600607. Yuste R, Denk W (1995). Dendritic spines as basic functional units of
Xu L, Anwyl R, Rowan MJ (1997) Behavioural stress facilitates the neuronal integration. Nature 375: 682684.
induction of long-term depression in the hippocampus. Nature Zakharenko SS, Patterson SL, Dragatsis I, Zeitlin SO, Siegelbaum SA,
387:497500. Kandel ER, Morozov A (2003) Presynaptic BDNF required for a
Xu L, Anwyl R, Rowan MJ (1998) Spatial exploration induces a per- presynaptic but not postsynaptic component of LTP at hip-
sistent reversal of long-term potentiation in rat hippocampus. pocampal CA1CA3 synapses. Neuron 39:975990.
Nature 394:891894. Zalutsky RA, Nicoll RA (1990) Comparison of two forms of long-
Yamagata M, Sanes JR, Weiner JA (2003) Synaptic adhesion mole- term potentiation in single hippocampal neurons. Science
cules. Curr Opin Cell Biol 15:621632. 248:16191624.
Yamamoto C, Sawada S, Kamiya H (1992) Enhancement of postsy- Zalutsky RA, Nicoll RA (1992) Mossy ber long-term potentiation
naptic responsiveness during long-term potentiation of mossy shows specicity but no apparent cooperativity. Neurosci Lett
fiber synapses in guinea-pig hippocampus. Neurosci Lett 138:193197.
138:111114. Zaman SH, Parent A, Laskey A, Lee MK, Borchelt DR, Sisodia SS,
Yamazaki M, Fukaya M, Abe M, Ikeno K, Kakizaki T, Watanabe M, Malinow R (2000) Enhanced synaptic potentiation in transge-
Sakimura K (2001) Differential palmitoylation of two mouse nic mice expressing presenilin 1 familial Alzheimers disease
glutamate receptor interacting protein 1 forms with different N- mutation is normalized with a benzodiazepine. Neurobiol Dis
terminal sequences. Neurosci Lett 304:8184. 7:5463.
Yang CH, Huang CC, Hsu KS (2004) Behavioral stress modies hip- Zamanillo D, Sprengel R, Hvalby O, Jensen V, Burnashev N, Rozov
pocampal synaptic plasticity through corticosterone-induced A, Kaiser KM, Koster HJ, Borchardt T, Worley P, Lubke J,
sustained extracellular signal-regulated/mitogen-activated Frotscher M, Kelly PH, Sommer B, Andersen P, Seeburg PH,
proein kinase activation. J Neurosci 24:1102911034. Sakmann B (1999) Importance of AMPA receptors for hip-
Yang CH, Huang CC, Hsu KS (2005) Behavioral stress enhances hip- pocampal synaptic plasticity but not for spatial learning. Science
pocampal CA1 long-term depression through the blockade of 284:18051811.
the glutamate uptake. J Neurosci 25:42884293. Zeng H, Chattarji S, Barbarosle M, Rondi-Reig L, Philpot BD,
Yang SN, Tang YG, Zucker RS (1999) Selective induction of LTP and Miyakawa T, Bear MF, Tonegawa S (2001) Forebrain-specic
LTD by postsynaptic [Ca2]i elevation. J Neurophysiol 81: calcineurin knockout selectively impairs bidirectional synaptic
781787. plasticity and working/episodic-like memory. Cell 107:617629.
Yang XD, Connor J, Faber D (1994) Weak excitation and simultane- Zhang W, Linden DJ (2003) The other side of the engram: experi-
ous inhibition induce long-term depression in hippocampal ence-driven changes in neuronal intrinsic excitability. Nat Rev
CA1 neurons. J Neurophysiol 71:15861590. Neurosci 4:885900.
Yasuda H, Barth AL, Stellwagen D, Malenka RC (2003) A develop- Zhu JJ, Qin Y, Zhao M, Van Aelst L, Malinow R (2002a) Ras and Rap
mental switch in the signaling cascades for LTP induction. Nat control AMPA receptor trafficking during synaptic plasticity.
Neurosci 6:1516. Cell 110:443455.
Ye B, Liao D, Zhang X, Zhang P, Dong H, Huganir RL (2000) GRASP- Zhuo M, Zhang W, Son H, Mansuy I, Sobel RA, Seidman J, Kandel
1: a neuronal RasGEF associated with the AMPA receptor/GRIP ER (1999) A selective role of calcineurin Aa in synaptic depo-
complex. Neuron 26:603617. tentiation in hippocampus. Proc Natl Acad Sci USA 96:
Yeckel MF, Kapur A, Johnston D (1999) Multiple forms of LTP in hip- 46504655.
pocampal CA3 neurons use a common postsynaptic mecha- Zucker RS, Regehr WG (2002) Short-term synaptic plasticity. Annu
nism. Nat Neurosci 2:625633. Rev Physiol 64:355405.
11 John OKeefe

Hippocampal Neurophysiology
in the Behaving Animal

One powerful approach to the study of the functions of a

11.1 Overview brain region is to correlate the electrical activity of its neurons
with some aspect of observed behavior or inferred cognition.
This chapter addresses a central question in the study of the This approach reveals what information is available to that
hippocampus: What information is represented in hippocam- part of the brain and when it is available. Furthermore, it is
pal electrical activity? When asking this question, we are sometimes possible to record both the inputs and the outputs
immediately confronted with one of the central problems in of an area, and under these conditions it may be possible to
neuroscience, the nature of how information is coded in the compute what transfer function is being performed.
nervous system. Historically, there have been two answers to Several patterns of electrical activity have been recorded
this question: On the one hand is the view that the nervous from the hippocampus, and they have been correlated with
system is composed of a large set of individual computing ele- behavioral or psychological states. During the mid-1960s,
ments, the neurons, and that these neurons interact with each Vanderwolf placed a relatively large electrode into the hip-
other by passing discrete bundles of information along their pocampus of the freely moving rat and recorded the EEG
axons. An alternative view is that the system is organized on activity during a wide range of behaviors (Vanderwolf, 1969).
more holistic principles, with large numbers of cells acting in He identied three distinct states: the rhythmical theta state,
concert perhaps reected in synchronous neuronal ring or in the large irregular amplitude activity (LIA) state, and the
rhythmical electroencephalographic (EEG) activity. small irregular amplitude activity (SIA) state. Theta, in turn,
The general position taken in this chapter is that the hip- could be classied into two subtypes on the basis of behav-
pocampus uses both strategies. There is ample evidence that ioral correlate and pharmacological sensitivity. In this chapter,
individual hippocampal pyramidal cells code for specic loca- these two types are called a-theta (for arousal/attention theta),
tions in an environment by a marked increase in ring rate; which is sensitive to anticholinergic drugs such as atropine,
but, equally, that part of the code for location involves the tim- and t-theta (for translation movement theta), which may be
ing of cell ring relative to a clock wave, represented by the serotoninergic and glutamatergic.
EEG theta wave and associated interneurons, and that groups The behavioral/psychological correlates of the two types of
of pyramidal cells can act cooperatively in an ensemble fash- theta activity have been characterized best in rats: In this ani-
ion. On this view, each pyramidal cell acts as an oscillator, and mal, the atropine-resistant component or t-theta occurs dur-
a sophisticated set of mechanisms exists for producing these ing a class of movements that may be loosely characterized as
oscillations. Each cell has the biophysical machinery to oscil- those that normally change the spatial relation between the
late in isolation from other cells or external inputs. These indi- animals head and the environment. The correlates of the
vidual oscillators are stabilized and synchronized by a set of atropine-sensitive theta, or a-theta, are less well dened and
inhibitory feedback circuits primarily involving inhibitory are best summed up as psychological states such as arousal,
interneurons. Neurons in the septum and brain stem provide attention, or intention to move. The behavioral correlates of
the neuromodulatory inputs necessary for hippocampal LIA are those that do not change the animals location in the
pyramidal cells to enter the oscillatory state and may also pro- environment: quiet sitting, eating, drinking, and grooming in
vide a driving signal that sets the overall frequency of the the absence of postural shifts. The SIA state occurs during
oscillations. behavioral transitions, often when the animal awakens from

475
476 The Hippocampus Book

slow-wave or rapid-eye-movement (REM) sleep or when it in a rhythmical bursting pattern during the EEG theta state.
abruptly stops running. It has not received much attention, Unlike the theta cells, however, the frequency of bursts is
and its behavioral correlates and physiological function are slightly higher than the gross EEG theta, causing each succes-
less well understood. sive burst to precess to earlier phases of the theta cycle as the
The rhythmical theta state reects the synchronous mem- animal moves through the place eld. This temporal code
brane oscillations of large numbers of pyramidal cells in the works together with the overall rate code to identify the ani-
CA1 eld and dentate gyrus that are locked into synchrony by mals location. In addition, variations in ring rate can signal
the inhibitory interneuronal network. One of its functions is aspects of behavior that occur in the place eld or the presence
to provide a clock signal against which the action potentials of (or absence) of objects encountered there. The same cells
the individual pyramidal cells can be timed. Another function often re in different environments, but the preferred loca-
might be to set up the optimal circumstances for the induc- tions are unrelated if the environments are sufficiently dis-
tion of long-term potentiation (LTP) (see Chapter 10). The similar to each other. One notable feature of these place cells
nonrhythmical LIA state has a more random, broader spec- is that in unconstrained open eldsenvironments in which
trum especially in the lower-frequency range and may repre- the animal is free to move in all directionsthe cells re in the
sent an inactive, relaxed state of the same network when it is place eld irrespective of the direction in which the animal is
not being driven. Alternatively, it may be an active state in its facing. In environments that constrain the animals behavior,
own right in which memories previously encoded in the hip- the cells become directionally sensitive and may be said to rep-
pocampus are strengthened and/or transferred to other resent the successive locations along a path. In addition to the
regions of the brain. LIA is characterized by sharp waves of animals location, some pyramidal cells signal the presence or
about 100 ms duration that occur randomly, with an aver- absence of particular objects within the place eld or the per-
age interval of 1 second. Associated with them are higher- formance of particular behaviors there.
frequency ripple oscillations of 100 to 200 Hz, which also In addition to representing locations and features of the
may reect the operation of the inhibitory network. In addi- environment, the place cells have been shown to learn about
tion to oscillations in the theta and ripple bands, oscillations new environments or changes in a familiar environment. For
at intermediate frequencies (beta: 1230 Hz; gamma: 30100 example, place cells initially treat similar environments as
Hz) have been recorded in association with various aspects of identical but can learn to differentiate between them with
olfactory behavior. repeated exposure.
Placing a microelectrode rather than a gross electrode into Place cells are typically recorded from the hippocampus
the CA1 hippocampal cell layer reveals a much richer and sur- proper but have also been recorded from other parts of the
prising set of behavioral correlates correlations between phys- hippocampal formation, namely the subiculum, presubicu-
iology and behavior. On the basis of the EEG recordings one lum, parasubiculum, and entorhinal cortex as well as the hip-
might expect to see cell-ring patterns that are correlated with pocampus. The properties of cells in these various regions
the animals movements, and indeed this is observed. vary, and it is still not clear how this diverse population of
However, it is not the whole story. Ranck (1973) placed micro- place cells is organized into a functional network or which
electrodes into the pyramidal cell layers of CA3 and CA1 and functions are performed by each region.
found two classes of cellular response. One type, which he Two other major classes of spatial cell have been found in
called the theta cell, red at frequencies ranging from 10 Hz the hippocampal formation: the head direction (HD) cell and
(when the animal sat quietly or engaged in other LIA behav- the grid cell. The HD cell is sensitive to the orientation of the
iors) to as high as 100 Hz (as it ran around the environment rats head with respect to the environmental frame, irrespective
or engaged in other theta behaviors). Furthermore, they of the animals location in that environment. These cells have
burst at the same frequency and showed consistent correla- been found in several regions, most notably the anterior thal-
tions with different phases of the various EEG waves. Here, at amus and dorsal presubiculum. Different cells have different
the level of the single cell, was one clear correlate of the EEG preferred directional orientations. The animals orientation is
patterns, banishing forever any lingering skepticism about given partly by environmental cues and partly by interoceptive
their functional signicance. The second class of cell, the com- cues derived from vestibular and/or proprioceptive inputs. In
plex spike cell, had a much lower baseline ring rate, and addition to directly controlling behavior based on environ-
many were effectively silent for long periods of time. Their mental directions, these cells may provide directional informa-
dening characteristic is the occasional short burst of action tion to the place cells. The third major class of spatial cell,
potentials with successively decreasing amplitudes. Their the grid cell, provides a metric for marking off distances in the
major behavioral correlate was rst identied by OKeefe and environment. These cells have been found in layers 2/3 of the
Dostrovsky (1971) as the animals location. They reported medial entorhinal cortex, which sends a major projection to
that these place cells were typically silent as the rat moved the hippocampus proper, and in the lower layers, which receive
around the environment until it entered a small patch of the inputs from the hippocampus. Each of these cells lays a grid-
environment when the cell began to re (the place eld). It is like pattern of ring on top of every environment the animal
now recognized that, within the place eld, these cells also re encounters. The orientation and spacing of the grid varies in a
 Hippocampal Neurophysiology in the Behaving Animal 477

systematic fashion from cell to cell and appears to depend on pocampal function, although none tells the whole story on its
information generated by the animals self motion. own. Theta and LIA have been studied most and so receive the
Place cells have also been described in primates including most attention here.
humans, as have a variety of other spatial cells: HD cells and
spatial view cells, which respond when the animal looks at a 11.2.1 Hippocampal EEG Can Be Classied
particular location. into Four Types of Rhythmical and Two
Nonspatial behavioral correlates of hippocampal complex Types of Nonrhythmical Activity
spike cells have also been reported in rodents and primates.
Simple sensory stimuli, such as tones or somatosensory stim- In the hippocampus of the freely moving rat, six prominent
uli, appear to be relatively ineffective in the untrained animal, EEG patterns have been identied: four rhythmical and two
although there are several reports that pyramidal cells respond nonrhythmical. Rhythmical patterns (and their frequency
to these stimuli following classic conditioning or discrimina- ranges) include theta (612 Hz), beta (1230 Hz), gamma
tion learning tasks. Furthermore, increased ring in complex- (30100 Hz), and ripple (100200 Hz) waves. Nonrhythmical
spike cells has been reported to correlate with different aspects patterns include LIA and SIA. Figure 111 shows examples of
of behavior in approach tasks, such as whether the animal is each. Some patterns can co-occur (e.g., LIA with ripples, theta
approaching an area containing cues to be recognized or dis- with gamma), whereas others appear to be mutually exclusive
criminated or containing a reward. Some authors have argued (e.g., theta, LIA, and SIA). The latter three waveforms appear
that these ndings support the idea that the hippocampus is to correspond to mutually exclusive states of hippocampal
involved in many types of relational processing in addition to functioning.
those in the spatial domain. As we shall see, it is sometimes Theta consists of rhythmical, often sinusoidal oscillations
difficult to decide whether the ring of a hippocampal cell in that vary in frequency from 6 to 12 Hz in the rat but can be as
a particular task is spatial or nonspatial. Whether these non- low as 4 Hz in the rabbit and cat (Fig. 111A, traces 14). The
spatial correlates can eventually be explained within a spatial frequency power spectrum is narrow (Fig. 111B, walking),
framework or alternatively, signal the need for an extension of with a sharp peak around 7 to 10 Hz and often associated with
the functions attributed to the hippocampus into nonspatial higher harmonics (second peak around 16 Hz in Fig. 111B,
domains is discussed. walking). The LIA pattern looks much more random and is
often characterized by sharp waves that resemble the spike and
wave of epileptiform tissue (see Fig. 111A, traces 6, 7). The
LIA power spectrum is atter, with fewer peaks than are seen
11.2 Hippocampal Electroencephalogram in theta and more power in the lower (15 Hz) frequencies
Can Be Classied into Distinct Patterns, (Figs. 111B, still and 112G). High-frequency 200-Hz ripples
with Each Providing Information About occur on the LIA sharp waves (Fig. 111C). SIA is a low-
an Aspect of Hippocampal Function amplitude pattern that contains a broad spectrum of high fre-
quencies and occurs only occasionally (Figs. 111A, 7, pencil
If an electrode is placed in the hippocampus and electrical tap and 112C). Beta waves occupy the frequency range 12 to
activity in the frequency range 1 to 200 Hz is recorded as the 20 Hz (Fig. 111D) and gamma waves the frequency range 20
animal goes about its daily business, distinct patterns of elec- to 100 Hz (Fig. 111E). Either can occur alone or in combina-
trical activity are seen that vary as a function of state of alert- tion with theta, LIA, or SIA.
ness, sensory stimulation, behavior, and anatomical location.
These patterns of electrical activity are known collectively as 11.2.2 Each EEG Pattern Has
the electroencephalogram, or EEG. The EEG reects the Distinct Behavioral Correlates
activity of large numbers of neurons and probably includes
contributions from action potentials in disparate cell types, The simplest behavioral correlates of the various EEG patterns
excitatory and inhibitory synaptic potentials, and dendritic occur in the rat, and here we follow Vanderwolf s general
and glial slow potentials. As such, it can provide a measure of description (Whishaw and Vanderwolf, 1973; Vanderwolf,
information about the overall function of a brain region but 2001). It should be noted, however, that theta, in particular,
not at the same level of precision as the activity of single units. has slightly different behavioral correlates in different species.
It is probably most useful for signaling when large numbers of In the awake rat, theta occurs primarily during movements
neurons are acting together synchronously. Because this is an that can loosely be described as translationalthose that
important mode of operation of cortical areas such as the hip- change the location of the animals head with respect to the
pocampus, it follows that different EEG patterns can serve as environment: walking, running, swimming, jumping,
a bridge between behavior on the one hand and single or mul- exploratory head movements, struggling (see examples in
tiple unit activity on the other. As we shall see in this chapter, Fig. 111A, 14). Theta also occurs during REM sleep and
there are several types of hippocampal EEG pattern, and each occasionally during immobile attention or arousal. In the rab-
type provides information about a different aspect of hip- bit and guinea pig, this immobile attention-related theta
478 The Hippocampus Book

A B
4.5

1
movement movement
Log
Walk
2 Power

22 inch jump
Still
3
1.5
placed in box 11 inch jump 0 50 100
Frequency (c/s)
4
picked up placed in water swim climb out

5
sitting still teeth chatter head turn sitting still C
1 Hz to 10 kHz
1
6
sleep 500 Hz to 10 kHz
1s 500 mV
2
7
sleep pencil tap

100 to 400 Hz
3
4 1 to 50 Hz

100 ms

D E
Mucosa P
OB

1 1
H 5.0 s
Movement
2 2
D
3 L Dentate

3
4 10-50 Hz
5 M R Dentate
S
6 4

s
30-80 Hz
5

Figure 111. Hippocampal electroencephalography (EEG) patterns. broad-band (trace 1) and ltered (24) recordings. Ripples (3) and
A. Theta during rapid-eye-movement (REM) sleep (trace 1), jump- sharp waves (4) are accompanied by bursts of action potentials (2)
ing (2, 3), and swimming (4). Large irregular amplitude activity in hippocampal interneurons (small spikes) and principal cells
(LIA) during quiet sitting (5) and slow-wave sleep (6, 7). Note the (large spikes). (Source: Buzsaki et al., 1992.) D. Beta activity in the
large amplitude sharp waves especially prominent during slow-wave olfactory bulb (OB) and dentate gyrus (D) during sniffing toluene
sleep. Small irregular amplitude activity (SIA) during brief arousal (thickening of s in trace 6). Note the absence of beta in the hip-
from slow-wave sleep after a pencil tap (7). (Source: Whishaw and pocampus (2, H) and the absence of gross movement (5, M). E.
Vanderwolf, 1973.) B. Frequency power spectrum during walking Gamma activity (3, 4) in the dentate during sniffing (s). Trace
and standing still. Note the peaks around 8 and 16 Hz and the gen- 5 shows the increase in breathing recorded during sniffing.
erally higher amplitude in the beta and gamma bands during walk- (Source: Vanderwalf, 2001.)
ing. (Source: Leung, 1992.) C. LIA ripples and sharp waves in
 Hippocampal Neurophysiology in the Behaving Animal 479

A Movement D Carbachol + Movement 8.6

30

20

10

0
0 5 10 15 20 25 30

B Immobile E Carbachol + Immobile 6.5

30

20

10

0
0 5 10 15 20 25 30

C F
6.6
30
Procaine Atropine + Movement
30

20

10

0
0 5 10 15 20 25 30

G Atropine + Immobile 30

er (db)
(db
20

power
10
10

0
0 5 10 15 20 25 30

frequency (hz)

Figure 112. EEG patterns and frequency power spectra during in D. F. t-Theta during movement is not blocked by an intraseptal
movement and immobility in the rat. A. Theta during spontaneous injection of the anticholinergic atropine. G. LIA during immobility
movement. B. LIA during quiet immobility. C. All theta is abolished following intraseptal atropine. All EEG traces are 3 s long except C,
by a procaine injection into the medial septum. D.Theta during which is 2 seconds. Arrows in D, E, and F indicate peak theta fre-
movement following intraseptal injection of the cholinomimetic quency in hertz. (Source: After Lawson and Bland, 1993, with per-
carbachol. E. a-Theta during immobility following intraseptal injec- mission.)
tion of the cholinomimetic carbachol has a lower frequency than

occurs more often. In the cat, it is related to active exploratory

eye movements. 11.3 Hippocampal Theta Activity
LIA occurs during behaviors that do not change the ani-
mals location, such as sitting quietly, eating, drinking, groom- 11.3.1 Hippocampal Theta Activity:
ing, and during slow wave sleep (SWS) (Fig. 111A, 57). SIA Historical Overview
is seen infrequently: in the awake rat when ongoing move-
ment is abruptly halted after a long train of theta, during sud- Theta activity and its behavioral correlates have been studied
den, motionless transitions from rest/sleep to alertness (Fig. extensively ever since its discovery in the rabbit hippocampus
111A, 7, pencil tap), and during periods of SWS after REM by Jung and Kornmuller in 1938 (Jung and Kornmuller,
episodes. Much less is known about the full range of behav- 1938). The initial skepticism about the possibility that such
ioral correlates of beta and gamma, but they both have been regular large amplitude waves could be generated in the brain
shown to be associated with different types of olfactory input. was overcome when Green and Arduini (1954) repeated the
Gamma occurs when the animal sniffs at a wide range of observations during the 1950s and showed that the theta
odors (Fig. 111E), whereas beta waves have been observed pattern correlated inversely with neocortical desynchroniza-
only during the sniffing of odors associated with predators tion, suggesting that it represented the hippocampal arousal
(Fig. 111D). pattern. Later, attempts were made to correlate the theta rhy-
480 The Hippocampus Book

thm with aspects of learning: Adey et al. (1960) in the United Theta is not uniformly distributed in the hippocampal for-
States reported frequency shifts during simple discrimination mation but varies in both phase and amplitude in different
learning, and Grastyan et al. (1959) in Hungary found high- parts. Whereas synchronous, large-amplitude theta waves are
frequency theta during orienting behavior early in auditory- prominent in the dentate gyrus and CA1 areas of the hip-
reward conditioning but a desynchronized SIA pattern later, pocampus, most mapping studies report the absence of theta
after the animal had learned to approach the reward. Studies in area CA3. Despite this, CA3 cells can show good phase cor-
by Gray (1971) in Britain suggested that particular frequencies relates to the EEG recorded elsewhere (e.g., in CA1 or the den-
of hippocampal theta activity (ca. 7.7 Hz) in the rat may also tate gyrus). The implication is that the presence or absence of
occur in association with reactions to nonreward. theta waves in the EEG is not a simple function of the under-
These differences of opinion from different laboratories lying single-cell activity. The presence of theta must depend, at
led to speculation that the behavioral correlate of theta varied least in part, on other factors, such as the anatomical arrange-
from species to species and perhaps from task to task. Some ment of the cells and the phase relations of their activity (see
measure of order was brought to the eld by Vanderwolf s Buzsaki, 2002 for a recent review of the cellular basis of theta).
careful observations of the strictly behavioral correlates of
hippocampal EEG patterns (Vanderwolf, 1969). He made the 11.3.3 Both Types of Theta Activity Are
important point that changes of the kind seen by Adey and Dependent on the Medial Septal/DBB but Only
Grastyan that appeared to be correlated with different phases t-Theta Is Dependent on the Entorhinal Cortex
of learning were also correlated with specic changes in
behavior at the same time. The changes in theta frequency As discussed in Chapter 3, the major cholinergic input to the
seen by Adey, for example, tended to be correlated with hippocampal formation arises from the medial septal nucleus
changes in motor behavior during the course of learning. and the nucleus of the diagonal band of Broca. The relative
Vanderwolf argued that it was the strictly behavioral correlate contributions of the medial septal nucleus and entorhinal cor-
of these EEG patterns that was primary. In various rodent tex to the overall hippocampal theta pattern have been dis-
species, he and his colleagues observed that theta activity was sected using a combination of lesions and pharmacological
associated with what he termed voluntary movements, manipulations (Bland and Oddie, 2001; Buzsaki, 2002).
whereas LIA occurred during stereotyped xed-action pat- Lesions of the medial septal nucleus and associated diagonal
terns. His subsequent observation that there were two types band eliminate both types of theta activity. Injection of vari-
of theta, each with different behavioral correlates, claried ous drugs into the septum reveals differences in the pharma-
matters further (Kramis et al., 1975). One type is assoc- cological basis of a- and t-theta (Lawson and Bland, 1993).
iated with arousal or attention, whether in association with Inactivation by intraseptal injections of the local anesthetic
movement or immobility, and the other with a class of move- procaine or the -aminobutyric acid (GABA)ergic drug mus-
ment alone. These two types are expressed to different degrees cimol eliminates all theta from the hippocampus (Fig. 112
in different species and depend on different neuromodula- C). Intraseptal injection of cholinergic antagonists (such as
tory inputs. The changes in hippocampal EEG during classic atropine) also block both thetas if the animal sits quietly (Fig.
conditioning in the rabbit are examined in more detail in the 112G) but leave movement-related t-theta intact (Fig.
section on nonspatial learning and memory (see Section 112F). Conversely, injections of cholinergic agonists such as
11.11). carbachol produce a state of continuous theta in the hip-
pocampus regardless of whether the animal is moving (Fig.
11.3.2 Hippocampal Theta Activity Is 112D) or sitting quietly (Fig. 112E). The frequency of the a-
Comprised of Two Components, a-Theta theta that occurs during immobility is about 2 Hz lower than
and t-Theta, Which Can Be Distinguished when the animal is moving (compare power spectra in Fig.
on the Basis of Behavioral Correlates 112, D and E).
and Pharmacology Lesions of the entorhinal cortex, in contrast, eliminate t-
theta but leave a-theta intact (Kramis et al., 1975). Following
In addition to differences in their behavioral correlates, the such lesions, injections of atropine eliminate the remaining
two components of theta recorded in freely moving animals theta. Chronic injections of para-chlorophenylalanine (PCPA)
also differ in their pharmacology. One is affected by drugs that or reserpine, both of which reduce the levels of serotonin in
act at cholinergic synapses, such as the antagonists atropine the brain, appear to eliminate t-theta. Because the N-methyl-
and scopolamine, the agonist carbachol, and the anticholin- D-aspartate (NMDA) receptor blocker ketamine also elimi-
esterase, eserine. Vanderwolf and colleagues (Kramis et al., nates t-theta and results in a depth prole similar to urethane,
1975) called this component atropine-sensitive theta, but we it is likely that the glutamatergic afferents from the entorhinal
refer to it here by the more psychological name, arousal- or cortex to the distal dendrites of CA1 and CA3 are responsible
attentional-theta (a-theta for short). The second component for this subcomponent of theta (Buzsaki, 2002).
of theta is correlated with translational movements and is The simplest explanation of these ndings is that t-theta
unaffected by cholinergic drugs. There is some evidence that depends on bers that pass through, or synapse in, the medial
it is dependent on the transmitters serotonin and glutamate. It septum on their way to the entorhinal cortex with onward
is called translation-movement theta (t-theta for short). connection to the hippocampus. On the other hand, a-theta
 Hippocampal Neurophysiology in the Behaving Animal 481

requires the integrity of the direct cholinergic projection from ing a noxious stimulus, in response to an aversive conditioned
the medial septal-diagonal band of Broca to the hippocampus stimulus, or when the animal is immobile but preparing
perhaps through the activation of networks of inhibitory to move (such as just before a jump). It also co-occurs with
interneurons (see Chapter 8). t-theta during movement. In rabbits and guinea pigs, it is
readily produced in immobile animals by innocuous visual,
11.3.4 t-Theta Occurs During auditory, or tactile stimuli. Repeated stimulation, however,
Movement Through Space leads to habituation of this response. Sainsbury and colleagues
(1987a,b) have shown that the ability of a stimulus to elicit
Following Vanderwolf s observational studies, the tight cou- hippocampal a-theta is dependent on the preexisting level of
pling between t-theta and certain types of movement was arousal. A relatively neutral stimulus such as a tone, which
strengthened by the results of theta-conditioning experi- ordinarily has little effect, readily elicits a-theta if the animal
ments by Black (1975). The idea was to reward animals for has previously been sensitized with an arousing stimulus
producing short trains of theta activity under two training such as the sound of an owl or the sight of a predator. One
conditions: In one they were allowed to move, but in the other possibility is that a-theta represents a subthreshold activation
they were forced to hold still. Training to produce theta with of the motor system. Experiments in support of this idea have
frequencies greater than 7 Hz (t- theta) was successful only been carried out by Sinnamon and his colleagues (Sinnamon,
when movement was allowed. Thus, movement is necessary 2000; Sinnamon et al., 2000). They recorded theta activity in
for t-theta, but what aspect of such movement is being sig- urethane-anesthetized rats before, during, and after hind-limb
naled by theta activity? stepping movements elicited by electrical stimulation of the
Vanderwolf s original suggestion that theta was correlated hypothalamus or pharmacological block of the midbrain
with some aspect of voluntary behavior is not operational raphe nucleus. Both manipulations elicited low-frequency
enough because it is not clear whether, in the case of animals, premovement (presumptive a- type) theta activity as well as
a movement can ever be said to be truly volitional. the higher-frequency (presumptive t- type) theta, which
Considering the movements during which theta occurs most accompanied the hind-limb stepping movements. It might be
readilywalking, running, swimming, jumpingthe likely that a- theta reects the activation of movement programs in
common factor is translation through space. The fact that the absence of the movement itself. Alternatively or in addi-
small head movements during exploratory sniffing are also tion, it might reect the organization of sensory inputs as
accompanied by theta suggests that the crucial factor is the reected in the correlation of single-unit activity in different
translation of the head through space or, more specically, sensory nuclei with hippocampal theta (see Section 11.3.7).
the generation of motor outputs that would normally result in
translation of the head through space. Changes in the fre- 11.3.6 Theta and Sleep
quency of theta activity are correlated with either the animals
speed of movement through the environment (Rivas et al., Theta occurs during the REM phase of sleep; LIA and SIA
1996; Slawinska and Kasicki, 1998) or the rapidity with which occur during slow-wave sleep; and SIA bursts frequently come
a movement is initiated. The latter has been shown in experi- at the termination of an REM episode. Pharmacological stud-
ments in which rats have been trained to jump up onto a ledge ies have shown that the theta recorded during the actual eye
or to initiate running in a runway (Whishaw and Vanderwolf, movements of the REM phase of sleep is unaffected by cholin-
1973). ergic drugs and therefore resembles t-theta, and that outside
The behavioral correlate of theta amplitude is less clear. of these episodes is a-theta.
Although Vanderwolf and colleagues failed to nd a positive
correlation between amplitude and speed, Rivas et al. (1996) 11.3.7 Theta Activity in Nonhippocampal Areas
reported a positive correlation between both the frequency
and the amplitude of theta and the speed of movement in the In the rat, the theta system is centred on the hippocampal for-
guinea pig when movements are initiated in simple runways. mation, taken here to include the septum, subicular area, and
On the other hand, there is no correlation in jumping experi- entorhinal cortex (Alonso and Garcia-Austt, 1987a,b;
ments between amplitude and speed of movement. It seems Brankack et al., 1993). However there are also reports of EEG
that there are many factors contributing to the amplitude of and cellular activity phase-locked to theta in the cingulate cor-
theta, such that a consistent correlate of amplitude is not tex (Leung and Borst, 1987), prefrontal cortex (Hyman et al.,
always seen. 2005; Jones and Wilson, 2005; Siapas et al., 2005), perirhinal
cortex (Muir and Bilkey, 1998), posterior hypothalamus
11.3.5 a-Theta Occurs During Arousal including the mammillary bodies (Kirk and McNaughton,
and/or Attention as well as Movement 1991; Bland et al., 1995; Slawinska and Kasicki, 1995; Kocsis
and Vertes, 1997), brain stem reticular formation (Nunez et
a-Theta may correlate with psychological states such as al., 1991), amygdala (Par and Gaudreau, 1996; Seidenbecher
arousal or attention. It rarely occurs in isolation in the rat but et al., 2003), and superior (Natsume et al., 1999) and inferior
is much more common in rabbits, guinea pigs, and cats. In (Pedemonte et al., 1996) colliculi. Some of these areas, such
rats, it occurs naturally only when the animal freezes follow- as the posterior hypothalamus and the brain stem reticular
482 The Hippocampus Book

formation, are involved in the circuitry that generates the that theta oscillations in humans are found during virtual
hippocampal theta rhythm and might be expected to show movement, exploratory search, and goal-seeking. One differ-
activity synchronized with theta, whereas others are part of ence from the rat is that human theta bursts tend to be of
the sensory systems (the colliculi) or the limbic system (amyg- shorter duration. These kinds of studies conrm the suspicion
dala and cingulate cortex). These latter correlations must rep- that the difficulty recording primate theta has more to do with
resent a fairly widespread function for theta, such as the inappropriate behavioral paradigms and recording techniques
synchronization or binding together of neurons in many sen- than with the absence of a theta system per se.
sory, motor, and emotional/motivational centers as well as There has also been recent interest in the related eld of
those involved in spatial perception and memory. frequency analysis of EEGs recorded from scalp electrodes.
For example, increases in the power present in the EEG at
11.3.8 Does the Hippocampal EEG in Monkeys theta frequencies have been shown to be related to the suc-
and Humans Have a Theta Mode? cessful encoding of new information (for a review see
Klimesch, 1999). However, these studies typically do not
Although theta patterns are readily observed in cats, dogs, and demonstrate the peak in the power spectrum at theta fre-
rodents, it has been difficult to establish whether a clear rhyth- quencies or the long continuous records of trains of theta
mic theta pattern can been recorded from either monkey or activity shown by Kahana and colleagues in their virtual real-
human hippocampus. There have been hints over the years ity study. Localizing the source of theta in scalp-recorded
that it might exist. For example, Stewart and Fox (1991) EEGs is even more problematical than with subdural elec-
reported a theta-like pattern (79 Hz) in the hippocampal trodes; one experiment where it was attempted implicated the
EEG of urethane-anesthetized monkeys. In an earlier study, anterior cingulate rather than the hippocampus (in a task
Watanabe and Niki (1985) reported the existence of the showing increased theta with increased working memory
rhythmical theta-like ring patterns in monkey hippocampal load) (Gevins et al., 1997). Spectral peaks at theta frequencies
cells. Similarly, rare experiments using depth electrodes in the have been found in experiments using magnetoencephalogra-
human hippocampus have observed activity at theta frequen- phy and have been interpreted as consistent with a generator
cies, although their behavioral correlates were not clear near the hippocampus (Tesche and Karhu, 2000), although
(Halgren et al., 1978; Arnolds et al., 1980). This paucity of data this technique also suffers from problems of accurate source
has led some authors to suggest that theta may not exist in localization. Nevertheless, these ndings, and particularly the
monkeys and primates or may not have the same behavioral subdural recordings of Kahana et al., clearly show that theta
correlates as in other mammals. Some reasons for the failure activity exists in the human brain and tempt one to speculate
to record prominent theta patterns in the primate EEG are that it might be related to the hippocampal system. There is a
examined in the next section. more extensive discussion of recording of the EEG and evoked
First, the existence of an oscillatory theta pattern in the potentials from the human brain in Chapter 12.
EEG depends not only on the rhythmical ring of cells but
also on the correct cytoarchitectonic orientation of pyramidal 11.3.9 Functions of Theta
cells to create the appropriate electrical dipole. As was seen in
the section on theta activity in eld CA3, cells can burst with Work on the rat hippocampus suggests three possible func-
a theta pattern in the absence of pronounced theta waves in tions for theta. First, it acts as a global synchronizing mecha-
the EEG. Different neuroarchitecture could account for a nism, essentially locking the entire hippocampal formation
theta system in primates in the absence of rodent-like theta into one global processing mode and organizing the activity in
patterns in the gross EEG. each hippocampal region with respect to the others.
A second problem is that most monkey and human record- Simultaneous recordings of the EEG in different hippocampal
ing is done while the subject is immobile, a condition that is locations have shown that theta activity at comparable loca-
not conducive to recording t-theta. Thirdly, most recordings tions (e.g., in the CA1 pyramidal layer) is in synchrony and
in humans have been from the scalp, and it is possible that the coherent across large areas of the hippocampal formation
skull and scalp are acting as lters, effectively screening out the (Mitchell and Ranck, 1980; Fox et al., 1986; Bullock et al.,
theta patterns. Some evidence to support the last two possibil- 1990). This means that if two cells have ring patterns that are
ities comes from recent ndings (Kahana et al., 1999) that systematically related to the local theta cycle, they have sys-
theta patterns could be recorded from electrodes placed on tematic temporal relations to each other, even if they are
the surface of the human neocortex while subjects navigated located far apart in the hippocampus. Although the theta
through a virtual reality maze. Theta activity was recorded rhythm is centered on the hippocampal formation, sensory
from several cortical regions, but temporal lobe theta showed and motivational areas are also brought under its sway. We
the best correlation with maze difficulty. More theta activity begin to see here evidence for a widespread system of oscilla-
was seen during traverses through more difficult 12-choice tions that organizes the activity of many disparate brain areas.
mazes than through simpler 6-choice mazes. The same group A second function of the theta oscillations is to provide a
has used depth as well as subdural electrodes in humans per- periodic clocking system for the timing of hippocampal
forming a virtual taxi driver task (Caplan et al., 2003) to show spikes. As set out in greater detail in Section 11.7.9, the phase
 Hippocampal Neurophysiology in the Behaving Animal 483

relation of each pyramidal cell measured against the concur- of the hippocampus as a whole or the absence of some aspect
rent theta activity is not constant but can vary from one cycle of hippocampal function such as the theta state; functionally,
to the next (OKeefe and Recce, 1993; Skaggs et al., 1996). As a it has been suggested that they represent a neural correlate of
rat runs through the ring eld of a spatially coded pyramidal memory consolidation. Each sharp wave lasts 50 to 100 ms
cell (the place eld), the cell res bursts of spikes at an inter- and has a maximum amplitude in the stratum radiatum that
burst frequency slightly higher than that of the concomitant can be as large as 1 mV or more. Their resemblance to the
EEG theta. This leads to a precession of the phase of ring to interictal spike and wave complex of epileptogenic cortex
earlier points on each successive cycle. Over the course of the may give some clues to the peculiar susceptibility of the hip-
ve to seven theta cycles that comprise the typical place eld, pocampus to seizure activity (Bragin, 1999; Draguhn et al.,
the phase of the EEG at which the cell res may precess 2000; Buzaki and Draguhn, 2004). They occur more or less
through a full 360, although smaller amounts of precession synchronously over large areas of the CA1 eld of the dorsal
are also seen. Furthermore, the phase of ring is highly corre- hippocampus: Recordings at different points along the septo-
lated with the animals location within the place eld (more so temporal axis of the hippocampus have shown that they are
than with the duration spent in the eld). Thus, temporal in phase over the entire extent (Buzsaki et al., 1992; Chrobak
variation in spike ring conveys information about the ani- and Buzsaki, 1996). They reverse polarity in the pyramidal
mals spatial location. An analysis of this phenomenon (Jensen cell layer and reach their maximum amplitude several hun-
and Lisman, 2000) shows that the temporal information pro- dred microns into the stratum radiatum (OKeefe and Nadel,
vided by the phase precession can improve localization of the 1978, pp. 150153). Buzsaki and colleagues have suggested
animals position by more than 40% compared to that that sharp waves originate in the CA3 eld and that the
obtained by the use of ring rates alone. sharp waves recorded in CA1 are the summated extracellular
A third function for theta is to provide temporal control excitatory postsynaptic potentials (EPSPs) of the Schaffer col-
over long-term potentiation (LTP) induction and, by infer- laterals that result from synchronous ring of the CA3 pyram-
ence, the storage and retrieval of information from the hip- idal cells (Csicsvari et al., 2000). Direct stimulation of these
pocampus. As noted in Chapter 10, theta-burst electrical bres results in evoked potentials with similar shape and
stimulation of hippocampal afferents is an effective way to depth proles.
induce LTP. Furthermore, there is some evidence that volleys Around the time of the negative peak of the sharp wave,
arriving at different phases of the ongoing theta are differen- there is a high-frequency oscillation of between 120 and 200
tially effective (Pavlides et al., 1988; Huerta and Lisman, 1995; Hz whose peak amplitude occurs in the CA1 pyramidal cell
Holscher et al., 1997; Hyman et al., 2003). Inputs arriving at layer (OKeefe and Nadel, 1978, pp. 150153; Buzsaki et al.,
the positive phase of CA1 theta result in synaptic potentiation, 1992) (Fig. 111C). During these ripples there are synchro-
whereas those arriving at the negative phase yield depotentia- nous bursts in almost all theta interneurons and in about 1 in
tion or depression. On the basis of this and other evidence, 10 of the complex-spike pyramidal cells. Intracellular record-
Hasselmo (2005) proposed that the various phases of theta ings from the soma of pyramidal cells during ripples reveals
oscillation represent different modes of operation. Speci- intracellular oscillations that mirror the extracellular pattern.
cally, the peak of the CA1 theta is the period during which Hyperpolarization results in a reduction of ripple amplitude
encoding of new information entering the hippocampus from at
70 mV, with a reversal at more negative potentials. This
the entorhinal cortex takes place, and the trough is the period sequence of events suggests that the sharp wave itself reects
during which retrieval of information from the hippocampus the synchronous barrage of afferent activity from CA3 cells
to the entorhinal cortex occurs. onto the apical dendrites of CA1 cells. The rapid activation of
inhibitory interneurons causes synchronized hyperpolarizing
inhibitory postsynaptic potentials (IPSPs) in the somata of
pyramidal cells.
11.4 Non-theta EEG Patterns in the Are sharp waves and ripples conned to the hippocampus
Hippocampal EEG: LIA, SIA, proper, or do they spread to other structures? Recordings in
Ripples, Beta, and Gamma the subiculum and the deep layers of the entorhinal cortex
show that sharp waves also occur in these structures at about
11.4.1 Sharp Waves, Ripples, and the same time as those in hippocampus. The sharp wave in lay-
Single Units During Large Irregular Activity ers V and VI of the entorhinal cortex, which receive inputs
from the hippocampus, followed those in the hippocampus by
During LIA, large sharp waves occur in the hippocampal EEG 5 to 30 ms (Chrobak and Buzsaki, 1996). In contrast, the pres-
(Fig. 111A, traces 57 and Fig.111C, trace 4). In CA1, the ence of hippocampal sharp waves is not reected in the activ-
sharp waves occur most frequently during slow wave sleep and ity of the upper layers (II and III) of the entorhinal cortex,
quiet sitting, less frequently during eating and drinking, and which project to the dentate gyrus, hippocampus, and subicu-
least frequently during grooming. They appear to occur dur- lum. Recordings from the dentate gyrus show that the granule
ing periods of low arousal and are often, but not always, inhib- cells are also inuenced by the sharp-wave activity of CA3. The
ited by arousing stimuli. They may represent a resting state anatomical basis for this retrograde effect is not entirely clear.
484 The Hippocampus Book

There are projections from CA3 into the polymorph layer, and sizable percentage of the pyramidal cells, when compared with
these bers may be ending on mossy cells, which in turn proj- their highly differentiated patterns of activity during the theta
ect to the dentate granule cells. In the ventral tip of the hip- state as the animal moved around an environment, would be
pocampus, there appear to be some CA3 pyramidal cells that unlikely to convey much information. Perhaps most tellingly,
project directly into the molecular layer of the dentate gyrus. it was found that lesions of the fornix decreased neither the
frequency nor the amplitude of ripples and sharp waves; if
11.4.2 Dentate EEG Spikes During LIA anything, they increased themagain suggesting release of
the hippocampus from an activated theta state. In this view,
Sharp-wave activity during LIA also occurs in the granule cell the LIA state is the passive activity of a system with extensive
layers of the dentate gyrus. It is also associated with a high- positive and negative feedback loops and other oscillatory
frequency ripple, although the frequency is not as high as that mechanisms.
of the CA1 ripple. Most granule cells do not re during den- An alternative hypothesis is that LIA is an active, rather
tate sharp waves, but intracellular recordings have shown that than an idling, state of the hippocampus whose function is to
they are nevertheless depolarized at this time. Surprisingly, strengthen synaptic modications that have occurred during
the overall effect of dentate spikes on CA3 cells is inhibitory. the immediately prior periods (Buzsaki, 1989). Pointing to the
This presumably reects the heavy innervation of interneu- similarity between the synchronous volleys of afferent activity
rons by the mossy bers (Acsady et al., 1998) (see Chapter 5, impinging on the dendrites of CA1 dendrites during LIA and
Section 5.7). the type of tetanic stimulation known to cause LTP, Buzsaki
suggested that synaptic potentiation occurs naturally during
11.4.3 Pharmacology of LIA LIA. He went on to propose that this potentiation acts as a
boost to synapses that had been only weakly modied during
Little is known about the pharmacology of LIA, the sharp the previous theta behaviors and perhaps plays a role in a
waves, or the ripples. Ripples are eliminated under halothane memory consolidation process. Evidence in support of this
anesthesia, and their frequency is reduced under urethane and idea comes from experiments by Skaggs and McNaughton
ketamine to around 100 Hz (Ylinen et al., 1995). In the mouse, (1996). They looked at pairs of CA1 pyramidal cells with over-
sharp waves and ripples can be elicited by application of KCl lapping place elds in freely moving rats that were running
to the dendrites of pyramidal cells following pharmacological around a small triangular runway. When identied cells were
block of GABAA receptors, making it unlikely that they are recorded before and after the animal had run around for
due to activity in networks of the inhibitory interneurons extended periods, they observed during LIA an increase in the
(Nimmrich et al., 2005). Synchronization between ripples cross-correlation between cells that had been simultaneously
appears to depend on gap junctions. Consistent with the idea active during the previous period. That is, cells with ring
that the LIA state is a passive absence of theta, procaine or elds that were close together in the environment were more
muscimol suppression of the medial septum, which inhibits likely to re close together in time during an ensuing period
theta activity, has only a small effect on the power of hip- of slow-wave sleep than cells with more distant elds. The
pocampal LIA (Bland et al., 1996). Recordings of medial sep- small number of collateral bers between CA1 pyramidal cells
tal and diagonal band cells during LIA show that the ring suggests that this effect might be due to an increase in the effi-
rates of most of these cells are reduced relative to theta, cacy of the common input from CA3 or the entorhinal cortex
strengthening this suggestion. On the other hand, cholinergic onto these cells, rather than to the direct connections between
activation of the septum completely suppresses hippocampal them. In the consolidation view of LIA, the uncoupling of the
LIA, replacing it with low-frequency theta even in the immo- hippocampus from the septum during LIA is merely a neces-
bile animal (Fig. 112E). Theta appears to be the active state sary condition for hippocampal consolidation to occur.
driven from the medial septum, and LIA is the passive state A related idea is that LIA is a period involving transfer of
that occurs in its absence. information from hippocampus to neocortex. Support for
this view comes from experiments (Siapas and Wilson, 1998)
11.4.4 Behavioral Correlates that showed a correlation between hippocampal sharp waves
and Functions of LIA and neocortical spindle waves during slow-wave sleep. The
hypothesis that information might be transferred from the
In their discussion of hippocampal sharp waves and ripples hippocampus to the neocortex as a result of LIA-associated
during LIA, OKeefe and Nadel (1978, pp. 150153) suggested sharp waves is consonant with evidence from behavioral
that one way to think about this EEG state was as an absence experiments addressing the question of long-term storage of
of the theta stateit is the hippocampus in a non-theta idling memory traces in or outside of the hippocampus. Chapter
mode. This suggestion was based partly on the observation 12 presents evidence that lesions to the hippocampus in ani-
that there was always a period of at least a few seconds mals and humans can sometimes cause temporally graded
between the onset of an LIA-associated pattern of behavior amnesia. One interpretation of such a gradient is that mem-
(such as immobility) and the onset of the sharp wave and rip- ory traces are stored only in the hippocampus for a short
ples. They also argued that the nearly synchronous bursts in a period before being sent to the neocortex for permanent stor-
 Hippocampal Neurophysiology in the Behaving Animal 485

age (see Chapter 13, Section 13.3 for a more extensive discus- quency 10- to 100-Hz waves have also been described. This
sion of consolidation). After this period, damage to the hip- band is further divided into beta (1020 Hz) (Leung, 1992)
pocampus would no longer result in memory loss. Although (Fig. 111D) and gamma (20100 Hz) (Fig. 111E) activity.
the sharp waves may be part of such a mechanism, there is Gamma waves were rst described in the cat amygdala by
some evidence against this general idea. For example, Leonard Lesse in 1955 and have since been reported in widespread
and colleagues (1987) have shown that LTP cannot be induced brain regions of animals and humans (Singer and Gray, 1995).
in the hippocampus during slow-wave sleep. If this nding It has been suggested that gamma synchrony between various
can be generalized to other parts of the brain, it would greatly regions of the neocortex binds together the simple elements of
reduce the attractiveness of the hypothesis that sharp waves a complex representation. In the hippocampal formation, they
could serve as a basis for consolidation because no new LTP- occur most clearly in the entorhinal cortex and dentate but
based learning could take place. There is clearly much more to have also been reported in the CA elds (Csicsvari et al.,
understand about ripple/sharp-wave states. A possible role in 2003). They may be related to 40-Hz oscillations that have
memory formation may not be restricted within the connes been reported in the olfactory (Freeman, 1975) and visual
of the currently dominant theory of intrahippocampal con- (Singer and Gray, 1995) systems.
solidation or hippocampo-neocortical trace transfer during Gamma activity is slightly depressed in the rat by both sep-
sleep. For example, ripple/sharp-wave events may have a tal lesions and cholinergic antagonists (Leung, 1985). In rab-
purely intrahippocampal housekeeping function. To take just bits, drugs that stimulate a-theta during immobility (e.g.,
one possibility, synaptic renormalization or overall gain con- physostigmine) increase the amount of this intermediate fre-
trol processes might occur during LIA. quency. However, this increase does not seem to occur in
immobile rats. Seizures cause a dramatic increase in the
11.4.5 Small Irregular Activity amount of beta/gamma activity, an effect blocked by cholin-
ergic antagonists and unaffected by the animals ongoing
Small irregular activity (SIA) is characterized by low- behavior (Leung, 1992). The extent to which this beta/gamma
amplitude irregular activity in the hippocampus and desyn- activity is a reection of the underlying behavior of neural ele-
chronization in the neocortical EEG. In 1967, Pickenhain ments (principal cells or interneurons) is unclear, although it
and Klingberg reported low-amplitude irregular activity in may be a correlate of the activity of hilar theta cells.
the hippocampus of rats during transitions to alertness where
no orienting movements were made, such as when a click 11.4.7 Olfactory Stimulation Can Elicit
awakened them from sleep. Vanderwolf and Whishaw Hippocampal Gamma and Beta Waves
(Whishaw and Vanderwolf, 1971) noted a similar pattern dur-
ing transitions to alertness but added the observation that In the rat hippocampal formation, both beta and gamma
SIA, as they called it, occurred when rats abruptly halt volun- waves occur preferentially during olfaction. The dentate
tary movement. gamma waves appear to occur during the sniffing of odors in
Jarosiewicz and colleagues (2002) have extended the char- general but do not occur during odorless sniffing or other
acterization of SIA to include periods during sleep. Sleep SIA sensory stimulation (Vanderwolf, 2001) (Fig. 111E). The
bursts occur repeatedly during all periods of slow-wave sleep behavioral correlates of the CA1/CA3 gamma have not been
and after nearly every REM episode. Each burst typically lasts established. In general, gamma and theta occur independ-
a few seconds, with a range from 200 ms to many seconds. A ently, but under some (undefined) circumstances they
brief tone presented in sleep routinely elicited an increase in become synchronized, with the gamma occurring preferen-
electromyographic (EMG) and neocortical arousal accompa- tially at the positive peak of the dentate theta waves. The
nied by hippocampal SIA, suggesting that it is a state interme- dentate, but not the CA1, gamma is dependent on the per-
diary between LIA and theta (Jarosiewicz and Skaggs, 2004b). forant path input from the entorhinal cortex, as lesions of the
Sleep SIA is characterized by the cessation of ring in most entorhinal cortex abolish the dentate gyrus gamma but
pyramidal cells. The 3% to 5% that continue to re do so enhance the CA1/CA3 gamma (Bragin et al., 1995). Dentate
actively and repeatedly over successive bursts. These place cells gamma waves may be part of the mechanism for synchroniz-
are probably continuing to represent the location where the ing the olfactory inputs arriving via the entorhinal cortex
rat fell asleep because rotating the platform and the sleeping with the hippocampal theta. Beta waves have a more restricted
rat away from a given cells place eld location relative to the olfactory correlate than gamma waves. They occur in the den-
testing laboratory did not have an effect on SIA ring tate gyrus in response to olfactory inputs that signal, or mimic
(Jarosiewicz and Skaggs, 2004a). those that signal, the presence of predators (Fig. 111D)
(Vanderwolf, 2001): compounds found in the anal scent gland
11.4.6 Beta/Gamma Activity in the Hippocampus secretions of weasels and foxes, most organic solvents includ-
ing toluene and xylene, and phytochemicals derived from
In addition to the low-frequency EEG activity seen during plants such as eucalyptol and salicylaldehyde. In general, these
theta and LIA, and the higher frequency ripple activity char- odors also elicit a fear response and behavioral avoidance.
acteristic of sharp waves and dentate spikes, intermediate fre- Other strong smells, such as ammonia, cadaverine, or
486 The Hippocampus Book

putrescine, which are either approached or not avoided, do day tasks such as eating, drinking, sleeping, and searching for
not elicit dentate beta waves. Vanderwolf has argued that food and water. This emphasis on naturalistic behavioral cor-
these olfactory correlates of gamma and beta waves in the relates led to several important discoveries. The rst was that
hippocampus suggest a primary olfactory function for this they noticed that there were two major classes of cells that
structure, but it is more likely that, like the theta waves, could be distinguished by differences in their anatomical and
they represent one aspect of a more complex overall func- physiological properties (e.g., ring rates, action potential
tion, such as cognitive mapping or associative memory for- width, and relative locations in the hippocampus). Ranck
mation. termed these two classes complex-spike and theta cells, and
these terms are still in general usage. We describe their prop-
erties in a subsequent section. Perhaps more importantly, they
discovered that the ring patterns of the two cell types had
11.5 Single-cell Recording in the Hippocampal repeatable behavioral correlates. Ranck had trained his ani-
Formation Reveals Two Major Classes of mals to approach one location to obtain food and another to
Units: Principal Cells and Theta Cells get water, and emphasized the relation of the complex-spike
cell ring pattern to the behavioral approach to reward.
Although EEG recordings provide some information about OKeefe was more impressed by the spatial correlate and
the circumstances under which large numbers of neurons in a named the cells place cells. It is now widely accepted that the
brain region become synchronously active, it is generally location of the animal in a familiar environment is the major
accepted that a complete understanding of function can be determinant of when such cells re.
gained only by looking at the behavioral correlates of single The second class of neurons, the theta cells, has less specic
units. One reason for this is that although neighboring neu- behavioral correlates. As the name implies, their behavioral
rons often share common functional properties they may not correlates are closely related to those of the gross EEG waves
always respond to the same specic stimulus or behavior. As and in particular theta. They tend to change rate during hip-
we shall see, this is particularly true of the hippocampus, pocampal EEG theta, and many display strong phase locking
where neighboring pyramidal cells share the property of rep- to the individual theta waves.
resenting places in an environment. However, because differ- Extracellular recordings in the freely moving rat enable
ent cells become active in different parts of an environment, complex-spike and theta cells to be distinguished on the basis
this property is not revealed in the hippocampal EEG. of differences in their wave shapes, ring rates, and other
The recording of individual neuronal responses in the hip- properties (Fig. 113). Complex-spike place cells have a much
pocampus of the awake freely moving rat began during the broader action potential than theta cells (Fig. 113BD), often
early 1970s with the work of Ranck (1973) and OKeefe display a complex-spike burst pattern in which the later spikes
(OKeefe and Dostrovsky, 1971). They both encouraged their in a burst are smaller in amplitude and of longer duration
animals to move around the environment, engaging in every- than the rst (Fig. 113A), and have a lower spontaneous

Figure 113. Complex-spike cells and theta cells have different theta cells (int). E. C-S cells have a narrower range of interspike
physiological properties. A. Theta cells have a steady ring rate and intervals (ISI) F. Three-dimensional plot of ring rate against
constant size amplitude spikes, whereas C-S cells sometimes emit a mean ISI and spike asymmetry (a-b in D) separates the overall
complex-spike burst in which the later action potentials in the burst population into two clusters. (Source: A,B. After Christian and
are lower in amplitude and broader. BD. Action potentials of C-S Deadwyler, 1986; CF. After Csicsvari et al., 1998 with permission.)
cells (pyr) are broader and have a larger initial hump than those of
 Hippocampal Neurophysiology in the Behaving Animal 487

background ring rate (generally about 1 Hz) (Fig. 113F). In has been reported. Some theta cells tend to re a few millisec-
contrast, theta cells have a much higher ring rate (10100 onds after a complex-spike cell recorded on the same tetrode
Hz) (Fig. 113F) with all action potentials being of the same (see Box 111), and this is reected in a short latency cross-
amplitude (Fig. 113A) and of shorter duration (Fig. 113B- correlation between their spike trains (Csicsvari et al., 1998)
D). An important caveat is that in the rabbit this separation of (Fig. 114B-D and Box 112).
spikes into two nonoverlapping classes is less clear as there The same generalizations can be made for rabbits.
appears to be a large subclass of complex-spike cells with However, there appears to be an additional group of interme-
spontaneous ring rates as high as 6 Hz. diate cells that exhibit complex spikes and that can be
It is highly likely that in the rat the complex spike cells are antidromically activated by stimulation of projection path-
pyramidal cells and the theta cells are one or more types of ways; but they have a resting ring rate that is intermediate
interneuron. Intracellular staining of neurons that display between the theta and complex-spike cell ring rates shown in
complex spikes in brain slices reveals they have the morphol- rats (Berger et al., 1983).
ogy of pyramidal cells, whereas those without complex spikes
are interneurons. Sometimes it is possible to activate complex 11.5.1 Distinctive Spatial CellsComplex-spike
spike cells antidromically from electrodes placed in hip- Place Cells, Head-direction Cells, and Grid
pocampal outow pathways, but the theta cells can only be CellsAre Found in Various Regions of the
driven orthodromically (Berger and Thompson, 1978; Fox Hippocampal Formation
and Ranck, 1981; Christian and Deadwyler, 1986). Another
strong piece of evidence in support of the idea that complex- Discovery of the place cells led to the development of the cog-
spike cells are pyramidal cells and theta cells are interneurons nitive map theory of hippocampal function by OKeefe and
comes from the temporal relation between their ring pat- Nadel (1978), which has guided much of the subsequent
terns. Most pyramidal cells innervate neighboring interneu- research and theorizing on hippocampal function (see
rons via an axon collateral, which should give rise to an Chapter 13, Section 13.4). Equally important for our under-
increased probability of ring in the theta cell shortly after standing of the spatial functions of the hippocampal forma-
an action potential in the pyramidal cell. Just such a relation tion was the discovery of two other classes of spatial cell: In

Box 111
Microelectrode Recording Technique

Microelectrodes placed in the extracellular space in the vicinity of a neuron detect the current
ow associated with action potentials. Experience has shown that relatively large electrodes
with at tips are preferable for recording in chronic animals because they do not puncture the
cells and therefore do less damage during small movements of the electrode relative to the
brain. One problem with the use of single electrodes in a structure such as the hippocampus is
that it is difficult to isolate single units on the basis of action potential amplitude and shape.
This is because of the close packing of the identically sized and shaped cells. Action potentials
from all cells on a sphere with the electrode at the center are identical. There is a danger that
all such spikes are considered to have come from one neuron. Template-matching algorithms,
which are so useful when different cell types are in close proximity, offer no way out of this
ambiguity and also have difficulty coping with the fact that a single pyramidal cell sometimes
res simple spikes and sometimes complex spike bursts in which the later action potentials in
the burst have reduced amplitude and broader waveforms than the initial spike (Fig. 113A).
One solution, introduced by McNaughton et al. (1983b), is to use multiple electrodes whose
tips are close enough to sense the action potentials from a group of neurons but, being spaced a
short distance apart, cannot both be at the center of a notional sphere. Each electrode tip is a
slightly different distance from a given cell and, consequently, records its action potential with
a slightly different amplitude and shape. These sometimes subtle differences can be used to
distinguish a multiunit recording from several electrodes into spikes emanating from different
cells. The principle is analogous to a stereophonic recording in which two or more micro-
phones are used to capture the sound of an orchestra; hence, the earliest electrodes of this
type used two wires and were called stereotrodes. In general, n  1 electrodes are necessary
to identify uniquely the action potential from a neuron in n-space. In the hippocampus, trio-
trodes with three tips oriented in the plane of the quasi-two-dimensional pyramidal cell layer
would probably suffice; but in practice, tetrodes are commonly used to ensure adequate
isolation (OKeefe and Recce, 1993; Wilson and McNaughton, 1993). Figure 114A shows the
prole of action potentials of complex-spike and theta cells on the four electrodes of a tetrode.
488 The Hippocampus Book

A C
pyr int

4 -50 0 50
0.2ms

B D
pyr int

2
0.2ms
4
-50 0 ms 50
Figure 114. Hippocampal interneurons sometimes re a few mil- cells in B shows the interneuron res at a peak latency of 2 ms fol-
liseconds after neighboring pyramidal cells. A. Waveforms of a lowing the spike in the pyramidal cell. D. Controlling for the repeti-
pyramidal (pyr) and an interneuron (int) recorded on the same tive rhythmical pattern of ring within each of the spike trains does
tetrode (traces 14). B. Multiple sweeps triggered on the pyramidal not eliminate the short-term causal relation between the two.
cell show that the interneuron often red at a short but variable Horizontal line in D indicates a signicance level above which the
latency shortly after, suggesting a monosynaptic coupling between correlation is highly signicant. (Source: After Csicsvari et al., 1998,
the two. C. Cross-correlogram between the ring trains of the two with permission.)

Box 112
Auto-correlation and Cross-correlation Functions

Correlation techniques are used to investigate the temporal relations of spike occurrences to
themselves (auto-correlation) (Fig. 113 E), to the occurrence of spikes in other neurons
(cross-correlation) (Fig. 114C,D), and to other brain events (e.g., theta waves) (Fig. 115) or
to sensory and motor events in the environment (e.g., peristimulus event histogram) (see Fig.
1117, later). The auto-correlation function is useful for revealing repetitive or rhythmical pat-
terns of ring. A graph is constructed in which the x-axis represents time intervals before and
after each spike event, and the y-axis represents the probability that the cell will re during each
interval before or after that spike event. A period of 50 ms is a useful period of time for looking
at the complex spike properties of principal cells in which the cell res repetitively within 10
ms (Fig. 113A,E); 500 ms is a useful length of time for demonstrating the rhythmical pattern
of theta cell ring. The cross-correlation function reveals the tendency of two cells to re with a
particular temporal relation to each other. Here one cell is chosen as the target, and its spikes
x the zero point on the time axis. The probability of the other cell ring at times earlier and
later is calculated and displayed as a histogram. This type of analysis is useful for identifying
potential synaptic relations between cells or the existence of common inputs to the cells. The
cross-correlogram between a complex-spike cell and a theta cell is shown in Figure 114C,D.
The peak at 2 ms suggests that there is a short-latency excitatory synaptic connection from the
pyramidal cell to the theta cell. The probability associated with this peak gives some indication
of the strength of this connection. The correlation between hippocampal units and the phase of
the global EEG theta signal is useful for showing the temporal relations of different types of
unit to the EEG and to each other (Fig. 115). A nal use of correlation techniques is to look
for a relation of spike ring to a sensory event or motor action in the external world or to
another brain event. Illustrated in Figures 1117 to 1119 are peristimulus histograms of hip-
pocampal unit responses to auditory stimuli as a result of conditioning experiments and in
Figure 1120 (see later) the increase in activity during different aspects of the behavioral learn-
ing paradigm.
 Hippocampal Neurophysiology in the Behaving Animal 489

theta
ripple
0 360 720
0.10

firing probability

firing probability
0.04

0.05
Pyramidal
Cells
Cells
0.00 0.00

1.0
firing probability

firing probability
0.4
0.20
PV Basket
Cells
0.00 0.0

0.30
0.6
firing probability

firing probability
0.15
Axo-axonic
Cells
0.0 0.00

0.16
firing probability

firing probability 0.08

O-LM
Cells 0.04

0.00 0.00
0 360 720 -1 0 1
theta phase normalized time

Figure 115. Phase relations of pyramidal cells and various classes of interneurons for theta
(top left) and sharp-wave ripples (top right). Note that the various interneurons have phase rela-
tions different from those of theta activity. Note also that pyramidal cells and basket cells re
during the ripples, axo-axonic cells re slightly before, and both they and the O-LM cells are
silent during the ripples. (Source: After Klausberger et al., 2003.)

1984, Ranck described the head direction units, and in 2005 The grid cells are found in the medial entorhinal cortex.
Hafting and colleagues discovered the entorhinal grid cells. Their ring maps lay down a regular pattern of locations
The availability of directional and distance information to the across all environments the animal encounters. They are well
mapping system was a prediction of the theory, and the dis- suited to provide the self-motion or idiothetic information for
covery of these cell types in the hippocampal formation pro- the construction of place cells, as well as the distances and
vides strong support for it. directions between them. They are considered in greater detail
Head direction cells were rst discovered in the dorsal pre- in Section 11.7.7.
subiculum (or postsubiculum) (Ranck, 1984; Taube et al., Not everyone who has recorded from principal cells in the
1990a). The activity of these cells complements that of the hippocampus agrees with this spatially conditioned way of
place cells: They do not take into account the animals location looking at the data. For example, there have been reports of
but signal the direction in which it is pointing relative to the temporal correlates between pyramidal cell ring with eyelid
environmental frame. They are rarely found in the hippocam- responses during nictitating membrane conditioning and
pus proper but mainly in another part of the hippocampal for- sensory, motor, and task-related correlates during olfactory
mation, the dorsal presubiculum, which has strong anatomical learning and memory tasks. These reports have suggested
connections to the hippocampus via its projections to the a function for the hippocampus broader than spatial memory
entorhinal cortex, perhaps to the grid cells located there. The and navigation, perhaps as a storage device for nonspatial
best estimate is that about 25% of the cells in this area are sen- as well as spatial relations. Several authors, most notably
sitive to head direction, whereas other cells there have angular Eichenbaum and Cohen (Cohen and Eichenbaum, 1993;
head velocity, running speed, locational, and both directional Eichenbaum and Cohen, 2001), have returned to the origi-
and locational correlates (Sharp, 1996; Cacucci et al., 2004). nal Ranck description of these cells as having behavioral
They are discussed in Section 11.9. approach as well as spatial and relational correlates. The next
490 The Hippocampus Book

sections focus on the extensive literature on place cells, theta ple oscillation; oriens/lacunosum-moleculare cells, with their
cells, and on other spatially related cells in particular, head- somata in the stratum oriens and their axons targeted on the
direction cells, and grid cells. Work on nonspatial correlates is distal dendrites of the pyramidal cells, re on the negative
described in later sections (see Section 11.11). phase of the theta wave and are silent during sharp waves; axo-
axonic cells, which target the axon hillock of the pyramidal
cells, re on the positive theta wave and are also silent on the
sharp waves (Fig. 115).
11.6 Theta Cells In the rabbit, the same cells take part in both a-theta and t-
theta. Typically, the ring rate during a-theta is lower than
The second major class of cells in the rodent hippocampal for- during t-theta even for the same frequency of theta. Within
mation is the theta cell. These cells are distinguishable from the each type of theta, the ring rates of the theta cells vary as a
complex-spike cells by having a briefer action potential, a function of the frequency of theta. In Section 11.7.9, we dis-
higher ring rate, and a different anatomical location (Fig. cuss in greater detail the temporal relation between the com-
113). They also have a different relation to the hippocampal plex-spike and theta cells and the EEG theta.
EEG. Given the predominance of theta EEG activity during
certain behaviors, it is not surprising to nd that many cells in 11.6.2 Pharmacology of Theta Cells
the hippocampus have a rhythmical pattern of ring that is
related to theta. Theta cells were originally dened by their As we have seen in Chapters 3 and 8, interneurons in the hip-
strong, consistent phase correlation to the EEG theta pattern pocampus receive both GABAergic and cholinergic inputs
and by their increased ring rate during theta, irrespective of from the septal nuclei, GABAergic and endorphin peptidergic
the animals location. Subsequently, the class was broadened to inputs from other inhibitory interneurons, and glutamatergic
include cells that decrease their ring rates during theta. In inputs from the pyramidal cells. Finally, they receive cate-
contrast, the timing of complex-spike action potentials relative cholaminergic inputs from the brain stem and might be
to the theta waves is more complex and changes as the animal expected to respond to transmitters such as noradrenaline
moves through the place eld (see Section 11.7.9). In the rest (norepinephrine). Firing rates of theta cells recorded in the
of this section, the relation of theta cells to the EEG, their phar- urethane-anesthetized rat decreased by up to 75% following
macology, and their correlation with behavior are described. iontophoretic application of atropine or scopolamine (Stewart
et al., 1992). There was, however, no change in the phase rela-
11.6.1 Theta Cells Fire with a tion between the remaining spikes and the EEG theta. This is
Consistent Phase Relation to EEG Theta markedly different from the effect of the same drugs on com-
plex-spike cells, which did not change their rate of ring but
The dening feature of theta cells is their close relation to the altered their patterns of ring from the usual complex-spike
hippocampal EEG. Bland and his colleagues (Colom and burst pattern to more continuous ring. This suggests that the
Bland, 1987) have identied four classes of theta cell, theta-on cholinergic input to the hippocampus from the medial septal
and theta-off cells, each of which is subdivided into phasic nucleus makes direct contact with the theta cells and increases
and tonic subtypes. Theta-on cells increase their ring rates their ring rate but has no inuence on their theta burst mode
during theta activity, whereas theta-off cells decrease their of ring. It further suggests that the pattern of activity in com-
activity. Phasic cells have strong constant phase relations to plex spike cells, in contrast to the rate, has only a modest inu-
theta; tonic ones do not. The theta-off cells are found much ence on the pattern of ring of the theta cell interneurons.
less frequently than theta-on cells and are particularly rare in Opiates have a different effect: They directly inhibit the
the freely moving rat. interneurons and indirectly disinhibit the pyramidal cells,
In the urethane-anesthetized rat, both CA1 and dentate increasing their discharge rate ring (Pang and Rose, 1989).
theta cells re close to the negative peak of the dentate theta. Finally norepinephrine has an effect that is broadly opposite
In contrast, in the awake animal, maximal ring is found to that of the opiates, exciting interneurons and inhibiting the
much closer to the positive peak of the dentate theta (Fox et principal cells in both the CA1 eld (Pang and Rose, 1987)
al., 1986). In addition, there is a broad range of preferred and the dentate gyrus (Rose and Pang, 1989).
phases in the various cells pointing to a heterogeneous popu-
lation of theta cells. It is likely that more than one class of 11.6.3 Hippocampal Theta Cells Have
interneuron is involved, corresponding to the different classes Behavioral Correlates Similar to
of hippocampal interneuron identied in Chapter 8. Different Those of the Hippocampal EEG
types of interneuron have different phase correlates to theta
and the LIA sharp waves. For example, Klausberger (2003) In the rat, theta-on cells increase their ring rates during the
reported that the basket cells, which have their cell body in the EEG theta and begin to re in a bursting pattern in phase with
pyramidal cell layer and inhibit the perisomatic region of the the theta rhythm. This normally occurs during behaviors such
pyramidal cells, preferentially re on the positive/negative as walking and swimming, which change the animals location
part of the theta wave and on each wave of the sharp-wave rip- in an environment. During LIA behaviors (e.g., slow-wave
 Hippocampal Neurophysiology in the Behaving Animal 491

sleep, quiet sitting, eating, drinking, grooming), theta cell r- brain regions outside the hippocampal formation and in par-
ing is lower in frequency and more random in pattern. ticular by cells in the septal region and in the thalamus.
Interestingly, low-frequency theta mode ring occurs during
the postural shifts of grooming, such as changes from face- 11.7.1 Place Cells Signal the Animals
washing to ank-grooming. Location in an Environment
Theta-off cells have the opposite behavioral correlates,
increasing their ring rate during LIA and decreasing their Place cells were discovered by OKeefe and Dostrovsky in 1971
rates during theta. There are differences between the behav- (OKeefe and Dostrovsky, 1971). After recording from rat CA1
ioral correlates of theta in the rat and rabbit (see above), and complex-spike cells during various behavioral tasks and in
similar differences are found in the correlates of theta cells in response to various types of sensory stimulation, they noticed
these two animals. For example, a-theta in the rabbit occurs that the activity of some cells was more closely related to the
much more readily in response to arousing stimuli in the animals location than to any aspect of the task in which it was
absence of movement. It is no surprise, then, that the theta engaged. OKeefe (1976) conrmed and extended these obser-
cells also re during arousing stimuli as well as during move- vations by recording from rats as they ran between the arms of
ments such as walking and hopping. As we shall see in the sec- a three-arm maze to obtain different rewards. He christened
tion on conditioned responses in single hippocampal units these cells place cells and called the location where each cell
(11.11.5), this a-theta related ring may help explain the red its place eld. There were several aspects of the ring pat-
involvement of the hippocampus in the timing of such terns of these cells that suggested they were signaling the
responses to the (arousing) conditioned stimulus. abstract concept of place rather than acting as simple sensory
An important pointer to one function of the theta cells cells. First, it was not possible to identify any single sensory
comes from an interesting observation by Nitz and stimulus that reliably controlled the cells activity. Second, after
McNaughton (2004). They found that a large number of CA1 the rat had some experience with running on the maze in the
theta cells turn off during the rst exposure to a novel envi- dark as well as in the light, many of the complex-spike cells
ronment, whereas the dentate interneurons increased ring continued to re in the appropriate location with the lights
rates. Perhaps the CA1 interneurons are part of a mechanism out. Third, on the broad-armed maze used by OKeefe, many
for identifying familiar and novel environments and for cells red equally well as the animal faced in any direction.
controlling the processes of learning that occur in CA1 Finally, place eld ring did not seem to depend on the ani-
pyramidal cells when the animal is confronted with environ- mals motivation or incentive for visiting a location. For exam-
mental novelty. ple, interchanging the food and water rewards at the ends of
the different arms of the maze had no effect on place cell ring.
Thus, complex-spike cells did not appear to be tuned to spe-
cic sensory stimuli, they tolerated radical changes in lighting,
11.7 Complex-spike Cells and they were omnidirectional and uninuenced by reward.
and Spatial Processing The notion that something more abstract was being signaled
such as locationseemed an appropriate conclusion.
The primary correlate of hippocampal complex-spike cell r- Some skepticism greeted the rst qualitative reports of the
ing is the animals location in an environment. For this reason properties of these cells. However, the introduction of photo-
they have been called place cells (Fig. 1121A, see color insert). graphic methods and, later, the development of computa-
This section begins with a description of the properties of tional methods of obtaining objective data gradually
place cells, followed by the factors that control the location, convinced the most ardent skeptics (Box 113 and Fig. 116).
size, and shape of place elds. Both exteroceptive sensory cues Over the years, improvements in single-cell isolation (see Box
and internal proprioceptive/vestibular cues play a role in 112), coupled to ever more sophisticated behavioral proce-
determining place eld structure and location. The frame of dures and unusual bits of apparatus (e.g., morph boxes with
reference for the spatial coordinates of place elds can vary walls that can be recongured), have helped unravel many of
depending on the environment. It is known that the spatial the properties and determinants of place cells.
code is conveyed by the temporal ring pattern of the com- OKeefe (1976) noted that although many complex-spike
plex-spike cells as well as by their absolute rate of ring. The cells could be classied as simple place cells, others had more
animals location in an environment is not given by the activ- complex properties. For these cells, the ring rate was depend-
ity of a single place cell but by the pattern of ring across a ent on factors in addition to location. For example, some cells
large number of such cells. It follows, therefore, that it is increased their ring rates if the animal experienced an object
important to look at the network properties of place cells. in their place eld or engaged in a particular type of behavior
Place cells have been found in areas outside of the hippocam- there (e.g., running or sniffing). The clearest example of this
pus, in particular in the subiculum and entorhinal cortex; and type of cell was one that red maximally when the animal
the properties of these cells and how they differ from those in went to a specic location on the maze and either failed to
hippocampus are discussed. The nal section examines the nd something that had been there often before or found
data suggesting that place cell ring is controlled by activity in something new. These complex spatial cells were called mis-
492 The Hippocampus Book

Box 113
Quantitative Recording of Place Fields

An objective description of the ring eld of a place cell is

achieved by constructing a map of ring frequencies across
the surface of an environment (see Fig. 116). The location
of a small light xed to the animals head is monitored by an
overhead television camera, and the animals position is
recorded throughout the trial. The environment is broken up
into a set of squares, and two independent measures are
computed for each square: the amount of time the animal
spent there and the number of times the cell red while the
animal occupied that square. A ring rate per square can be
calculated by dividing the number of spikes in each square by
the amount of time the animal spent there. Spatial ring
rates are often portrayed as false color or grayscale maps in
which different colors or shades of gray represent different
A ring rates. In some laboratories, the data are spatially aver-
aged and smoothed before this step. The colors or grayscales
allocated to the different ring rates are usually adjusted rela-
tive to each cells peak ring rate (auto-scaling) to give com-
parable eld pictures from high- and low-ring cells. When
elds are created for the same cell across different conditions,
it is often useful to x all of the color map levels to the rst
one to facilitate comparisons. In some experiments, two dif-
ferently colored lights or infrared lights of different intensi-
ties are xed at different positions on the animals head so
the computer monitoring the animals movements can com-

7.2
pute the instantaneous heading direction as well as location.
The best estimates of spatial ring are achieved during tasks
in which the animal visits each section of the environment
for an equal amount of time.

B
questioned whether all of these cells are best described as hav-
Figure 116. Firing eld of a CA1 place cell. A. Raw data from a rat ing spatial elds. This work is described in more detail later
foraging for food in a square box for 10 minutes. Gray line traces (see Section 11.11).
the animals path through the environment; black boxes show ring
of place cell. B. Place ring eld as a grayscale map where darker
11.7.2 Basic Properties of Place Fields
colors represent higher ring rates. Inset number in white gives the
peak ring rate. (Source: After Wills et al., 2005, with permission.)
The size and shape of place elds vary with the shape and per-
haps also the size of the testing enclosure as well as with the
place cells. Whether the ring of these misplace cells was part of the hippocampus from which the recordings are taken.
due to the absence of the expected object or reward or to the In the dorsal hippocampus, elds on small, open platforms
myostatial (exploratory) sniffing elicited by the mismatch have a tendency to be located along the edges or just off them
was not clear. Ranck (1973) noted similar behavioral corre- (seen when the animal peers over the edge), whereas those in
lates of hippocampal complex-spike cells but did not identify closed cylinders or rectangular boxes tend to be equally dis-
their spatial properties. He labeled one class approach- tributed around the environment (Muller et al., 1987) (Fig.
consummate cells because they red as the animal ran toward 117). On mazes, elds may be located anywhere, with no ten-
the food or the water reward. Another class he called dency for them to cluster in particular arms or in particular
approach-consummate mismatch cells because they red parts of the maze such as at choice points (OKeefe, 1976;
maximally when the animal sniffed around the reward loca- Olton et al., 1978; McNaughton et al., 1983a; OKeefe and
tion when the reward was withheld. It now seems likely that Speakman, 1987). There are, however, occasional reports that
many and perhaps all of Rancks rst class of cells are synony- provide exceptions to this generalization: Hetherington and
mous with OKeefes category of place cells and his second Shapiro (1997) reported that elds tended to cluster closer to
class synonymous with the misplace cells. Eichenbaum and his walls than in the center of rectangular boxes, and Hollup et al.
colleagues have recorded cells during odor discrimination (2001) found a greater representation of elds in the region of
tasks with properties similar to those described by Ranck and the goal in an annular-shaped swimming pool.
 Hippocampal Neurophysiology in the Behaving Animal 493

Place fields of 32 simultaneously recorded complex-spike cells

Figure 117. Firing elds of 32 place cells simultaneously recorded the lower left, and so on. In reality there is no topographical relation
while a rat foraged for food in a 62-cm2 box. The place eld maps between the location of cells in the hippocampus and the location
are arranged topographically so elds in the northwest of the box of their elds in an environment. (Source: After Lever et al., 2002,
are located at the upper left, elds in the southwest are located at with permission.)
494 The Hippocampus Book

One distinctive feature of the representation of any partic- of the hippocampus were not extensively explored in these
ular environment by the hippocampal place cells is the studies, but a small number of cells recorded there (Maurer et
absence of topographical mapping of function onto anatomy, al., 2005) appear to have elds twice the size of those in the
such as is seen in the neocortex. That is, place cells located middle region. It seems reasonable to assume, therefore, that
next to each other in the hippocampus are no more likely to eld sizes increase as one goes from the dorsal to the ventral
have elds located next to each other than those far away hippocampus; and some elds in the most ventral hippocam-
(OKeefe et al., 1998). In contrast, there have been some pus might be very large, covering areas the size of the usual
reports that neighboring cells have elds closer than expected laboratory testing enclosures or even larger. This wide range
by chance (Eichenbaum et al., 1989) or that cells in the hip- of eld sizes has important implications for analysis of the
pocampus are functionally organized in bands separated by sensory correlates of hippocampal cells and for analysis of the
300 to 400 m along the long axis (Hampson et al., 1999). A differences in behavioral functions of the dorsal versus ventral
denitive resolution of the question seems to have been given hippocampus. Some authors have claimed that hippocampal
by Redish and colleagues (2001). Using tetrodes to ensure ade- complex-spike cells respond to specic sensory inputs (e.g., an
quate spike isolation, they collected a large number of record- odor) in the absence of a spatial correlate because the cells did
ings of pairs of complex-spike cells during tasks in which the not re at higher rates in one part of the testing apparatus
animals ran along circular or linear tracks. They compared the than in any other (e.g., Wood et al., 1999). Clearly, however, if
place elds of cells recorded from the same tetrode (see Box some spatial ring elds can be as large as or even larger than
112) and therefore anatomically neighboring each other, the size of the apparatus, this conclusion is not warranted. At
with those of cells from different tetrodes and therefore more the very least, it is necessary to test sensory responses in two
distant from each other. They found no tendency for the place enclosures located in different laboratories. Similarly, when
elds of neighboring cells to be closer to each other than trying to discern the differences in function between the dor-
would be expected for cells located at any two points in the sal and ventral hippocampi, it is important to take this aver-
hippocampus. They also recorded the activation of the imme- age eld size difference into account. Fields that cover an
diate early gene Arc, which identies active cells, as the animal entire testing enclosure might be described as spatial context
explored two different environments. In neither case did they neurons and in conditioning paradigms, for example, might
nd any evidence for clustering of cells with similar spatial or have the primary function of distinguishing between two test-
temporal correlates. One is tempted to speculate that there ing boxes rather than identifying different locations within the
must exist some mechanism for preventing neighboring cells same testing box.
with their overlapping inputs from adopting similar place How quickly are place elds formed, and how do they
elds in any environment. Perhaps one function of the develop? The answers to these questions can provide some
inhibitory interneuronal networks is to allow some pyramidal indication of whether some or all of the representation of an
cells to capture a territory in an environment and to exclude environment is pregured and what rules govern the plastic-
their neighbors from ring in that region. ity involved in the initial development of the spatial represen-
Place cells rarely have more than one eld in a single envi- tation. Several studies have provided reasonably good
ronment (typical testing environments are less than 1 m2). answers. On a small sample of 15 place cells, Hill (1978) found
The best current estimate is that cells with double elds com- that 11 had fully edged elds on the rst entry into the rele-
prise no more than about 5% to 10% of the total. The sizes of vant part of the environment. On a larger sample of cells and
place elds in the dorsal hippocampus vary considerably. To using tetrode recordings, Wilson and McNaughton (1993)
some extent, eld size is a function of the threshold used allowed their rats to explore one-half of an open eld box
for separating signal from noise and the location of the until it was familiar; they then removed a partition wall of the
recording site along the dorsal septo-temporal axis of the hip- box, giving access to the whole. They reported that all place
pocampus. Using a 1-Hz threshold as the minimum rate elds had stabilized within 10 to 15 minutes of entry into this
within a place eld, eld sizes in a cylinder of 76 cm radius new part of the enclosure. Frank et al. (2004) looked at the
ranged from a minimum of 4% of the surface area to a maxi- development of elds on a three-arm maze when one of the
mum of 62%, with a median size of 18% (Muller et al., 1987). usual arms was closed and a new one with a different angular
(These data were obtained with single-electrode recording orientation was opened. They found that about one-quarter
techniques; it is likely that the actual place eld sizes are even of the new elds on the novel arm developed rapidly within
smaller.) As can be seen from Figure 117, there appears to be the rst 2 minutes and that most elds had stabilized after 5 to
a slight tendency for the elds in the center of the environ- 6 minutes of experience with this arm. Interestingly, cells that
ment to be larger than those toward the periphery, but this has had less than 4 minutes of novel experience on the rst day
not been quantied. showed considerable plasticity on the second day, whereas
The average size of the place eld varies with location those with more than 4 minutes experience were much more
along the long axis of the hippocampus. As the recording elec- stable. The ring in some elds strengthened in the initial
trode is moved to more ventral portions of the hippocampus, location, that of others weakened, and some shifted to a new
the size of place elds expands. Fields in the middle region of location altogether. Still others were initially silent on the arm,
the hippocampus are almost twice as large as those in the sep- but after several minutes of exposure to the new arm turned
tal hippocampus (Jung et al., 1994). The most ventral regions on from this zero baseline and began to re. The authors
 Hippocampal Neurophysiology in the Behaving Animal 495

pointed out that the latter phenomenon is incompatible with reduction in ring rate during the rst entry into the novel
a role for postsynaptic spike timing-dependent plasticity in environment, although according to Frank et al. this lasted
the recorded cells as the mechanism underlying place eld only for the rst minute of exploration. In a subsequent sec-
development. This mechanism suggests that presynaptic tion, evidence is presented in support of OKeefes (OKeefe,
activity coincident with postsynaptic spiking strengthens the 1976) original idea that place cells could be activated in the
active synapses, and the absence of postsynaptic cell ring place eld by sensory cues experienced there or by path inte-
prior to the development of the full-blown place eld rules it gration signals that compute the distance and direction from
out. Of course, it could be that the changes are taking place in other locations. McNaughton et al. (1996) has speculated that
an earlier part of the circuit with postsynaptic cells that are when an animal rst enters a novel environment the ring of
initially active. Our own studies (Lever, Cacucci, Burton, place cells is driven by the path integration inputs and that
OKeefe, unpublished) have examined CA1 place eld dynam- only subsequently are the environmental stimuli experienced
ics during the rst exposure to an entirely novel environment: in each location associated with these place elds. A plausible
a 60-cm sided square box located in a curtained environment alternative is that the small number of seed cells are driven by
of which the animal had no prior experience. In agreement the appropriate combination of environmental cues from the
with Frank and colleagues, we found that place eld establish- rst exposure (perhaps including the animals own urine),
ment was rapid and could occur within 2 to 3 minutes. Most and the rest of the cells are subsequently co-opted by the path
of the CA1 cells that developed elds during the rst exposure integration system as the animal moves around the environ-
to an environment maintained these elds, but some stopped ment relative to these known locations. Further experimenta-
ring or changed eld location. This is reminiscent of the tion is needed to distinguish between these alternatives.
process of place eld development upon exposure to a new
part of a familiar environment, described above. Setting up a 11.7.3 Place Fields are Nondirectional in
representation of a new environment or part of an otherwise Unrestricted Open-eld Environments
familiar environment seems to involve a competitive compe- but Directional When Behavior Is
tition among a group of neighboring cells perhaps mediated Restricted to Routes
by the inhibitory interneuronal networks. Initially, a small
number of seed cells have place elds from the moment the Place eld ring sometimes depends on the direction in which
animal enters the environment, although the eld ring rate the animal faces and sometimes it does not. In open-eld
may increase or decrease with continued experience. These environments, the place elds are nondirectional (Muller et
cells prevent neighbors from ring in their territories through al., 1994) (Figs. 118A; see Fig. 1121A, color insert); on radial
activation of inhibitory interneurons. Other cells gradually ll arm mazes (McNaughton et al., 1983a; Muller et al., 1994;
in the unclaimed territory until the entire space is represented, Markus et al., 1995) and narrow linear tracks (OKeefe and
at which point no other cells are allowed to develop place Recce, 1993; Gothard et al., 1996a), they are highly directional
elds in that environment and the representation stabilizes. (Fig. 118B). Whereas in the rst situation each cell may be
Both Wilson and McNaughton (1993) and Frank and col- said to represent a location, in the second it might more prop-
leagues (2004) noted that some theta cells showed a marked erly be described as representing a serial position along a path.

Figure 118. Place elds are nondirectional in open-eld environ- Firing elds of three cells recorded simultaneously as the rat runs
ments but directional on linear tracks. A. Central panel shows the back and forth on a 1.5-m linear track for food reward at each end.
ring eld of a place cell in a 78 cm diameter cylinder regardless of Cells 1 and 2 red as the animal ran from left to right but not in the
the direction in which the rat is facing. The four outer panels show opposite direction; cell 3 red during right to left runs. (Peak rates
the eld when the animal is facing in the four cardinal directions: shown in black: 9 top, 10 middle, 7 bottom.) (Source: Courtesy of
northward at the top, eastward to the right, and so on. (Peak rates OKeefe.)
are shown in black: 12 center, 10 north, 12 east, 9 south, 9 west.) B.

cell 1
all
cell 2

cell 3

eastward runs westward runs

496 The Hippocampus Book

The difference between the two situations appears to be due, work of the stable room cues in each rats hippocampus, the
in part, to the constraints that the shape of the testing appara- more likely the animal was to be successful at the task. The
tus places on the animals behavior. It is not a matter of sam- relation of place elds to different environmental reference
pling from two cell populations because the same cells can frames is considered at greater length in a subsequent section.
have directional or nondirectional elds in different testing One major difference between directional and nondirec-
apparatuses or under different behavioral constraints (Muller tional eld ring is the inuence on the ring rate within the
et al., 1994; Markus et al., 1995). The most important variables eld of the overall trajectory of the path in the former but not,
appear to be whether the animal can turn around in any given so far as is known, in the latter. During behavioral tasks that
location and if it approaches the same location from many promote directional ring and require the rat to traverse the
directions. If it can, cells are relatively nondirectional. same part of the apparatus as part of two different paths, place
However, if the animal always runs through the same location cells re differently in the place eld on the two paths (Frank
along the same one or two paths, place cells tend to be direc- et al., 2000; Wood et al., 2000; Ferbinteanu and Shapiro, 2003).
tional. Note that one can induce narrow and stereotypical For example, Wood and colleagues (2000) trained animals on
pathtaking in open elds by reward shaping. Markus et al. a continuous alternation task in a T maze that had been tted
(1995) originally trained rats to perform random foraging in with return tracks from each goal to the start of the stem so
an open cylinder and then retrained them to take direct paths the animal could run in a continuous gure-of-eight path by
in the same cylinder among four reward locations in either turning left or right alternatively at the choice point of the T
clockwise or anticlockwise sequences. The proportion of place junction. Activity in two-thirds of the place elds in the stem
elds dened as directional doubled from about 20% in the of the T varied markedly depending on whether the animal
random foraging condition to 40% in the direct path condi- entered the stem following a return from the left-hand or
tion. An important study has reexamined directionality in right-hand goal. In the Wood et al. study, it was not possible
CA1 cells (Battaglia et al., 2004). These authors trained rats to to distinguish between ring owing to the previous part of the
run on narrow linear and circular tracks of the kind that path (retrospective coding) and the future part of the path
would normally induce a high proportion of unidirectio- (prospective coding). Two of the three studies (Frank et al.,
nal place elds. Their nding across experiments was that 2000; Ferbinteanu and Shapiro, 2003), however, did allow for
placing various proximal multimodal cues along the tracks this distinction and reported that, in addition to the retro-
signicantly increased the proportion of bidirectional place spective effect of the previous path on place cell ring, the
elds. This is puzzling because there are usually no distinctive subsequent path (i.e., the upcoming turn to be made at the T
local cues in open elds where the cells are typically nondirec- junction) also had an effect on ring in the stem (a prospec-
tional. The dynamics of (non)directionality need further tive effect), although the effect was seen in only a few hip-
study; there are no published reports examining directionality pocampal cells in the Frank et al. study and was much more
on rst exposures in linear tracks. Does omnidirectionality prominent in cells in the entorhinal cortex.
appear as soon as a place eld develops, or do the elds start These effects appear to be due to the animal repeatedly
out with a directional bias and only lose this directionality running in a continuous trajectory in the same direction
with further experience of the environment (Kali and Dayan, through the entire path and thus activating a xed sequence of
2000)? place cells. These prospective and retrospective effects have
The existence of both directional and nondirectional not been reported in nondirectional elds in open-eld envi-
modes of spatial representation raises the question of whether ronments, suggesting they have something to do with the
the hippocampus is providing fundamentally different types stereotyped ballistic nature of the paths taken on linear tracks.
of information in these two situations. Specically, does the In the one study in which the effect was not seen, the animals
spatial localization system, of which the place cells are a part, were trained to alternate on a Y maze with broad arms (Lenck-
know that the rat is in the same location when, say, traveling Santini et al., 2001a). They had to return to the start arm fol-
east as when traveling west? Perhaps it is only acting to its lowing each visit to a goal arm, resulting in a start arm left
fullest capacity when the cells are omnidirectional and is goal start arm right goal start arm, etc. alternation
showing limited functionality when they are unidirectional. pattern. This behavioral pattern would activate the same cells
Against this is the recent demonstration (Rosenzweig et al., in a different order on successive runs, and the wide arms
2003) of a good correlation between the degree to which place would allow the rat to take slightly differing paths on each
cells recorded on a linear track used a particular reference run. Interestingly, one observation in the Ferbineatu and
frame and the animals ability to localize an unmarked loca- Shapiro study suggests that the effect is not due solely to the
tion dened within that framework. Place cells were recording activation of the immediately prior place cell in the sequence
on a linear track where locations could be identied in one of or the execution of the immediately preceding turn. On error
two conicting frameworks: in the framework of (1) a goal trials in which the animal rst visited an incorrect arm before
box that changed location along the track from trial to trial or entering the correct goal arm for that trial, about one-half of
(2) stable room cues. The target location remained xed rela- the cells with place elds in the goal arm continued to re cor-
tive to the stable room cues but moved from trial to trial rela- rectly. This shows that small deviations from the standard
tive to the moving goal box. The authors showed that the path through the environment do not disrupt the effect. The
higher the percentage of place cells that related to the frame- authors suggested that it is the overall journey from one place
 Hippocampal Neurophysiology in the Behaving Animal 497

to another that is important, not the actual sequence of loca- are silent most of the time, making it difficult to estimate the
tions traversed. Buzsaki (2005) has speculated that the ring population number. Thompson and Best (1989) tried to get an
of any given place cell may be inuenced by place cells from accurate estimate of the population by recording during slow-
early stages in the path and not just by the immediately pre- wave sleep or while the rat was under light barbiturate anes-
ceding cell in the sequence. At this stage, it is not clear how thesia. Both procedures are known to enhance the
much these effects are due to the beginning and end points of spontaneous ring rates of complex-spike cells and may there-
the path and how much to particular motor behaviors, such as fore provide a better estimate of the total population under the
body turns made during the path. One possibility is that the recording electrode than can be obtained during the active,
ring of many cells on these tracks is driven by the path inte- awake state. They also recorded complex spikes in three envi-
gration system (see Section 11.7.7), and activation of the ronments: an elevated eight-arm radial maze, an enclosed
motor, proprioceptive, and vestibular systems at the turn into cylinder, and an enclosed rectangular environment. They con-
the stem of the T maze marks an important location on the cluded that slightly more than a third of the cells that could be
path and is carried over to the rest of the path. This would identied under barbiturate anesthesia had place elds in at
explain some of the retrospective effects but would not least one of the three environments. Almost all of the other
explain the prospective effects unless there is activation of the cells recorded had very low spontaneous rates in all environ-
turning machinery some period of time prior to reaching the ments (and were thus termed silent cells), with many ring
turn itself. no spikes for the entire period of testing. Only 14% of cells had
These studies have been interpreted by their authors as elds in two environments, and only 1% had elds in all three
pointing to a place cell basis for the episodic functions of the environments. That is, many cells with elds in one environ-
hippocampus (see Chapter 13, Section 13.4). OKeefe and ment acted like silent cells in the other environments. On the
Nadel (1978, 1979) originally suggested that the spatial func- basis of the physiological similarities between the place cells
tions of the rat hippocampus could be elaborated into a spa- and the silent cells, and the distribution of elds across the
tiotemporal episodic memory in humans by adding a three boxes, Thompson and Best suggested that if enough
temporal time-stamp signal, which would allow each set of environments were tested every complex-spike cell would have
events occurring in the same location to be distinguished a eld in at least one of them. In several studies, McNaughton
from each other on the basis of their time of occurrence. The and his colleagues (Wilson and McNaughton, 1993a; Gothard
present results, however, do not provide evidence for this et al., 1996a) have sampled large groups of complex-spike cells
unique time-stamp signal, which would enable different runs and nd that 30% to 70% have place elds in a given environ-
to be identied as independent episodes. Rather, it seems to be ment. Their group (Guzowski et al., 1999) has used the imme-
more appropriately interpreted as evidence of the organiza- diate early gene Arc to label all of the CA1 cells active in two
tion of place cells into integrated, ordered sequences, as might groups of animals, each exploring a different environment. In
be expected for the representation of paths. one group 44% of CA1 cells were active and in the other group
To conclude this section, it is clear that the place cells can 45%in broad agreement with the results from single-unit
be directional or nondirectional depending on the specic cir- recordings. In animals that were allowed to explore both boxes
cumstances in which they are recorded. It is important to and in which double staining was carried out, 22% of the cells
emphasize that the nondirectional mode of place cell activity were active in one environment, 23% in the other, and 16% in
rules out the possibility that they are simply responding to a both. These studies make it clear that a large proportion of
particular conguration of distal sensory cues. They cannot complex-spike cells have a place eld in some environments,
be responsive just to a particular view or scene in front of the consistent with the view that spatial representation is one of
animal (sometimes called the local view), as this would the primary functions of the hippocampus.
change radically as a function of the heading direction. Is there any relation between the size or shape of the place
Moreover, as these experiments routinely control for local elds that a cell displays in one environment and those
sensory inputs such as intramaze smells and textures, nondi- obtained in another? The answer depends on the similarity
rectionality is evidence in favor of the idea that these cells can between the environments and the amount of experience
signal something quite abstract. the animal has had with them. When the environments are
sufficiently different, such as the eight-arm radial maze with-
11.7.4 What Proportion of Complex-spike out walls and a small enclosed rectangular box used in the
Cells Are Place Cells? study of Thompson and Best (1989), and the animal has had
considerable experience in both, the elds are very different.
To estimate the proportion of complex-spike cells that are In contrast, when both environments have walls and the
place cells, it is necessary to record from the same cells in sev- animal is inexperienced, most elds are initially similar in
eral environments. This is because most place cells do not have shape and location despite differences of shape or color be-
a place eld in every environment. For example, early studies tween boxes. With experience, the place cells begin to differ-
(OKeefe and Conway, 1978; Kubie and Ranck, 1983) reported entiate between the two environments, and after a period of
that whereas some complex-spike cells had elds in two or time most cells re differently in the two boxes (Bostock et al.,
more recording environments others had elds in one but not 1991; Lever et al., 2002; Wills et al., 2005). This phenomenon
the other. A second problem is that many complex-spike cells is called remapping and is a good example of a type of learn-
498 The Hippocampus Book

ing that is reected in hippocampal cell ring. Most of the In boxes with high walls and limited views of the room,
remapping that occurred between square and circular envi- place cells often use the box itself as the frame of reference.
ronments involves the cessation of ring in one of the two Rotation of the box or rotation of a cue-card suspended
environments; more rarely, there is a shift in eld location prominently on the interior wall of the box, generally causes
between the two (Lever et al., 2002). One important factor equal rotation of the place elds (Kubie and Ranck, 1983;
inuencing whether a complex-spike cell has a place eld in Muller and Kubie, 1987). On multi-arm mazes with high
any given environment might be the overall level of inhibition walls, the elds sometimes take their reference from specic
experienced by that cell in that environment. Thompson and arms and not from the entire maze. Under these circum-
Best found higher background rates in environments in which stances, interchanging arms can result in the elds following a
the cell had a place eld than in environments in which it did specic arm of the maze irrespective of its allocentric location
not. A reasonable speculation, therefore, is that silent cells are (Shapiro et al., 1997). In this situation, the cells do not re all
potential place cells whose level of inhibition is high in the over the surface of the arm but maintain a localized ring eld
silent environment perhaps due to enhanced excitatory relative to the frame of the arm and can thus still be said to
inputs to its inhibitory interneurons from place cells with be spatially coded. It should not be thought, however, that
maintained elds in that environment. there is a sharp distinction between extramaze and intramaze
cues and frameworks or that any given cell is forced to choose
11.7.5 Frame of Reference of Place Fields between them. In an experiment described in more detail
later, OKeefe and Burgess (1996) found that although most
If place cells identify an animals location, they must do so place elds were controlled by either the animals location rel-
ultimately on the basis of sensory information. In this section, ative to the walls of the testing box or relative to the laboratory
this issue is addressed from the perspective of frames of refer- frame, some were responsive to combinations of one wall of
ence. In the next, the role of specic sensory modalities is the testing box and an extramaze cue, such as the wall of the
examined. The framework to which place elds are referenced testing room. Bures and colleagues (Rossier et al., 2000) devel-
might be egocentric or allocentric. An egocentric framework oped an elegant behavioral paradigm for dissociating intra-
is one that is anchored to the animals body (or some part of maze from extramaze frames of reference. Rats are trained to
it, such as the trunk, head, or eye) and that travels with the avoid a prohibited area of a static circular platform with a
animal as it moves through the environment. An allocentric good view of the surrounding laboratory. Under these cir-
framework, on the other hand, is one that is xed to some part cumstances, the prohibited area can be dened with reference
of the environment. It is therefore a framework in which the to the extramaze laboratory reference frame or the intramaze
animals position changes as it traverses the environment. The frame of the platform itself. Recording from place cells with
fact that complex-spike cells recorded in an open eld envi- the platform rotating slowly showed that the ring elds of
ronment re in their place eld irrespective of the direction in some are xed to the extramaze room cues, whereas those of
which the animal is facing appears to rule out a purely ego- others remain with the rotating intramaze cues. Turning off
centric reference framework. On the other hand, the fact that the lights increased the number of intramaze elds. A large
complex-spike cells can sometimes be directional (e.g., on lin- proportion of place elds disintegrated when the two frames
ear tracks) shows that under some circumstances they can be of reference were dissociated, suggesting that information
egocentric or that the inputs to these cells may be coded in from both was required.
egocentric frameworks. In a subsequent experiment, Zinyuk et al. (2000) demon-
In open elds, where place cells appear to be anchored strated that the proportion of cells related to the extramaze
within an allocentric reference frame, it has been shown that frame, the intramaze frame, or both depended on whether the
this frame can be a reference to the room, the testing box, or animal was trained to navigate within the environment. The
the maze in which the animal has been placed. When record- animals in the navigation group were trained to use the extra-
ing takes place on symmetrical elevated mazes or open elds maze environmental cues to go to an unmarked location in
with low side-walls, the room generally provides the overall the static open-eld box to receive food pellets. Each entry
framework. This is readily revealed by rotating the testing into this target zone triggered the release of a food reward into
apparatus. The usual result is that place elds follow cues a random location for which the rats had to forage. When the
located outside the apparatus (OKeefe, 1976; Olton et al., intra- and extramaze frames of reference were dissociated by
1978) (see Fig. 1112B, later). This is not to say that the intra- rotating the box at 1 rpm, 38% of the place elds stayed with
apparatus, or what are commonly referred to as intramaze, the extramaze framework, 9% with the rotating intramaze
cues are without inuence. They ordinarily may contribute to framework, and 31% with the conjunction of both. Cells in
place eld size or location but be overshadowed in importance this last category red only at that point of the cycle when the
by more distal, or extramaze, cues when the two types are two frameworks coincided. The ring elds of only one-fth
put in opposition. Maze rotation, however, sometimes gives of the cells were disrupted during rotation. In contrast, when
rise to locational ambiguity. A cell may develop two place another group of animals carried out a random foraging task
elds immediately after the rotation, but they quickly resolve in the same environment without the navigational compo-
into one eld with repeated experience of the new congura- nent, almost three-fths of the cells were disrupted by rotat-
tion (OKeefe, 1976; Thompson and Best, 1989). ing the platform. Of the small number of intact elds, 21%
 Hippocampal Neurophysiology in the Behaving Animal 499

stayed with the extramaze environment framework, 14% with navigational system which calculates where an animal
the rotating intramaze framework, and 7% with both. This is in an environment independently of the stimuli
study makes several important points. First, it conrms that impinging on it at that moment. The input from the
place cells can be anchored by the extramaze environmental navigational system gates the environmental input,
cues, the intramaze box cues, or both, in agreement with pre- allowing only those stimuli occurring when the animal
vious ndings. Second, it shows that training the animals to is in a particular place to excite a particular cell. One
pay attention to the extramaze cues increases the proportion possible basis for the navigational system relies on the
of place cells that use these cues, alone or in combination, fact that information about changes in position and
from less than one-third to more than two- thirds. Third, direction in space could be calculated from the ani-
somewhat surprisingly, training to attend to the extramaze mals movement. When the animal had located itself
cues markedly increases the proportion of cells responsive to in an environment (using environmental stimuli),
the conjunction of the relevant extramaze cues and the irrele- the hippocampus could calculate subsequent posi-
vant intramaze framework. It appears that in the absence of tions in that environment on the basis of how far,
explicit training to attend to the spatial aspects of the envi- and in what direction the animal had moved in the
ronment many place cells take their inputs from combinations interim. . . . In addition to information about distance
of a small number of weak distal and local intra-maze cues traversed, a navigational system would need to know
which are easily disrupted by the rotation. In animals trained about changes in direction of movement either rela-
to attend to the distal cues the cells incorporate much more tive to some environmental landmark or within
robust information especially about the extra-maze cues and the animals own egocentric space. . . . (OKeefe, 1976,
are less easily disrupted when these cues are placed into pp. 107108)
conict. Support for this interpretation comes from experi-
ments in the same testing paradigm using blockade of one Both hypotheses have received experimental support: Place
hippocampus by injection of tetrodotoxin. This has no effect elds can be controlled by exteroceptive information from the
on the ability of rats to avoid a place successfully on the basis environment or by movement-generated proprioceptive and
of either extramaze or intramaze cues alone but blocks their vestibular stimuli. This section looks at the external sensory
ability to avoid a place on the basis of extramaze cues in information arising from the environment and asks: Are some
the face of conicting intramaze cues (Wesierska et al., 2005). sensory modalities privileged over others? Which cues deter-
The effects of unilateral blockade are probably due to the dis- mine the angular orientation of a eld in a symmetrically
ruptive effect of the procedure on the other intact hippo- shaped environment, and do these differ from those that con-
campus (Olypher et al., 2006). These results should serve trol its distance from the borders of the environment? The fol-
as a warning that some of the conclusions drawn from studies lowing section examines the role of internal cues generated by
on animals foraging in the open elds in the absence of the animals own behavior in providing estimates of distance
an explicit requirement to encode the spatial aspects of and direction traveled.
the environment might not be providing an accurate pic- We begin by asking: Are some sensory modalities privi-
ture of the full range of the spatial capacities of the hip- leged over others? There are two ways to identify the role of
pocampus. sensory inputs in controlling place elds. The rst follows the
Interestingly, this group has also shown that training rats pioneering methods of Honzik (1936) in his study of the sen-
to use room-based but not arena-based cues to navigate sory control of maze learning. Here, a specic sensory modal-
increases the amount of total variance in place cell ring that ity is eliminated by lesion or occlusion and the consequences
is accounted for by position (Olypher et al., 2002). examined. The second approach is to construct articial envi-
Accordingly, they speculated that some of the variance in ronments in which there are a few controlled cues that can be
place cell ring not accounted for by position (Fenton and experimentally manipulated. Various manipulations have
Muller, 1998) is attributable to attentional switches between been explored: Cues have been rotated; the distance between
alternative spatial reference frames. Their analysis may pro- them changed; the geometry of the environment altered; asso-
vide an interesting window into ostensibly unstable coding of ciative value or cue meaning taken into account. The results
place by hippocampal cells in some experiments. show that many place cells receive information from more
than one sensory input; they can use subsets of this total input
to identify correctly the preferred place; and geometrical
11.7.6 Place Fields Can Be Controlled
information about distance to features of the environment is
by Exteroceptive Sensory Cues
particularly important.
What sensory information does a place cell use to determine A simple manipulation to test the role of vision is to turn
where it res? OKeefe (1976) suggested that two sources of out the room lights (OKeefe, 1976; Quirk et al., 1990; Markus
information were available. et al., 1994). Some cells are unaffected, but others show a more
interesting response. For example, the eld may disappear on
Each place cell receives two different inputs, one con- the rst visit to the preferred area but return on the second or
veying information about the large number of envi- third visit. In a cylindrical open-eld apparatus, place elds
ronmental stimuli or events, and the other from a tend to remain intact if the animal is in the environment when
500 The Hippocampus Book

the lights are extinguished. However, if the animal is removed amount. Neither the shape of the eld nor its distance from
from the environment and then put back into it in the dark, the walls of the box was affected, but there was a slight loss of
about half of the elds change (Quirk et al., 1990). New elds place specicity.
may appear, and they are maintained even if the lights are now The location of the controlling cues within the environ-
turned on. In contrast, there are many fewer cells that retain ment can be important. Cressant et al. (1997) found that three
similar elds in the light and the dark on the radial arm maze. objects placed in a triangular conguration in the center of a
Furthermore, of the cells ring in both light and dark, the r- cylinder did not control the orientation of place elds, but
ing in the light was more reliable. Such observations offer yet they did do so when these same objects were moved to the
another indication that place cells act in distinct ways in dif- periphery. This result was interpreted as suggesting that distal
fering environments. It appears that, in a cylindrical arena or peripheral landmarks are more important for xing the ori-
rectangle, once a pattern of place elds is set up for a particu- entation of the place elds, perhaps because the animals
lar environment it is relatively stable to alterations of extero- movement relative to centrally placed local cues causes con-
ceptive cuesunless the animal is removed from the stant reordering of their positions relative to each other within
environment. In contrast, place cells are more sensory-bound the egocentric frameworks of the distal sensory modalities,
on radial arm mazes and in other situations that restrict an whereas movements relative to distal cues do not. Central
animals movements to a linear path. object A is sometimes seen to the left of central object B and
Eliminating sensory modalities by occlusion or using sometimes to its right.
lesions indicates that there are still place cells after the elimi- If the angular orientation of a place eld in a symmetrical
nation of olfactory, visual, or both visual and auditory inputs. environment is controlled by distal cues, what determines its
Save and colleagues (1998) found that rats blind from birth shape and size and its radial distance from the walls of the
have completely normal place elds when tested as adults. The environment? The rst clue came from an experiment (Muller
only discernible difference was a decreased ring rate of place and Kubie, 1987) that showed that doubling the dimensions of
cells compared to controls. Hill and Bests study (1981) of ani- a cylinder or a rectangular box had two effects: In many cells,
mals deprived of both vision and audition during adulthood the elds remapped to unpredictable shapes and locations; in
revealed they also had apparently normal place fields. about one-third of the place cells, however, eld size increased
However, rotating the arms of the maze showed that, on aver- to about twice the original size without signicant effect on
age, these animals were much more bound to the local cues eld shape. This is less than the fourfold increase that would
of an arm. Further experiments on the small group of cells have represented an exact proportion to the increased area of
that had elds anchored to the allocentric room frame of ref- the recording chamber. Factors other than simple scaling
erence showed that rapid rotation of the animal for 20 to 30 seemed to be at work. A related experiment (Wilson and
seconds prior to running on the maze caused their elds also McNaughton, 1993b) looked at eld changes after a partition
to come under the control of the intramaze cues. On the was removed, converting a square box to a rectangular box
assumption that the rotation primarily affected the vestibular whose long dimension was twice that of its short one. Here
system, these results suggest that, in the absence of distal cells with elds in the square were maintained, whereas other
visual and auditory cues, the angular orientation of place cells cells began to re in the new section of the box. It appears,
can be controlled by olfactory/tactile cues on the maze and by then, that under some circumstances the eld sizes are inu-
vestibular cues. The latter may be mediated by inputs from the enced by the size of the testing environment, but under others
head direction system (see Sections 11.9 and 11.10). they are not. The factors controlling this difference are still
Several studies have attacked the problem of cue control by unclear, but they may relate to whether the animal is, or is not,
constructing environments in which the distal stimuli were in the box when the change is made. If in it, the animal is bet-
explicitly identied and controlled. OKeefe and Conway ter placed to see the change happening in front of him, which
(1978) trained animals on a T maze located within a set of triggers exploration. Under these circumstances the existing
curtains that excluded visual cues from the rest of the labora- elds are maintained intact and new ones created to represent
tory. Four objects were hung around the periphery of a T the new part of the environment. However, when the animal
maze inside the curtains, and the animal was taught to choose is removed from a recording chamber with which it is famil-
an arm of the maze. The four cues were rotated from trial to iar and placed in a dilated version of it, the system may be
trial. Place elds always rotated with the cues (see Fig. 1112B, tricked into treating this as the same environment with the
later, for a similar result). Removal of one or more cues on resulting alteration of existing elds.
each trial showed that a few elds depended on one specic OKeefe and Burgess (1996) employed the second proce-
cue, but most were maintained so long as any two of the four dure to look at the effect of changing box size and shape on
cues were present. Muller and colleagues (Muller and Kubie, established place elds. Combining features of the preceding
1987) showed similar cue dependence of place elds recorded studies, the same cell was studied while the rat foraged for
from rats foraging for food in cylinders or rectangular boxes. food in four differently shaped boxes: two squares, one with
A single white cue-card xed to the wall exerted control over double the dimensions of the others; and two rectangles, each
the angular location of the eld but not its distance from the with its small side taken from the dimensions of the small
wall. Rotation of the card rotated the elds by the same square and its large side from the large square (Fig. 119). One
 Hippocampal Neurophysiology in the Behaving Animal 501

A B
Plank SS HR

4.6 4.1
VR LS

4.9 5.0
Camera view Enclosure

C D

Figure 119. Place elds in four rectilinear enclo-

sures reveal that each eld is comprised of several 5.1 3.5 20.2 5.5
subcomponents. A. Layout of the testing area shows
how the small square (61 cm2) is constructed from
four wooden planks set on their edges. The same
set of planks can be joined in different congura-
tions to make rectangular (61 by 122 cm) and large
square (122 by 122 cm) enclosures. B. Place eld
with strong inputs from the left and bottom walls 6.7 6.5 13.0 4.9
does not change its relative location in the various
boxes. C. A different place eld with strong inputs
from the top, left, and right walls is elongated in the
E
horizontal rectangle. D. A eld similar to that
shown in C breaks into two components in the hor-
izontal rectangle. E. A cell with elds in the small 17.0 0.0
square and vertical rectangle (two left-hand panels)
shows steadily diminishing ring rates in the boxes
intermediate to the vertical rectangle and the large
square. Numbers in each panel represent the peak
eld ring rate (shown in black). (Source: After
OKeefe and Burgess, 1996, with permission.) 19.0 15.7 6.9 0.4 0.1

rectangle was oriented with its long side parallel to the long reduce their rates in an incremental way in boxes intermediate
dimension of the testing laboratory and the other with its long to these two. A small number of cells had one of their eld
side perpendicular to this dimension. The same construction dimensions determined by the distance to the room walls and
materials were used, and the oor covering was interchanged the other by a wall of the box. One important observation was
frequently, eliminating local olfactory and tactile cues as that no cells used the framework of the shape of the box itself.
determinants of eld shape and location. The important nd- If this had been the case, one would expect to see cells with
ing was that for most cells the eld location and shape was elds that rotated by 90 in the two rectangular congurations
determined by the distance to two or more walls usually of the or that red only in the small and large squares. These pat-
box but, more rarely, of the room itself. Some cells were con- terns were never observed. The experiment shows that in envi-
trolled by the distance to two box walls in orthogonal direc- ronments in which the animal has sufficient distal
tions (e.g., the south and west walls) (Fig. 119B); others were information to x its directional orientation it is not the
sensitive to the distance to three or four walls (e.g., Fig. geometry of the testing box that is being captured by the place
119C,D). The expansion of eld width along one dimension eld but the intersection of distances to two or more elon-
sometimes resulted in double peaks or even the apparent gated features of the environment (see Chapter 13 for a fur-
splitting of a eld into two (Fig. 119D). Some cells (e.g., Fig. ther discussion of the geometry of the spatial representation
119E) that had a eld in one box (vertical rectangle) but were and Chapter 14 for a discussion of a computational model
silent in an adjacent box (big square) could be shown to that accounts for these ndings).
502 The Hippocampus Book

Gothard and her colleagues obtained similar results cues. The distance from each set of cues could be computed
(Gothard et al., 1996b). Place cells were recorded while the rat on the basis of sensory inputs from that cue or from path inte-
attempted to nd a reward hidden at a xed distance and gration signals (see Chapter 14, Section 14.3 for a more
direction from two large objects. Each trial began with the rat detailed discussion of this model).
leaving a small box and ended when it returned to it where it Fenton et al. (2000a,b) investigated the problem of cue-
was also rewarded. The location of the box, like that of the control of place elds in a different experimental paradigm.
objects, was moved around the room from trial to trial. Cells They recorded while rats foraged for food pellets in cylinders
were found that red as the rat entered or exited the start/goal with two distinct cue cards on the wall and varied the angle
box; others red relative to one or other of the goal objects; between the cards by small amounts. Removal of either card
and a third group maintained their eld location relative to had no effect on place elds, and rotation of the remaining
the framework of the room, regardless of the location of the card caused equal rotation of the place elds, demonstrating
goal or start box. This study complemented that of Burgess independent control over the elds by each card. In contrast,
and OKeefe in showing that relative location to objects or changing the angle between the cards from 135 to either 160
walls is important. It also reinforced the notion that different or 110 caused subtle changes in the location and shape of the
place cells recorded at the same time may use two or more elds. In general, elds shifted and transformed in the direc-
frames of reference in a single environment. tion of card movement. Fields closer to the cards were more
In another study, Gothard et al. (1996a) examined the inuenced than those farther away. An interesting nding was
frame of reference issue further by recording cells in rats that the peak ring rate was highest for the original card con-
trained to run back and forth on a linear track for food reward guration and decreased as the cards were brought closer
at each end. The goal at one end of the track remained xed together or placed farther apart. Although carried out under
relative to the room, whereas that at the other end was a box different experimental circumstances and interpreted some-
that could be moved between one of ve locations along the what differently by the authors, these results are broadly in
track from trial to trial. Recall that on linear tracks or narrow- line with those of OKeefe and Burgess and of Gothard et al.
armed radial mazes many place cells tend to have directional in showing that place elds are determined by distances from
elds. In the present experiment just over 42% red in one of specic environmental features.
the two directions, another 45% in the opposite direction, and An overview of all the experiments reviewed in this section
just over 12% in both directions and these usually did not bear suggests that there are at least two independent exteroceptive
a close spatial relation to each other. Place elds closer to the determinants of place elds. The directional orientation of a
goal box were tied to it and moved with it, whereas those place eld in a symmetrical enclosure is controlled by polariz-
closer to the other end of the track tended to stay xed with ing distal cues in the room or at the periphery of the testing
respect to the room and the xed goal location. A number of box. On the other hand, the location and shape of the eld is
cells were inuenced by both the moving box and the xed determined by the distance of the animal from two or more
goal. In these cases, as the box was moved farther away from walls or by other features of the environment. In addition, it is
the other end of the track there was evidence for the stretch- clear that interoceptive cues can inuence place eld shape
ing and splitting of the elds described by OKeefe and and location, which we discuss in the next section.
Burgess in their expanding two-dimensional world.
Two interpretations of these results are possible. Gothard 11.7.7 Idiothetic Cues Can Control Place Fields
and her colleagues suggest that there are two representations
or charts of the space on the track: one related to the frame- Place elds can also be located on the basis of idiothetic cues
work of the box and the other to the room. As the animal runs generated by an animals own movements, which consist of
along the track, it switches from one map or chart of the envi- interoceptive stimuli such as head, neck and limb propriocep-
ronment to the other. The animal was, in some sense, treating tors, vestibular signals, and motor reafference signals from
this single runway as two separate environments: a world in intended movements together with exteroceptive stimuli
which it was located relative to the goal box while it was in its derived from optic ow and whisker-detected airow. We
vicinity and a world in which it was located in the room frame noted earlier that rats blind from birth have completely nor-
as it got farther from the moving box. These authors further mal place elds when tested as adults (Save et al., 1998).
interpreted the constant distance of the box- related elds Detailed analysis of the behavior of these animals suggested
from the box as evidence for a path integration mechanism. that they used olfactory and tactile information to recognize
OKeefe and Burgess favor the alternative view: that under objects in the recording environment and then updated the
the conditions of these experiments, the normal map-like locations of place elds on the basis of interoceptive cues asso-
omnidirectionality and connectedness between place cells has ciated with their own movements. Whereas some 80% of
been lost, and they are essentially acting as isolated individu- place cells in the sighted control rats in this study red in the
als. Each individual cell is inuenced by the box and the xed appropriate location before the animal had made contact with
goal in proportion to the distance of its eld from each. Cells any of the objects, none of the place cells in the blind rats did
close to the box are almost entirely controlled by a box, so. However, after contact with one of the objects, 60% of cells
whereas those far from it are controlled by the xed room red appropriately in a single place, and 75% did so after
 Hippocampal Neurophysiology in the Behaving Animal 503

Cell 1 Cell 1 Cell 2 Cell 3

13Hz
Cell 2

Figure 1110. Entorhinal cortex grid cells.

Raw data (left) and ring elds (right) 7Hz
for three grid cells. Note the regularity of
Cell 3
the ring, which forms a triangular grid
pattern in each cell. Numbers indicate the
peak rate. Peaks of the three cells (top
in grayscale, bottom as numbers 13)
are slightly offset from each other so when
superimposed they tend to tesselate the
space. (Source: After Hafting et al., 2005, 1m 23Hz
by permission.)

exploring two objects. Completely normal ring patterns were elds when the rat was replaced into the enclosure (Jeffery et
seen after contact with all three objects. al., 1997). In both experiments the rotations were brief, and
Maintaining spatial specicity in this way represents a their effects were subsequently assessed in a stationary envi-
form of memory. It is presumably an active or working ronment. Similar ndings with head direction cells (Blair and
memory that is maintained by some form of inertial naviga- Sharp, 1996) raise the possibility of a coupling between the
tion or path-integration mechanism. Once an animal had spatial localization and head direction system. The interac-
identied its location in an environment on the basis of exte- tions between the two are discussed in Section 11.10.
roceptive cues, a path integration mechanism could continu- A movement-generated estimate of distance and direction
ously update its position by calculating the changes in would explain the instances of place cell stability in the shift
distance and direction from the original position resulting from light to dark in memory tasks and following lesions that
from the animals movements. This type of mechanism might limit access to exteroceptive cues. The source of this path-
explain the short-term memory properties of place cells, integration signal is either input from interoceptive cues or
which are discussed in greater detail in Section 11.8. collateral discharges arising from motor structures actively
Further evidence for idiothetic inuences on place elds generating movements. This information arrives at the hip-
comes from experiments in which animals have been passively pocampus via several routes. Information about the animals
rotated. As we saw in the experiment of Hill and Best, rotating heading direction is carried by the head direction cells and
animals deprived of vision and audition before placing them enters the hippocampal system through the presubiculum pro-
on a radial arm maze results in the rotation of place elds. jection to the medial entorhinal cortex. We discuss the proper-
Several groups (Sharp et al., 1995; Wiener et al., 1995; Bures et ties of the head direction cells in greater detail in Section 11.9
al., 1997; Jeffery et al., 1997) have studied the effect of rotating and their interaction with the place cells in Section 11.10.
the testing enclosure, or parts of it, with the animal inside it. When it arrives to the medial entorhinal cortex, head direction
Sharp et al. (1995) rotated the cylinder walls or oor sepa- information is combined with distance information in a set of
rately at a fast or slow speed and in the light or dark. It was grid cells (Hafting et al., 2005). One of the functions of the grid
assumed that the slow rotations were below the speed cells is to convey information about distances in specic envi-
detectable by the vestibular system and would not be com- ronmental directions. Each grid cell res in several locations in
pensated for: The elds rotated with the rotating enclosure an environment, with the locations forming a regular pattern
relative to the laboratory frame. Both visual and vestibular sig- as though they were nodes on a triangular grid (Fig. 1110; see
nals inuenced the angular location of place elds to about also Fig. 1121D, color centerfold). Different cells recorded at
the same extent. Slow rotation of the rat in a separate cham- the same location have the same grid spacing and orientation
ber outside of the testing enclosure also led to rotated place relative to the environment. They differ, however, in the loca-
504 The Hippocampus Book

tion of the nodes such that the ring peaks of one cell are mated that place elds can be maintained for only 1 to 2 min-
slightly shifted from those of its neighbor. The multiple inter- utes on the basis of path integration information alone (Save
digitated elds of several such cells together cover the environ- et al., 2000).
ment (Fig. 1110, right). For each cell, the size of the grid
appears to be independent of the size or shape of the environ- 11.7.8 Are Place Cells Inuenced
ment. The orientation of the grid relative to the environment, by Goals, Rewards, or Punishments?
however, is dependent on the location of a polarizing visual
cue on the wall of the enclosure in much same way as are the If place cells are part of a navigational system that can guide
postsubicular head direction (Taube et al., 1990b) and the hip- an animal to locations containing desirable objects such as
pocampal place cells (Muller and Kubie, 1987). It seems likely, food and water or to avoid dangerous places, one might expect
then, that the orientation of each grid is controlled by the head to nd goal cells whose activity was sensitive to goal location
direction cells of the presubiculum. Cells located at increasing or navigation to a goal or, more generally, a change in place
depths from the postrhinal border form grids whose nodes cell ring following a shift in the valence of parts or all of an
have elds of increasing size and spacing. In summary, the grid environment. These goal or valence cells might exhibit any
cells probably do not form a map of a given environment by of several characteristics: The location of their place elds
themselves but provide the Euclidean distance and direction might alter when the valence of the environment changes, and
metric postulated by the cognitive map theory of hippocampal it might happen in a way that reects the goal location in an
function (OKeefe and Nadel, 1978). This idiothetic informa- environment; there might be a disproportionate number of
tion is combined with sensory information about each envi- place cells with elds located at reward sites; or there might be
ronment to create the specic map of that environment in the nonlocational cells whose ring rate depended on which goal
hippocampus. was being sought or navigated toward. Several experiments
To maintain the appropriate distance between grid points have searched for the rst type by shifting goal locations:
as the animal moves around the environment, the grid cells Some reported no effect (OKeefe, 1976; Speakman and
must take its speed into account. Information about speed OKeefe, 1990; Zinyuk et al., 2000), while others found con-
may arise in posterior hypothalamic areas that are known to comitant place eld shifts (Breese et al., 1989; Kobayashi et al.,
provide theta-related inputs to the hippocampus and where 1997; Hollup et al., 2001). Here we need to bear in mind that
electrical stimulation produces running or jumping, with the rewards have both sensory and incentive properties, and eld
speed of the movement increasing as a function of the stimu- shifts might be due to the former rather than the latter. Three
lation intensity (Bland and Vanderwolf, 1972). Integration of a recent studies employed experimental paradigms in which the
speed signal over time as the animal ran in a constant direction goal was not marked by any physical stimulus. The rats in the
would give a measure of distance traveled during that time. Hollup et al (Hollup et al., 2001) experiment searched for a
There is direct evidence that information about an animals hidden platform in a modied annular watermaze and the
speed of movement is available to the hippocampus. A small elds of some cells shifted when the goal location was
number of nonprincipal speed cells have been recorded in changed. Kobayashi used rewarding brain stimulation of the
the hippocampus. The ring rates of these cells directly corre- hypothalamus in order to designate particular locations as
lates with the animals speed of running regardless of direction goals and reported that six of 31 place cells either developed
or location (OKeefe et al., 1998). Second, higher running additional new elds in the rewarded location or shifted their
speeds tend to increase place cell ring rates (McNaughton et place eld to that location {Kobayashi et al., 1997}. Finally,
al., 1983a; Wiener et al., 1989; Zhang et al., 1998). Further- Zinyuk et al. (2000) used a paradigm similar to that of
more, the ring frequencies of many place cells from a sta- Kobayashi et al. but rewarded the animals with random food
tionary animal running in a wheel located in the place eld pellets rather than electrical brain stimulation; they found no
were positively correlated with the speed of running (Czurko changes in place elds when the reward location was moved.
et al., 1999; Hirase et al., 1999). Some cells asymptoted within Because there was no physical reward at the goal locations in
the range of speeds reached in the wheel. Cells that wholly or these experiments, the explanation of the discrepancies can-
partially code for speed have also been found in the pre- not rely solely on the perceptibility of the goal or its role as a
subiculum (Sharp, 1996), which projects to the hippocampus sensory cue. We note that many of the cells that shifted elds
via the medial entorhinal cortex. In all, we can conclude that in the Kobayashi et al. experiment had very low ( 1 Hz) place
information about both an animals heading direction and its eld rates, below those normally accepted in these types of
speed of movement through the environment is available to experiment. These considerations, however, do not apply to
the hippocampal formation, and this information most likely the Hollup ndings (Hollup et al., 2001).
is combined in the medial entorhinal grid cells. Moita et al. (2004) used aversive electrical stimulation of
The path integration system suffers from the fact that the orbit of the eye to condition rats to an auditory condi-
errors accumulate rapidly as the animals heading direction tioned stimulus (CS) or to the background context. They
and distance from the original location are continuously found that as a result of training some place elds were
updated. On the basis of data from several studies in which altered. More place cells changed their ring elds in the con-
attempts were made to remove exteroceptive information text than in the cue-conditioned group and more in the con-
after initial localization or make it irrrelevant, it has been esti- ditioning box than in a different control box. Furthermore,
 Hippocampal Neurophysiology in the Behaving Animal 505

they found that following cue conditioning the place cells landing zone, where the pellets initially dropped. Most pellets,
began to re with a short latency response to the auditory cue however, were retrieved and eaten elsewhere. One-fourth of
but only if the animal was in the place eld of that cell when the cells in the prelimbic/infralimbic areas had place elds and
the CS was delivered (Moita et al., 2003). This result is dis- these were about three to four times larger than those found
cussed further in Section 11.11.3. in the hippocampus. The centers of a large percentage of these
Evidence for the second type of goal representation comes elds were concentrated in the trigger (36%) and landing
from Hollup et al. (2001), who found twice as many cells rep- (42%) zones. Rotating the cue card on the wall of the enclo-
resenting the unmarked goal in their annular watermaze than sure rotated the animals representation of the trigger zone
would be expected by chance. The third type of active goal cell location, as judged by its behavior, but had no effect on the
was reported in a study of temporal and frontal cells recorded landing zone place elds. Conversely, changing the location of
in human epileptic patients while they played a taxicab game the pellet dispenser and thus the location of the landing zone
in which they searched for passengers in a small virtual reality caused a shift in the landing zone elds to the new area but
town and took them to their destinations in the form of spe- had no effect on the behavioral approach to the trigger zone.
cic storefronts (Ekstrom et al., 2003). About one-fth of the The large size of the prefrontal place elds and the concentra-
cells recorded were sensitive to the goal being sought, almost tion of their centers at goal regions to which the animal navi-
three-fourths of which responded while searching for a spe- gated make them much better potential candidates for goal
cic location. An additional 7% were involved as the subject cells than hippocampal neurons. They bear some resemblance
searched for more than one store and 22% while searching for to the goal cells postulated in the models of spatial navigation
passengers. A small number of cells showed a place by goal of Burgess and colleagues (Burgess et al., 1994; Burgess and
interaction, ring in a particular location if the subject OKeefe, 1996) (see Chapter 14, Section 14.4).
crossed it en route to a particular goal. These goal cells were
located throughout the temporal and frontal lobes and not 11.7.9 Temporal Patterns of Place Cell Firing
concentrated in the hippocampus or parahippocampal gyrus
as were the place and spatial-view cells (see below) recorded in Complex-spike cells do not re in a continuous pattern when
the same study. the rat runs through a place eld but burst with a frequency
It seems clear, then, that changing the valence of an envi- close to that of the EEG theta rhythm. In several studies (Fox
ronment or regions within that environment can cause the r- and Ranck, 1975; Buzsaki et al., 1983) recordings were made
ing fields of some place cells to shift. The functional while the rats were running on a treadmill. Unfortunately,
signicance of this is not clear, unless it turns out that the cells pyramidal cells do not re at their maximal rate in such a sit-
that changed had elds where the animal was located during uation unless the animal is in that part of the environment
the occurrence of the rewarding or punishing event. At pres- that the cell represents (its place eld). On a treadmill, CA1
ent, it is not clear exactly how goals are represented or whether complex-spike cells often re at a low rate and display a pref-
goal location is stored in the hippocampus or outside of it. erence for the positive peak of the dentate gyrus theta. Phase
When no distal sensory information is available, the sensory correlates have also been studied on narrow tracks as the rat
qualities of rewards may be used to locate place elds. In a runs through a cells place eld (OKeefe and Recce, 1993;
richer environment, food and water may be categorized as Skaggs et al., 1996; Harris et al., 2002; Mehta et al., 2002;
objects that are potentially unstable over time. Food sources Yamaguchi et al., 2002; Huxter et al., 2003), and a different,
become depleted and new ones become available, and this more interesting pattern has emerged. OKeefe and Recce rst
happens over a time scale quite different from that of other noted that instead of remaining correlated to a constant phase
cues such as trees or bushes. Nonetheless, the locations of of the EEG theta cycle, as on treadmills, the phase of ring
reward have to be stored somewhere. However, on balance, the changed in a systematic way. When the rat entered the cells
experimental results, although not conclusive, suggest that the place eld, the cell began ring at a particular phase of theta.
information about goal location is probably not stored in any However, as the animal progressed through the eld, the
simple fashion in the CA3 or CA1 areas of the hippocampus bursts of unit ring occurred on an earlier phase of each suc-
itself. Regions such as the lateral septum subiculum, nucleus cessive theta cycle (Fig. 1111C). The phase of ring correlates
accumbens, or prefrontal cortex, which receive inputs from with the animals location in the place eld (Fig. 1111B,D),
the hippocampal formation, might be the site of such place- and this correlation is higher than for time after entry into the
reward cells. In support of this hypothesis is a recent study eld (Fig. 1111E) or instantaneous ring rate (Fig. 1111F).
strongly suggesting that one type of goal cell can be found in This phase precession phenomenon is partly explained by cells
the prelimbic/infralimbic areas situated in the rats medial ring rhythmically at a frequency higher than that of theta.
prefrontal cortex (Hok et al., 2005). Rats were trained in a Dentate granule cells also phase shift but by a lesser amount,
cylinder to spend a short period of time in a localized but and the onset of ring is at an earlier phase of the theta cycle
unmarked region to receive a pellet of food elsewhere. The than that seen in the CA1 cells (Skaggs et al., 1996).
pellets dropped from an overhead dispenser into a localized Is there information in the phase correlate of place cell r-
zone but then bounced elsewhere, ending up all over the ing beyond that contained in the ring rate? If there is, does it
enclosure. There were thus two localized but unmarked zones, add greater precision to the locational information contained
a goal zone, which upon entry triggered the reward, and a in the rate, or are the two coding for independent variables
506 The Hippocampus Book

0 37.4 Hz
B rate code

C
Temporal code

200 ms

D E
r = -0.562 r = -0.379
Figure 1111. Temporal coding of loca-
Phase (deg)

Phase (deg)

tion in hippocampal pyramidal cells.

300 300
A. Rat shuttled from left to right and
200 200 then back again to receive a food reward
at each end wall of a linear track. B.
100 100 Firing rate eld map of the cell based on
multiple runs. C. Spikes on a single run
0 0 through the eld show phase precession,
-150 -140 -130 0 0.5 1
moving to earlier phases of the EEG
Position (cm) Time in field (s) theta cycle with each successive wave.
Spikes are shown in black; lighter tick

F G marks above identify the zero crossing

point of the EEG. D. Plot of theta phase
r = -0.250 versus the position for multiple runs
Mean pearsons r

0.5 through the eld of an example cell. The

Phase (deg)

300 0.4 correlation between phase and position

(D) is higher than the correlation with
200 0.3 the time in eld (E) or the instantaneous
100 ring rate (F) for an average cell. G.
0.2 Correlation of phase with position, time,
0 0.1 and instantaneous ring rate across a
0 20 40 population of 94 place elds. Best corre-
Inst. firing rate (spikes/s) 0.0 lation is with position. (Source: After
pos time ifr Huxter et al., 2003, by permission.)

(OKeefe, 1991)? Jensen and Lisman (2000) have used data the correlation between phase and a xed proportion of the
collected by Skaggs et al. (1996) and shown that, with certain total eld size. Whether phase and rate can represent variables
assumptions, the accuracy of locating an animals position on other than location and speed is an open question. It has been
a narrow track can be increased by more than 40% when known for some time that the ring rate in the place eld can
phase is taken into account in addition to rate. In this view, vary considerably from one traverse to the next (see above),
phase acts like a vernier, permitting ner-grain location of the and there is some evidence that different odors located in a
animal within the place eld. Evidence for the dual coding place can be coded by different rates (Wiebe and Staubli, 1999;
hypothesis comes from a linear track study showing that Wood et al., 1999a). Some evidence in support of the idea that
phase and rate could vary independently (Huxter et al., 2003). phase can represent a nonspatial as well as a spatial variable
The rate within the eld varied as a function of the animals comes from the running wheel experiments of Hirase et al.
running speed, and phase coded for the proportion of the (1999) and Harris et al. (2002). They found that whereas the
place eld the animal had traversed. Changing the size of the rate of ring of the place cells was a function of the animals
eld by changing the distance between the walls at the ends of speed of running, the phase with respect to theta was relatively
the track increased the rate of phase precession to maintain constant at low ring rates but occasionally changed at higher
 Hippocampal Neurophysiology in the Behaving Animal 507

rates in the absence of changed location. One variable that (1998) studied the oscillations inside the dendrites of CA1
caused a change in phase in some cells was the directional ori- cells during theta and showed that they are 180 phase-shifted
entation of the wheel, and thus the animal, to the room. with respect to those in the soma, as predicted by the wave
One important consequence of the phase precession effect interference hypothesis. They also found that depolarization
is that the spatial overlap between the ring elds of two place of the dendrites causes an increase in the frequency of these
cells is represented within each theta cycle by the amount of oscillations. This and other models of the phase shift are con-
time the bursting pattern in the cell of the rst eld entered sidered in greater detail in Chapter 14.
precedes that of the second eld. This temporal difference can Alternatively, the site of phase shift generation may ulti-
be demonstrated in the cross correlation between the ring mately be found in areas outside of the hippocampus but pro-
patterns of the two cells (Skaggs et al., 1996). The farther apart jecting to it, such as the entorhinal cortex. A study by Zugaro
the eld borders, the larger is the temporal gap between ring and colleagues (2005) lends strong support to the idea of an
peaks within each theta cycle. extrahippocampal origin. They trained rats to shuttle between
Part of the phase precession effect might be explained by a two ends of a linear track and recorded the phase precession
coupling between the amount of cellular depolarization and in hippocampal place cells. During some runs through the
the time of ring of the cell (Harris et al., 2002; Mehta et al., place eld, a single 0.1 ms duration electric shock was deliv-
2002). As the animal advances farther toward the center of the ered to the bers of the ventral hippocampal commissure.
eld, the cell would be more depolarized, re more spikes on This had the effect of resetting the phase of the hippocampal
each cycle, and begin ring at an earlier phase of the cycle. theta and inhibiting all recorded cells for approximately 200 to
This cannot be the whole story, however, as the phase contin- 250 ms. They found that when the cells recovered and began
ues to precess in the latter part of the eld when the ring rate ring again they did so at the correct phase of the (reset) theta
(and presumably the level of depolarization) is falling. An to code correctly for the animals location at that point. This
alternative explanation for phase precession invokes a mecha- nding appears to argue against a primary role for the hip-
nism based on the interaction of two theta-like waves (OKeefe pocampus itself in the phase precession and points in the
and Recce, 1993; Lengyel et al., 2003). These waves would nor- direction of structures afferent to the hippocampus that
mally be of the same frequency but 180 phase-reversed so might not have been affected by the brief intervention. The
that, when added together, they cancel each other. A slight most likely source of this preserved positional information is
increase in the frequency of one of these oscillators relative to the entorhinal cortex, which has direct projections to the CA3
the other would result in an interference pattern when the two and CA1 pyramidal cells as well as to the dentate granule cells
waves are added together. If it is assumed that the extracellu- and which was probably not affected by the electrical shock to
lar EEG represents the lower frequency of the two oscillators, the ventral hippocampal commissure. Evidence that it escaped
and that place cells re on the peaks of the interference pat- comes from an observation by Zugaro et al. They looked for
tern, the precession phenomenon would be seen. The dual but did not see a rebound evoked potential in the hippocam-
oscillator model predicts that the size of the place eld is pus in response to the shock, suggesting that the entorhinal-
dependent on the difference between the frequencies of the hippocampal pathways were not greatly affected via the
two oscillators. The closer the two frequencies are together, the CA1-to-entorhinal return pathway. As we have seen, the
larger is the eld. An article by Maurer et al. (2005) provides medial entorhinal cortex contains grid cells that are theta-
experimental evidence in support of this predicted relation modulated and might be able to identify changes in the ani-
between eld size and the oscillatory frequency of the pyram- mals location in the environment independent of the
idal cell. They showed that the average sizes of place elds are hippocampus.
not uniform along the long (septo-temporal) axis of the hip-
pocampus but increase in the temporal direction. In parallel 11.7.10 Place Fields in Young and Aged Animals
with this eld expansion, place cells in the dorsal hippocam-
pus have a higher (intrinsic) frequency of oscillation than In altricial animals, which depend on their mother after birth
those located more ventrally despite no differences in the fre- (such as the rat), the brain and in particular the hippocampus
quency of the EEG theta or the theta cells in the two regions. continues to develop for considerable periods of time after
The dual oscillator interference model suggests there is a birth. For example, the dentate gyrus continues to generate
causal link between these two phenomena, with the dorsoven- large numbers of new granule cells for several weeks after
tral gradient in intrinsic frequency (higher in the dorsal hip- birth and at a lesser rate throughout life (see Chapter 9), and
pocampus) leading to the gradient in the spatial extent of the inhibitory processes do not reach their full adult level of func-
ring elds (smaller elds in the dorsal hippocampus). tioning until postnatal age 28 days (P28). Tests of spatial
The origin of the two separate but interacting theta memory and navigation suggest that the hippocampus is not
rhythms and the site of their interaction are unknown. The fully functioning until P40. Schenk (1985) trained rats of
latter may be in the hippocampus proper. One may be inter- different ages on two versions of the Morris watermaze: the
nal to CA3 and CA1 pyramidal cells, which can oscillate under standard version in which the platform was hidden and the
the cholinergic inuence of the septal nuclei. The other theta animal had to use distant spatial cues throughout learning
rhythm might arise from the direct projection from the and a version in which the platform was hidden but its loca-
entorhinal cortex to the CA3 and CA1 areas. Kamondi et al. tion was marked by a visible proximal cue. This meant the task
508 The Hippocampus Book

could be learned by a hippocampal or a nonhippocampal that the older animals had larger ring elds (Barnes et al.,
strategy. Which of these methods was used was probed by tri- 1983). Subsequent studies, however, have not been able to
als in which the proximal cue was removed. Schenk found that replicate this change but have found that the eld sizes of the
animals aged P28 took longer to learn a new platform task aged animals were normal or, under some circumstances, even
than did adults and reached adult levels of performance only more compact than those of the adult controls (Mizumori et
at ages greater than 40 days. Animals at age P35 could perform al., 1996; Tanila et al., 1997a). There is also disagreement as to
the version of the task with the proximal visual cue; but when whether the place elds in aged animals are more or less reli-
the cue was removed, they failed to show transfer to the distal able than in younger animals. Whereas Barnes and colleagues
cues. This nding suggests that they had developed all of the (1983, 1997) found that the CA1 place elds in aged animals
abilities necessary to solve the watermaze task except forma- are less reliable, Tanila et al. (1997a) found that, if anything,
tion of the distal cue-based spatial representation needed to they were more reliable. Mizumori et al. (1996) found that
navigate to the goal in the absence of the visible proximal cue. they were more reliable in the CA1 pyramidal cells but less
A longitudinal study in which rats were originally trained on reliable in the hilar cells. Barnes et al. (1997) has suggested a
the hidden platform version of the watermaze at age 21 and reconciliation of these apparently conicting results. They
then given daily trials for 69 days put the development of spa- also found no difference in eld size between groups but
tial navigation ability somewhat earlier (Clark et al., 2005). found that older animals spontaneously remapped between
They found that the performance of the animals rose above trials in the absence of any environmental changes. In the ear-
chance level after the second day of training and steadily lier Barnes et al. study, data were averaged across trials. If
increased over the next 10 days to asymptote at adult levels at spontaneous remapping occurred between trials, it might lead
approximately P35. It is reasonable to conclude that the abil- to the impression that the ring elds were more dispersed
ity to form allocentric spatial memories and to perform spa- and less reliable. The cause of the spontaneous remapping is
tial navigation does not fully develop until around 35 to 40 not clear but may signal weakening of control of the head
days of age in rats. direction system over the hippocampal place elds with age.
In the only study of the development of place cells in young Spatial behavior in the watermaze also appears to reect a
animals, Martin and Berthoz (2002) recorded complex-spike decrease in the consistency of the spatial representation in
cells from animals at ages P27, P29, P34, P40, and P52 and older animals in that they sometimes head directly for the
above while they were searching for random food pellets on hidden platform and on some trials take much longer to get
an open eld platform. They found that the place elds of there, resulting in a bimodal distribution of scores (Barnes et
younger animals were larger and more diffuse than those of al., 1997).
adults and became more compact with age, nally reaching Tanila and colleagues (1997a) looked at whether cue con-
adult values at about P52. Furthermore, the place elds trol over place elds changed with aging. Aged animals were
recorded on successive 10-minute trials were unstable in divided into spatial memory-impaired and spatial memory-
younger animals, shifting location from one trial to the next, unimpaired groups following testing in the Morris water-
and reached adult levels of stability only at age P52. In con- maze. They were then trained on the four-arm radial maze,
trast, a small number of head direction cells, which signal the with both distal visual cues and local visual, tactile, and olfac-
orientation of the animals head relative to environmental tory cues available on the arms of the maze. Their place cells
cues (see Section 11.9), were recorded from the cingulum at were then recorded. Probe trials in which either set of cues was
P30 and appeared to be indistinguishable from those reported rotated or scrambled were also conducted to determine which
in the adult. Previous work on the development of hippocam- set of cues was controlling the elds. About two-fths of the
pal theta had suggested that theta waves could be recorded on elds of young animals remapped when the distal and local
the hippocampal EEG as early as P10, developing to adult lev- cues were rotated by 90 in opposite directions, and approxi-
els over the next 2 weeks (Leblanc and Bland, 1979). It seems mately another third followed the distal cues. A smaller per-
reasonable to conclude from both the behavioral and the elec- centage (about one-fth) followed the local cues. In contrast,
trophysiological data that the rat hippocampus continues to more than three-fourths of the place elds in the aged animals
develop over the rst month of life and reaches the mature with memory decits followed the distal cues with fewer than
adult level of functioning only at ages 40 to 50 days after birth. one-fth remapping. Scrambling the distal cues caused the
The head direction and the theta systems, on the other hand, latter group of place cells to become responsive to the local
appeared to be functioning at earlier ages, reaching maturity cues, showing that the predominant inuence of the distal
by P30. cues in the double rotation probes was not due to a sensory
At the other end of the life cycle, when animals get older, decit. It appears as though the cells of the younger animals
their spatial learning abilities decrease (for a review see were relying on all of the cue information, whereas those of
Rosenzweig and Barnes, 2003) and their hippocampal place the memory-impaired aged animals were selectively attending
cells appear to undergo changes. There seems to be little over- to the distal visual cues. It is not clear why this shift should
all agreement as to whether place elds get larger or become occur with age or how to explain it. It is especially puzzling
less reliable as animals age. In the rst study that compared the because the aged animals were selected on the basis of their
sizes of place elds in adult and aged animals, it was reported decits in the Morris watermaze, which requires attention to
 Hippocampal Neurophysiology in the Behaving Animal 509

distal cues for its solution. The authors noted that all of the provides powerful cholinergic and GABAergic inputs to the
animals had adopted a strategy of entering adjacent arms of hippocampal formation and has a major inuence on hip-
the maze to solve the task and therefore may not have been pocampal theta activity. As we shall see in Section 11.12.6,
using a spatial strategy. Whether forcing them to use a spatial many medial septal cells are theta cells that have a good phase
strategy would have made a difference is not clear. relation with the hippocampal theta. Its role in the control of
Rosenzweig and colleagues (2003) looked at environmen- place eld ring has been studied by Mizumori et al., 1989).
tal control of place elds in a different way and compared it They trained rats in an eight-arm radial maze and recorded
with the animals performance in a spatial task. They trained from place cells in CA1 and CA3 during performance of the
animals in a task where they (and their place elds) could task. Surprisingly, blocking the medial septum with the anes-
locate themselves within one of two frameworks (Gothard et thetic procaine left place elds intact despite disrupting maze
al., 1996a). The goal at one end of a linear track was xed rel- performance. Subsequent experiments have suggested that r-
ative to the room cues, whereas the goal at the other end was ing in the subiculum is disrupted and that this area of the hip-
located inside a box that moved along the track from one run pocampal formation thus contributes to the control of
to the next. Place elds located close to the moving box tended behavior in this task. Leutgeb and Mizumori (1999) replicated
to move with the box, whereas those distant from it stayed this observation but found a difference between the place cells
xed relative to the room framework. As the animal ran from in the lesioned animals and controls when the animals were
the box to the xed end of the track, the population of cells placed in a new environment or faced with altered spatial
could be said to switch frameworks. Rosenzweig et al. trained cues. Interestingly the place elds of the lesioned animals
the rats to slow down at an unmarked location on the track to showed less transfer from light to dark and slightly greater
receive positively rewarding electrical brain stimulation. This transfer between rooms.
goal location stayed xed relative to the room cues and thus Lesions or temporary inactivation of other regions have
would be better located after the population response of the been shown to affect place elds in different ways. Lesions of
cells had switched into the room frame of reference. They the perirhinal cortex, which projects to the entorhinal cortex
found that the adult animals learned the task better than the and therefore might be providing sensory information to the
aged animals and that as a group the adult animals switched hippocampal place cells, have no effect on basic eld parame-
from the box framework to the room framework earlier on the ters but reduce the consistency of locational ring (Muir and
track than the aged animals. Furthermore, on an animal-by- Bilkey, 2001): In contrast to the stability of eld locations in
animal basis there was a good correlation between the point at the control animals, eld centers in the lesioned rats fre-
which the ensemble of place cells recorded switched and the quently moved from one exposure to the testing box to the
ability of the animal to distinguish between the goal location next. A similar effect on the stability of hippocampal place
and a control location. Although it was not formally tested, it elds was found after lesions of the prefrontal cortex (Kyd and
seems reasonable to conclude that the ability to switch into the Bilkey, 2005), and this may have been mediated by frontal pro-
xed room framework was at least in part dependent on the jections to the entorhinal/perirhinal cortex.
ability of room cues to inuence the place elds and that this Temporary inactivation of the retrosplenial cortex (Cooper
was decient in the aged animals. and Mizumori, 2001) disrupt an animals performance on the
radial arm maze in the dark and during the initial learning
11.7.11 Hippocampal Place Cell Firing Is phase in the light. During inactivation, the location of place
Inuenced by Other Areas of the Brain elds shift on the maze. There are head direction (HD) cells
and more complex HD, location, and movement cells in ret-
Lesions of the septal or entorhinal projections to the hip- rosplenial cortex (Chen et al., 1994; Cho and Sharp, 2001).
pocampus have been reported to cause a decrease in the num- This suggests that the retrosplenial cortex is more involved in
ber of place cells that are found and, in the case of entorhinal nonvisual (perhaps path integration) control of place cells and
lesions, to a shift in the stimulus control of place eld ring spatial behavior. On the other hand, the role of objects placed
(Miller and Best, 1980). Following entorhinal cortex lesions, at the periphery of the environment in controlling the orien-
the elds rotated with the maze, unlike the elds of normal tation of place elds in a symmetrical environment seems to
animals, which remain anchored to extramaze cues. These be mediated, at least in part, by the visual and the parietal cor-
ndings are consistent with the idea that CA1 and CA3 cells tices. After lesions of the visual cortex, 70% of place elds did
have access to distal sensory input via the entorhinal cortex. not rotate in step with the rotation of the landmarks. In com-
The importance of direct entorhinalCA1 connections has parison, 100% of place elds in the controls did so (Paz-
been highlighted by Brun et al. (2002), who removed the input Villagran et al., 2002) (see Section 11.7.6 regarding how place
from CA3 (and thus the dentate gyrus as well) onto CA1 cells elds can be controlled by exteroceptive sensory cues). When
and showed that CA1 place cells could still form well dened, the objects were removed, the elds in both lesioned and con-
stable place elds in repeated exposures to a familiar environ- trol animals remained in the standard position xed relative
ment. to the room. The parietal cortex also seems to be involved in
The inuence of the medial septal nucleus is somewhat the control of place cell orientation relative to the landmark
different from that of the entorhinal cortex. Recall that it objects but in a different way (Save et al., 2005). Following
510 The Hippocampus Book

lesions to the parietal cortex, most cells (78%) still rotated and 5 (12%) to object identity. The remaining 21 responded
with the objects but, unlike in control animals, did not main- to both. Most of the responses were inhibitory. Interestingly,
tain this rotated location when the objects were subsequently the spatial neurons were more heavily concentrated in the
removed. This type of short-term memory for the visual loca- posterior hippocampus, which is the analogue of the dorsal
tion of landmarks is discussed in greater detail in Section 11.8 hippocampus in rodents. Ringo and colleagues (Ringo et al.,
and appears to be dependent on the integrity of the parietal 1994; Sobotka et al., 1997; Nowicka and Ringo, 2000) investi-
cortex. gated the role of eye movements in hippocampal single unit
responses. Monkeys were trained to look to one of ve loca-
11.7.12 Primate Hippocampal Units tions in the light or in the dark to receive rewards. In an early
also Exhibit Spatial Responses study, about one-third of the units changed their rate during
saccades; in a subsequent study, 13% of the cells were shown
The spatial properties of single units in the hippocampus of to be sensitive to position and another 17% were shown to be
primates including humans have been studied. This is impor- sensitive to direction. Because the animals head was xed in
tant because it establishes the generality of the spatial nature these experiments, it is not possible to say whether the effec-
of the hippocampus and shows that the ndings in rodents are tive saccades were to locations in the laboratory or in a head-
not unique. However, if there is broadening of the function of centered framework. Remembering the location of a stimulus
the human hippocampus to include episodic memory as well on a computer screen might not be the same type of spatial
as spatial memory, it might be expected that the cells in task as locating oneself in an environment and might more
humans and, more generally, primates might have a broader easily be solved using egocentric spatial strategies dependent
spectrum of response properties than those found in rodents. on parietal than hippocampal cortex (Burgess et al., 1999).
Furthermore, the standard approach to recording single units Wirth et al. (2003) studied the responses of hippocampal neu-
in primates is markedly different from that found in the freely rons during a task that required the monkeys to learn the asso-
moving rat. Primates are usually restrained with their heads ciation between specic scenes and specic locations. Novel
xed, and stimuli are often presented in ways that make it dif- pictures of scenes were presented on a VDU screen, and the
cult to identify a spatial correlate or to dissociate the differ- animals task was to learn to move its eyes to one of four loca-
ent frames of reference that might be used for the localization. tions on the screen at the end of the presentation to obtain a
It is also known that restraining rats severely depresses the reward. Each scene was presented for 0.5 second followed by a
locational correlate of the hippocampal complex-spike cells 700-ms delay during which the screen was blank before the
(Foster et al., 1989). These constraints have been partially animal was allowed to move its eyes to the required location.
overcome in a few studies by giving the animals increased In all, 61% of the hippocampal cells recorded had ring rates
mobility either in movable chairs or carts or by allowing them that were signicantly altered during the presentation of one
to move freely around a large cage. We should not be sur- of the scenes, during the delay that followed it, or both.
prised, then, if there are differences between the unit/behav- Furthermore, there was a good correlation between the altered
ioral correlates found in primates and those in rodents. The ring during the trial and the learning of the behavioral
spatial responses are described in this section; the nonspatial response. Changes in ring rate preceded the behavioral
responses are described separately in Sections 11.11. 1 and 8. learning by a small number of trials for 14 cells, occurred at
Several types of spatial response have been found in pri- the same time in 4 cells, and followed learning in 19 cells. This
mate hippocampal units. Responses to the location of the suggests that although the ring rate of some cells may have
stimulus on a VDU screen, spatial-view cells that respond to been related to the learning of the scene-location association
the location at which the animal is looking, and place cells it may have been involved in learning other aspects of the
similar to those in the rodent have all been described. When scene in others. Control trials involved the presentation of
the response of hippocampal units to the identity of a stimu- familiar scenes that required the same eye movement as the
lus is compared to its spatial location, considerably higher novel scene. This rules out the possibility that the hippocam-
percentages of units are found to the latter variable. Rolls and pal ring was related to a particular eye movement or to a spe-
his colleagues (1993) tested monkeys on object recognition cic location on the screen. Although it is possible to describe
tasks and compared unit responses to object familiarity with this task as learning an arbitrary association between a scene
those to object location. They reported that about 9% of cells and an eye movement to a location dened relative to the
in the hippocampal region responded differentially to the VDU screen, it is also possible that the animals were learning
location of the stimulus on a display screen. In contrast, only to attend to a particular location in each new scene, to remem-
a small percentage (2%) of cells responded to familiar objects ber the scene and the location during the delay, and to look at
(see Section 11.11.8). Colombo et al. (1998) recorded units in the location when allowed to do so after the delay. On this
the primate hippocampus during a similar delayed matching view, these cells would be closely related to the spatial-view
to sample task in which either the spatial location or the object cells, which respond when an animal looks at a location rather
identity of the sample stimulus had to be remembered. They than goes there (Rolls et al., 1997) (see next paragraph).
found 41 neurons that responded during the delay period, and Several groups have also looked for spatially coded cells in
of these 15(37%) were related exclusively to spatial position the hippocampus of monkeys free to move around the envi-
 Hippocampal Neurophysiology in the Behaving Animal 511

ronment. Rolls and colleagues (1997) have found a type of monkeys were also trained in a two-dimensional screen-based
spatial cell not described in the rodent: the spatial-view cell. task in which they had to move a pointer to different parts of
Spatial-view cells respond selectively when the animal looks at the screen. A subset of cells were recorded in both the virtual
particular locations in the testing room irrespective of where reality task and the screen-based task. Of the cells with spatial
the animal is situated in the room when it looks at that loca- responses in either or both tasks, about one-third had spatial
tion. For example, one cell increased its ring rate markedly activity in both, one-half in the virtual reality task alone, and
when the animal looked at a particular corner of the room only 15% in the screen-based task. It appears that large-scale
regardless of where it was located in the room itself and allocentric spatial tasks are better for activating primate hip-
regardless of the orientation of gaze required to look there. pocampal neurons than are screen-based egocentric tasks.
Some of these cells continued to respond when the target If we are to compare hippocampal physiology in rodents
location is screened off by curtains, suggesting that the cells and primates appropriately, more recordings are needed from
are not responding to particular sensory features in that loca- primates whose heads are unrestrained and who are free to
tion. This group has also reported the existence of whole body locomote around a complex environment (see Section 11.2).
motion cells (OMara et al., 1994), which may be related to the There has been some interesting progress along these lines. In
speed cells in the rat described above. In contrast, this group the rst study in completely freely moving monkeys, rodent-
has looked for but not found place cells comparable to those like place cells with high signal-to-noise ratios in the hip-
seen in rats. The existence of spatial-view cells might be an pocampus proper were found in squirrel monkeys performing
indication that primates have developed the ability to identify a spatial memory task in three dimensions (Ludvig et al.,
places and their contents without physically visiting those 2004). Interestingly, many of these cells had elds that
places, an important step in the evolution of the spatial map- involved the walls of the wire-mesh testing cage, areas that
ping system. would not have been sampled in the oor-bound experiments
Ono and his colleagues (1993) found place-coded neurons of Rolls and Ono.
in the hippocampal formation of monkeys that could visit dif- Another approach has been to record units from human
ferent locations in an environment while performing different epilepsy patients (awaiting determination of seizure foci). In
tasks. The monkeys were trained to sit in an enclosed cart and early studies the responses to faces, objects, and scenes were
to move it to nine different locations in the testing environ- studied (see Section 11.11.9). Of relevance here is a study of
ment by pressing a lever. About 13% of the cells red more unit activity during virtual locomotion in a taxi-driver game
when the animal was at one location than when it was in other (Ekstrom et al., 2003). This study recorded units from the
locations. In a subsequent task, the animals were required to temporal lobe and frontal cortex and looked for evidence of
perform an object-in-place discrimination. The cart was cell activity responsive to the variables of place, view, and
moved to a particular location and the view window was goal-seeking and of conjuctions between these variables. Place
opened, allowing the animal a sight of an object at that loca- responses clustered in the hippocampus to a greater extent
tion. About one-fourth of cells responded when objects were than in the parahippocampus (24% of hippocampal cells
shown to the animal in this task. A subset of these cells (5% of being pure place cells versus 8% in the parahippocampus),
the total) were object-in-place cells that responded differen- and pure location-independent spatial-view cells clustered
tially when the animal was shown an object in a particular in the parahippocampus (17% vs. 5% in the hippocampus).
location and not in other locations. These cells were also not The goal cells were distributed evenly throughout the tempo-
interested in a different object in the preferred location. They ral and frontal cortices (see above). The place-responsive cells
appear to have properties similar to those of the place and were found to be nondirectional, as would be predicted from
object-in-place cells described in the rat. Further evidence the rodent literature, given that the subjects were free to move
that the cells are appropriately described as place cells comes through areas in the virtual town from different directions. In
from an experiment by Nishijo et al. (1997) in the same labo- an important control, Eckstrom et al. looked for unit
ratory. Here the animal in the cart was moved backward as responses to isolated landmarks before the patients learned to
well as forward through the environment, and the cells con- use them to navigate in a virtual reality environment; they
tinued to re at the same location. This manipulation reverses failed to nd any.
the cues in egocentric space but leaves allocentric cues
unchanged. This group has also used virtual reality environ-
ments to provide testing environments more related to the
open eld tasks used with rodents (Hori et al., 2005). Monkeys 11.8 Place Cells Are Memory Cells
were trained to use a joystick to move between ve reward
locations in a large 100 m virtual diameter space containing a A role for the hippocampal formation in memory is suggested
20 m diameter arena surrounded by landmarks such as a tree, by activity-dependent synaptic plasticity, such as long-term
a house, and the building. Almost one-third of the hippocam- potentiation (LTP) and long-term depression (LTD) (see
pal units displayed spatial elds. When the distal cues were Chapter 10) and the profound amnesia associated with dam-
rearranged, two-thirds of the cells tested remapped by either age to the medial temporal lobe (see Chapters 12 and 13). Is
ceasing to re or by changing the spatial pattern of ring. The this mnemonic capability also reected by place cells? This
512 The Hippocampus Book

chapter presents evidence that these cells can maintain their tive and interoceptive cues other than these might be capable
ring elds over a period of a few minutes in a working mem- of maintaining it (see Section 11.9). As we saw in Section
ory task, that place eld characteristics can change with expe- 11.7.11, a role for the parietal cortex in this short-term spatial
rience over the course of a day, that sensory control of place memory has been demonstrated (Save et al., 2005).
cell ring can become more differentiated with experience A nal observation on these cells was made during control
over days and weeks or can shift from exteroceptive to intero- trials in which the animal was not placed in the start arm until
ceptive cues when the animal learns that the former are unsta- after the controlled cues had been removed, preventing the
ble, that place cell activity during sleep reects experience in animal from knowing which arm contained the reward.
the prior waking period and may be involved in consolidation Under these circumstances, choice performance falls to
of recently acquired spatial memories, and nally that several chance levels, but the pattern of place cell ring still stayed
properties of normal place cell behavior in familiar and novel appropriate for the animals choice of goal (Fig. 1112D). This
environments, such as place eld stability, are based on mem- constitutes strong evidence that these cells are not ring to the
ory and depend on the NMDA receptor implicated in LTP. actual environmental location but to where the animal
thinks it is. Other studies have also found a good relation
11.8.1 Hippocampal Place Cells Remember between place eld activity and behavioral choices (Zinyuk et
the Animals Location for Several Minutes al., 2000; Lenck-Santini et al., 2001a, 2002; Rosenzweig et al.,
During a Spatial Working Memory Task 2003) but there has also been failure to nd this (Jeffery et al.,
2003). In the latter experiment, rats continued to perform a
When discussing the properties of place cells, we saw that they hippocampus-dependent spatial task, albeit at a reduced level,
displayed memory properties such as continuing to re in following a change in the testing box that caused remapping
appropriate places when the lights have been turned off. The of most of the CA1 place elds. Here, it is plausible to assume
cells use environmental cues to set up their ring pattern, but that place cells in other areas of the hippocampal formation
once established the pattern is sometimes surprisingly free of did not remap and continued to support the behavior.
environmental inuences. The clearest demonstration of this
memory phenomenon was in experiments in which the 11.8.2 Place Field Plasticity During
environmental cues controlling place elds have been identi- Unidirectional Locomotion
ed; and it was shown that place elds were maintained fol-
lowing the removal of these cues (Muller and Kubie, 1987; One interesting line of investigation has involved changes in
OKeefe and Speakman, 1987; Save et al., 2005). We concen- place eld characteristics over time as rats run along a track.
trate here on the OKeefe and Speakman experiment in which When animals are run on narrow tracks, many of the elds are
place cell ring pattern was shown to be correlated with the directional, ring in one direction of movement but not the
animals behavior. other. The experiments described in this section test models
OKeefe and Speakman (1987) recorded cells during a spa- describing the effects of temporally asymmetrical LTP and the
tial memory task in a cue-controlled environment (Fig. encoding of sequence-related information (see discussion in
1112). Rats were rst trained on a plus shaped elevated maze Chapter 14; see also Section 11.7.3). Mehta and colleagues
to go to a goal dened by a set of cues within the curtained (1997, 2000) have shown that over the course of a few dozen
environment (Fig. 1112A). Rotation of the cues and goal traverses in the same direction along a track, CA1 place elds
between trials ensured that the rats learned to use these cues. shift backward relative to that direction of motion and
Place elds rotated in step with the cue rotations (Fig. become larger. They also found that an experience-dependent
1112B). On some trials, the cues were removed while the rat skewness develops in the place elds; that is, on the animals
was still in the start arm before it was allowed to choose. Tests rst runs of the day a given cells place eld is symmetrical, but
on normal rats had established that well trained animals could with continued exposure the place cell tends to re at a higher
remember the location of the goal for periods as long as 30 rate when the animal leaves the place eld than as it enters it.
minutes after the cues were removed. Of 30 cells with elds on However, this increased skewness in CA1 place elds has not
the maze, 27 maintained these elds following removal of the been found by others (Huxter et al., 2003). The development
cues (Fig. 1112C). Cells with elds on the nonstart arms for of place eld expansion and backward shift has been shown to
a particular trial also red in the correct place, demonstrating depend on NMDA receptors (Ekstrom et al., 2001). An
that place cell memory is more than the continuous persist- intriguing aspect of the phenomenon in CA1 is that the effect
ence of a trace set up during the registration period; the sys- resets overnight, as if the cells do not remember the previous
tem can, in addition, compute the correct location on the days experience on the track (Mehta et al., 2000). Evidence
maze of place elds for cells that are inactive in the start arm. from Lee et al. (2004) conrms and extends this nding but in
One explanation for this memory phenomenon points to a addition shows a difference betweeen the CA3 and CA1 elds:
role for the distal cues at the periphery of an environment in CA3 elds develop skewness and shift backward immediately
orienting the head direction system. Once this has been set by on exposure to a newly altered track and maintain these
the orientation of the control cues on a given trial, exterocep- changes on subsequent days, whereas CA1 elds shift back-
 Hippocampal Neurophysiology in the Behaving Animal 513

A Plus maze with 6 distal cues B Place fields rotate with distal cues
north goal

goal
west goal all goals east goal

pen

south goal

C Place cells remember locations D Place cells predict spatial choice

goal goal goal

perception perception memory

Perception

goal
goal goal

control control
memory exp goal rat goal

Figure 1112. Place cell activity during a spatial working memory ing the perceptual (top) and memory (bottom) phases of the exper-
task. A. Layout of the testing environment showing an elevated iment. D. A different place cell res strongly in the
90 arm and
plus-shaped maze surrounded by six cues and a set of curtains. B. weakly in the 90 arm on both perceptual and memory trials
Place cell with a eld on the arm 90 counterclockwise to the goal (upper left and right). During control trials, it fails to re with cor-
(
90, inner panel). The ring eld remains constant with respect rect relation to the (unknown) goal (lower left) but res with cor-
to the distal cues and goal as they are rotated by multiples of 90 rect relation to the animals choice of goal arm at the end of the
relative to the laboratory frame (outside panels). Trials have been trial (lower right). In B and C, each contour  0.7 spikes s-1; in D,
rotated so the goal is always shown at the top. More contours indi- each contour  1.5 spikes s-1. (Source: After OKeefe and Speakman,
cate a higher ring rate. C. Same cell as in B maintains its eld dur- 1987, with permission.)

ward only from the second day of testing and do not maintain 11.8.3 Cue Control over Hippocampal Place Cells
the changes. Lee et al. interpreted these ndings as suggesting Can Change as a Function of Experience
that the CA3, but not the CA1, network can store sequence
information in the long term, and that CA1 is more involved Experience with an environment can change which features of
in novelty detection. Note that CA1 cells can show long-term that environment control place elds. Such features include
plasticity in other paradigms (see Section 11.8.3). the color of a polarizing cue and the geometrical shape of the
514 The Hippocampus Book

environment. Bostock et al. (1991) studied place elds in two Tanila et al., 1997b), latent because it does not necessarily
recording chambers that were identical except for a white or manifest in behavior, and incidental because it takes place in
black polarizing card at the periphery. They found that when the absence of explicit reward.
the black card was rst substituted for the white card most of In the Wills et al. study (2005a), the boxes were more dif-
the place elds remained the same. However, as the rat con- ferent from each other to begin with, differing not only in
tinued to experience the white and black card environments, shape but also in the color, odor, and texture of the walls.
the place cells began to distinguish between them. The Original training took place in a white wooden circle and a
remapped elds fell into three classes: (1) the eld in the brown morph square. Under these circumstances, the remap-
black card environment was a rotation of that in the white ping took place over a period of minutes, and most of the cells
card environment, or (2) the cell ceased to re in the black had differentiated between the two environments by the end
card environment, or (3) the elds in the two environments of the rst few exposures. Furthermore, the remapped cells act
were completely different in location and shape. The second in a unitary ensemble fashion, as was revealed by challenging
and third classes were described as complex remapping and them with shapes intermediate between the circle and the
were found to be all-or-none: If one place cell showed square. After several days experience with the original train-
complex remapping, so did another simultaneously recorded ing enclosures the animals were transferred to circles and
cell. Learned remapping can also occur between environ- squares constructed from the same brown morph box. It was
ments of different geometrical shapes (Muller and Kubie, then possible to probe the basis for each cells differentiation
1987). between circle and square by recording in a series of octagons
Lever et al. (2002) and Wills et al. (2005) studied the devel- that vary systematically from the circle-like (regular octagon,
opment of shape remapping in CA1 place cells in detail. Both adjacent side ratio 4:4) to the square-like (adjacent side ratio
experiments studied remapping between square and circular 7:1) (Fig. 1113). Most of the remapped cells shifted abruptly
enclosures. In the Lever et al experiment, the enclosures dif- from a square to circle ring pattern (for example, by treating
fered only in shape and were identical in color, texture, and the 6:2 octagon as a square and the 5:3 octagon as a circle).
odor; in the Wills experiment they differed across all four Furthermore, in all animals tested on this probe, all of the
dimensions. The activity of several hippocampal place cells simultaneously recorded cells show the abrupt shift at exactly
was recorded on each day of the experiment; often they were the same point in the sequence of octagons. This suggests that
different cells, but in some cases recordings were taken from there are two mechanisms at work: pattern separation and
the same cells over many consecutive days. Remapping was pattern association (see Chapter 14). The rst would permit
rapid and coherent in the latter experiment but slow and indi- the abrupt switch from the circle to square pattern despite a
vidualistic in the former. As in the Bostock et al. experiment very small change in the geometry of the boxes. The second
described above, Lever found that on rst exposure to a circu- would be based on some type of cooperation among the active
lar and a square environment most place elds were identical subset of place cells, perhaps revealing the operation of an
in the two shapes. With continued exposure, however, the cells attractor network and would account for the coherent behav-
began to differentiate between the environments. Unlike ior of the network.
Bostock et al., they found that different cells differentiated
between the environments at different rates, so after a few days 11.8.4 Control of the Angular Orientation
some of the place elds were still similar in the two environ- of Place Cells in a Symmetrical Environment
ments whereas others had clearly remapped. After 1 to 3 weeks Can Be Altered by the Animals Experience
of exposure to both environments, most of the cells had dif- of Cue Instability
ferent elds in the two. Recording from individual cells
through their remapping transition period showed that it In a symmetrically shaped environment such as a cylinder, an
took one of three forms. The most common form involved a animals sense of direction can be controlled by distal polariz-
gradual reduction of ring rate in one of the two shapes until ing cues, such as a card on the wall of the environment, or by
it reached zero. Other forms of remapping involved the devel- internal path integration cues, such as the amount it has
opment of a new eld in one of the shapes in tandem with the turned around. Commonly, the visual information provided
decline of the original eld in that shape or, more rarely, a shift by a distal cue overshadows the internal cues, and rotating the
of the eld in one shape to a new location. Once the cells had cue causes an equal rotation of the elds (see Section 11.7.6).
learned to differentiate between the two different-shaped Jeffery and collegues (Jeffery et al 1997; Jeffery 1998) showed
boxes, rats were placed back in their home cages for delay that given certain experiences the animal could learn to give
periods of up to 39 days without any exposure to the shapes; priority to internal path integration signals over the external
upon reexposure, most of the cells continued to discriminate cue. They tested the effectiveness of the cue card by rotating it
the two shaped boxes. CA1 place cells thus demonstrated when the animal was out of the enclosure and the effective-
long-term memory for this learned discrimination. Note that ness of the internal path integration system by slow rotations
in both this and the Bostock et al. study the rats were not (subvestibular threshold) of the animal relative to the frame-
trained to discriminate the environments but were equally work of the enclosure. As expected, they found that, initially,
rewarded in both. Thus, the learning is incremental (see also the visual cue card controlled place field orientation.
 Hippocampal Neurophysiology in the Behaving Animal 515

SQUARE 1:7 2:6 3:5 4:4 CIRCLE

A G M E J C H F L D K B I N
CELL
0.5 0.9 0.5 0.3 0.4 0.6 0.6 8.3 9.5 8.6 8.7 4.6 9.4 6.9
1

0.1 0.0 0.1 0.1 0.1 0.1 0.2 5.0 6.9 5.8 4.3 5.7 5.8 5.2
2

0.7 0.7 0.8 0.8 1.5 0.7 3.0 6.9 6.3 7.3 5.1 7.5 5.3 4.6
3

0.4 0.0 0.4 0.2 0.5 0.2 0.5 9.7 11.2 8.3 10.2 11.4 7.7 6.4
4

5
0.4 0.1 0.7 0.0 0.0 0.2 0.7 12.6 13.1 11.7 4.7 14.4 11.2 4.4

6
0.2 0.2 0.1 0.6 0.5 0.7 0.6 9.8 14.3 8.7 9.5 11.9 15.0 15.8

7 0.3 0.6 0.7 0.3 0.3 0.2 0.7 5.1 7.2 9.3 5.6 7.0 6.5 3.6

0.5 0.6 0.4 0.7 0.2 0.5 0.4 2.5 2.7 9.8 5.6 3.8 3.9 3.0
8

9
10.7 4.0 3.1 3.8 1.6 11.1 1.3 0.3 0.2 0.3 0.4 0.1 0.4 0.5

10
3.0 3.4 6.8 4.7 5.1 9.1 3.1 0.2 0.2 0.6 0.5 0.1 0.6 0.4

7.7 8.4 6.1 12.1 2.3 10.5 1.1 0.7 0.3 0.3 0.3 0.4 0.8 0.3
11

13.1 7.8 11.8 6.5 12.2

12 0.3 0.4 0.5 0.4 0.4 0.6 0.5
14.4 6.9

7.2 3.9 3.3 7.1 4.9 3.8 5.7 6.0 6.7 5.9 4.0 11.3 4.9 11.0
13

7.4 6.6 8.6 2.1 7.9 7.7 6.9 4.9

14
12.6 8.4 4.2 7.3 9.5 6.4

0.7 3.7 4.2 3.4 4.4 4.1 3.2 9.4 9.7 10.5 10.2 6.9
15 7.3 12.4

6.0 6.6 10.0 6.3 7.4 3.6 7.3

16
4.6 2.9 6.7 6.4 10.4 4.2 2.4

7.1
17
8.8 11.1 18.6 11.5 10.3 5.7 11.9 4.1 6.1 13.9 4.1
3.1 7.3

18
3.2 3.7 2.2 2.7 3.5 5.6 3.5 2.5 4.5 0.5 2.1 0.4 2.0 1.1

10.4
19 15.7 12.3 1.7 12.8 9.7 10.7 7.2 7.6 21.2 18.0 18.5 4.3 10.8

13.1 9.1 7.1 11.1 5.8 12.6 7.5 7.5 4.2 6.8 6.3 10.3 6.3 4.2
20

Figure 1113. Abrupt coherent switch from square-like pattern to pattern at the same point in the series between the 2:6 and 3:5 octa-
circle-like pattern in an incrementally changing series of octagons. gon (arrow at top). The order in which the octagons were experi-
Fields of 20 simultaneously recorded place cells following fast enced is indicated by the letters A to N in the second row. Numbers
remapping training. Cells 117 met the criterion for remapping, in panels are peak rates. (Source: After Wills et al., 2005, with per-
and all switched abruptly from the square-like to circle-like mission.)

However, following experience in which the rat could see the enclosure, demonstrating that they were under the cont-
the cue card move relative to the background environment rol of the vestibular system. Rotations of the cue when the
and therefore was not a stable landmark, control of place eld animal could not see it move did not produce the shift away
orientation passed from this visual cue to the interoceptive from cue control. It seems likely that the site for this cue-
path integration system. Now the elds were no longer con- instability learning is in the head-direction system itself (see
trolled by rotations of the visual cue but could be shifted Sections 11.9 and 11.10), perhaps in the presubiculum. This is
by slow rotations of the animal relative to the framework of suggested by two key features of the data. First, all the cells
516 The Hippocampus Book

behave as a map-like ensemble, following either exterocep- sleep. Firing sequence replay has been replicated by Wilsons
tive cues or interoceptive cues. Second, once the path integra- group for slow-wave sleep (Lee and Wilson, 2002) and for
tion system surplants the visual cue card in the control of ori- REM sleep (Louie and Wilson, 2001). It should be emphasized
entation, it generalizes to novel situations. Following that the replay time scales relative to the waking spatial expe-
remapping induced by changing the color of the environ- rience differ markedly for these types of sleep, involving about
ments walls, the newly remapped population of CA1 cells 20-fold compression during slow-wave sleep and basically no
immediately follows the interoceptive rather than the visual compression during REM sleep. Second, it seems that slow-
cues, without needing to relearn the cue instability wave sleep replays sequences immediately after they were
(Chakraborty et al., 2004). experienced, whereas REM sleep replays experience at least a
day old. Finally, cross-correlational analysis showed that
11.8.5 Complex-spike Cell Firing and increased correlations were found between pairs of neocorti-
Connectivity During Sleep Is Modulated cal cells as well as pairs of hippocampal cells and, most
by Prior Spatial Learning Experiences intriguingly, between pairs of hippocampal and neocortical
cells (Qin et al., 1997; Siapas and Wilson, 1998). The latter
It has been suggested that consolidation of memories might nding may be taken as evidence for increases in the func-
take place during sleep and, in particular, during REM sleep tional connectivity between, as well as within, structures.
(Maquet, 2001; Stickgold et al., 2001). Therefore, one might In all, these results indicate selective, orderly reactivation of
expect neuronal ring patterns to reect this consolidation cells that have recently taken part in a spatial experience. They
activity. Furthermore, if consolidation involves the transfer of also suggest that co-activation of place cells during the waking
information from hippocampus to neocortex (an idea devel- experience might lead to increased connectivity and further
oped further in Chapters 12 and 13), one might expect there consolidation during sleep states. These studies have involved
to be increased interaction between cells in these brain areas CA1 cells. A future goal is to compare CA3 and CA1 cells and
during and after sleep. Experiments have suggested that the relate these ndings to the experience-dependent place eld
cell activity during ripples and sharp waves, or that related to changes during unidirectional track running described above
the theta waves of REM sleep, could be the neural signature of in this section. The mechanisms that immediately produce
this information transfer/consolidation process. sequence replay during sleep in CA1 appear to reset (Lee and
Pavlides and Winson (1989) recorded pairs of place cells in Wilson, 2002), like the CA1 place eld changes such as skew-
rats during sleep sessions that followed experience on a radial ness and backward shift. Given that these are maintained on
arm maze. The rats were allowed to visit the place eld of subsequent days in CA3 cells (Lee et al., 2004), one might pre-
one of the two cells prior to each sleep session but not the dict that experience-dependent changes in CA3CA3 connec-
other. Remarkably, there was a selective increase in ring dur- tivity in sleep would also be less transient.
ing both the slow-wave sleep and REM sleep periods of What is needed to take these results forward and demon-
the cell that had been allowed to be active in its place eld strate a causal relation between the sleep phenomenon and
prior to sleep. The other cell showed no change in ring rate. subsequent memory capacity is a physiological or pharmaco-
In another study following place cell activation on a runaway logical technique for selectively disrupting postexperience LIA
during waking, the theta phase at which these cells red or REM theta activity or, even better, altering the patterning of
during subsequent REM sleep shifted to the positive peak (Poe place cell activity during these periods.
et al., 2000). Because LTP is more likely to occur at this phase,
this nding might be an indication that recently experi- 11.8.6 NMDA Receptor Confers
enced cells are more susceptible to plastic modication. Mnemonic Properties on Place Cell Firing
Evidence for such changes comes from studies on changes in
cell connectivity. Wilson and McNaughton (1994) examined Several studies have pointed to a role for the NMDA receptor
the effects of spatial experience on changes in functional con- (NMDAR) in conferring mnemonic properties on place cell
nectivity between CA1 cells. Using cross-correlation tech- function (reviewed by Nakazawa et al., 2004). McHugh et al.
niques (see Box 112), they found an increase in the (1996) studied the place cells of mice in which the NMDA
correlation between pairs of cells with overlapping place elds receptor subunit NR1 had been knocked out only in CA1
in slow-wave sleep after, compared to before, the environmen- pyramidal cells. Deletion of this subunit renders the receptor
tal experience; these correlations were more pronounced dur- inoperable. Rotenberg et al. (1996) looked at the elds of place
ing ripples (see also (Kudrimoti et al., 1999). There was cells in animals that transgenically expressed a mutated cal-
evidence from another study (Qin et al., 1997) that some cium-independent form of CAMKII, part of the intracellular
of this increased connectivity takes place between cells that signaling pathway that mediates the effects of the NMDA
were already connected prior to the environmental experi- receptor. Both groups of animals had decits in LTP and spa-
ence. Skaggs and McNaughton (1996) ran rats in one direc- tial memory (Rotenberg et al., 1996; Tsien et al., 1996). In both
tion on a narrow triangular track and found that the temporal experiments fewer place cells were found in the mutants and
ordering of ring between cells with partially overlapping the quality of the mutant place elds was degraded, a result
elds was replicated in a compressed form during slow wave that has also been seen in other mutant mice with disrupted
 Hippocampal Neurophysiology in the Behaving Animal 517

NMDAR-dependent plasticity (Cho et al 1998). In a different month. Some of them are activity-dependent and temporary,
mutant. Rotenberg et al found that place elds were less stable lasting only the length of a single trial in a working memory
over repeated exposures to the environment (Rotenberg et al., task or a session on a maze, whereas others appear to be per-
2000). In the McHugh et al. experiment, the place cells showed manent, such as the differentiation between two boxes follow-
an interesting loss of the usual temporal relation between the ing slow remapping. Some of these changes clearly depend on
ring patterns of cells with overlapping place elds. As we saw the integrity of the NMDA receptor in different parts of the
in Section 11.7.9, place cells re with a theta-like bursting pat- hippocampus, but others may not. Indeed, for many of these
tern as the animal runs through the ring eld, which is changes it has not yet been rmly established that the under-
revealed by strong periodicity in the cross-correlogram of the lying synaptic plasticity takes place in the hippocampus itself
two spike trains (see Box 112). The absence of this periodic- rather than in structures afferent to the hippocampus and is
ity in the McHugh et al. experiment suggests a fundamental merely being passively reected by the cells there. Finally, there
breakdown in the temporal ring pattern of the place cells or is substantial evidence that some type of consolidation process
the temporal relations between cells lacking a functional takes place during sleep following an experience, and it is
NMDA receptor. reected in place cell ring. Whether this involves intrahip-
The global role of the NMDA receptor in the long-term pocampal processes, such as the strengthening of synapses
stability of place elds was studied by Kentros et al. (1998). between cells with overlapping elds, or involves the transfer
They showed that systemic administration of the competitive of information from hippocampal stores to neocortical ones,
NMDA channel blocker CPP had no effect on established must await further experimentation. It is clear that there are
place elds in an environment long familiar to the rats or on sufficient examples of plasticity reected in place cell ring to
the formation of newly remapped place elds in a novel make them one of the more fruitful targets in the study of the
environment, nor on the short-term stability of these new, neural basis of memory. In a subsequent section (see Section
remapped place elds, as measured over a couple of hours. 11.11), we examine the evidence for changes in hippocampal
However, on the second day of exposure to the novel environ- cell activity during nonspatial learning and memory tasks.
ment, the previous days place elds had been forgotten:
Again the cells remapped the new environment, but the new
patterns of ring bore no relation to those of the previous day.
This nding suggests that the NMDA receptor is more impor- 11.9 Head Direction Cells
tant for the stability of place elds in the long term than it is
in their initial establishment or their short-term maintenance. Another well characterized class of spatial cells recorded in the
Highly informative studies have used conditional CA3- hippocampal formation of freely moving awake animals is the
restricted NMDA receptor knockout mice to explore the con- head-direction (HD) cells, found in the dorsal presubiculum
tribution of long-term CA3 plasticity to CA1 place cell ring and regions connected with the presubiculum, such as the
and spatial behavior (Nakazawa et al., 2002, 2003). These anterior thalamus (Ranck, 1984; Taube et al., 1990a; Taube,
studies suggest two important roles for CA3 plasticity: pattern 1995a, 1998). As already noted, HD cells re whenever the
completion (see Chapter 14), whereby only a subset of cues rats head points in a specic direction relative to the environ-
are sufficient to reinstate a pattern of ring originally associ- ment, irrespective of its location or whether it is moving or
ated with the full cue set; and rapid, single-exposure, learning. still (see Fig. 1121C, color centerfold). The primary correlate
Pattern completion is suggested by the nding that although is the azimuthal orientation of the head in the horizontal
CA1 cells in the mutants showed normal ring in full-cue plane. Pitch and roll appear to be relatively unimportant, as is
reexposures to an environment they showed signicantly the orientation of the rest of the body. Figure 1114A shows
reduced levels of ring (albeit in the appropriate locations) in the ring rate for three of these HD units plotted as a function
partial-cue reexposures. Impaired behavior was also seen in of the direction of heading. Each cell has a single preferred
probe trials during the watermaze task under partial-cue con- direction, and ring falls off rapidly, symmetrically, and
ditions. A CA3 plasticity-dependent role in rapid environ- almost linearly as the head direction rotates away from the
mental learning is suggested by the poorer quality of the preferred direction. We refer to these directions in terms of
mutants CA1 place elds in novel environments and by compass headings but do not imply that these cells are sensi-
impaired performance in the delayed match-to-place version tive to geomagnetism. The portion of the 360 circle covered
of the watermaze (see Chapter 13). by a given cell ranges from about 60 to 140, with the average
being about 90 (Fig. 1114A). An impressively Euclidean
11.8.7 Summary of Place Cell Plasticity property of HD cells is that the preferred direction is remark-
ably independent of position (Taube et al., 1990a; Burgess et
The evidence summarized in this section indicates that place al., 2005); the preferred direction vectors in the various parts
cells show several forms of plasticity, some of which can be of an environment appear not to converge (e.g., upon a salient
correlated with the animals behavior in spatial memory tasks. distal cue, as one might have expected) but are parallel. The
These various forms of plasticity last for markedly different distribution of peak ring directions across the population of
periods of time, ranging from a few minutes to longer than a cells is uniform, with no direction preferred over any other.
518 The Hippocampus Book

A 3 Head Direction Cells Firing Fields

6 25 100

5 20 80
firing rate (spikes s-1)

4
15 60

3
10 40
2

5 20
1

0 0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360 0 60 120 180 240 300 360
heading direction (degrees)

B Fields rotate with cue card C Fields shift after cue card removal
25
25
20 cell 1 cell 1

20
15 standard standard
configuration 15 configuration

10
cell 2 10 cell 2

5
5
0
0 60 120 180 240 300 360
0
0 60 120 180 240 300 360

25

cue card 25
20 rotated 180

20
15
no cue card
15
10

10
5

5
0
0 60 120 180 240 300 360
0
0 60 120 180 240 300 360

Figure 1114. Head direction cells in the dorsal presubiculum. lines. C. Removal of the cue card shifts ring elds of two cells
A. Firing elds of three cells as a function of the animals heading recorded at the same time. Again, both elds shift by approximately
direction. Note that each ring eld subtends an angle of about 90 40 and maintain a constant difference in their preferred heading
but that the ring rates differ. (Source: After Taube et al., 1990a.) directions. The x axes in the bottom gures of B and C have been
B. Rotation of a cue card on the wall of the cylindrical enclosure by extended beyond 360, and part of the eld is duplicated to show
180 rotates the angular orientation of two simultaneously recorded the entire shape of the ring elds around 0/360. (Source: B, C.
elds by a similar amount. Note that the relative angle between the Afer Taube et al., 1990b, both with permission.)
two elds tends to remain the same, as shown by the dashed parallel

Similarly, there does not appear to be any topography to the 11.9.1 Head Direction Cells are
directions represented by neighboring neurons, an arrange- Controlled by Distal Sensory Cues
ment reminiscent of the lack of environmental topography in
the layout of place cells. Importantly, unlike hippocampal Taube et al. (1990a) found that rotations of a cue card xed
place cells, which signal location in some locations but are to the apparatus wall by multiples of 90 produced almost
silent in others, HD cells re in all environments tested. commensurate rotations of the directional ring eld (Fig.
 Hippocampal Neurophysiology in the Behaving Animal 519

1114B). Slight deviations from perfect rotation, however, blindfolded rats also had a small effect that subsequent tests
suggested that in the experimental setup used the card was revealed was mostly attributable to the oor. There was also
not the only environmental cue exerting stimulus control over some evidence that the preferred direction in blindfolded rats
the preferred direction. This was conrmed in tests in which was less stable.
the card was removed altogether. Card removal revealed two
interesting properties of these cells (Fig. 1114C). First, the 11.9.2 Angular Distance Between any
width of the directional ring eld remained the same, as did Given Pair of Head Direction Cells
the peak ring rate. This suggests that these properties depend Always Remains Constant
on factors intrinsic to the cell itself or to the network in which
it is embedded, not on environmental inputs. Second, the pre- An important property of HD cells is obligatory coupling
ferred direction shifted to a new unpredictable compass point the angular distance between the preferred directions of
in almost two-thirds of the cells. Shifts ranged from 108 in pairs of HD cells is remarkably resistant to alteration (Fig.
the clockwise direction to 66 in the counterclockwise direc- 1114B,C). Regardless of the cue-control manipulation, both
tion. The preferred direction appeared to remain approxi- cells of a simultaneously pair are always found to rotate by the
mately the same in the remaining one-third of cells. Zugaro et same amount. Figure 1114C illustrates this nicely: Following
al. (2001) tested the relative inuence of distal and proximal cue removal, the preferred head direction of cell 1 rotates by
cues in determining the orientation of HD cells. As Cressant approximately 40 accompanied by a similar rotation in cell 2.
et al. (1997) had shown with place eld orientation, Zugaro The important implication of this coupling of preferred direc-
found that the rotation of three distinct objects at the periph- tions between cells is that there must be some kind of hard-
ery of a platform, bounded by a cylindrical enclosure, caused wiring of the network of HD cells such that the population of
an equal rotation in the preferred directions of all anterior cells ring at any one time gives an accurate and, above all,
thalamus HD cells tested. The cylinder prevented the rats unambiguous representation of heading direction. The main-
from seeing more distal background cues, such as geometri- tenance of a constant angular distance between HD cell pairs
cally congured black curtains along the walls of the room. in the face of environmental change is well modeled by attrac-
When the same object-set rotation was performed in the tor networks (e.g., Redish and Touretzky, 1996; Zhang, 1996)
absence of the cylinder, the preferred directions of all the HD (see Chapter 14).
cells were essentially unaltered. Although the cylinder-absent,
background-visible condition was always performed second, 11.9.3 Head Direction Cells Can
the consistency of the responses provides further evidence also Be Controlled by Idiothetic Cues
that the orientation system is preferentially controlled by the
most distal cues available. In a fashion similar to the place cells, some HD cells maintain
In addition to cue card shifts, Taube and colleagues their preferred direction following removal of the cue card.
(1990b) found that changing the shape of the testing enclo- One interpretation of this nding is that the sense of direction
sure from a cylinder to a square or rectangle also caused is continually updated in the absence of environmental cues
changes in the preferred direction. In both right-angled boxes, on the basis of idiothetic or inertial navigation cues includ-
the polarizing cue card occupied the same location relative to ing those from the vestibular and proprioceptive systems.
the environment as it had in the cylinder. When tested in the Consistent with this possibility is the nding that vestibular
rectangular enclosure, most cells (8/10) shifted directions by lesions cause cells to lose their preferred direction (Stackman
large amounts; but on subsequent testing of these cells in and Taube, 1997). It must be stressed, however, that idiothetic
the square enclosure, fewer than half (3/8) had preferred cues alone are almost certainly insufficient to maintain a con-
directions shifted relative to those they had displayed in the stant preferred direction over a long period of nonstereotyped
cylinder. As with card removal, the peak ring and angular locomotion. For instance, in darkness, olfactory signals from
width remained constant. Golob and Taube (1997) found the cylinder and/or oor are probably required in tandem
that only 2 of 11 cells changed their preferred direction by with idiothetic cues for stable orientation, as suggested by the
more than 18 from the standard cylinder to a square box. place cell results of Save et al. (2000) (Section 11.7.7).
As previously, more radical shape changes (i.e., from the cylin- Additional evidence for the role of vestibular cues on pre-
der to triangular or rectangular enclosures) caused large shifts ferred heading direction comes from experiments in which
in directional preference; all 17 cells tested shifting by at least the oor (and therefore the rat) and the black-and-white
36. (One minor caveat: This study tested HD cell activity in striped walls of the recording chamber were rotated together
hippocampal-lesioned animals only.) in tandem or independently (Blair and Sharp, 1996). There
Goodridge et al. (1998) looked at the role of sensory moda- were two rates of rotation, one above and the other below
lities other than vision on the ring of head direction cells. that assumed to be detectable by the vestibular system. These
Whereas a simple auditory cue, such as a localized series of manipulations were carried out both in the light and
clicks or bursts of noise, was ineffective, a localized smell did in the dark in an attempt to dissociate visual motion cues
exert a small but signicant control over the preferred direc- from vestibular effects. In the light, slow rotation of the wall
tion. Rotation of the walls and oor of the testing chamber in and oor together resulted in the comparable rotation of the
520 The Hippocampus Book

preferred heading direction in all cells tested. In contrast, rotation through intermediate directions. In familiar environ-
rapid rotation of the wall or oor individually or together was ments, control of the head direction system by visual stimuli
ignored by most cells, which maintained their preferred direc- can be rapid indeed. Perhaps the slower gures are due to the
tion relative to the laboratory frame. A few cells did show par- time it takes for the animal to notice the changed position of
tial or complete rotations under these conditions, revealing the cue card.
some inuence of idiothetic cues. The picture was different in
the dark. Now both fast and slow rotations changed the pre- 11.9.4 Head Direction Cells Are Found in
ferred heading direction for many but not all of the cells. This Different Anatomically Connected Brain Areas
pattern of results suggests that the preferred direction is con-
trolled by several factors, which include visual information Head-direction cells have been recorded in areas of the brain
from the laboratory itself and visual motion and vestibular in addition to the dorsal presubiculum, where they were
information derived from the animals movements. The latter rst discovered. These areas include the anterior dorsal
become more effective in the absence of the distal visual cues. thalamic nucleus (Taube, 1995a), lateral mammillary nucleus
Disorientating or disruptive rotations of the animal carried (Stackman and Taube, 1998), lateral dorsal thalamic nuclei
out before it is placed in the environment have also been (Mizumori and Williams, 1993), retrosplenial cortex (Chen et
examined. If these are carried out on a routine daily basis for al., 1994; Cho and Sharp, 2001), and striatum (Wiener, 1993).
several weeks prior to recording, HD cells have a less stable Apart from the striatum, these areas are all strongly intercon-
relation to the frame of the laboratory both within and across nected. The HD cells in the lateral mammillary nucleus, ante-
recording sessions. They are also less well controlled by rior thalamus, and dorsal presubiculum have been most
explicit visual cues (Knierim et al., 1995). Even in animals that studied, and this section concentrates on the differences in
had not been disorientated in this way and that displayed their properties. Do they tell us anything about the way in
strong stable control by visual cues, subsequent introduction which the head direction signal is constructed? What is the
of the disorienting procedure prior to each daily recording contribution of each part of the circuit? In addition to the
session caused the HD cells to become progressively uncou- characterization of the properties of the HD cells in each area,
pled from strong visual control. Knierim and colleagues sug- two additional approaches to these questions have been used.
gested that this is evidence that path integration navigation The rst looks at the relative timing of the signal in the vari-
cues predominate when an animal rst enters a new environ- ous areas, and the second asks what the effect of a lesion in
ment and that environmental cues gain control over the head one area is on the activity in another. With the rst approach,
direction system only after a period in which they maintain the best temporal correlation between the HD cell ring and
a stable relation to the path integration system. Stability, it the animals heading direction is computed. The idea is to see
is thought, initially derives from the path integration system. whether cell ring is better related to the animals current
On this argument, the association of the head direction sys- heading or to its heading in the immediate past or future. The
tem to environmental cues would provide corrections for the assumption is that if the best correlated cell ring precedes the
inevitable accumulation of errors to which the path integra- current heading direction, it is more likely to reect some
tion system is subject. This requires that the system, which aspect of neural activity in the motor system that is producing
originally used the path integration system as the basis for the head movements; conversely, if the best correlated cell r-
assigning stability (or a direction) to the visual cue, would ing lags behind the behavior, it is more likely to reect sensory
then be able to use that cue to correct the drift in the system. feedback generated by the movement. The latter assumption
This boot-strap operation appears to require rapid association is not infallible, however, because it is equally possible that
of specic cues (e.g., those provided by a cue card) to an oth- lagged cell ring represents aspects of the neural control of
erwise stable but preexisting framework. The amount of movement shifted by a time delay.
exposure time for a visual cue card to gain control over pre- The second use of time shift correlation analyses is to com-
ferred orientation was studied in passing in earlier studies. pare the temporal relations between the areas in which HD
Goodridge et al. (1998), for instance, found that 8 minutes cells are found. If the cells in one area show a ring pattern
was sufficient for all cells tested, but that as little as 1 minute that is earlier relative to the current heading direction than
sufficed for some cells. Zugaro et al. (2000) found that pre- those of a second area, it is reasonable to suppose that the rst
ferred directions shifted to a new, fairly stable orientation brain area makes computations that come earlier in the cir-
within 15 seconds of a cue card rotation. More recently, an cuit than the second. In a modular system with strictly serial
interesting study designed specically to examine this issue connections between the modules, the relative latencies with
(Zugaro et al., 2003) has shown that reorientation induced by respect to an external event may be taken as an indication of
90 rotation of a peripheral cue card in the dark can occur the functional and perhaps causal connectivity between brain
with a latency on the order of 100 ms after the cue-card shift areas.
becomes visually apparent. The authors reasonably concluded There is a clear consensus that, on average, the ring of HD
that such latencies are more compatible with an abrupt jump cells in the dorsal presubiculum is approximately in synchrony
from one preferred direction to another than with a gradual with the current direction of heading. HD cells in the lateral
 Hippocampal Neurophysiology in the Behaving Animal 521

mammillary nucleus lead those of the anterior dorsal thala- 11.9.5 Dorsal Tegmental Nucleus of Gudden
mus (by about 6070 ms), and the anterior dorsal thalamus Provides Information About the Direction and
leads the dorsal presubiculum (by about 2030 ms) (Blair et Angular Velocity of the Animals Head Rotation
al., 1997; Stackman and Taube, 1998; Taube and Muller, 1998).
Although it is clear that there are average time shifts between The midbrain nucleus called the dorsal tegmental nucleus of
the areas, there is also a wide distribution of shifts within Gudden (DTN) is reciprocally connected to the lateral mam-
any area, resulting in an overlap of shifts between areas. millary nucleus and contains cells whose ring rate correlates
The pattern of temporal correlations suggests a functional with the angular velocity of the head (Bassett and Taube, 2001;
pathway that originates in the lateral mammillary nucleus, Sharp et al., 2001). Integration of the angular velocity over
passes information to the anterior dorsal thalamic nucleus, time would produce a signal proportional to the change in
and thence to the dorsal presubiculum. Remarkably, this is angular direction, which could be used to update the current
part of the classic Papez circuit originally believed to provide heading direction. There appear to be several types of angular
the neural substrate for emotions (see Chapter 2), and there is velocity cell in the DTN: Many show strong correlations with
abundant evidence of its anatomical basis from numerous velocity of movement regardless of the direction of the move-
tract tracing studies (see Chapter 3). It is important to ment, increasing their ring rates in both directions, whereas
remember, however, that there are also substantial connec- others are asymmetrical, increasing their ring rates in one
tions in the opposite direction and, as we shall see in the next direction and decreasing them or ring at a constant rate in
section, good reason to be cautious when putting forward any the other. In addition to these angular velocity cells, the DTN
simple serial theory of the elements of the head direction sys- contains a small number of HD cells, and there is evidence
tem in rodents. that some of the angular velocity cells also show head direc-
A similar story emerges from lesion experiments. Lesions tion or head pitch correlates.
of the anterior dorsal thalamic nucleus abolish the head direc-
tion signal in the dorsal presubiculum (Goodridge and Taube,
1997). In contrast, lesions in the dorsal presubiculum do not
abolish the anterior dorsal thalamic nucleus signal but have 11.10 Interactions Between Hippocampal
more subtle effects on it. The clearest of these is an increase in Place Cells and Head Direction Cells
the amount of time by which anterior dorsal thalamic nucleus
HD cell ring anticipates the animals heading direction. This What is the relation between place cells and HD cells? In the
suggests that a contribution of the feedback from the dorsal original formulation of the cognitive map theory, OKeefe and
presubiculum to the anterior dorsal thalamic nucleus is to Nadel suggested that the map of an environment consisted of
reduce the anticipatory interval expressed in the anterior dor- a set of place representations bound together by information
sal thalamic nucleus leg of the system. A second effect of dor- about the direction and distance between them (OKeefe,
sal presubiculum lesions is to abolish the control of a cue card 1976; OKeefe and Nadel, 1978) (see Section 11.7). In this
over the preferred heading direction of anterior dorsal thala- view a place representation could be activated either by direct
mic nucleus cells (Goodridge and Taube, 1997) and to reduce sensory information impinging on the animal when it occu-
the consistency and stability of the preferred heading direc- pied that place or by activation of a different place representa-
tion between recording sessions. tion together with the appropriate distance and direction
Blair and colleagues (1999) investigated the effects of bilat- inputs between that place and the target place. The theory was
eral or unilateral lesions of the lateral mammillary nuclei on extended in 1991 (OKeefe, 1991a,b) by the suggestion that
the properties of HD cells in the anterior thalamus. Following place cells were dependent on the input from two or more
unilateral lesions, thalamic cells still had preferred directions, HD cells for directional information and that rotation of the
but they sometimes differed from prelesion ones, the peak r- HD system relative to the environmental frame would pro-
ing rates were often reduced, and the turning curves were duce the rotation of place elds seen in these experiments.
broader. In general, the HD cell ring properties shifted to McNaughton and colleagues (1996) suggested that the place
resemble those in the mammillary bodies. Following bilateral cell ring eld was determined by the distance to a single
lesions, no directional cells could be found in the anterior object in a specic direction and that this directional signal
thalamus. was provided by the HD system. Evidence in favor of the idea
The overall pattern of changes following lesions of vari- that the place system is dependent on the HD system comes
ous nuclei support the notion that directional signals travel from the following.
from the hypothalamus through the thalamus to the cortex. Both place and HD cells respond similarly to rotation and
Furthermore, they are consistent with the idea that the dorsal removal of polarizing cues, such as the cue card in the stan-
presubiculum is the site at which visual sensory information dard cylinder (Muller and Kubie, 1987; Taube et al., 1990b;
gains access to the HD system, and the lateral mammillary Cressant et al., 1997; Zugaro et al., 2001). Vestibular lesions
nucleus/anterior dorsal thalamic nucleus pathway is the source abolish anterior thalamic directional ring and location-
of path integration control based on vestibular information. specic ring in hippocampal place cells (Stackman and
522 The Hippocampus Book

Session 1 Session 2 Session 3 Session 4

Place cell 1

Place cell 2

Head direction cell

Figure 1115. Place and head direction cells recorded simultane- same time, which also rotates. Both place elds and preferred direc-
ously maintain the same angular orientation despite both rotating tion of the HD cell are stable before (sessions one and two) and
relative to the environment. Top two rows show the place elds of after (sessions three and four) the intervention. Note that the elds
two cells that rotate by about 135 following disorientation of the of both place cells and the HD cell maintain a constant angular
animal by gentle spinning between sessions two and three. Bottom relation to each other. (Source: After Knierim et al., 1995, with
row shows the eld of a head-direction (HD) cell recorded at the permission.)

Taube, 1997; Stackman et al., 2002; Russell et al., 2003). In the variability was also reliably seen in anterior thalamus-lesioned
one experiment (Knierim et al., 1995) in which HD cells and animals, the effect was mild in these cases. Similarly, removing
place cells were recorded simultaneously, they stayed in regis- the cue card caused large shifts in the place elds angular
ter even under circumstances where the control by environ- location in the dorsal presubicular animals but minor shifts in
mental stimuli was lost: Rotating the animal prior to the the anterior thalamic lesioned animals.
recording session often led to unpredictable rotation of both In contrast to the large effect that dorsal presubicu-
the preferred heading direction of the HD cell and the angu- lar lesions have on hippocampal place cells, there is little evi-
lar orientation of the place cell, but they maintained their dence that hippocampal lesions have a major effect on the
xed relation to each other (Fig. 1115). basic ring characteristics of HD cells. Golob and Taube
Lesions to the anterior thalamus or the dorsal presubicu- (1997) reported that about the usual number of HD cells
lum signicantly increased the directionality of place elds were recorded in the anterior thalamic nuclei and dorsal pre-
recorded in the cylinder in comparison with control animals subiculum following hippocampal lesions, and that they were
(Calton et al., 2003). Recall that in normal rats the place elds under control of the visual cue card to the same extent.
are essentially nondirectional in this testing apparatus (Muller Furthermore, changes in the preferred ring direction in novel
et al., 1994) (see Section 11.7). Such lesions also reduce the enclosures of different shapes were broadly similar to previous
spatial coherence of the place elds. The relation between the studies of nonlesioned animals. In a follow-up experiment, the
standard cue card and the place elds depends on the lesion effect of combined lesions of the hippocampus and overlying
locus. Note that in intact animals, as discussed above, rotation neocortex on the control of the HD cells by idiothetic path
of the card causes the preferred direction of most HD cells integration signals was examined (Golob and Taube, 1999).
to reorient rapidly so as to maintain a xed relation to the The preferred head direction was monitored as the animals
card. Calton et al. (2003) found that in animals with lesions moved from a familiar environment to an unfamiliar one. In
of the dorsal presubiculum the angular locations of hip- normal animals, the self-movement cues provided sufficient
pocampal place elds were not controlled by the cue card but information to maintain the heading direction in the new
shifted unpredictably from trial to trial. Although intertrial environment consonant with that in the old familiar one, con-
 Hippocampal Neurophysiology in the Behaving Animal 523

sistent with previous work (Taube and Burton, 1995). HD cells than the processing and storage of specically spatial infor-
in lesioned animals, however, were not able to do this and, fur- mation. We return to this question in Chapter 13 on lesion
thermore, took up to 4 minutes to reach a stable preferred ori- results, as a successful theory must account for the data from
entation in the new environment. On the other hand, a cue several experimental domains. Suffice it to mention here that
card in the new environment was capable of establishing con- nonspatial responses in hippocampal units do not per se argue
trol over the HD preferred orientation, showing that there was against a spatial function for the hippocampus. A spatial sys-
no decit in this part of the system. The primary inuence of tem would need to incorporate information about the loca-
the hippocampus on the head direction system appears to be tions of objects, rewards, and dangers as well as using
to maintain a consistent preferred direction in the HD cells as nonspatial information in the construction of place represen-
an animal moves between different enclosures in the same lab- tations.
oratory, presumably on the basis of path integration signals or The data on nonspatial unit responses fall into two pri-
context information about the laboratory. The only caveat here mary classes: those suggesting a role in nonspatial sensory
is that lesions of the overlying neocortex resulted in effects that processing and those showing a correlation with some aspect
were similar but of lesser magnitude, and it is not possible of a nonspatial learning process. They are discussed in sepa-
therefore to rule out a role for the neocortex or a more general, rate sections.
nonspecic effect of the lesions.
As we shall see in Chapter 13, there is evidence that hip- 11.11.1 Hippocampal Cells Have Been
pocampus-lesioned rats can learn an allocentric spatial mem- Implicated in the Processing of Nonspatial
ory task in which they are required to go to a location dened Sensory Information
by its distance from a single object in a specic environmental
direction. This capability appears to depend on the HD system Studies by McLean, Ranck, and OKeefe during the 1960s and
but not, because it depends on a single vector, on the hip- early 1970s looked for, but could not nd, selective sensory
pocampal system. inputs to the hippocampal complex-spike cells. On the other
In summary, the evidence strongly suggests that the place hand, Vinogradova (1977) and colleagues reported a nonspe-
system relies on the HD system to provide it with the direc- cic effect of sensory stimulation in the awake rabbit; how-
tional basis of a xed framework, which acts as the scaffolding ever, it appears likely from the ring rates of her cells, the
for the representation of an environment. Two HD cell char- absence of complex spikes, and subsequent work on the cor-
acteristics are crucial to this: First, the angular distance relates of theta cells in the rabbit (Sinclair et al., 1982) that
between HD cells remains constant; and second, the vectors many of her cells were theta cells. As pointed out above, theta
created by the signaling of each HD cell ring in different por- cells in the rabbit increase their ring during theta EEG
tions of an environment are parallel. Whereas environmental episodes, and in the rabbit these episodes occur during non-
changes can cause the distance between different place elds movement arousal as well as during movement.
to shift relative to each other, the HD cells appear to be rigidly Ranck (1973) and OKeefe (1976) studied the role of
xed relative to each other, and it is only the relation between sensory inputs in the freely moving rat and, aside from a
the total constellation of cells and the environment that can be few olfactory responses reported by OKeefe, neither found
altered. Rotation of the HD system rotates the entire place sys- much evidence for these inputs. However, some of the unit
tem. The most likely route for the inuence of the HD system responses to the cues used in more recent learning experi-
on the hippocampal place cells is via the medial entorhinal ments could be interpreted as unlearned responses to the
cortical grid cells whose orientation relative to the environ- stimuli themselves. For example, Wood et al. (1999a) reported
ment is also controlled by distal cues (see Section 11.7.7). The that 8% of cells in the rat hippocampus responded to the
contribution of the hippocampal place system to the HD sys- olfactory cues in an olfactory recognition task. Likewise,
tem is less clear. It may be the origin of information about the Tamura et al. (1992) reported that 10% of units in the hip-
wider context that allows the animal to maintain a constant pocampus and surrounding regions of the monkey responded
heading direction relative to distant cues as it moves from a to specic objects. Creutzfeldt and colleagues (Vidyasagar et
familiar part of a territory to an unfamiliar one. al., 1991; Salzmann et al., 1993) found that as many as 38% of
hippocampal and parahippocampal units in the monkey
changed activity in response to arousing stimuli such as the
presentation of a raisin or the sight of the experimenter. It is
11.11 Hippocampal Complex-spike not clear whether these are perceptual, learned, or general
Cells Have Been Implicated in Nonspatial arousal responses. These studies are examined in more detail
Perception and Learning in the section on nonspatial learning, below.
Perceptual responses to stimuli have also been reported
In addition to the widely reported spatial and movement cor- in studies on the human hippocampal formation. In one
relates of hippocampal cells, there have been reports of other experiment (Kreiman et al., 2000b), subjects viewed a series
correlates from several laboratories. They have suggested to of pictures drawn from nine classes: household objects,
some that the functions of the hippocampus are more general unknown faces portraying different emotions, famous faces,
524 The Hippocampus Book

spatial layouts including the facades of houses and natural (CR) such as an eyeblink or suppressed heart rate. In a more
scenes, animals, drawings of famous people or cartoon char- complicated differential conditioning paradigm, two stimuli
acters, cars, food items, and abstract drawings. Overall, 14% of (CS and CS
) are interspersed randomly, one followed by
units from medial temporal lobe sites showed a visual the US and the other not. In this latter paradigm, the animal
response to one or more stimuli. Interestingly, of the 32 hip- learns to discriminate between the stimuli, coming eventually
pocampal units responsive to visual stimuli, 29% responded to respond to the rst but not the second. Typically, during the
selectively to spatial layouts, 12% to famous faces, and less early stages of learning, both stimuli elicit responses and only
than 10% to stimuli drawn from the other categories. subsequently does the animal learn to inhibit its response to
Therefore, although hippocampal cells responded to famous the CS
. In addition, there is a learned arousal response that
faces and other stimuli, pictures portraying the layout of envi- is not specic to the hippocampus but can be reected in the
ronments or buildings were by far the most effective of these activity of hippocampal cells.
visual stimuli. The human hippocampus, like that of the Delacour (1984) trained rats on an aversive differential
rodent and nonhuman primate, appears to prefer information conditioning paradigm. During slow-wave sleep two different
about spaces over faces and objects. tones served as the conditioning stimuli and mild electric
shock to the neck as the unconditioned stimulus. He recorded
11.11.2 Hippocampal Unit Activity May increased neck muscle tone as the conditioned response (CR)
Show Correlations with Different Aspects and the cortical EEG as an independent measure of arousal.
of Nonspatial Learning Tasks During the early stages of training, both the CS and CS

elicited a muscle activation CR that was associated with corti-

In pioneering studies conducted during the 1960s, Jim Olds cal arousal; during the later stages, there was a decrease in the
and Menahem Segal recorded unit activity in the hippocam- cortical arousal to both stimuli as well as the development of
pus during classical conditioning to a tone. They saw an a differential neck muscle response, with the response to the
increase in ring to the tone following conditioning. These CS remaining high and that to the CS
steadily declined to
studies were the rst of a number that have looked at the baseline (Fig. 1116A). The initial increase in the EMG acti-
responses of hippocampal neurons during or after nonspatial vation to both stimuli was thought to reect general arousal,
learning, primarily using conditioning paradigms or discrim- whereas the later differentiation between them reected
ination learning. Some of these studies were carried out using movement to the CS. Unit responses to the CSs of single hip-
multiunit recording techniques, which, like EEG, register only pocampal complex-spike cells, hippocampal theta cells, and
the collective properties of a group of neurons. As we have dentate granule cells were monitored. During the early phase
seen, the spatial coding of environmental features in complex- of training, all cell types increased activity to both stimuli in
spike cells is carried out at the single-cell level, and group parallel with EMG activation, the complex-spike cells some-
recordings would not reveal this spatial code. We therefore what faster than the theta cells. During the latter phases, how-
concentrate primarily on studies that have recorded single- ever, they acted differently: The hippocampal complex-spike
unit activity and refer to the multiunit studies only when they cells ceased ring to both positive and negative conditioning
add something different. stimuli, paralleling the decrease in cortical arousal. Like the
Two types of behavioral paradigm have often been used cortical arousal, at no time during training did they differen-
to look at nonspatial learning: (1) classical conditioning in tiate between the two CSs (Fig. 1116B). In contrast, the
which the animals response has no effect on the stimulus- responses of the hippocampal theta and dentate granule cells
reward contingencies and (2) signaled operant conditioning followed the pattern of the EMG response and began to dif-
in which it does. A good example of the former are experi- ferentiate between the stimuli as the behavioral differentiation
ments on classical conditioning of the rabbit nictitating mem- developed (Fig. 1116C). As a control for the specicity of
brane (NM) response; a good example of the latter would be these responses, Delacour recorded from thalamic cells as well
a nose poke, go/no-go instrumental task taught to a rat. In the and found that they also divided into two classes: Those in the
sections that follow, we turn rst to classical conditioning centre median acted similiar to the complex spike cells,
experiments in the rat and rabbit and then examine the some- whereas those recorded in the dorsomedial nucleus resembled
what more extensive literature on operant conditioning in rats the theta cells. On the basis of these results, Delacour argued
and monkeys. rather convincingly that the response of the complex-spike
cells was a reection of general arousal, whereas the pattern
11.11.3 Hippocampal Unit Activity of activity of the theta/granule cells was a reection of the
During Aversive Classical Conditioning learned differential increase in neck muscle activation to
the two conditioned stimuli. The latter response may reect
Hippocampal unit activity has been recorded during aversive the movement correlates of the hippocampal theta cells. He
classical conditioning: a single conditioned stimulus (CS) fol- further argued that these were not specic responses but were
lowed after a short period by an unconditioned stimulus (US) representative of more general responses in arousal and move-
consisting of a shock or other noxious stimulus. After suffi- ment systems that could be recorded elsewhere in the brain as
cient pairings, the CS comes to elicit a conditioned response well. Laroche and Bloch (Bloch and Laroche, 1981; Laroche et
 Hippocampal Neurophysiology in the Behaving Animal 525

effect was greater following conditioning to the context of the

box than when there was an explicit auditory stimulus. These
70
results may give some insight into the role of the hippocam-
60
pus in this behavioral paradigm, at least in the rat. It is possi-
% emg CR

50 A
40
ble that, in addition to the conditioning of arousal responses,
30
some conditioned responses in hippocampal units in immo-
3 bile animals reect the fact that during conditioning some
complex spike unit activity

20
10 cells shift their place elds to the animals location and begin
0 2 B to respond to the CS in that location. Against this interpreta-
tion is the fact that restraining the rat abolishes the ring in
the place eld (Foster et al., 1989). However the possibility
1
3 remains that restraint may not abolish the gating function on
conditioned stimuli. We return to this possibility following a
0
discussion of unit activity during the nictitating conditioning
theta unit activity

2 C paradigm.

1 11.11.4 Nictitating Membrane Conditioning

in the Rabbit: Role of Theta
0 2 3 4 5 6
h 1
Conditioning of the rabbit nictitating membrane (NM) has
blocks of trials been used successfully as a model system for studying the neu-
Figure 1116. Two types of response in hippocampal unit activity ral bases of learning and memory since it was introduced by
during classic conditioning of the arousal response to positive Richard Thompson in 1976 (Thompson, 1976). In this para-
and negative tones. A. Neck electromyography (EMG) increases digm, the CS (e.g., a 6 KHz tone) is paired with a US, which
during the early trials and then begins to differentiate between can be a puff of air to the eye or a small electric shock to the
the CS (black stars) and CS (open stars) by the last two blocks
orbit. By varying the temporal relation between the CS and
of trials (arrows). B. Principal hippocampal neurons increase their
US, one can investigate various types of classical conditioning.
ring rate during the early part of the trial in parallel with the
EMG response but then fall off to both stimuli. C. Hippocampal In delay conditioning, the US overlaps the last portion of the
theta cells and dentate granule cells also show an initial increase CS, so both terminate together. In trace conditioning, the CS
but then differentiate between the CSs in a manner that parallels ends before the US begins, and there is a temporal gap
the EMG activity. Solid symbols are CS responses, and open between the two. As we saw with classical conditioning in the
symbols are CS
responses in all panels. h, habituation trials. rat (Delacour, 1984), there is an important role for arousal in
(Source: Delacour, 1984.) this form of learning. There is clear evidence that theta activ-
ity gets conditioned to various aspects of the task during clas-
sical conditioning experiments in the rabbit. Powell and
al., 1983) found a similar increase to a CS in the dentate mul- Joseph (1974) conditioned the corneoretinal potential in the
tiple unit response after conditioning. rabbit using a mild electric shock to the eye as the US. This
Moita and colleagues (2003, 2004) have carried out an was preceded by a CS. A second stimulus was not followed by
important classical conditioning experiment in the rat that the US and served as a CS
. During the early stages of learn-
may throw further light on the role of hippocampal cells in ing, before differential responses to these two stimuli had been
these tasks. They recorded the responsiveness of complex- established, there was a high incidence of theta to both CSs;
spike and theta cells to an auditory stimulus before and after after differential conditioning, when the CS but not the CS

delay classical conditioning in which the stimulus was paired consistently elicited the US, a considerably higher amount of
with a brief electric shock to the eyelid. No complex-spike theta occurred to the CS. During this second phase of condi-
cells responded with a short latency to the tone prior to con- tioning there was a differential response of the neck EMG to
ditioning, but many did so after conditioning. However, they the CS. This pattern of responses is similar to that seen in the
did so only if the animal was located in the place eld of the rat (see above) and indicates that the early theta activity was
cell during CS presentation (Fig. 1117). As Figure 1117 related to general arousal, and the later theta activity was
shows, the same cell might give a strong, brisk response to the related, at least in part, to the motor response. Experiments
auditory stimulus when the animal was in the place eld but that manipulate the animals arousal have shown that it has a
respond little or not at all to the same stimulus when the ani- strong effect on learning rates and that this effect may be
mal was outside the eld. The animals location in the place mediated in part via its effect on the baseline rate of theta
eld appeared to gate the response to the CS rather than sim- activity. Berry and colleagues (Berry, 1989) showed that the
ply summate with it. Moita and colleagues (2004) also found pretraining background amount of hippocampal theta was a
that place cells can develop a place eld or can shift eld loca- good predictor of NM conditioning rates and that this vari-
tion in the conditioning box following conditioning, and this able was strongly inuenced by the level of arousal. Following
526 The Hippocampus Book

A cell 1 cell 2
out of field out of field
field in field field in field
excluded excluded
24 24

pip response (Hz)

pip response (Hz)

20 20
16 16
12 12
8 8
4 4
0 0

B
2.0 1.0
normalized response
normalized response

1.5 0.8

1.0 0.6
(Z-score)
(Z-score)

0.5 0.4

0 0.2

-0.5 0
Z in Z out

Figure 1117. Response of hippocampal complex spike cells to an immediate zone not considered as part of either. Both cells
auditory CS following classic eyeblink conditioning. The cells showed a strong response to the stimulus inside the eld and no
respond only when the animal is in the place eld. A. Responses in (cell 1) or few (cell 2) spikes outside the eld. B. Population
two individual cells. On the left of each panel, place elds are shown response histograms to stimulus presented inside (dark shading)
with the higher ring rates in the darker colors; on the right of each and outside (light shading) the eld of each cell. Across the popula-
panel are histograms of the unit responses while the animal is inside tion there is a signicant difference between the in-the-eld and
the place eld (black color) and outside the eld (light color). The out-of-eld responses (histogram, right). The x-axes in the unit
middle panels show the portion of the environment included as response histograms are 250 ms in total, which was the period of
part of the eld, the portion considered to be outside the eld, and CS presentation. (Source: Moita et al., 2003.)

mild water deprivation, there was a faster rate of learning. A cells were typically activated by the CS to re a series of theta-
similar pattern was found with trace conditioning (Kim et al., like bursts, which often continued throughout the trial and
1995). in some cases continued beyond the termination of the
trial. Berger and colleagues also reported a third category of
11.11.5 Single-unit Recording cells, silent cells, which constituted 11% of the neurons
in the Hippocampus During Nictitating recorded; they had exceptionally low spontaneous ring rates
Membrane Conditioning of Rabbits ( 0.2/second), were not activated by fornix stimulation, and
did not participate in the conditioned response. As we shall
Several laboratories have recorded the activity of single units see below, this estimate of the percentage of cells in this cate-
and multiple units from the hippocampus of rabbits during gory may be low.
classical conditioning of the NM response. Berger and col- Weiss et al. (1996) recorded single units from rabbit hip-
leagues (1983) recorded single units in the rabbit hippocam- pocampus during trace conditioning of the NM, a version of
pus during simple delay NM conditioning. Pyramidal cells the conditioning task that is sensitive to hippocampal lesions:
were identied by their antidromic activation from fornix Either the CR is abolished, or its timing is altered (see Chapter
stimulation. They comprised the largest proportion of cells 12). They also reported the existence of the same three classes
recorded; and following conditioning, many increased their of cells reported by Berger and colleagues in their delay con-
ring rates during the CS period (Fig. 1118AC). They typi- ditioning experiments; but in other respects the results dif-
cally emitted one or more bursts of spikes during each trial, fered markedly. Weiss et al. found a much larger percentage of
some showing a pattern of activity that closely modeled the cells that did not participate in the conditioning (40% in con-
NM response (Fig. 1118A) whereas others were more selec- trast to Bergers 11%) and relatively few units that were signif-
tive, ring during different time epochs of the trial (Fig. icantly excited (in contrast to inhibited) during the CS or trace
1118B,C). One type of theta cell increased its overall level of period in the conditioned animals in comparison with
activity during the trial (Fig. 1118D,E), whereas another type unpaired controls. Thus, in CA1, 14% of pyramidal cells were
showed an overall decrease in activity (Fig. 1118F,G). Theta excited during the CS period in contrast to 9% of the unpaired
 Hippocampal Neurophysiology in the Behaving Animal 527

Pyramidal cells Theta cells

A
nm

unit D F

cs ucs

Figure 1118. Firing patterns of hippocampal units E G

following nictitating membrane conditioning in the
rabbit. AC. Pyramidal cells. D, E. Excited theta cells.
F, G. Inhibited theta cells. Top trace in each panel is C
the nictitating membrane record; bottom trace is the
histogram of the average unit response. The rst
arrow indicates CS onset and the second arrow UCS
onset, an interval of 250 ms. Note the theta bursting
at approximately 8 Hz following the CS onset in the
theta cells. (Source: Berger et al., 1983.)

controls; and 11% were excited during the trace period in whereas trace conditioning, during which few such unit
contrast to 9% of the controls. In contrast to the ndings of responses are found, does require an intact hippocampus.
Berger and colleagues, the increase in inhibitory responses rel- How might the hippocampus be involved in trace but
ative to the controls was twice as large as the increase in exci- not delay conditioning? There are three distinct but related
tatory responses. McEchron and Disterhoft (1997) obtained possibilities. The rst suggests that the hippocampus pro-
similar results in animals conditioned in a trace paradigm vides the information that the animal is in a frightening place,
when recordings were taken after asymptotic performance the second that the animal is in a place where frightening
had been reached. In addition, they recorded from some ani- events happen, and the third that it provides information
mals during the earlier stages of learning. They found that the about the timing of the unpleasant event. The rst two make
maximal activation in complex spike units occurred on the clear links to the spatial functions of the hippocampus,
trials just prior to the onset of learning; as behavioral condi- whereas the third does not necessarily do so. We deal with
tioning proceeded and the conditioned responses appeared these in turn.
more frequently, these unit responses actually diminished. Our rst two mechanisms relate to ones suggested by
Finally, when looked at on a trial-by-trial basis, there was no Nadel and colleagues (1985). According to them, the spatial
correlation between the unit activity and the occurrence of the functions of the hippocampus might lead to its being involved
conditioned response. This pattern suggests that the hip- in conditioning in two ways: as the substrate for a direct asso-
pocampus may not be involved directly in the generation or ciation between the background cues and the US or less
timing of the motor response. Rather, it may be involved only directly as the basis for an association between the CSUS
indirectly, perhaps playing a role in the creation of a tempo- event and the overall context. Evidence for a hippocampal role
rary behavioral state that precedes the motor learning but is in conditioning to the background cues (the fearful context
a necessary condition for it to occur. Lesions of the hip- hypothesis) comes from experiments showing that animals
pocampus do not generally affect delay conditioning but do with hippocampal lesions do not condition to the context
affect trace conditioning, changing the timing of the condi- (Phillips and LeDoux, 1994; Kim et al., 1995). This could
tioned response in the trace version of the task (for a review come about either because there is an association of fear with
see OKeefe, 1999). Thus, if we compare the effects of hip- the entire testing box or with specic locations in it. Recall
pocampal lesions on the NM conditioning with the results that the ventral hippocampus contains place cells with large
of single-unit recording experiments, we are left with a para- elds that may extend to an entire testing environment, and
dox. Delay conditioning, in which there are a sizable number they could provide the basis for context conditioning. An
of pyramidal cells whose temporal activation prole precedes alternative is that the more localized place elds in the dorsal
and models the CR, does not require the hippocampus, hippocampus could shift following context conditioning so
528 The Hippocampus Book

the overall pattern of place cell ring was different, perhaps gap. Theta oscillations, with their relatively constant period,
signifying a dangerous environment (Moita et al., 2004). could act as a clock over short intervals. During conditioning
The second possibility is that the cells showing conditioned of the rabbit NM response, this clock signal can be reset by the
responses are place cells whose elds coincide with the loca- CS and could therefore be used to predict the occurrence of
tion of the testing box, and the responses to the CS are signal- the US. Even here there might be a secondary role for the spa-
ing the occurrence of an event in that place. As we have seen tial functions of the hippocampus. Recall that a-theta in the
above in the experiments of Moita and colleagues (2003), rabbit can be driven by arousing stimuli (e.g., the CS) much
complex-spike cells in the rat, which are normally not respon- more easily than in the rat (see Section 11.3). That required
sive to auditory stimuli, begin to respond to the CS following level of arousal might be based on conditioning to the appa-
delay conditioning but only if the animal occupies the ring ratus and other nonspatial cues, but it might also be driven by
eld of that cell. If the same changes are occurring in the rab- hippocampus-mediated fear conditioning to the background
bit, this might explain the increased responsiveness of cells context as well. Under circumstances where the arousal level is
with elds in the conditioning location; moreover, the elds of appropriate, pyramidal cells may be able to count theta cycles
some cells might shift to that location following conditioning. and use this clock signal to predict the timing of the US. This
With this interpretation, the increased responsiveness might timing signal would then be available to brain stem regions to
reect the fact that the hippocampus is now signaling that the control the occurrence of the CR. In the absence of this signal
animal is in a dangerous location and, furthermore, that the (e.g., following hippocampal damage), conditioning would
auditory cue predicts the onset of danger in that place. More still occur but the timing of the CR would be controlled by
evidence for a role for the hippocampus in contextual gating other factors. Some support for this view comes from an
of CSUS events comes from a context shift experiment. experiment on trace conditioning of the heart rate in rabbits
Although lesions of the hippocampu have no obvious effect (McEchron et al., 2003). Pairing a CS with an aversive US
on simple delay conditioning, they do affect the role the back- results in slowing of the heart rate in anticipation of the US;
ground cues (i.e., the room in which conditioning takes place) this can be conditioned with trace intervals as long as 20 sec-
play in that conditioning. Penick and Solomon (1991) showed onds. McEchron and colleagues used trace intervals of 10 and
that simple delay conditioning was disrupted in normal rab- 20 seconds in two groups of animals and found that one-
bits when the animal was moved into a new room after condi- fourth of complex-spike cells showed a burst of activity timed
tioning was completed but hippocampal-lesioned animals to coincide with the end of the trace interval. The effect
were not affected. This result ts nicely with the idea that the was weak on any given trial; but when summed over trials
hippocampus provides spatial contextual information that there was a discernible response. The problem for the theta-
gates the conditioning of fear to the CS. This information counting hypothesis is that 10 seconds is a long time to count
might not be necessary for simple delay conditioning, but it theta cycles. For further discussion of the contributions that
might be essential for trace conditioning. contexts make to learning, see Chapter 13.
The third possibility is that the hippocampus provides
information about the timing of the US. The interpretation of 11.11.6 Hippocampal Unit Recording
the altered timing response in trace conditioning following During Operant Tasks
dorsal hippocampal lesions is unclear. Why, for that matter,
does the conditioned unit response occur just prior to the During operant conditioning tasks, the animal must emit a
unconditioned stimulus in normals? The original rationale of response to gain a reward or avoid punishment. Often the
the NM learning paradigm was to rule out any instrumental availability of reinforcement is signaled by a sensory stimu-
contribution to the learning and in particular the possibility lus such as a tone. Christian and Deadwyler (1986) recorded
that the conditioned response would protect the eye from the from complex-spike and theta cells during an appetitive oper-
US or otherwise attenuate its impact (Thompson, 1976). If, on ant conditioning task. They trained thirsty rats to poke their
the other hand, one accepts that the conditioned response in noses into a small antechamber in the wall of a box to receive
these conditioning paradigms is a reection of the prediction a water reward. For some animals, the availability of reward
of the US, the timing of the CS is important; and the lesion was signaled by a tone, and no sensory discrimination was
results suggest that the hippocampus is involved in setting up required; for others, a differential CS/CS
procedure was
the conditions under which the short-term prediction of used. Following successful conditioning to the single tone
stimuli can occur. Both Rawlins (1985) and Wallenstein and stimulus (Fig. 1119A), theta cells showed a consistent in-
colleagues (1998) have suggested that the hippocampus is crease in ring rate during the 200 ms following tone onset
needed to bridge a temporal gap between two stimuli to be (Fig. 1119A, theta). In contrast, no change from the back-
associated. This role would be particularly important in para- ground rate was seen in the complex spike cells (Fig. 1119A,
digms such as trace conditioning, where there is a CSUS complex spike). During two-tone differential conditioning,
interval and normal animals generate a CR just prior to the the theta cells showed an increase to both stimuli with a
US. What might the underlying mechanism be? One possibil- greater increase to the CS (Fig. 1119B, theta). In contrast,
ity is that hippocampal a-theta is being used as a timing the pyramidal cells registered a marginal but nonsignicant
mechanism to allow a short-term signal to bridge the CSUS change to the CS (Fig. 1119B, complex-spike). Recordings
 Hippocampal Neurophysiology in the Behaving Animal 529

CS+ CS-
30 20 10
A theta cell B 40
50
theta cell
0
**
** **
60 **
50
** **
** ** *
50
50
40
** **
** ** 40
40
30
30
20
30
mean firing rate (spikes s-1)

20

mean firing rate (spikes s-1)

20
10 10

0
10 0
-200 200 400 600 800 ms -200 200 400 600 800 ms

0 tone onset tone onset

-200 200 600 800 ms 3
-200 200 400 600 800 ms
3 complex spike cell complex spike cell *

* 3
3 2
2
2 1
0
2
1
1

1
0 -200 200 400 600 8000ms
-200 200 400 600 800 ms
-200 200 400 600 800 ms
0 tone onset tone onset

Figure 1119. Theta and complex spike cell responses following paradigm (B). Single asterisks indicate a signicant difference from
nose-poke conditioning to tones. Theta cells (top) show increased pre-tone ring rates at the 0.05 level of signicance and double
ring above the pre-tone background rate in response to the tone in asterisks at the 0.01 level of signicance. The apparent increase in
both simple (A) and compound (B) conditioning paradigms. In the ring rate in the complex spike cells 800 ms following the tone in
latter, the unit response to the CS is more marked than to the CS
. both A and B is an artifact from the reward dispenser. (Source:
Complex spike cells (bottom) do not respond in either the simple Christian and Deadwyler, 1986.)
conditioning paradigm (A) or the complex compound conditioning

taken from animals while they acquired the task showed that pyramidal cells. The initial short-latency response of hip-
the changes in theta cell ring occurred in parallel to acquisi- pocampal interneurons and granule cells to either a CS or
tion of the conditioned EMG response and disappeared with a CS
is presumably activation reecting an arousal input
subsequent extinction. Again no changes were seen in com- from the brain stem; at about the same time, a small percent-
plex-spike cells during the course of acquisition. age of pyramidal cells, in some studies, also show an arousal
In a subsequent experiment from the same laboratory, response. Depending on the stage and type of training, this
Foster and colleagues (1987) did nd a small but signicant can either be an inhibitory or a weak excitatory response. The
increase in ring in their population of complex-spike cells to later phase ( 200 ms after CS onset) of the unit activity is
both conditioned stimuli but still no differential activity to the related to the behavioral response. For the theta and granule
CS. A more detailed look at the differential response that did cells there is prolonged activation continuing throughout the
occur in the theta cell group revealed that the difference in CS period but no such response to the CS
. The pyramidal
response to the two stimuli was due to an initial increase in cells do not participate in this second longer-latency phase.
ring to both stimuli, which peaked at about 80 to 100 ms Eichenbaum and his colleagues have studied the behav-
after tone onset and was then maintained throughout the ioral correlates of hippocampal cells in various olfactory
1-second tone period to the CS but fell back to baseline in recognition and discrimination tasks. They recorded two
response to the CS
. major behavioral correlates: Some cells red when the ani-
The simplest explanation for the pattern of results mal sniffed at the odor cues, whereas others changed their r-
observed in these studies is that there are two independent ing rates during various stages of the approach to the cues
factors operating during conditioning: arousal and prepara- or to the goal. In a successive go/no-go discrimination task
tion for the motor response. Both contribute to the ring of (Eichenbaum et al., 1987), the rat was presented with one
theta cells, but only one of them, arousal, inuences the odor of a pair and had to poke its nose into the single-odor
530 The Hippocampus Book

port in response to the CS but not to the CS

. Water reward water nearby was signaled by a mismatch between the current
was available on the other side of the testing box, requiring the and previous stimulus. Again, cells that had peak ring rates
animal to shuttle continuously between opposite sides of the at different points in the task were found, including 12% that
box. Three behavioral correlates of unit response were identi- peaked during the cue sampling period. Somewhat disap-
ed: cells that red when the animal was sniffing at the odor pointingly however, only eight cells (3%) had ring patterns
(cue sampling 15%); cells that red when the animal that could be unambiguously classied as signaling a mis-
approached the sniff port or ran from the sniff port to the match between successive stimuli. As we shall see below in
water cup at the other end of the box (reward/port approach the section on primate studies of delayed nonmatch to
60%) (Fig. 1120A); and theta cells (10%). Many of the cue sample, there are also very few cells in these studies that
sampling units had ring patterns that were maximally syn- responded selectively to the familiarity or unfamiliarity of a
chronized to the onset of cue sniffing. No evidence was found stimulus.
for cells that preferred one odor over others. Almost all cue In a forced-choice discrimination trial, Wiener and col-
sampling cells red more to the positive stimulus that signaled leagues (1989) used simultaneous rather than successive odor
the availability of water reward in another part of the envi- discrimination. Both the CS and CS
were presented at the
ronment. In addition, there was evidence that in the trials that same time from two adjacent odor ports. In total, 22% of
followed a CS
trial the cells gave a larger response than in complex-spike cells had increased activity during cue sam-
those that followed a CS. A follow-up study (Otto and pling (Fig. 1120A, center panel). However, closer examina-
Eichenbaum, 1992) looked at complex-spike cell ring during tion of these data showed that only 13% of cue-sampling cells
a continuous recognition olfactory memory task in which any discriminated between odors irrespective of location, and
one of 32 odors could be presented, and the availability of 44% took the stimulus location into account. Figure 1120B

Figure 1120. Hippocampal cell ring correlates with various cell that responded maximally when the animal sniffed at a particu-
aspects of an odor discrimination task. A. Firing rate histograms of lar odor (odor 1) on the left side of the odor sampling port and a
cells that red best when the animal approached the odor sampling second (different) odor (odor 2) on the right side. Top trace: best
port (top trace), sniffed at the odors (middle trace), and approached response to odor 1 on the left and odor 2 on the right; second trace:
the reward cup (bottom trace). Arrows in the top two traces indi- same odors from the opposite ports gives a lower response; third
cate initiation of the behavior; arrow in the bottom trace indicates a and fourth traces: different odors are also less effective. Dotted line
nose-poke into the reward cup. B. Firing pattern of an odor/spatial marks initiation of the trials. (Source: Wiener et al., 1989.)

A B
port approach

odor 1/odor 2

odor 2 /odor 1
cue sampling

odor 3/odor 4

cup approach

1 sec
odor 4/odor 3
 Hippocampal Neurophysiology in the Behaving Animal 531

shows an example of a cell that had a strong spatial compo- the maze. OKeefe and Recce (1993) used a linear track with
nent. It red best to a particular odor pair when one was pre- identical food reward at both ends and found that some cells
sented on the left and the other on the right. Somewhat more re in one direction and other cells red in the opposite direc-
surprising, 44% of cup-approach cells signaling approach to tion. Even if one were tempted to describe these cells as goal
the reward cup on the opposite side of the box also had approach cells, it is not the approach to the food or the food
responses that depended on the position of the prior nose container per se that is being signaled but their locations at
poke or the odor/position interaction. It is difficult to inter- different ends of the track. It does not seem warranted to
pret these data, but one possibility is that the rats may have describe the cells reported in the experiments of Eichenbaum
been turning in different directions away from the odor port and colleagues as cue-sampling or goal-approach cells in the
as they headed toward the reward. If this is true, the cup absence of similiar manipulations.
approach response in this ostensibly nonspatial task may Two experiments have provided evidence about the effi-
depend on whether the animal turns toward the left or right cacy of cues and goals in isolation from their location (Wiebe
on its exit from the sniff port and thus passes through a place and Staubli, 1999; Wood et al., 1999). Wood et al. recorded
eld on one side of the sniff port as it turns in one direction from hippocampal cells during a variant of a continuous
but not in the other. Alternatively, this may be an example of olfactory recognition task in which they tried to dissociate
the dependence of some place cells ring on the prior turn location, odor, and the match/mismatch aspects of the task.
taken by the animal before entry to the place eld (see Section Rats were trained on an open platform to approach a small
11.7.3). One can conclude that a small number of complex- cup containing sand scented with one of nine odors. In each
spike cells respond to the CS in a differential go/no-go dis- trial the cup was placed in one of nine locations. If a cup had
crimination but that, in general, few hippocampal cells code an odor different from that of the previous one, it contained
for the specic odor quality, and that many cells take the loca- food for which the animal could dig (nonmatch); if the odor
tion of the odor into account. was the same, there was no food (match), and digging went
On the basis of these and other results (see Chapter 13), unrewarded. Cells were recorded during the 1 second prior to
Eichenbaum and his colleagues proposed that the hip- arrival at each cup. An analysis of variance showed that of the
pocampus stores information about a wide range of rela- total 127 cells, 8% responded to odor in the absence of any
tions, both nonspatial and spatial. Unit responses to odors other correlate, 11% solely to location, and 10% solely to the
and approaches to the odor port or reward dispenser might match/mismatch aspect of the task. The remainder of the
be taken as evidence of nonspatial representations and responsive cells took interactions between these variables into
unit responses signaling the identity between two successive account. In all, 20% of cells had nonspatial correlates, and
stimuli as evidence of one type of nonspatial relationship: the 32% took location into account. (Another 20% changed their
identity relation between two stimuli experienced at different ring rate as the animal approached any of the cups. The lat-
times. ter may simply be movement- or speed-relatedbecause this
As described earlier in the chapter (see Section 11.7.1), was not measured, it is not discussed further.) These results
many place cells re more when the animal sniffs at a location are notable as a higher percentage of cells with a match/mis-
when the stimulus in that location has been altered in some match correlate irrespective of location was found than in the
way, and these results from odor discrimination paradigms previous reports from the same laboratory (see above).
may provide additional information about the conditions Furthermore, there are fewer purely spatially coded cells than
under which this mismatch response occurs. To identify cell are usually found. This may be because the experimental
responses as being due to the odor cue independently of its design required the animal to approach each cup from a dif-
location, it is necessary to present the same odor in two or ferent angle during different trials and therefore via different
more locations. The demonstration that place elds can be locations; this might lead to a signicant underestimation of
rather large, encompassing a large part of the testing box (see the number of place cells. Approaching a cup from the north
Section 11.7.2), means that it may be necessary to test the would not force the animal to cross the same region of space
same odors in two different laboratories as well as in two test- as approaching the same cups from the south. It is also not
ing boxes in the same laboratory before concluding that hip- clear how much of the difference between cellular match and
pocampal cells do not have a spatial correlate. Similarly, the nonmatch responses can be attributed to the different behav-
cells that re selectively during port approach and cup iors of digging and turning away from the cup that were used
approach have much in common with place cells recorded on as the response measures, rather than the relational judgment
linear tracks. Recall that under conditions that constrain the itself. Nevertheless, taken at face value, these results provide
animal to move along narrow pathways, the place cells are evidence for the more general relational theory. Signicantly,
unidirectional. To distinguish a goal-oriented response from a this task has been shown by the authors of the Wood et al.
motivationally neutral place response, it is necessary to have study themselves (Dudchenko et al., 2000) not to be disrupted
two identical goals in the environment or two different goals by selective hippocampal damage.
that can be interchanged. OKeefe (1976) reported that inter- Wiebe and Straubli (1999) used a Y maze task that forced
changing the water and food at the end of a three-arm maze the animal to traverse the same locations on the approaches to
did not change the location of complex-spike ring elds on the goals and came to a conclusion different from that of
532 The Hippocampus Book

Wood and colleagues. Animals were trained on a delayed non- weak representation of odor or performance in the test arms,
match-to-sample odor task with each trial comprised of three there was a strong representation of the animals location. Of
discrete periods: a sampling period during which the animal the 571 cells with behavioral correlates in the test arms either
sniffed one of two odors in the start arm, a delay period dur- alone or as an interaction with another correlate, 70% had a
ing which it was conned to the start arm and had to remem- spatial correlate, whereas only 27% had an odor correlate and
ber the recently smelled odor, and a test period during which 21% a performance correlate. More signicantly, 37% of the
it was required to choose one of the other two arms contain- cells with spatial correlates were pure place cells with no other
ing the odor it had not recently experienced. During the rst correlates, whereas most of the cells with olfactory and per-
two periods, everything happened within the same arm, and formance correlates also had spatial correlates (92% and 93%,
cellular activity could be correlated only with the odor pre- respectively). It is difficult to escape the conclusion that when
sented on that trial and with subsequent correct or incorrect the testing situation allows them to be identied spatial
performance on the choice part of the trial. During the third responses predominate either as pure place responses or as
(test) period the animal could enter one of two goal arms, and place combined with some other aspect of the task, even
therefore a correlation with location as well as odor and per- in tasks where the spatial component is irrelevant to the
formance could be sought. Overall, most of the 1101 cells solution.
recorded from the dentate gyrus, CA3, and CA1 had signi- A potential criticism of the Wiebe and Straubli experiment
cant correlations with one or more aspects of the task. Some is that they only used two odors, but there might be more than
of the results are shown in Table 111. Several conclusions can one location in each of the test arms. This points up an impor-
be drawn from these results. First, during the sample phase, tant methodological problem in these studies. To equate two
which always took place in the same arm and therefore did not tasks on the number of odors and locations, it is important to
allow the contribution of location to be examined, a small have an accurate measure of these items. It might be thought,
percentage of cells had signicant changes in ring rate to the for example, that the number of place elds in each arm of the
specic odor presented (5%), to whether the subsequent maze is greater than the number of odors, but the odors used
choice would be correct or incorrect (2%), or to the interac- in the tasks are typically compounds and may be distin-
tion between two (4%). None of these ring rate changes, guished on the basis of many different elemental aspects of the
however, carried over into the delay phase of the trial (Table compound. In the absence of more information about the
111, delay) suggesting that the hippocampus does not main- elemental components of locations and smells in any given
tain an active trace of the correct odor choice, as it does of task, the best one can do is try to vary each variable systemat-
location in a comparable spatial memory task (OKeefe and ically to see what effect it has on the results.
Speakman, 1987). Second, in marked contrast to what was
seen in the sample phase, during the test phase only a tiny per- 11.11.7 Comparison of Hippocampal
centage of cells red to the odor, to the correctness of the Cells During Operant Conditioning
choice, or to the interaction between them (all  1%). This and Place Tasks in Rats
represents a large discrepancy with the odor and performance
correlates found in the sample arm. Third, in contrast to the Are the neurons that show place responses the same as those
that take part in nonspatial learning tasks conditioning, or
are there two separate populations of cells? Several studies
Table 111.
have compared the response of the same hippocampal neu-
Percentages of Total Number of Recorded Cells
with Signicant Correlates to Various Aspects of an
rons during place tasks and conditioning tasks. Three studies
Olfactory Delayed Nonmatch-to-Sample Task have asked whether complex-spike cells might be specically
involved in the place task and the theta cells might be speci-
Sample Delay Test phase cally involved in the conditioning tasks. Christian and
phase phase phase
Deadwyler (1986) used antidromical stimulation of projec-
Parameter (%) (%) (%)
tion pathways to identify complex-spike cells as pyramidal
Odor 4.8 0 0.9 cells. They found a clear double dissociation of the cells
Performance 2.1 0 0.6 involved in the two tasks. In total, 81% of complex-spike cells
Odor  performance 4.3 0 0.2 had place elds in the place task in comparison with none of
Position 26.8 the theta cells. Conversely, 81% of the theta cells participated
Position  odor 8.4 in the conditioning task whereas none of the complex-spike
Position  performance 5.6
cells did. In two experiments, Eichenbaum and his colleagues
Position  odor  performance 4.5
(Eichenbaum et al., 1987; Wiener et al., 1989) asked whether
Position exclusively 13.5
Odor exclusively 2.4 0.8 cells that were related to events in their olfactory discrimina-
Performance exclusively 2.1 0.4 tion task also had place elds in the same or different tasks. In
the rst study, they found that 43% of cells with correlates in
After Wiebe & Staubli, 1999. the odor discrimination task had place elds in the same
 Hippocampal Neurophysiology in the Behaving Animal 533

environment; in the second study, they recorded some com- known to depend on the perirhinal/parahippocampal cortices
plex-spike cells in a spatial task, some in the odor discrimina- and only minimally (Alvarez et al., 1995) or not at all (Murray
tion task, and a third group in both tasks. They found 75% of and Mishkin, 1998) on the hippocampus itself (see Chapter
the complex-spike cells had place elds in the spatial task 13) (Aggleton and Brown, 1999). In keeping with the lesion
compared to 58% with correlations in the odor discrimina- results, it has been found that only a small number of cells
tion task. Of cells collected in both tasks, 85% had place elds show a differential response to the familiarity of the stimuli
in contrast to 54% that had correlates in the odor task. (0%2.3% in various studies (Riches et al., 1991; Tamura et
We can conclude that in conditioning tasks involving rats al., 1992; Rolls et al., 1993; Salzmann et al., 1993; Brown and
that were not required to move around the environment to Xiang, 1998). In contrast, Wilson et al. (1990) found that 40%
any great extent, most or all hippocampal cells taking part in of their hippocampal cells in this task correlated with the
the conditioning response were theta cells. Complex-spike response, signaling whether the animal was reaching to the left
cells, by contrast, either did not take part or showed slight or right position during their task. Salzmann and Creutzfeldt
inhibition of their resting ring rate. In discrimination condi- (Salzmann et al., 1993) reported that 28% of hippocampal
tioning tasks in which the animal is required to move, the and 33% of parahippocampal neurons responded to both pre-
complex-spike cells also become engaged. The clearest exam- sentations of the visual stimuli in a delayed match-to-sample
ple of cells that may not have a spatial correlate are the cue- task but that even more cells (38% in each area) red when
sampling cells reported in the Eichenbaum studies. However, the animal was presented with a raisin. Vidyasagar and
even some of these responses may be covert place responses, Creutzfeldt (Vidyasagar et al., 1991) reported similar results
as clearly suggested by the Wiebe and Straubli experiments. but that arousing events such as the cage door being opened
or closed or the experimenter entering or leaving the room
11.11.8 Hippocampal Units During Nonspatial also produced a response in these cells. On the basis of these
Learning in Nonhuman Primates ndings, they attributed hippocampal single-unit activity to
the animals behavioral state rather than to any specic mem-
Tamura et al. (1991) reported that 10% of cells in the hip- ory. These studies provide no evidence for the representation
pocampal formation responded differentially to the presenta- of stimulus familiarity in the hippocampus.
tion of three-dimensional objects, some of which had been In contrast and in keeping with the lesion results, there is
conditioned to rewarding (3%) or aversive (2%) stimuli. In a evidence that cells in the rhinal cortex are involved in famil-
follow-up study (Tamura et al., 1992), they showed that 61% iarity judgments. Brown and colleagues (Brown and Xiang,
of object-responsive cells tested in the same apparatus varied 1998) reported that about one-fourth of cells in the rhinal
their response as a function of the location of the stimulus in cortex and area TE showed a decreased response to the second
egocentric or allocentric space. The latter study indicates that presentation of the stimulus in a delayed match-to-sample
apparently perceptual responses in the hippocampus may be task.
spatially modulated. Such responses may underlie object-in- Rolls and his colleagues (Rolls et al., 1993) have tested
place associativity, which like odor-place associativity, is often monkeys on object recognition tasks and compared unit res-
shown to be hippocampus-dependent, unlike odor-object ponses to object familiarity with those to object location. Only
associativity, which is hippocampus-independent (Gilbert a small percentage (2%) of cells responded to familiar objects
and Kesner, 2002). (see above). In contrast, they reported that about 9% of cells
Hippocampal units have been recorded from primates in the hippocampal region responded differentially to the
while they performed in one of the paradigmatic relational location of the stimulus on a display screen. As part of a study
tasks: delayed match or non-match-to-sample. On these tasks, of the place properties of primate hippocampal neurons, Ono
the animal must signal whether two successive stimuli sepa- and his colleagues (1993) found that 17% of cells responded
rated by an interval are the same or different. It is now widely to objects such as a moving experimenter or apple when pre-
accepted that these tasks can be solved in two ways. The ani- sented in a particular part of the visual eld. Just over half of
mal can either assess the strength of the familiarity of the two these were object-in-place cells, which responded differen-
stimuli presented during the test phase and choose the least tially when the animal was shown the object in a particular
familiar, or it can remember which stimulus was previously location and not in other locations. They appear to have prop-
presented in the present situation or context and choose the erties similar to those of the misplace cells described in the rat
one it has not experienced there before. The second strategy is (OKeefe, 1976).
clearly both spatial and relational, but it could be argued that
the rst strategy is also relational although clearly nonspatial. 11.11.9 Hippocampal Units During
To judge a stimulus as familiar or novel, it is necessary to com- Nonspatial Learning in Humans
pare it to the representation of a previous stimulus and to
decide whether the relation between the two is one of equal- Single units have been recorded from medial temporal lobe
ity. Recognition tasks of this sort were originally thought to regions in patients with intractable epilepsy during both
depend on the integrity of the hippocampus but are now recognition and recall memory tasks (Heit et al., 1988; Fried et
534 The Hippocampus Book

al., 1997, 2002; Cameron et al., 2001) as well as during the per- itive fashion. The cells increase ring to both remembered and
ceptual categorization tasks described in Section 11.11.1. In forgotten stimuli but the increase is smaller to the subse-
general, a sizable percentage of hippocampal units was found quently remembered stimuli than to those forgotten.
to participate in both types of memory task. In a recognition
memory task, Fried and colleagues (2002) presented a set of 11.11.10 Conclusions
male and female faces portraying various emotions; and after
an interval of 1 to 12 hours they showed the same faces again In all species tested, a small percentage of hippocampal cells
together with an equal number of foils. A signicant number have been reported to re to nonspatial stimuli. The data are
of hippocampal neurons responded to faces during encoding clearest in humans, where about 12% of cells responded to the
(25%) and a somewhat larger percentage during the recogni- sight of famous faces, but this fell well below the proportion
tion phase (41%). Perhaps surprisingly, a large proportion of (29%) that responded to spatial stimuli such as pictures of
the hippocampal responses (62%) consisted of ring rate houses or interiors. Hippocampal unit responses to stimuli
decreases. Almost one-half the hippocampal neurons with an used as discriminanda in learning experiments have been
excitatory response but only one-sixth of those with an reported in all species tested: rats, rabbits, monkeys, and
inhibitory response showed some selectivity for gender or humans. There is some suggestion that the responses in the rat
emotion. Similar ndings were reported by Heit and col- may be due to a place-gated mechanism because the one study
leagues (1988), who used lists of words to be remembered and that tested this carefully found that the conditioned stimuli
found most hippocampal units selective for one of the words only elicited a pyramidal cell response after conditioning if
on the list. When tested with a second list of different words, it was presented while the animal was in the place eld of
slightly less than half of these cells (13/30) were also selective the cell (Moita et al., 2003). Theta cells are much less sensitive
for a word in that list. As the authors remarked, this suggests to location and may account for a large percentage of the
that rather than being preordained to respond to specic responses reported in the early studies, which did not distin-
stimuli the cells were tuning into the current context and guish between cell types. In general, monkey experiments
selecting one aspect of it (such as position along a list?). The reported few cells responding to objects per se but more to
result is reminiscent of the way in which a given pyramidal cell location or object-in-places. In keeping with the lesion results
in the rat hippocampus may have unrelated place elds in two (see Chapter 13), units responding to objects and their famil-
or more environments. Hippocampal units have also been iarity were more plentiful in the rhinal cortex than in the hip-
recorded during a paired-associate recall task (Cameron et al., pocampus.
2001). A series of word pairs were presented, and after 1 to 2 There are clearly cells in the human medial temporal lobe
minutes the subject had to respond with the second of the pair that respond to words as well as faces and other stimuli in
when prompted with the rst: 29% of hippocampal units both recognition and recall paradigms. Furthermore, in one
responded during stimulus encoding and 44% during recall. study of verbal paired associate learning, hippocampal unit
These percentages are similar to those found in the entorhinal responses during encoding of a pair predicted the success of
cortex and amygdala in the same study and in the face recog- subsequent recall (Cameron et al., 2001). Somewhat surpris-
nition study of Fried et al. (2002) but signicantly higher than ingly, there was only a hint of lateralization of these responses
the response to faces and other nonspatial visual stimuli in the to the left hemisphere, as might be expected from the lesion
perceptual classication study of Kreiman et al., 2000a), sug- data. It is clear that the human hippocampus is involved in the
gesting a specic role for learning. An interesting aspect of the storage of verbal as well as nonverbal material.
study was that one-fth of the hippocampal cells showed a In general then, there is evidence that nonspatial informa-
differential ring rate during encoding between word pairs tion is represented in the hippocampus but that the number
that were subsequently remembered correctly and those that of cells involved is considerably smaller than those involved in
were forgotten. Most intriguingly, most of these cells showed spatial information and often the nonspatial responses are
a smaller increase in ring rate to the word pairs that were secondary to a primary spatial correlate. The larger percentage
subsequently remembered than to the ones that were forgot- of responses to nonspatial stimuli reported in humans is con-
ten. Whatever the role being played by the hippocampus in sonant with the wider function attributed to the human hip-
these tasks, it cannot simply be that increased ring in hip- pocampus in narrative and episodic memory (OKeefe and
pocampal neurons during encoding leads to a stronger mem- Nadel, 1978; OKeefe, 1996, 2001; Burgess et al., 2002).
ory trace and therefore to better recall.
Although one must be cautious when comparing across
studies, even from the same laboratory, the results can be
summarized as follows. (1) The percentage of cells responding 11.12 Other Distinctive Cells in the
to faces and words is higher when they are part of a memory Hippocampal Formation and Related Areas
task than when they are simply perceived; (2) these stimuli are
equally well represented in both recognition and recall mem- As we have seen, the hippocampal formation is part of a wide-
ory tasks; and nally (3) the response of hippocampal units to spread system of anatomically related regions involved in
a stimulus during the encoding phase of a recall test correlates memory and navigation. In addition to the CA elds and
with subsequent recall of that stimulus but in a counterintu- the dentate gyrus, the hippocampal formation contains two
 Hippocampal Neurophysiology in the Behaving Animal 535

anatomically distinct regions: the subicular complex and functional circuit diagram of the interconnections between
the parahippocampal cortex. Response properties of cells in these regions of the hippocampal formation that underlies
some of these other regions have been touched on at different spatial behavior and memory.
points in this chapter where appropriate. For example in
Sections 11.9 and 11.10, we summarized the properties of 11.12.1 Subicular Region Has Fewer Place
head-direction (HD) cells found in the presubiculum and, in Cells than the Hippocampus Proper,
addition, regions of the Papez circuit, which forms the inputs and Their Properties Differ
to these cells. In this section, we briey summarize the prop-
erties of cells in some of these other related brain regions and Place cells and theta cells have been recorded from several
touch upon the question of whether these cells are afferent to subdivisions of the subicular region. We consider the similar-
the hippocampal formation, providing spatial information, or ities and differences between cells in the next sections. The
alternately should be thought of as efferent output structures, subiculum contains both place cells and theta cells (Sharp
receiving spatial and memory information from the hip- and Green, 1994; Sharp, 1997). The percentage of theta cells
pocampal formation. is approximately the same as found in the hippocampus
There are three ways this question might be approached. proper (10%), and most of the remaining cells have strong
The rst is to look at the effects of lesions elsewhere in the cir- spatial signals. There are no complex-spike cells, but one
cuit on hippocampal place cell ring (Brun et al., 2002; Calton group (29% of total) shows a bursting pattern with a peak
et al., 2003) and presubicular HD cell ring (Golob and interspike interval of 2 to 4 ms. These cells do not seem to dif-
Taube, 1997) as described in Sections 11.7.4 and 11.9/11.10, fer in any other respect from the nonbursting types except
respectively. The second examines the relative latency of the that they have a stronger theta modulation. The eld sizes and
neurons in each area to respond to the stimulus or behavior average ring rates of the subicular place cells are much larger
under study. This approach depends on the assumption that when compared to those in the hippocampus. A typical subic-
neurons in a region that projects to a second region in a serial ular cell res over the entire environment with an average rate
fashion, on average, re earlier in response to the appropriate of 5 to 15 Hz (in contrast to 0.52.5 Hz in the hippocampus)
stimulus or during the relevant behavior than the neurons in and has one or more areas of increased ring. Like the place
that second region. On the assumption that many cells in the cells of the hippocampus, subicular place cells recorded in a
hippocampal formation and related areas have spatial aspects cylinder are under the inuence of a cue card on the wall
of behavior as their primary correlate, it is possible to com- and rotate in step with it. It has been claimed that the subicu-
pare the temporal correlation of cells in different hippocam- lar cells have a stronger directional component than the
pal regions in response to manipulations of these spatial hippocampal cells, but this claim must be treated with cau-
variables to estimate the temporal relation between cells in tion. The amount of variance in subicular ring rate that is
different regions indirectly. For example, one might study the explained by direction is actually very small (Sharp and
latency to onset of ring of HD cells with the same preferred Green, 1994).
orientation in different parts of the Papez circuit. Ideally, this One important difference from hippocampal place cells is
ought to be done with simultaneous recording from the that subicular cells are far less likely to remap across environ-
two (or more) regions in question to control for variations ments that differ in shape (Sharp and Green, 1994), shape and
between rats in behavior and location of the tracking lights visual markings (Sharp, 1997), or size (Sharp, 1999b). For
on the head. The third method is based on a comparison of instance, whereas many of the hippocampal cells displayed
the place correlates of the neurons in the two areas. If one markedly different elds in a cyclinder and a rectangle, the
has a serial model of place eld construction, it might be subicular elds were similar in the two environments. This is
possible to attribute some neuronal responses to earlier stages reminiscent of what is found in the spatially coded cells in the
in that process than others. For example, Barnes and col- supercial layers of the entorhinal cortex (see below) and to
leagues (1990) compared the eld sizes of place cells in differ- hippocampal place cells during the animals initial experiences
ent regions of the hippocampal formation and subiculum. of the two (see Section 11.8.3). A study of the timing of subic-
As we shall see, some generalizations arise from this third ular cell ring relative to the animals location suggests that
type of analysis. The locational code in the CA elds of subicular cells re slightly earlier on average than CA1 cells
the hippocampus is more environment-specic than in other (70 vs. 40 ms), perhaps indicating that they come earlier in the
regions. Hippocampal place cells often re in distinct patterns computational chain (Sharp, 1999a). This result may seem
in different environments (remapping); indeed, many hip- paradoxical in light of the strong connections from CA1 to the
pocampal cells are or become silent in a given environment subiculum, but it is consistent with the data showing that
(see Section 11.7.4). By contrast, the locational region(s) sig- subicular cells do not remap the environments that CA1 cells
naled by nonhippocampal cells are much more likely to be in do.
similar places across dissimilar environments. Moreover, in In conclusion, various approaches to the functional rela-
other regions, such as the subiculum and supercial entorhi- tion between the subiculum and the hippocampus proper
nal cortex, principal cells are never silent. suggest that the subicular place cells cannot receive their spa-
Taking the results of all three methods into account, it tial properties solely by virtue of their inputs from the hip-
is difficult but not impossible to make suggestions about the pocampus proper. They may be more dependent on direct
536 The Hippocampus Book

inputs from the entorhinal cortex and/or they themselves may tional and directional signals are stable and robust in TPD
provide a source of spatial inputs to the entorhinal cells. cells, with the directional signal carrying more information.
The directional and locational signals are dissociable in differ-
11.12.2 Presubiculum Contains Several ent environments. The preferred direction of a given TPD
Classes of Spatial Cell cell appears to be environment-invariant, being similar across
different locations in the testing room (like the HD cells) and
The dorsal presubiculum (or postsubiculum in some authors across different enclosures in the same location (cylinder ver-
terminology) contains other types of spatial cell in addition to sus an open circular platform). In contrast, the locational
the HD cells described above (see Section 11.9). A survey elds tend to differ more between environments; notably,
study (Sharp, 1996) of presubicular cells in the standard TPD cells generally have different locational elds in the cylin-
cylinder also reported cells with correlates for angular veloc- der and open circular platform. TPD cell ring represents
ity, running speed (see also Lever et al., 2003), location and an integration of location-related and direction-related infor-
direction, and location. The nature of these correlates suggests mation. It is possible that directional information may have to
an important role for the presubiculum in navigation and spa- be incorporated into and distributed in theta wave packets to
tial memory. One class of these cells, called theta-modulated be used by the navigation system. Neither anterodorsal thala-
place-by-direction (TPD) cells, has been examined in detail mic nor presubicular HD cells show theta modulation.
(Cacucci et al., 2004) (Fig. 1121B, see color insert). TPD cells Cacucci et al. provided evidence that a robust orientation
generally have a strong tendency to re at or just before the signal operating in theta mode exists in the hippocampal for-
trough of the local theta oscillation. Their depth of theta mation.
modulation is high and comparable to that of cells in the There have been no published reports of cells recorded
medial septum/diagonal band of Broca (King et al., 1998). The from the ventral portions of either the presubiculum or the
quantitative analyses of Cacucci et al. indicate that both loca- parasubiculum. By analogy with the differences between dor-

Figure 1121. Four spatial cell types in the hippocampal forma- direction cell does not have localized ring but has strong direc-
tion. False color ring eld maps (left) show the ring rate as a tional selectivity. (Source: AC. Courtesy of Francesca Cacucci.)
function of the animals location in cylindrical environments irre- D. Grid cell has multiple place elds and no directional selectivity.
spective of heading direction; directional polar plots (right) show Directional grid cells also exist. Numbers associated with the ring
the ring rate for the same cell as a function of the animals heading rate plots represent the maximum ring rate (red regions in ring
direction irrespective of location. A. Place cell has single localized rate maps and peak x and y values in polar plots). Diameter of
eld and no directional selectivity. B. Place by directional cell has cylinders in A C is 79 cm; diameter in D is 2 m. (Source: Edvard
single localized eld and strong directional selectivity. C. Head and May-Britt Moser.)
 Hippocampal Neurophysiology in the Behaving Animal 537

sal and ventral hippocampus (Jung et al., 1994), these ventral the overall set of elds of several such cells covers the entire
regions might be expected to contain lower levels of spatial environment. For each cell, the size of the grid appears to be
signaling and/or larger spatial elds. independent of the size or shape of the environment. As one
goes more ventral in the medial entorhinal cortex, the size of
11.12.3 Parasubiculum the grid gets larger. This grid system appears to be based on
path integration inputs generated by the animals own move-
The parasubiculum contains location-specific neurons, ments and may provide the hippocampal mapping system
although in much lower percentages (ca. 10%) than are found with distance and directional information on which to con-
in the hippocampus proper or the subiculum (Taube, 1995b). struct maps of individual environments.
In contrast, it contains a much higher percentage of theta Frank and colleagues (2000) recorded from neurons in the
cells (41%). No complex-spike cells have been found there. In hippocampus proper and in the supercial and deep layers of
comparison with the hippocampus, the ring elds of the the entorhinal cortex while rats ran along single or double U-
parasubicular place cells are signicantly larger and have less shaped tracks (the latter termed W-shaped tracks). On the
spatial information content. In common with all place and single U-shaped track, the animal shuttled back and forth
HD cells recorded thus far, their place elds rotate in step with between the two prongs of the U to receive rewards at each
the rotation of the cue card on the wall of the environment. end; on the W track it shuttled rst between the left-hand and
Analysis of the temporal relations of the ring elds of these middle prongs and then between the middle and right-hand
cells to the animals location showed a variety of relations, but ones and vice-versa. Similar to the ndings of Quirk et al. (see
on average the cell ring preceded the animals location by above), they found that: (1) EC cells re at rates about ve
about 60 ms. This is less than the average of about 120 ms for times those of CA1 cells; (2) the locational elds of cells
hippocampal cells, suggesting that the parasubicular neurons in both the supercial and deep layers of the entorhinal
may be later in the circuit than the hippocampal cells. cortex were larger (by about three times) and contained
less locational information than those in CA1. Deep EC cells
11.12.4 Spatial Cells in the Entorhinal Cortex contained more locational information than supercial EC
cells. Some cells in all three regions were sensitive to the place
Quirk and colleagues (1992) recorded from cells in layers 2 from which the animals had recently come (retrospective cod-
and 3 of the medial entorhinal cortex under conditions simi- ing) or to which they were intending to go (prospective cod-
lar to those in which hippocampal place cells have been ing) (cf. Section 11.7.3 for a discusssion of retrospective and
recorded (Muller et al., 1987). They found both place-coded prospective coding in hippocampal cells). However, the deep
and theta cells. Two properties of the spatial cells were differ- entorhinal cortex contained a signicantly higher proportion
ent from those displayed by the hippocampal place cells but of prospectively coding cells than the CA1 or the supercial
similar to those found in the subiculum. The entorhinal cor- entorhinal cortex. Cells in the deep entorhinal layers were also
tex (EC) elds were larger and less spatially compact; and more likely to reect environment-invariant aspects of the
unlike the hippocampal cells but like the subicular cells, they paths taken along these tracks. For instance, one deep EC cell
showed less sensitivity to the shape of the enclosure. Whereas red on both the U and W tracks as the animal ran away from
many of the hipppocampal place cells displayed markedly dif- a prong and turned left into the adjacent arm. These ndings
ferent elds in the cyclinder and the square, the EC cell elds suggest that the deeper layers of the entorhinal cortex, which
were similar in the two environments. Unlike the hippocam- receive strong inputs from the CA1 eld and the subiculum,
pus, where this similarity between elds in the two boxes may be using spatial information provided by these regions to
occurs only during the initial experience of the environments, construct routes between known locations. Alternatively, the
the EC cells appear not to learn to discriminate. As we saw ear- results may be related to the regular eld structure of grid cells
lier (Mizumori et al., 1992; Jeffery et al., 1995), the theta cells reported by Hafting and colleagues (see above).
in the EC are dependent on the integrity of the medial septum
in the same way as the hippocampal theta cells. 11.12.5 Cells in the Perirhinal Cortex
As discussed in Section 11.7, work by Fynn, Hafting, and Code for the Familiarity of Stimuli
colleagues (Fyhn et al., 2004; Hafting et al., 2005) suggests that
at least one class of cells in the supercial layers of the medial Cells in the perirhinal cortex of the monkey and rat (Brown et
entorhinal cortex lays a grid-like structure on every environ- al., 1987) are sensitive to the familiarity of the stimulus. The
ment the animal visits. Each grid cell res in several locations amplitude of the cells response to the second and subsequent
in each environment, with the locations forming a regular repetitions of an object is smaller than that to the rst presen-
pattern as though they were nodes on a triangular grid (see tation, and this decreased response recovers as a function of
Figs 1110 and 1121D, see color insert). Different cells the time and number of intervening items since the previous
recorded at the same location have the same grid spacing and exposure to that stimulus. This ts with the suggestion that
orientation relative to the environment but differ in the loca- the decit in delayed non-match-to-sample in the monkey or
tion of the nodes, such that the ring peaks of one cell are human following large lesions in the mesial temporal lobes is
slightly shifted from those of its neighbor. The result is that due primarily to the involvement of these rhinal structures
538 The Hippocampus Book

(Brown and Aggleton, 2001) (see Section 11.11.8 and Chapter level: In any single environment, half or more of the cells in
13 for further discussion). the hippocampus proper are silent. Outside the hippocampus
proper, principal cells are generally not silent and indeed tend
11.12.6 Cells in the Medial Septum to re at higher rates than hippocampal place cells. Further
Are Theta Cells study is required to disentangle the particular role played by
each of the regions in the hippocampal formation.
Cholinergic and GABAergic cells in the medial septum and Relative to hippocampal place cells, cells in some neigh-
diagonal band of Broca (DBB) supply the driving inputs that boring extrahippocampal regions tend to, have larger elds,
set the frequency of hippocampal theta. As we saw in Section and are less sensitive to environmental changes such as alter-
11.3.3, lesions of this region abolish hippocampal theta. ations in the shape of the enclosure. It seems reasonable to
Ranck (1973) recorded from cells in the medial septum of suggest that the presubicular HD cells are afferent to the
freely moving rats and conrmed earlier reports by Petsche medial EC grid cells, which in turn are afferent to the hip-
and Stumpf that cells there had a rhythmical bursting pattern pocampal place cells; and the anatomical connections
that was phase-locked to hippocampal theta. About 50% of between these regions are consonant with this functional rela-
the septal cells showed this pattern. Attempts have been made tion. Less clear is the relation between the subiculum and the
to classify septal cells into those likely to be cholinergic or hippocampus. The greater susceptibility of hippocampal place
GABAergic on the basis of their waveform (Matthews and Lee, cells to environmental change and the latency data suggest
1991; King et al., 1998). Cells have also been classied on the that the subiculum provides environmental inputs to the hip-
basis of the strength of their correlation to the hippocampus pocampus perhaps in the form of the distance to one or more
theta. In one study (King et al., 1998) , 47% showed a strong boundaries; on the other hand, the anatomy suggests a strong
phase relation to the hippocampal theta with strong peaks in projection from the CA1 eld to the subiculum that is not
the auto-correlation at the theta period. Each cell had its own reciprocated. The perirhinal and lateral entorhinal cells most
individual preferred phase, which was constant over days. The likely provide information about environmental landmarks
rest of the septal cells show weaker relations to theta. There and other objects to the hippocampus where it is integrated
is, however, only a slight preference for a particular phase with the place information to support behavior based on
across the whole population with all phases represented. object-in-place knowledge. The deeper layers and the lateral
Putative cholinergic cells (on the basis of wave shape) tended septum are the major efferent targets of the CA1 and CA3 pyr-
to be concentrated in the less rhythmical class than putative amids, respectively. Little is known about the behavioral cor-
GABAergic cells. On the other hand, GABAergic medial septal relates of deep layers of the entorhinal cortex and the
neurons, which selectively target hippocampal inteneurons, prefrontal cortex, but they may be the rst stages in the trans-
appear to form two distinct populations tightly coupled either formation of the hippocampal signal into environmentally
to the trough (178) or the peak (330) of hippocampal theta specic route information or where it is used to program the
waves (Borhegyi et al., 2004). approach to goal locations. The hippocampal projection to
As we saw in Section 11.3.4 on the theta rhythms and the the EC may stabilize the grid cell location relative to a given
EEG, there is some evidence that the frequency of theta is cor- environment.
related with the speed with which an animal moves in an envi-
ronment. A similar correlation was found in the rhythmical
theta cells of the septum (King et al., 1998). Rhythmical cells
showed a burst frequency that was correlated with the speed 11.13 Overall Conclusions
of movement on a linear track. The slope of the regression lin-
ear curve was about 0.9 Hz/m/s with an intercept at 8.4 Hz. A This chapter has focused on the neural correlates of hip-
small number of rhythmical septal cells had an interesting pocampal EEG and single-unit activity. We found that the
directional response. They red with a strong rhythmicity EEG could be categorized into several frequency bands: theta,
when the animal ran in one direction but lost their rhythmic- beta, gamma, and the high-frequency ripples. Each of these
ity when it ran in other directions. has a different behavioral correlate and is reected by different
ring patterns in hippocampal interneurons. In rodents, the
11.12.7 Summary of Extrahippocampal beta and gamma oscillations relate primarily to olfactory
Place Field Properties stimuli, but theta and the ripples have much broader corre-
lates. There are two types of theta, which are differentially sen-
The ring of extrahippocampal cells tends to be more envi- sitive to cholinergic drugs: a-Theta is activated during periods
ronment-invariant than hippocampal place cell ring, gener- of arousal or attention, and t-theta is related to movements
alizing across the various environments explored and routes (e.g., walking, swimming, jumping) that translate the animals
taken. Thus, cells in the subiculum and entorhinal cortex may position relative to the environment. Hippocampal interneu-
signal broadly equivalent locations, or similar points along a rons re in synchrony with the concurrent theta; and various
route, in different environments. Whereas extrahippocampal types of interneuron, targeting different parts of the pyrami-
cells generalize, hippocampal place cells basically discriminate dal cell, preferentially re on different phases of theta. One
(especially after experience). This is revealed at the population function of theta activity is to coordinate and perhaps bind
 Hippocampal Neurophysiology in the Behaving Animal 539

together neural activity in different parts of the nervous sys- the ongoing hippocampal EEG theta activity; events and stim-
tem. Theta-related ring patterns have been found in cells in uli occurring in that location are signaled by variations in the
such disparate regions of the nervous system as the prefrontal ring rate above the baseline. The existence of temporal cod-
cortex, amygdala, and inferior colliculus. Theta also organizes ing strategies in the nervous system means that one must be
the temporal relation between the ring of cells in the hip- cautious when interpreting the results of recording techniques
pocampus such that inputs to a cell at one phase of theta pro- that reect only relative rates of neural activity in contrast to
duce larger synaptic modications (LTP) than at other phases. the timing of action potentials. This is especially true when
Finally, each theta cycle acts as a timing cycle against which the the representation is distributed across a population of cells
phase of hippocampal pyramidal cell ring can be measured. with some cells increasing and others decreasing their ring
This phase coding complements the locational signal carried rates in addition to shifting phases.
in the gross ring rate and allows rate modulations above the Several types of learning have been demonstrated in
baseline to code for variables such as the speed of running in hippocampal pyramidal cells: short-term changes in the ori-
addition to location (see below). Dual phase and ring rate entation of place elds relative to the environment lasting
coding may be a general strategy for binding together the rep- for the duration of the trial; longer changes in the shape of
resentations of different aspects of the world in the train of the place eld lasting for the duration of the testing session;
action potentials of a single cell. The high-frequency ripples and long-term discrimination between similar environments
occur in conjunction with the large irregular activity (LIA) that appears to be permanent. One type of environmental dis-
state of the hippocampal EEG, which occurs during nontrans- crimination appears to involve the collective behavior of
lational activities such as sleeping, quiet resting, grooming, pyramidal cells, perhaps operating as a discrete attractor in
drinking, and eating. Theta and LIA/ripples appear to be which the behavior of neighboring pyramidal cells, in addi-
mutually exclusive states of the hippocampus, never occurring tion to environmental inputs, is taken into account in the nal
at the same time. It has been suggested that the synchronized representation.
bursts of activity that occur in hippocampal neurons during The location of place elds depends on factors in addition
the ripples may reect the transfer of information from the to the animals physical location in an environment. Under
hippocampus to the neocortex as part of a memory consoli- certain circumstances, elds may depend on the animals pre-
dation process. vious behavior, such as the turn that it has just made at a
Three classes of cells with spatial correlates have been choice point or one that it intends to make. Training on aver-
reported in the hippocampal formation: the place cells in the sive tasks such as eyeblink conditioning in the rat has also
hippocampus itself (Fig. 1121A, see color insert); the head been shown to shift the location of elds. Whether this means
direction (HD) cells in the dorsal presubiculum (Fig. 1121C, that the animal conceives of these environments as different is
see color insert); and the grid cells in the medial entorhinal not clear at present.
cortex (Fig. 1121D, see color insert). In addition, there are It has also been reported that pyramidal cells respond to
cells that combine two types of information, such as the place nonspatial stimuli such as the odors used in a running recog-
by direction cells found in the presubiculum (Fig. 1121B, nition task or the auditory cues in a nictitating membrane
see color insert). Place cells signal the animals location, HD conditioning task. The existence of such responses together
cells signal the animals direction of heading, and grid cells with decits in some types of conditioning tasks (e.g., trace
provide information about distances moved in particular conditioning following hippocampal lesions) has suggested a
directions. It seems reasonable to assume that the place cell broader function for the hippocampus than purely spatial. On
ring patterns are constructed from combinations of grid cell the other hand, several studies have suggested that when the
inputs, which in turn get their directional orientation from locations in which these stimuli occur are varied, these non-
the head direction input. Together these three cell types pro- spatial responses are gated by the animals location. In a rat
vide the information required to construct a mapping system eyeblink conditioning experiment, the auditory conditioned
that identies the animals location in an environment and stimulus only elicited short latency response in hippocampal
relates these locations to each other on the basis of the dis- cells when the animal was in the place eld of that cell. In
tance and direction from one to the other. It should also allow an olfactory delayed match-to-sample task, cells appeared
the animal to move from one location in that environment to to respond to the olfactory cues alone when position was
another along any available path, supporting exible naviga- not varied in the start arm of a Y maze; in contrast, only olfac-
tion. As we mentioned above, hippocampal pyramidal cells tory responses gated by position were seen in the two goal
also code for events or stimuli that occur in particular places, arms where position was available as a factor. To rule out
laying the basis for the more general episodic memory system position as a contribution to hippocampal unit responses,
seen in the human hippocampus. A true episodic memory it is important to provide the stimulus in more than one
system would incorporate the time of occurrence of events as location.
well as their location. Studies of single-unit activity in monkey hippocampus
Encoding of events as well as locations in hippocampal suggest that although place responses also dominate there is
pyramidal cells is made possible by a dual coding strategy: evidence of nonspatial inputs as well. As primate research
Location is conveyed by a combination of increased ring rate moves in the direction of the use of freely moving monkeys,
above a low-level baseline and the phase of ring relative to and especially when the animal can explore the walls as well as
540 The Hippocampus Book

the oor of the environment, we will begin to get a better pic- Barnes CA, Suster MS, Shen J, McNaughton BL (1997) Multistability
ture of the relation between these nonspatial and spatial of cognitive maps in the hippocampus of old rats. Nature
inputs. There are also a small number of studies of hip- 388:272275.
pocampal unit activity in humans. Units responding to pic- Bassett JP, Taube JS (2001) Neural correlates for angular head veloc-
ity in the rat dorsal tegmental nucleus. J Neurosci 21:57405751.
tures of faces and household objects have been reported, but
Battaglia FP, Sutherland GR, McNaughton BL (2004) Local sensory
the percentage of cells responding to pictures of houses and
cues and place cell directionality: additional evidence of pros-
locations in the same studies has been higher. Responses to pective coding in the hippocampus. J Neurosci 24: 45414550.
words have been reported frequently, especially when they are Berger TW, Thompson RF (1978) Identication of pyramidal cells as
presented in a learning paradigm. This may be an indication the critical elements in hippocampal neuronal plasticity during
that the function of the human hippocampus is broader than learning. Proc Natl Acad Sci USA 75:15721576.
purely spatial. This is in line with the original suggestion of Berger TW, Rinaldi PC, Weisz DJ, Thompson RF (1983) Single-unit
the cognitive map theory that the incorporation of verbal and analysis of different hippocampal cell types during classical
temporal inputs into its original spatial functions would result conditioning of rabbit nictitating membrane response.
in employment of the hippocampus as an episodic and narra- J Neurophysiol 50:11971219.
tive memory system as well a spatial one. Berry SD, Swain RA (1989) Water deprivation optimizes hippocam-
pal activity and facilitates nictitating membrane conditioning.
Behav Neurosci 103:7176.
Black A (1975) Hippocampal electrical activity and behavior. In: The
ACKNOWLEDGMENTS hippocampus: a comprehensive treatise ( Isaacson RL, Pribram
KH, eds), pp 129167. New York: Plenum Press.
I am grateful to my co-editors for extensive and helpful comments on Blair HT, Sharp PE (1996) Visual and vestibular inuences on head-
earlier drafts of the chapter. I also thank Dr. Colin Lever (Leeds direction cells in the anterior thalamus of the rat. Behav
University) who made extensive comments and suggestions on an Neurosci 110:643660.
earlier version. Support for the research from my own laboratory Blair HT, Lipscomb BW, Sharp PE (1997) Anticipatory time intervals
cited in this chapter came from the Medical Research Council UK, of head-direction cells in the anterior thalamus of the rat: impli-
Wellcome Trust, and the BBSRC. cations for path integration in the head-direction circuit.
J Neurophysiol 78:145159.
Blair HT, Cho J, Sharp PE (1999) The anterior thalamic head-direc-
REFERENCES tion signal is abolished by bilateral but not unilateral lesions of
the lateral mammillary nucleus. J Neurosci 19:66736683.
Acsady L, Kamondi A, Sik A, Freund T, Buzsaki G (1998) GABAergic Bland BH, Oddie SD (2001) Theta band oscillation and synchrony in
cells are the major postsynaptic targets of mossy bers in the rat the hippocampal formation and associated structures: the case
hippocampus. J Neurosci 18:33863403. for its role in sensorimotor integration. Behav Brain Res
Adey WR, Dunlop CW, Hendrix CE (1960) Hippocampal slow waves: 127:119136.
distribution and phase relations in the course of approach Bland BH, Vanderwolf CH (1972) Diencephalic and hippocampal
learning. Arch Neurol 3:7490. mechanisms of motor activity in the rat: effects of posterior
Aggleton JP, Brown MW (1999) Episodic memory, amnesia, and hypothalamic stimulation on behavior and hippocampal slow
the hippocampal-anterior thalamic axis. Behav Brain Sci 22: wave activity. Brain Res 43:6788.
425490. Bland BH, Konopacki J, Kirk IJ, Oddie SD, Dickson CT (1995)
Alonso A, Garcia-Austt E (1987a) Neuronal sources of theta rhythm Discharge patterns of hippocampal theta-related cells in
in the entorhinal cortex of the rat. I. Laminar distribution of the caudal diencephalon of the urethane-anesthetized rat.
theta eld potentials. Exp Brain Res 67:493501. J Neurophysiol 74:322333.
Alonso A, Garcia-Austt E (1987b) Neuronal sources of theta rhythm Bland BH, Trepel C, Oddie SD, Kirk IJ (1996) Intraseptal microinfu-
in the entorhinal cortex of the rat. II. Phase relations between sion of muscimol: effects on hippocampal formation theta eld
unit discharges and theta eld potentials. Exp Brain Res 67: activity and phasic theta-on cell discharges. Exp Neurol
502509. 138:286297.
Alvarez P, Zola-Morgan S, Squire LR (1995) Damage limited to the Bloch V, Laroche S (1981) Conditioning of hippocampal cells: its
hippocampal region produces long-lasting memory impair- acceleration and long-term facilitation by post-trial reticular
ment in monkeys. J Neurosci 15:37963807. stimulation. Behav Brain Res 3:2342.
Arnolds DE, Lopes da Silva FH, Aitink JW, Kamp A, Boeijinga P Borhegyi Z, Varga V, Szilagyi N, Fabo D, Freund TF (2004) Phase seg-
(1980) The spectral properties of hippocampal EEG related regation of medial septal GABAergic neurons during hip-
to behavior in man. Electroencephalogr Clin Neurophysiol 50: pocampal theta activity. J Neurosci 24:84708479.
324328. Bostock E, Muller RU, Kubie JL (1991) Experience-dependent modi-
Barnes CA, McNaughton BL, OKeefe J (1983) Loss of place speci- fications of hippocampal place cell firing. Hippocampus
city in hippocampal complex spike cells of senescent rat. 1:193205.
Neurobiol Aging 4:113119. Bragin A, Engel J Jr, Wilson CL, Fried I, Buzsaki, G (1999) High-
Barnes CA, McNaughton BL, Mizumori SJ, Leonard BW, Lin LH frequency oscillation in human brain. Hippocampus 9: 137142.
(1990) Comparison of spatial and temporal characteristics of Bragin A, Jando G, Nadasdy Z, Hetke J, Wise K, Buzsaki G (1995)
neuronal activity in sequential stages of hippocampal process- Gamma (40100 Hz) oscillation in the hippocampus of the
ing. Prog Brain Res 83:287300. behaving rat. J Neurosci 15:4760.
 Hippocampal Neurophysiology in the Behaving Animal 541

Brankack J, Stewart M, Fox SE (1993) Current source density analysis lesions of the head direction cell network. J Neurosci
of the hippocampal theta rhythm: associated sustained poten- 23:97199731.
tials and candidate synaptic generators. Brain Res 615:310327. Cameron KA, Yashar S, Wilson CL, Fried I (2001) Human hip-
Breese CR, Hampson RE, Deadwyler SA (1989) Hippocampal place pocampal neurons predict how well word pairs will be remem-
cells: stereotypy and plasticity. J Neurosci 9:10971111. bered. Neuron 30:289298.
Brown MW, Aggleton JP (2001) Recognition memory: what are the Caplan JB, Madsen JR, Schulze-Bonhage A, Aschenbrenner-Scheibe
roles of the perirhinal cortex and hippocampus? Nat Rev R, Newman EL, Kahana MJ (2003) Human theta oscillations
Neurosci 2:5161. related to sensorimotor integration and spatial learning.
Brown MW, Xiang JZ (1998) Recognition memory: neuronal sub- J Neurosci 23:47264736.
strates of the judgement of prior occurrence. Prog Neurobiol Chakraborty S, Anderson MI, Chaudhry AM, Mumford JC, Jeffery
55:149189. KL (2004) Context-independent directional cue learning by
Brown MW, Wilson FA, Riches IP (1987) Neuronal evidence that hippocampal place cells. Eur J Neurosci 20: 281292.
inferomedial temporal cortex is more important than hip- Chen LL, Lin LH, Green EJ, Barnes CA, McNaughton BL (1994)
pocampus in certain processes underlying recognition memory. Head-direction cells in the rat posterior cortex. I. Anatomical
Brain Res 409:158162. distribution and behavioral modulation. Exp Brain Res
Brun VH, Otnass MK, Molden S, Steffenach HA, Witter MP, Moser 101:823.
MB, Moser EI (2002) Place cells and place recognition main- Cho J, Sharp PE (2001) Head direction, place, and movement corre-
tained by direct entorhinal-hippocampal circuitry. Science lates for cells in the rat retrosplenial cortex. Behav Neurosci
296:22432246. 115:325.
Bullock TH, Buzsaki G, McClune MC (1990) Coherence of com- Cho YH, Giese KP, Tanila H, Silva AJ, Eichenbaum H (1998) Abnor-
pound eld potentials reveals discontinuities in the CA1- mal hippocampal spatial representations in alphaCaMKIIT286A
subiculum of the hippocampus in freely-moving rats. and CREBalphaDelta mice. Science 279:867869.
Neuroscience 38:609619. Christian EP, Deadwyler SA (1986) Behavioral functions and hip-
Bures J, Fenton AA, Kaminsky Y, Rossier J, Sacchetti B, Zinyuk L pocampal cell types: evidence for two nonoverlapping popula-
(1997) Dissociation of exteroceptive and idiothetic orientation tions in the rat. J Neurophysiol 55:331348.
cues: effect on hippocampal place cells and place navigation. Chrobak JJ, Buzsaki G (1996) High-frequency oscillations in the out-
Philos Trans R Soc Lond B Biol Sci 352:15151524. put networks of the hippocampal- entorhinal axis of the freely
Burgess N, OKeefe J (1996) Neuronal computations underlying the behaving rat. J Neurosci 16:30563066.
ring of place cells and their role in navigation. Hippocampus Clark RE, Broadbent, Squire LR (2005) Imparied remote spatial
6:749762. memory after hippocampal lesions despite extensive training
Burgess N, Recce M, OKeefe J (1994) A model of hippocampal func- beginning early in life. Hippocampus 15:340346.
tion. Neural Networks 7:10651081. Cohen NJ, Eichenbaum H (1993) Memory, amnesisa and the hip-
Burgess N, Jeffery KJ, OKeefe J (1999) Intergrating hippocampal and pocampal system. Cambridge, MA: MIT Press.
parietal functions: a spatial point of view. In: The hippocampal Colom LV, Bland BH (1987) State-dependent spike train dynamics of
and parietal foundations of spatial cognition (Burgess N, Jeffery hippocampal formation neurons: evidence for theta-on and
KJ, OKeefe J, eds), pp 329. Oxford, UK: Oxford University theta-off cells. Brain Res 422:277286.
Press. Colombo M, Fernandez T, Nakamura K, Gross CG (1998) Functional
Burgess N, Maguire EA, OKeefe J (2002) The human hippocampus differentiation along the anterior-posterior axis of the hip-
and spatial and episodic memory. Neuron 35:625641. pocampus in monkeys. J Neurophysiol 80:10021005.
Burgess N, Cacucci F, Lever C, OKeefe J (2005) Characterizing mul- Cooper BG, Mizumori SJ (2001) Temporary inactivation of the ret-
tiple independent behavioral correlates of cell ring in freely rosplenial cortex causes a transient reorganization of spatial
moving animals. Hippocampus 15:149153. coding in the hippocampus. J Neurosci 21:39864001.
Buzsaki G (1989) Two-stage model of memory trace formation: a Cressant A, Muller RU, Poucet B (1997) Failure of centrally placed
role for noisy brain states. Neuroscience 31:551570. objects to control the ring elds of hippocampal place cells.
Buzsaki G (2002) Theta oscillations in the hippocampus. Neuron J Neurosci 17:25312542.
33:325340. Csicsvari J, Hirase H, Czurko A, Buzsaki G (1998) Reliability and
Buzsaki G (2005) Theta rhythm of navigation: link between path state dependence of pyramidal cell-interneuron synapses in the
integration and landmark navigation, episodic and semantic hippocampus: an ensemble approach in the behaving rat.
memory. Hippocampus 15:827840. Neuron 21:179189.
Buzsaki G, Draguhn A (2004) Neuronal oscillations in cortical net- Csicsvari J, Hirase H, Mamiya A, Buzsaki G (2000) Ensemble patterns
works. Science 304: 19261929. of hippocampal CA3-CA1 neurons during sharp wave-
Buzsaki G, Leung LW, Vanderwolf CH (1983) Cellular bases of hip- associated population events. Neuron 28:585594.
pocampal EEG in the behaving rat. Brain Res 287:139171. Csicsvari J, Jamieson B, Wise KD, Buzsaki G (2003) Mechanisms of
Buzsaki G, Horvath Z, Urioste R, Hetke J, Wise K (1992) High- gamma oscillations in the hippocampus of the behaving rat.
frequency network oscillation in the hippocampus. Science Neuron 37:311322.
256:10251027. Czurko A, Hirase H, Csicsvari J, Buzsaki G (1999) Sustained activa-
Cacucci F, Lever C, Wills TJ, Burgess N, OKeefe J (2004) Theta- tion of hippocampal pyramidal cells by space clamping in a
modulated place-by-direction cells in the hippocampal forma- running wheel. Eur J Neurosci 11:344352.
tion in the rat. J Neurosci 24:82658277. Delacour J (1984) Two neuronal systems are involved in a classical
Calton JL, Stackman RW, Goodridge JP, Archey WB, Dudchenko conditioning in the rat. Neuroscience 13:70515.
PA, Taube JS (2003) Hippocampal place cell instability after Draguhn A, Traub RD, Bibbig A, Schmitz (2000) Ripple (approxi-
542 The Hippocampus Book

mately 200-Hz) oscillations in temporal structures. J Clin Fried I, Cameron KA, Yashar S, Fong R, Morrow JW (2002)
Neurophysiol 17:361376. Inhibitory and excitatory responses of single neurons in the
Dudchenko PA, Wood ER, Eichenbaum H (2000) Neurotoxic hip- human medial temporal lobe during recognition of faces and
pocampal lesions have no effect on odor span and little effect on objects. Cereb Cortex 12:575584.
odor recognition memory but produce signicant impairments Fyhn M, Molden S, Witter MP, Moser EI, Moser MB (2004) Spatial
on spatial span, recognition, and alternation. J Neurosci representation in the entorhinal cortex. Science 305:12581264.
20:29642977. Gevins A, Smith ME, McEvoy L, Yu D (1997) High-resolution EEG
Eichenbaum H, Cohen N (2001) From conditioning to conscious mapping of cortical activation related to working memory:
recollection. Oxford, UK: Oxford University Press. effects of task difficulty, type of processing, and practice. Cereb
Eichenbaum H, Kuperstein M, Fagan A, Nagode J (1987) Cuesampl- Cortex 7:374385.
ing and goal-approach correlates of hippocampal unit activity Gilbert PE, Kesner RP (2002) Role of the rodent hippocampus in
in rats performing an odor-discrimination task. J Neurosci paired-associate learning involving associations between a stim-
7:716732. ulus and a spatial location. Behav Neurosci 116:6371.
Eichenbaum H, Wiener SI, Shapiro ML, Cohen NJ (1989) The organ- Golob EJ, Taube JS (1997) Head direction cells and episodic spatial
ization of spatial coding in the hippocampus: a study of neural information in rats without a hippocampus. Proc Natl Acad Sci
ensemble activity. J Neurosci 9:27642775. USA 94:76457650.
Ekstrom AD, Meltzer J, McNaughton BL, Barnes CA (2001) NMDA Golob EJ, Taube JS (1999) Head direction cells in rats with
receptor antagonism blocks experience-dependent expansion of hippocampal or overlying neocortical lesions: evidence for
hippocampal place elds. Neuron 31:631638. impaired angular path integration. J Neurosci 19:71987211.
Ekstrom AD, Kahana MJ, Caplan JB, Fields TA, Isham EA, Newman Goodridge JP, Taube JS (1997) Interaction between the postsubicu-
EL, Fried I (2003) Cellular networks underlying human spatial lum and anterior thalamus in the generation of head direction
navigation. Nature 425:184188. cell activity. J Neurosci 17:93159330.
Fenton AA, Muller RU (1998) Place cell discharge is extremely vari- Goodridge JP, Dudchenko PA, Worboys KA, Golob EJ, Taube JS
able during individual passes of the rat through the ring eld. (1998) Cue control and head direction cells. Behav Neurosci
Proc Natl Acad Sci USA 95:31823187. 112:749761.
Fenton AA, Csizmadia G, Muller RU (2000a) Conjoint control of Gothard KM, Skaggs WE, McNaughton BL (1996a) Dynamics of
hippocampal place cell ring by two visual stimuli. I. The effects mismatch correction in the hippocampal ensemble code for
of moving the stimuli on ring eld positions. J Gen Physiol space: interaction between path integration and environmental
116:191209. cues. J Neurosci 16:80278040.
Fenton AA, Csizmadia G, Muller RU (2000b) Conjoint control of Gothard KM, Skaggs WE, Moore KM, McNaughton BL (1996b)
hippocampal place cell ring by two visual stimuli. I. A vector- Binding of hippocampal CA1 neural activity to multiple refer-
eld theory that predicts modications of the representation of ence frames in a landmark-based navigation task. J Neurosci
the environment. J Gen Physiol 116:211221. 16:823835.
Ferbinteanu J, Shapiro ML (2003) Prospective and retrospective Grastyan, E., Lissak, K., Madarasz, I, Donhoffer, H. (1959) Hip-
memory coding in the hippocampus. Neuron 40:12271239. pocampal electrical activity during the development of
Foster TC, Christian EP, Hampson RE, Campbell KA, Deadwyler SA conditioned reflexes. Electroencephalogr Clin Neurophysiol
(1987) Sequential dependencies regulate sensory evoked 11:409430.
responses of single units in the rat hippocampus. Brain Res Gray JA (1971) Medial septal lesions, hippocampal theta rhythm and
408:8696. the control of vibrissal movement in the freelymoving rat.
Foster TC, Castro CA, McNaughton BL (1989) Spatial selectivity of Electroencephalogr Clin Neurophysiol 30:189197.
rat hippocampal neurons: dependence on preparedness for Green JD, Arduini AA (1954) Hippocampal electrical activity in
movement. Science 244:15801582. arousal. J Neurophysiol 17:533557.
Fox SE, Ranck JB Jr (1975) Localization and anatomical identica- Guzowski JF, McNaughton BL, Barnes CA, Worley PF (1999)
tion of theta and complex spike cells in dorsal hippocampal for- Environment-specic expression of the immediate-early gene
mation of rats. Exp Neurol 49:299313. Arc in hippocampal neuronal ensembles. Nat Neurosci
Fox SE, Ranck JB Jr (1981) Electrophysiological characteristics of 2:11201124.
hippocampal complex-spike cells and theta cells. Exp Brain Res Hafting T, Fyhn M, Molden S, Moser MB, Moser EI (2005)
41:399410. Microstructure of a spatial map in the entorhinal cortex. Nature
Fox SE, Wolfson S, Ranck JB Jr (1986) Hippocampal theta rhythm 436:801806.
and the ring of neurons in walking and urethane anesthetized Halgren E, Babb TL, Crandall PH (1978) Human hippocampal for-
rats. Exp Brain Res 62:495508. mation EEG desynchronizes during attentiveness and move-
Frank LM, Brown EN, Wilson M (2000) Trajectory encoding in the ment. Electroencephalogr Clin Neurophysiol 44:778781.
hippocampus and entorhinal cortex. Neuron 27:169178. Hampson RE, Simeral JD, Deadwyler SA (1999) Distribution of spa-
Frank LM, Stanley GB, Brown EN (2004) Hippocampal plasticity tial and nonspatial information in dorsal hippocampus. Nature
across multiple days of exposure to novel environments. 402:610614.
J Neurosci 24:76817689. Harris KD, Henze DA, Hirase H, Leinekugel X, Dragoi G, Czurko A,
Freeman WJ (1975) Mass action in the nervous system. San Diego: Buzsaki G (2002) Spike train dynamics predicts theta-related
Academic Press. phase precession in hippocampal pyramidal cells. Nature
Fried I, MacDonald KA, Wilson CL (1997) Single neuron activity in 417:738741.
human hippocampus and amygdala during recognition of faces Hasselmo ME (2005) What is the function of hippocampal theta
and objects. Neuron 18:753765. rhythm? Linking behavioral data to phasic properties of
 Hippocampal Neurophysiology in the Behaving Animal 543

field potential and unit recording data. Hippocampus tional control of hippocampal place elds. Exp Brain Res
15:936949. 117:131142.
Heit G, Smith ME, Halgren E (1988) Neural encoding of individual Jeffery KJ, Gilbert A, Burton S, Strudwick A (2003) Preserved per-
words and faces by the human hippocampus and amygdala. formance in a hippocampal-dependent spatial task despite
Nature 333:773775. complete place cell remapping. Hippocampus 13:175189.
Hetherington PA, Shapiro ML (1997) Hippocampal place elds are Jensen O, Lisman JE (2000) Position reconstruction from an ensem-
altered by the removal of single visual cues in a distance- ble of hippocampal place cells: contribution of theta phase cod-
dependent manner. Behav Neurosci 111:2034. ing. J Neurophysiol 83:26022609.
Hill AJ (1978) First occurrence of hippocampal spatial ring in a new Jones MW, Wilson MA (2005) Theta rhythms coordinate hippocam-
environment. Exp Neurol 62:282297. pal-prefrontal interactions in a spatial memory task. PLoS Biol
Hill AJ, Best PJ (1981) Effects of deafness and blindness on the spa- 3:e402.
tial correlates of hippocampal unit activity in the rat. Exp Neurol Jung MW, Wiener SI, McNaughton BL (1994) Comparison of spatial
74:204217. ring characteristics of units in dorsal and ventral hippocampus
Hirase H, Czurko HH, Csicsvari J, Buzsaki G (1999) Firing rate and of the rat. J Neurosci 14:73477356.
theta-phase coding by hippocampal pyramidal neurons during Jung R, Kornmuller AE (1938) Eine methodik der ableitung
space clamping. Eur J Neurosci 11:43734380. lokalisierter Potentialschwankungen aus subcorticalen Hirnge-
Hok V, Save E, Lenck-Santini PP, Poucet B (2005) Coding for spatial bieten. Arch Psychiatry 109:130.
goals in the prelimbic/infralimbic area of the rat frontal cortex. Kahana MJ, Sekuler R, Caplan JB, Kirschen M, Madsen JR (1999)
Proc Natl Acad Sci USA 102:46024607. Human theta oscillations exhibit task dependence during vir-
Hollup SA, Molden S, Donnett JG, Moser MB, Moser EI (2001) tual maze navigation. Nature 399:781784.
Accumulation of hippocampal place elds at the goal location Kali S, Dayan P (2000) The involvement of recurrent connections in
in an annular watermaze task. J Neurosci 21:16351644. area CA3 in establishing the properties of place elds: a model.
Holscher C, Anwyl R, Rowan MJ (1997) Stimulation on the positive J Neurosci 20:74637477.
phase of hippocampal theta rhythm induces long-term potenti- Kamondi A, Acsady L, Wang XJ, Buzsaki G (1998) Theta oscillations
ation that can be depotentiated by stimulation on the negative in somata and dendrites of hippocampal pyramidal cells in vivo:
phase in area CA1 in vivo. J Neurosci 17:64706477. activity-dependent phase-precession of action potentials.
Honzik CH (1936) The sensory basis of maze learning in rats. Comp Hippocampus 8:244261.
Psychol Monogr 13:1113. Kentros C, Hargreaves E, Hawkins RD, Kandel ER, Shapiro M, Muller
Hori E, Nishio Y, Kazui K, Umeno K, Tabuchi E, Sasaki K, Endo S, RV (1998) Abolition of long-term stability of new hippocampal
Ono T, Nishijo H (2005) Place-related neural responses in the place cell maps by NMDA receptor blockade. Science
monkey hippocampal formation in a virtual space. 280:21212126.
Hippocampus 15:991996. Kim JJ, Clark RE, Thompson RF (1995) Hippocampectomy impairs
Huerta PT, Lisman JE (1995) Bidirectional synaptic plasticity the memory of recently, but not remotely, acquired trace eye-
induced by a single burst during cholinergic theta oscillation in blink conditioned responses. Behav Neurosci 109:195203.
CA1 in vitro. Neuron 15:10531063. King C, Recce M, OKeefe J (1998) The rhythmicity of cells of the
Huxter J, Burgess N, OKeefe J (2003) Independent rate and tempo- medial septum/diagonal band of Broca in the awake freely mov-
ral coding in hippocampal pyramidal cells. Nature 425:828832. ing rat: relationships with behavior and hippocampal theta. Eur
Hyman JM, Wyble BP, Goyal V, Rossi CA, Hasselmo ME (2003) J Neurosci 10:464477.
Stimulation in hippocampal region CA1 in behaving rats yields Kirk IJ, McNaughton N (1991) Supramammillary cell ring and hip-
long-term potentiation when delivered to the peak of theta and pocampal rhythmical slow activity. Neuroreport 2:723725.
long-term depression when delivered to the trough. J Neurosci Klausberger T, Magill PJ, Marton LF, Roberts JD, Cobden PM,
23:1172511731. Buzsaki G, Somogyi P (2003) Brain-state- and cell-type-specic
Hyman JM, Zilli EA, Paley AM, Hasselmo ME (2005) Medial pre- firing of hippocampal interneurons in vivo. Nature
frontal cortex cells show dynamic modulation with the hip- 421:844848.
pocampal theta rhythm dependent on behavior. Hippocampus Klimesch W (1999) EEG alpha and theta oscillations reect cognitive
15:739749. and memory performance: a review and analysis. Brain Res
Jarosiewicz B, Skaggs WE (2004a) Hippocampal place cells are not Brain Res Rev 29:169195.
controlled by visual input during the small irregular activity Knierim JJ, Kudrimoti HS, McNaughton BL (1995) Place cells, head
state in the rat. J Neurosci 24:50705077. direction cells, and the learning of landmark stability. J Neurosci
Jarosiewicz B, Skaggs WE (2004b) Level of arousal during the small 15:16481659.
irregular activity state in the rat hippocampal EEG. Kobayashi T, Nishijo H, Fukuda M, Bures J, Ono T (1997) Task-
J Neurophysiol 91:26492657. dependent representations in rat hippocampal place neurons.
Jarosiewicz B, McNaughton BL, Skaggs WE (2002) Hippocampal J Neurophysiol 78:597613.
population activity during the small-amplitude irregular activ- Kocsis B, Vertes RP (1997) Phase relations of rhythmic neuronal r-
ity state in the rat. J Neurosci 22:13731384. ing in the supramammillary nucleus and mammillary body to
Jeffery KJ (1998) Learning of landmark stability and instability by the hippocampal theta activity in urethane anesthetized rats.
hippocampal place cells. Neuropharmacology 37:677687. Hippocampus 7:204214.
Jeffery KJ, Donnett JG, OKeefe J (1995) Medial septal control of Kramis R, Vanderwolf CH, Bland BH (1975) Two types of hip-
theta-correlated unit ring in the entorhinal cortex of awake pocampal rhythmical slow activity in both the rabbit and the
rats. Neuroreport 6:21662170. rat: relations to behavior and effects of atropine, diethyl ether,
Jeffery KJ, Donnett JG, Burgess N, OKeefe JM (1997) Direc- urethane, and pentobarbital. Exp Neurol 49:5885.
544 The Hippocampus Book

Kreiman G, Koch C, Fried I (2000a) Category-specic visual Lever C, Wills T, Cacucci F, Burgess N, OKeefe J (2002) Long-term
responses of single neurons in the human medial temporal lobe. plasticity in hippocampal place-cell representation of environ-
Nat Neurosci 3:946953. mental geometry. Nature 416:9094.
Kubie JL, Ranck JB Jr (1983) Sensory-behavioral correlates of indi- Louie K, Wilson MA (2001) Temporally structured replay of awake
vidual hippocampal neurons in three situations: space and con- hippocampal ensemble activity during rapid eye movement
text. In: Neurobiology of the hippocampus (Siefert W, ed), pp sleep. Neuron 29:145156.
433447, New York: Academic Press. Ludvig N, Tang HM, Gohil BC, Botero JM (2004) Detecting location-
Kudrimoti HS, Barnes CA, McNaughton BL (1999) Reactivation of specic neuronal ring rate increases in the hippocampus of
hippocampal cell assemblies: effects of behavioral state, experi- freely-moving monkeys. Brain Res 1014:97109.
ence, and EEG dynamics. J Neurosci 19:40904101. Maquet P (2001) The role of sleep in learning and memory. Science
Kyd RJ, Bilkey DK (2005) Hippocampal place cells show increased 294:10481052.
sensitivity to changes in the local environment following pre- Markus EJ, Barnes CA, McNaughton BL, Gladden VL, Skaggs WE
frontal cortex lesions. Cereb Cortex 15:720731. (1994) Spatial information content and reliability of hippocam-
Laroche S, Falcou R, Bloch V (1983) Post-trial reticular facilitation of pal CA1 neurons: effects of visual input. Hippocampus
associative changes in multiunit activity; comparison between 4:410421.
dentate gyrus and entorhinal cortex. Behav Brain Res 9:381387. Markus EJ, Qin YL, Leonard B, Skaggs WE, McNaughton BL, Barnes
Lawson VH, Bland BH (1993) The role of the septohippocampal CA (1995) Interactions between location and task affect the spa-
pathway in the regulation of hippocampal eld activity and tial and directional ring of hippocampal neurons. J Neurosci
behavior: analysis by the intraseptal microinfusion of carbachol, 15:70797094.
atropine, and procaine. Exp Neurol 120:132144. Martin PD, Berthoz A (2002) Development of spatial ring in the
Leblanc MO, Bland BH (1979) Developmental aspects of hippocam- hippocampus of young rats. Hippocampus 12:465480.
pal electrical activity and motor behavior in the rat Exp Neurol Matthews RT, Lee WL (1991) A comparison of extracellular and
66:220237. intracellular recordings from medial septum/diagonal band
Lee AK, Wilson MA (2002) Memory of sequential experience in the neurons in vitro. Neuroscience 42:451462.
hippocampus during slow wave sleep. Neuron 36:11831194. Maurer AP, Vanrhoads SR, Sutherland GR, Lipa P, McNaughton
Lee I, Rao G, Knierim JJ (2004) A double dissociation between hip- BL (2005) Self-motion and the origin of differential spatial
pocampal subelds: differential time course of CA3 and CA1 scaling along the septo-temporal axis of the hippocampus.
place cells for processing changed environments. Neuron Hippocampus 15:841852.
42:803815. McEchron MD, Disterhoft JF (1997) Sequence of single neuron
Lenck-Santini PP, Save E, Poucet B (2001a) Evidence for a relation- changes in CA1 hippocampus of rabbits during acquisi-
ship between place-cell spatial ring and spatial memory per- tion of trace eyeblink conditioned responses. J Neurophysiol
formance. Hippocampus 11:377390. 78:10301044.
Lenck-Santini PP, Save E, Poucet B (2001b) Place-cell ring does not McEchron MD, Tseng W, Disterhoft JF (2003) Single neurons in CA1
depend on the direction of turn in a Y-maze alternation task. hippocampus encode trace interval duration during trace heart
Eur J Neurosci 13:10551058. rate (fear) conditioning in rabbit. J Neurosci 23:15351547.
Lenck-Santini PP, Muller RU, Save E, Poucet B (2002) Relationships McHugh TJ, Blum KI, Tsien JZ, Tonegawa S, Wilson MA (1996)
between place cell ring elds and navigational decisions by Impaired hippocampal representation of space in CA1-specic
rats. J Neurosci 22:90359047. NMDAR1 knockout mice. Cell 87:13391349.
Lengyel M, Szatmary Z, Erdi P. (2003) Dynamically detuned oscilla- McNaughton BL, Barnes CA, OKeefe J (1983a) The contributions
tors account for the coupled rate and temporal code of place cell of position, direction, and velocity to single unit activity in
ring. Hippocampus 13:700714. the hippocampus of freely-moving rats. Exp Brain Res 52:
Leonard BJ, McNaughton BL, Barnes CA (1987) Suppression of hip- 4149.
pocampal synaptic plasticity during slow-wave sleep. Brain Res McNaughton BL, OKeefe J, Barnes CA (1983b) The stereotrode:
425:174177. a new technique for simultaneous isolation of several single
Leung LW (1985) Spectral analysis of hippocampal EEG in the freely units in the central nervous system from multiple unit records.
moving rat: effects of centrally active drugs and relations to J Neurosci Methods 8:391397.
evoked potentials. Electroencephalogr Clin Neurophysiol McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW,
60:6577. Knierim JJ, Kudrimoti H, Qin Y, Skaggs WE, Suster M, Weaver
Leung LS (1992) Fast (beta) rhythms in the hippocampus: a review. KL (1996) Deciphering the hippocampal polyglot: the hip-
Hippocampus 2:9398. pocampus as a path integration system. J Exp Biol 199(Pt
Leung LW, Borst JG (1987) Electrical activity of the cingulate cortex. 1):173185.
I. Generating mechanisms and relations to behavior. Brain Res Mehta MR, Barnes CA, McNaughton BL (1997) Experience-
407:6880. dependent, asymmetric expansion of hippocampal place elds.
Leutgeb S, Mizumori SJ (1999) Excitotoxic septal lesions result in Proc Natl Acad Sci USA 94:89188921.
spatial memory decits and altered exibility of hippocampal Mehta MR, Quirk MC, Wilson MA (2000) Experience-dependent
single-unit representations. J Neurosci 19:66616672. asymmetric shape of hippocampal receptive elds. Neuron
Lever C, Cacucci F, Wills TJ, Burton S, McClelland, Burgess N, 25:707715.
OKeefe J (2003) Spatial coding in the hippocampal formation: Mehta MR, Lee AK, Wilson MA (2002) Role of experience and oscil-
input, information type, plasticity, and behavior. In: The neuro- lations in transforming a rate code into a temporal code. Nature
biology of spatial behavior (Jeffrey KJ, ed), pp 199225. Oxford, 417:741746.
UK: Oxford University Press. Miller VM, Best PJ (1980) Spatial correlates of hippocampal unit
 Hippocampal Neurophysiology in the Behaving Animal 545

activity are altered by lesions of the fornix and endorhinal cor- wave-ripple complexes in the absence of synaptic inhibition.
tex. Brain Res 194:311323. J Physiol 563(Pt 3):663670.
Mitchell SJ, Ranck JB Jr (1980) Generation of theta rhythm in Nishijo H, Ono T, Eifuku S, Tamura R (1997) The relationship
medial entorhinal cortex of freely moving rats. Brain Res between monkey hippocampus place-related neural activity and
189:4966. action in space. Neurosci Lett 226:5760.
Mizumori SJ, Williams JD (1993) Directionally selective mnemonic Nitz D, McNaughton B (2004) Differential modulation of CA1 and
properties of neurons in the lateral dorsal nucleus of the thala- dentate gyrus interneurons during exploration of novel envi-
mus of rats. J Neurosci 13:40154028. ronments. J Neurophysiol 91:863872.
Mizumori SJ, Barnes CA, McNaughton BL (1989) Reversible inacti- Nowicka A, Ringo JL (2000) Eye position-sensitive units in hip-
vation of the medial septum: selective effects on the sponta- pocampal formation and in inferotemporal cortex of the
neous unit activity of different hippocampal cell types. Brain macaque monkey. Eur J Neurosci 12:751759.
Res 500:99106. Nunez A, de Andres, I, Garcia-Austt E (1991) Relationships of
Mizumori SJ, Ward KE, Lavoie AM (1992) Medial septal modulation nucleus reticularis pontis oralis neuronal discharge with sensory
of entorhinal single unit activity in anesthetized and freely mov- and carbachol evoked hippocampal theta rhythm. Exp Brain Res
ing rats. Brain Res 570:188197. 87:303308.
Mizumori SJ, Lavoie AM, Kalyani A (1996) Redistribution of spatial OKeefe J (1976) Place units in the hippocampus of the freely mov-
representation in the hippocampus of aged rats performing a ing rat. Exp Neurol 51:78109.
spatial memory task. Behav Neurosci 110:10061016. OKeefe J (1991a) The hippocampal cognitive map and navigational
Moita MA, Rosis S, Zhou Y, LeDoux JE, Blair HT (2003) Hippocam- strategies. In: Brain and space (Paillard J, ed), pp 273295.
pal place cells acquire location-specic responses to the condi- Oxford, UK: Oxford University Press.
tioned stimulus during auditory fear conditioning. Neuron OKeefe J (1991b) An allocentric spatial model for the hippocampal
37:485497. cognitive map. Hippocampus 1:230235.
Moita MA, Rosis S, Zhou Y, LeDoux JE, Blair HT (2004) Putting fear OKeefe J (1996) The spatial prepositions in English, vector grammar
in its place: remapping of hippocampal place cells during fear and the cognitive map theory. In: Language and space (Bloom P,
conditioning. J Neurosci 24:70157023. Peterson M, Nadel L, Garrett M, eds), pp 277316. Cambridge,
Muir GM, Bilkey DK (1998) Synchronous modulation of perirhinal MA: MIT Press.
cortex neuronal activity during cholinergically mediated (type OKeefe J (1999) Do hippocampal pyramidal cells signal non-spatial
II) hippocampal theta. Hippocampus 8:526532. as well as spatial information? Hippocampus 9:352364.
Muir GM, Bilkey DK (2001) Instability in the place eld location of OKeefe J (2001) Vector grammar, places, and the functional role of
hippocampal place cells after lesions centered on the perirhinal the spatial prepositions in English. In: Axes and vectors in lan-
cortex. J Neurosci 21:40164025. guage and space (van der Zee E, Slack J, eds). Oxford, UK:
Muller RU, Kubie JL (1987) The effects of changes in the environ- Oxford University Press.
ment on the spatial ring of hippocampal complex-spike cells. OKeefe J, Burgess N (1996) Geometric determinants of the place
J Neurosci 7:19511968. elds of hippocampal neurons. Nature 381:425428.
Muller RU, Kubie JL, Ranck JB Jr (1987) Spatial ring patterns of hip- OKeefe J, Conway DH (1978) Hippocampal place units in the freely
pocampal complex-spike cells in a xed environment. J Neurosci moving rat: why they re where they re. Exp Brain Res
7:19351950. 31:573590.
Muller RU, Bostock E, Taube JS, Kubie JL (1994) On the directional OKeefe J, Dostrovsky J (1971) The hippocampus as a spatial map:
firing properties of hippocampal place cells. J Neurosci preliminary evidence from unit activity in the freely-moving
14:72357251. rat. Brain Res 34:171175.
Murray EA, Mishkin M (1998) Object recognition and location OKeefe J, Nadel L (1978) The hippocampus as a cognitive map.
memory in monkeys with excitotoxic lesions of the amygdala Oxford, UK: Oxford University Press.
and hippocampus. J Neurosci 18:65686582. OKeefe J, Nadel L (1979) Precis of OKeefe and Nadels The
Nadel L, Willner J, Kurz E (1985) Cognitive maps and environmental Hippocampus as a Cognitive Map. Behav Brain Sci 2:487533.
contex. In: Context and learning (Balsam P, Tomie A, eds), pp OKeefe J, Recce ML (1993) Phase relationship between hippocam-
385406. Hillside, NJ: Erlbaum. pal place units and the EEG theta rhythm. Hippocampus
Nakazawa K, Quirk MC, Chitwood RA, Watanabe M, Yeckel MF, Sun 3:317330.
LD, Kato A, Carr CA, Johnston D, Wilson MA, Tonegawa S OKeefe J, Speakman A (1987) Single unit activity in the rat hip-
(2002) Requirement for hippocampal CA3 NMDA receptors in pocampus during a spatial memory task. Exp Brain Res 68:127.
associative memory recall. Science 297:211218. OKeefe J, Burgess N, Donnett JG, Jeffery KJ, Maguire EA (1998)
Nakazawa K, Sun LD, Quirk MC, Rondi-Reig L, Wilson MA, Place cells, navigational accuracy, and the human hippocampus.
Tonegawa S (2003) Hippocampal CA3 NMDA receptors are Philos Trans R Soc Lond B Biol Sci 353:13331340.
crucial for memory acquisition of one-time experience. Neuron OMara SM, Rolls ET, Berthoz A, Kesner RP (1994) Neurons
38:305315. responding to whole-body motion in the primate hippocam-
Nakazawa K, McHugh TJ, Wilson MA, Tonegawa S (2004) NMDA pus. J Neurosci 14:65116523.
receptors, place cells and hippocampal spatial memory. Nat Rev Olton DS, Branch M, Best PJ (1978) Spatial correlates of hippocam-
Neurosci 5:361372. pal unit activity. Exp Neurol 58:387409.
Natsume K, Hallworth NE, Szgatti TL, Bland BH (1999) Hippocam- Olypher AV, Lansky P, Fenton AA (2002) Properties of the extra-
pal theta-related cellular activity in the superior colliculus of the positional signal in hippocampal place cell discharge derived
urethane-anesthetized rat. Hippocampus 9:500509. from the overdispersion in location-specic ring. Neuroscience
Nimmrich V, Maier N, Schmitz D, Draguhn A (2005) Induced sharp 111:553566.
546 The Hippocampus Book

Olypher AV, Klement D, Fenton AA (2006) Cognitive disorganization Ranck JB (1984) Head-direction cells in the deep cell layers of the
in hippocampus: a physiological model of the disorganization dorsal presubiculum in freely moving rats. Soc Neurosci Abstr
in psychosis. J Neurosci 26:158168. 10:599.
Ono T, Nakamura K, Nishijo H, Eifuku S (1993) Monkey hippocam- Rawlins JNP (1985) Associations across time: the hippocampus as a
pal neurons related to spatial and nonspatial functions. temporary memory store. Brain Behav Sci 8:479496.
J Neurophysiol 70:15161529. Redish AD, Touretzky DS (1996) Modeling interactions of the rats
Otto T, Eichenbaum H (1992) Neuronal activity in the hippocampus place and head direction systems. In: Neural information pro-
during delayed non-match to sample performance in rats: evi- cessing systems 8 (Touretzky DS, Mozer MC, Hasselmo ME, eds),
dence for hippocampal processing in recognition memory. pp 6167. Cambridge, MA: MIT Press.
Hippocampus 2:323334. Redish AD, Battaglia FP, Chawla MK, Ekstrom AD, Gerrard JL, Lipa
Pang K, Rose GM (1987) Differential effects of norepinephrine on P, Rosenzweig ES, Worley PF, Guzowski JF, McNaughton BL,
hippocampal complex-spike and theta-neurons. Brain Res Barnes CA (2001) Independence of ring correlates of anatom-
425:146158. ically proximate hippocampal pyramidal cells. J Neurosci
Pang K, Rose GM (1989a) Differential effects of methionine5- 21:RC134.
enkephalin on hippocampal pyramidal cells and interneurons. Riches IP, Wilson FA, Brown MW (1991) The effects of visual stimu-
Neuropharmacology 28:11751181. lation and memory on neurons of the hippocampal formation
Par D, Gaudreau H (1996) Projection cells and interneurons of the and the neighboring parahippocampal gyrus and inferior tem-
lateral and basolateral amygdala: distinct ring patterns and dif- poral cortex of the primate. J Neurosci 11:17631779.
ferential relation to theta and delta rhythms in conscious cats. Ringo JL, Sobotka S, Diltz MD, Bunce CM (1994) Eye movements
J Neurosci 16:33343350. modulate activity in hippocampal, parahippocampal, and
Pavlides C, Winson J (1989) Inuences of hippocampal place cell r- inferotemporal neurons. J Neurophysiol 71:12851288.
ing in the awake state on the activity of these cells during subse- Rivas J, Gaztelu JM, Garcia-Austt E (1996) Changes in hippocampal
quent sleep episodes. J Neurosci 9:29072918. cell discharge patterns and theta rhythm spectral properties as a
Pavlides C, Greenstein YJ, Grudman M, Winson J (1988) Long-term function of walking velocity in the guinea pig. Exp Brain Res
potentiation in the dentate gyrus is induced preferentially on 108:113118.
the positive phase of theta-rhythm. Brain Res 439:383387. Rolls ET, Cahusac PM, Feigenbaum JD, Miyashita Y (1993) Responses
Paz-Villagran V, Lenck-Santini PP, Save E, Poucet B (2002) Properties of single neurons in the hippocampus of the macaque related to
of place cell ring after damage to the visual cortex. Eur recognition memory. Exp Brain Res 93:299306.
J Neurosci 16:771776. Rolls ET, Robertson RG, Georges-Francois P (1997) Spatial view cells
Pedemonte M, Pena JL, Velluti RA (1996) Firing of inferior colliculus in the primate hippocampus. Eur J Neurosci 9:17891794.
auditory neurons is phase-locked to the hippocampus theta Rose GM, Pang KC (1989) Differential effect of norepinephrine upon
rhythm during paradoxical sleep and waking. Exp Brain Res granule cells and interneurons in the dentate gyrus. Brain Res
112:4146. 488:353356.
Penick S, Solomon PR (1991) Hippocampus, context, and condition- Rosenzweig ES, Barnes CA (2003) Impact of aging on hippocampal
ing. Behav Neurosci 105:611617. function: plasticity, network dynamics, and cognition. Prog
Phillips RG, LeDoux JE (1994) Lesions of the dorsal hippocampal Neurobiol 69:143179.
formation interfere with background but not foreground con- Rosenzweig ES, Redish AD, McNaughton BL, Barnes CA (2003)
textual fear conditioning. Learn Mem 1:3444. Hippocampal map realignment and spatial learning. Nat
Pickenhain, L., and Klingberg, F. (1967) Hippocampal slow wave Neurosci 6:609615.
activity as a correlate of basic behavioural mechanisms in the Rossier J, Kaminsky Y, Schenk F, Bures J (2000) The place preference
rat. In Progress in brain research, Vol. 27 (Adley WR, Tokizane T, task: a new tool for studying the relation between behavior and
eds), pp 21827. Elsevier, Amsterdam. place cell activity in rats. Behav Neurosci 114:273284.
Poe GR, Nitz DA, McNaughton BL, Barnes CA (2000) Experience- Rotenberg A, Mayford M, Hawkins RD, Kandel ER, Muller RU (1996)
dependent phase-reversal of hippocampal neuron ring during Mice expressing activated CaMKII lack low frequency LTP and
REM sleep. Brain Res 855:176180. do not form stable place cells in the CA1 region of the hip-
Powell DA, Joseph JA (1974) Autonomic-somatic interaction and pocampus. Cell 87:13511361.
hippocampal theta activity. J Comp Physiol Psychol 87:978986. Rotenberg A, Abel T, Hawkins RD, Kandel ER, Muller RU (2000)
Qin YL, McNaughton BL, Skaggs WE, Barnes CA (1997) Memory Parallel instabilities of long-term potentiation, place cells,
reprocessing in corticocortical and hippocampocortical neu- and learning caused by decreased protein kinase A activity.
ronal ensembles. Philos Trans R Soc Lond B Biol Sci 352: J Neurosci 20:80968102.
15251533. Russell NA, Horii A, Smith PF, Darlington CL, Bilkey DK (2003)
Quirk GJ, Muller RU, Kubie JL (1990) The ring of hippocampal Long-term effects of permanent vestibular lesions on hip-
place cells in the dark depends on the rats recent experience. pocampal spatial ring. J Neurosci 23:64906498.
J Neurosci 10:20082017. Sainsbury RS, Harris JL, Rowland GL (1987a) Sensitization and hip-
Quirk GJ, Muller RU, Kubie JL, Ranck JB Jr (1992) The positional r- pocampal type 2 theta in the rat. Physiol Behav 41:489493.
ing properties of medial entorhinal neurons: description and Sainsbury RS, Heynen A, Montoya CP (1987b) Behavioral correlates
comparison with hippocampal place cells. J Neurosci 12: of hippocampal type 2 theta in the rat. Physiol Behav
19451963. 39:513519.
Ranck JB (1973) Studies on single neurons in dorsal hippocampal Salzmann E, Vidyasagar TR, Creutzfeldt OD (1993) Functional com-
formation and septum in unrestrained rats. I. Behavioral corre- parison of neuronal properties in the primate posterior
lates and ring repertoires. Exp Neurol 41:461531. hippocampus and parahippocampus (area TF/TH) during dif-
 Hippocampal Neurophysiology in the Behaving Animal 547

ferent behavioral paradigms involving memory and selective Skaggs WE, McNaughton BL (1996) Replay of neuronal ring
attention. Behav Brain Res 53:133149. sequences in rat hippocampus during sleep following spatial
Save E, Cressant A, Thinus-Blanc C, Poucet B (1998) Spatial ring experience. Science 271:18701873.
of hippocampal place cells in blind rats. J Neurosci 18: Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta
18181826. phase precession in hippocampal neuronal populations and
Save E, Nerad L, Poucet B (2000) Contribution of multiple sensory the compression of temporal sequences. Hippocampus 6:
information to place eld stability in hippocampal place cells. 149172.
Hippocampus 10:6476. Slawinska U, Kasicki S (1995) Theta-like rhythm in depth EEG activ-
Save E, Paz-Villagran V, Alexinsky T, Poucet B (2005) Functional ity of hypothalamic areas during spontaneous or electrically
interaction between the associative parietal cortex and hip- induced locomotion in the rat. Brain Res 678:117126.
pocampal place cell ring in the rat. Eur J Neurosci 21:522530. Slawinska U, Kasicki S (1998) The frequency of rats hippocampal
Schenk F (1985) Development of place navigation in rats from wean- theta rhythm is related to the speed of locomotion. Brain Res
ing to puberty. Behav Neural Biol 43:6985. 796:327331.
Seidenbecher T, Laxmi TR, Stork O, Pape HC (2003) Amygdalar and Sobotka S, Nowicka A, Ringo JL (1997) Activity linked to externally
hippocampal theta rhythm synchronization during fear mem- cued saccades in single units recorded from hippocampal,
ory retrieval. Science 301:846850. parahippocampal, and inferotemporal areas of macaques.
Shapiro ML, Tanila H, Eichenbaum H (1997) Cues that hippocampal J Neurophysiol 78:21562163.
place cells encode: dynamic and hierarchical representation of Speakman A, OKeefe J (1990) Hippocampal complex spike cells do
local and distal stimuli. Hippocampus 7:624642. not change their place elds if the goal is moved within a cue
Sharp PE (1996) Multiple spatial/behavioral correlates for cells in the controlled environment. Eur J Neurosci 7:544555.
rat postsubiculum: multiple regression analysis and comparison Stackman RW, Taube JS (1997) Firing properties of head direction
to other hippocampal areas. Cereb Cortex 6:238259. cells in the rat anterior thalamic nucleus: dependence on
Sharp PE (1997) Subicular cells generate similar spatial ring pat- vestibular input. J Neurosci 17:43494358.
terns in two geometrically and visually distinctive environ- Stackman RW, Taube JS (1998) Firing properties of rat lateral mam-
ments: comparison with hippocampal place cells. Behav Brain millary single units: head direction, head pitch, and angular
Res 85:7192. head velocity. J Neurosci 18:90209037.
Sharp PE (1999a) Comparison of the timing of hippocampal and Stackman RW, Clark AS, Taube JS (2002) Hippocampal spatial repre-
subicular spatial signals: implications for path integration. sentations require vestibular input. Hippocampus 12: 291303.
Hippocampus 9:158172. Stewart M, Fox SE (1991) Hippocampal theta activity in monkeys.
Sharp PE (1999b) Subicular place cells expand or contract their spa- Brain Res 538:5963.
tial ring pattern to t the size of the environment in an open Stewart M, Luo Y, Fox SE (1992) Effects of atropine on hippocampal
eld but not in the presence of barriers: comparison with hip- theta cells and complex-spike cells. Brain Res 591:122128.
pocampal place cells. Behav Neurosci 113:643662. Stickgold R, Hobson JA, Fosse R, Fosse M (2001) Sleep, learning,
Sharp PE, Green C (1994) Spatial correlates of ring patterns of sin- and dreams: off-line memory reprocessing. Science 294:
gle cells in the subiculum of the freely moving rat. J Neurosci 10521057.
14:23392356. Tamura R, Ono T, Fukuda M, Nishijo H (1991) Role of monkey hip-
Sharp PE, Blair HT, Etkin D, Tzanetos DB (1995) Inuences of pocampus in recognition of food and nonfood. Brain Res Bull
vestibular and visual motion information on the spatial ring 27:457461.
patterns of hippocampal place cells. J Neurosci 15:173189. Tamura R, Ono T, Fukuda M, Nakamura K (1992) Spatial respon-
Sharp PE, Tinkelman A, Cho J (2001) Angular velocity and head siveness of monkey hippocampal neurons to various visual and
direction signals recorded from the dorsal tegmental nucleus of auditory stimuli. Hippocampus 2:307322.
gudden in the rat: implications for path integration in the head Tanila H, Shapiro M, Gallagher M, Eichenbaum H (1997a) Brain
direction cell circuit. Behav Neurosci 115:571588. aging: changes in the nature of information coding by the hip-
Siapas AG, Wilson MA (1998) Coordinated interactions between hip- pocampus. J Neurosci 17:51555166.
pocampal ripples and cortical spindles during slow-wave sleep. Tanila H, Shapiro ML, Eichenbaum H (1997b) Discordance of spatial
Neuron 21:11231128. representation in ensembles of hippocampal place cells.
Siapas AG, Lubenov EV, Wilson MA (2005) Prefrontal phase locking Hippocampus 7:613623.
to hippocampal theta oscillations. Neuron 46:141151. Taube JS (1995a) Head direction cells recorded in the anterior thala-
Sinclair BR, Seto MG, Bland BH (1982) Theta-cells in CA1 and den- mic nuclei of freely moving rats. J Neurosci 15:7086.
tate layers of hippocampal formation: relations to slow- Taube JS (1995b) Place cells recorded in the parasubiculum of freely
wave activity and motor behavior in the freely moving rabbit. moving rats. Hippocampus 5:569583.
J Neurophysiol 48:12141225. Taube JS (1998) Head direction cells and the neuropsychological
Singer W, Gray CM (1995) Visual feature integration and the tempo- basis for a sense of direction. Prog Neurobiol 55:225256.
ral correlation hypothesis. Annu Rev Neurosci 18:555586. Taube JS, Burton HL (1995) Head direction cell activity monitored
Sinnamon HM (2000) Priming pattern determines the correlation in a novel environment and during a cue conict situation.
between hippocampal theta activity and locomotor stepping J Neurophysiol 74:19531971.
elicited by stimulation in anesthetized rats. Neuroscience Taube JS, Muller RU (1998) Comparisons of head direction cell activ-
98:459470. ity in the postsubiculum and anterior thalamus of freely mov-
Sinnamon HM, Jassen AK, Ilch C (2000) Hippocampal theta activity ing rats. Hippocampus 8:87108.
and facilitated locomotor stepping produced by GABA injec- Taube JS, Muller RU, Ranck JB Jr (1990a) Head-direction
tions in the midbrain raphe region. Behav Brain Res 107:93103. cells recorded from the postsubiculum in freely moving rats.
548 The Hippocampus Book

I. Description and quantitative analysis. J Neurosci 10: Wiener SI, Korshunov VA, Garcia R, Berthoz A (1995) Inertial, sub-
420435. stratal and landmark cue control of hippocampal CA1 place cell
Taube JS, Muller RU, Ranck JB Jr (1990b) Head-direction cells activity. Eur J Neurosci 7:22062219.
recorded from the postsubiculum in freely moving rats. II. Wills TJ, Lever C, Cacucci F, Burgess N, OKeefe J (2005) Attractor
Effects of environmental manipulations. J Neurosci 10: 436447. dynamics in the hippocampal representation of the local envi-
Tesche CD, Karhu J (2000) Theta oscillations index human hip- ronment. Science 308:873876.
pocampal activation during a working memory task. Proc Natl Wilson FA, Riches IP, Brown MW (1990) Hippocampus and medial
Acad Sci USA 97:919924. temporal cortex: neuronal activity related to behavioral
Thompson LT, Best PJ (1989) Place cells and silent cells in the hip- responses during the performance of memory tasks by pri-
pocampus of freely-behaving rats. J Neurosci 9:23822390. mates. Behav Brain Res 40:728.
Thompson RF (1976) The search for the engram. Am Psychol Wilson MA, McNaughton BL (1993) Dynamics of the hippocampal
31:209227. ensemble code for space. Science 261:10551058.
Tsien JZ, Huerta PT, Tonegawa S (1996) The essential role of hip- Wilson MA, McNaughton BL (1994) Reactivation of hippocampal
pocampal CA1 NMDA receptor-dependent synaptic plasticity ensemble memories during sleep. Science 265:676679.
in spatial memory. Cell 87:13271338. Wirth S, Yanike M, Frank LM, Smith AC, Brown EN, Suzuki WA
Vanderwolf CH (1969) Hippocampal electrical activity and volun- (2003) Single neurons in the monkey hippocampus and learn-
tary movement in the rat. Electroencephalogr Clin Neurophysiol ing of new associations. Science 300:15781581.
26:407418. Wood ER, Dudchenko PA, Eichenbaum H (1999) The global record
Vanderwolf CH (2001) The hippocampus as an olfacto-motor mech- of memory in hippocampal neuronal activity. Nature
anism: were the classical anatomists right after all? Behav Brain 397:613616.
Res 127:2547. Wood ER, Dudchenko PA, Robitsek RJ, Eichenbaum H (2000)
Vidyasagar TR, Salzmann E, Creutzfeldt OD (1991) Unit activity in Hippocampal neurons encode information about different
the hippocampus and the parahippocampal temporobasal asso- types of memory episodes occurring in the same location.
ciation cortex related to memory and complex behavior in the Neuron 27:623633.
awake monkey. Brain Res 544:269278. Yamaguchi Y, Aota Y, McNaughton BL, Lipa P (2002) Bimodality of
Vinogradova OS, Brazhnik ES (1977) Neuronal aspects of septo- theta phase precession in hippocampal place cells in freely run-
hippocampal relations Ciba Found Symp 145177. ning rats. J Neurophysiol 87:26292642.
Wallenstein GV, Eichenbaum H, Hasselmo ME (1998) The hip- Ylinen A, Bragin A, Nadasdy Z, Jando G, Szabo I, Sik A, Buzsaki G
pocampus as an associator of discontiguous events. Trends (1995) Sharp wave-associated high-frequency oscillation (200
Neurosci 21:317323. Hz) in the intact hippocampus: network and intracellular
Watanabe T, Niki H (1985) Hippocampal unit activity and delayed mechanisms. J Neurosci 15:3046.
response in the monkey. Brain Res 325:241254. Zhang K (1996) Representation of spatial orientation by the intrinsic
Weiss C, Kronforst-Collins MA, Disterhoft JF (1996) Activity of hip- dynamics of the head-direction cell ensemble: a theory.
pocampal pyramidal neurons during trace eyeblink condition- J Neurosci 16:21122126.
ing. Hippocampus 6:192209. Zhang K, Ginzburg I, McNaughton BL, Sejnowski TJ (1998)
Wesierska M, Dockery C, Fenton AA (2005) Beyond memory, navi- Interpreting neuronal population activity by reconstruction:
gation, and inhibition: behavioral evidence for hippocampus- unied framework with application to hippocampal place cells.
dependent cognitive coordination in the rat. J Neurosci J Neurophysiol 79:10171044.
25:24132419. Zinyuk L, Kubik S, Kaminsky Y, Fenton AA, Bures J (2000) Under-
Whishaw IQ, Vanderwolf CH (1971) Hippocampal EEG and behav- standing hippocampal activity by using purposeful behavior:
ior: effects of variation in body temperature and relation of EEG place navigation induces place cell discharge in both task-rele-
to vibrissae movement, swimming and shivering. Physiol Behav vant and task- irrelevant spatial reference frames. Proc Natl Acad
6:391397. Sci USA 97:37713776.
Whishaw IQ, Vanderwolf CH (1973) Hippocampal EEG and behav- Zugaro MB, Tabuchi E, Wiener SI (2000) Inuence of conicting
ior: changes in amplitude and frequency of RSA (theta rhythm) visual, inertial and substratal cues on head direction cell activ-
associated with spontaneous and learned movement patterns in ity. Exp Brain Res 133:198208.
rats and cats. Behav Biol 8:461484. Zugaro MB, Berthoz A, Wiener SI (2001) Background, but not fore-
Wiebe SP, Staubli UV (1999) Dynamic ltering of recognition mem- ground, spatial cues are taken as references for head direction
ory codes in the hippocampus. J Neurosci 19:1056210574. responses by rat anterodorsal thalamus neurons. J Neurosci
Wiener SI (1993) Spatial and behavioral correlates of striatal neurons 21:RC154.
in rats performing a self-initiated navigation task. J Neurosci Zugaro MB, Arleo A, Berthoz A, Wiener SI (2003) Rapid spatial reori-
13:38023817. entation and head direction cells. J Neurosci 23:34783482.
Wiener SI, Paul CA, Eichenbaum H (1989) Spatial and behavioral Zugaro MB, Monconduit L, Buzsaki G (2005) Spike phase precession
correlates of hippocampal neuronal activity. J Neurosci persists after transient intrahippocampal perturbation. Nat
9:27372763. Neurosci 8:6771.
12 Craig Stark

Functional Role of the Human Hippocampus

divided into temporopolar and posterior regions), and

12.1 Overview parahippocampal cortices.
As is discussed here and in Chapter 13, there have been a
This chapter examines data from human amnesic patients, number of attempts to arrive at a description of hippocampal
data from electrophysiological recordings in humans, and function. One class of theories proposes a functional distinc-
data from functional neuroimaging studies that attempt to tion that qualitatively sets the hippocampal region apart from
shed light on the question, What does the human hippocampus the adjacent structures in the medial temporal lobe, dening
do? In broad terms, we have learned a great deal about the what it is that the hippocampal region specically does that
kinds of memory in which the hippocampus is and is not other structures do not and vice versa. Such theories have
involved. The hippocampus is part of a system that plays a linked the function of the hippocampal region to conjunctive
critical role in the encoding and retrieval of long-term mem- (e.g., Sutherland and Rudy, 1989; OReilly and Rudy, 2000),
ory for facts and events. As a whole, the system is vital for this relational (e.g., Cohen et al., 1997), and/or episodic, recollec-
declarative or explicit form of memory but is not involved tive, or associative (e.g., Brown and Aggleton, 2001) forms of
in other forms of long term memory, in nonmnemonic memory. A second theory (the declarative memory hypothe-
aspects of cognition, or in immediate (or working) memory. sissee Chapter 13) is more quantitative in nature in sug-
Furthermore, its involvement in declarative memory is not gesting that the hippocampal region combines and extends
permanent but is time-limited in nature (the bases for each of the processing of the adjacent cortical structures that together
these conclusions are laid out later in the chapter). form the medial temporal lobe memory system (Squire and
We therefore know a great deal about the role of this sys- Zola-Morgan, 1991).
tem in human memory. Our understanding of the roles The conclusion on this question is that the data from
played by the individual components of the system is much human studies do not unambiguously support any of the
less complete. In the human, the system is often described as existing theories. In particular, the data suggest differentiation
consisting of two major sets of structures. One set of struc- of function within the medial temporal lobe but not along
tures, often dened as the hippocampal region, consists of versions of any of the binary dissociations proposed. The dif-
the CA elds of the hippocampus itself, the dentate gyrus, and ferentiation of function is likely to be a complex and graded
the subiculum. For obvious reasons, most of the studies in one that is most consistent with the second theorythat the
humans do not contain histological analyses. Unfortunately, hippocampal region combines and extends the processing of
without histological analyses, it is quite difficult to isolate the adjacent cortical structures. However, this theory is still
these components from each other (see Sections 12.3.3 and underspecied, as we do not have the data at hand to under-
12.3.5), so most studies that report data from the hippocam- stand the precise nature of the way in which the hippocampus
pus include data from the more extended hippocampal combines and extends adjacent processing or of the dynamic
region. The second set of structures consist of the adjacent interplay among the structures. Thus, although we have cer-
cortical structures that lie along the parahippocampal gyrus. tainly learned a great deal about the human hippocampus and
This set consists of the entorhinal, perirhinal (occasionally its role in memory, but it would be a mistake to believe that we

549
550 The Hippocampus Book

have a full and satisfactory answer to the fundamental ques- other patients with varying degrees and locations of resection.
tion: What does the human hippocampus do? This seminal work concluded with the following text.

Bilateral medial temporal lobe resection in man results

in a persistent impairment of recent memory when-
12.2 Patient H.M. ever the removal is carried far enough posteriorly to
damage portions of the anterior hippocampus and
It would be almost impossible to begin a discussion of the role hippocampal gyrus. In two cases in which bilateral
of the hippocampal region in human memory without con- resection was carried to a distance of 8 cm posterior
sidering the patient H.M. In a successful attempt to relieve to the temporal tips the loss was particularly severe.
otherwise intractable epilepsy, H.M. underwent bilateral The memory loss in these cases of medial temporal
resection of the medial portions of the temporal lobe (Fig. lobe excision involved both anterograde and some ret-
121). Although scattered hints in the literature prior to this rograde amnesia, but left early memories and technical
suggested that damage to the hippocampal region might affect skills intact. There was no deterioration in personality
memory, the detailed study of H.M. had an inuence on both or general intelligence, and no complex perceptual dis-
the study of memory and on neurosurgical practice that was turbance such as is seen after a more complete bilateral
unprecedented. Following a brief initial report (Scoville, temporal lobectomy. It is concluded that the anterior
1954) describing the profound memory loss that resulted hippocampus and [para]hippocampal gyrus, either
from the surgery, Scoville and Milner (1957) presented a separately or together, are critically concerned in the
detailed neuropsychological assessment of H.M. and nine retention of current experience. It is not known

Figure 121. Top. Extent of H.M.s lesion is shown by comparing are identied, and all show signs of damage in H.M. Bottom.
the coronal magnetic resonance imaging (MRI) scan of H.M. (left) Summary of the extent of the lesion. Both hemispheres are dam-
with that of a matched healthy volunteer (right). The hippocampus aged, with one shown intact for comparison only. (Source: Corkin et
(H), amygdala (A), collateral sulcus (CS), perirhinal cortex (PR), al., 1997, with permission. 1997 Society for Neuroscience.)
entorhinal cortex (EC), and medial mammillary nucleus (MMN)
 Functional Role of the Human Hippocampus 551

whether the amygdala plays any part in this mecha- nition. Third, she noted that short-term memory remained
nism, since the hippocampal complex has not been intact, indicating clear dissociation between immediate, or
removed alone, but always together with uncus and working, memory and permanent, long-term memory.
amygdala. (Scoville and Milner, 1957, p. 21) Finally, she noted that damage to the medial portions of the
temporal lobe did not abolish all forms of long-term learning
Subsequent testing of H.M. was able to reveal not only the and memory, indicating that the medial temporal lobes were
breadth and severity of H.M.s memory impairment but, crit- not required for at least some forms of long-term memory
ically, it also showed a vast array of cognitive and mnemonic (Fig. 122).
function that was unaffected by his large bilateral lesion. The Overall, subsequent data from severely amnesic patients
pattern of data allowed Milner (Milner et al., 1968; Milner, such as H.M. and data from animal models of amnesia (see
1972) to make several conclusions that expanded on those Chapter 13) have supported Milners basic conclusions rather
noted above and helped lay the foundation for subsequent well (although there is still no consensus as to whether each
investigation into the amnesic syndrome. claim is entirely or only largely true). In a more recent exten-
First, Milner noted that damage to the medial portions of sion of these conclusions, Squire and Zola-Morgan (1991)
the temporal lobe results in profound inability to acquire identied what they referred to as a medial temporal lobe
long-term memory for new facts or events (anterograde memory system (MTL), consisting of the hippocampal region
amnesia) although memories from early life appear to be (dened as the CA elds of the hippocampus proper, the den-
intact. Conversely, there was some clear loss of memory of tate gyrus, and the subiculum) and the adjacent entorhinal,
information acquired for some time prior to the operation perirhinal, and parahippocampal cortices (which together
(retrograde amnesia). Together, these observations indicated form the parahippocampal gyrussee Chapter 3). Together,
that the medial temporal lobe might not be a permanent this system is posited to be critically involved in the acquisi-
repository for memory but that it plays a time-limited, albeit tion of new fact (semantic) and event (episodic) memory.
critical, role in memory. Second, she noted that there was Notably, this system is not the permanent storage site for this
no loss in general intellect or perceptual ability, indicating declarative or explicit memory, nor is it the locus of other
clear dissociation between memory and other aspects of cog- cognitive functions or other forms of memory. For example, it

Figure 122. Patient H.M.s performance on the Rey-Osterreith the mirror-drawing task (bottom). A typical stimulus is shown on
gure-drawing task (top). With the original drawing in front of the left. Participants view the shape in a mirror and attempt to
him, his direct copy is accurate, demonstrating intact perceptual trace the shape while staying inside the lines (gray line shows the
abilities and a range of intact cognitive abilities. After an hour delay, authors rst attempt). H.M.s performance improved steadily both
H.M. was unable to produce any drawing and did not remember within testing sessions and across 3 days of testing (right), despite
having previously seen and copied the gure. (Data provided by having no conscious memory of having performed the task on pre-
Suzanne Corkin, personal communication, March 6, 2005.) In vious days. (Source: Adapted from Milner et al., 1998, with permis-
contrast, H.M. showed learning that lasted over multiple days in sion. 1998 Elsevier.)
552 The Hippocampus Book

is not involved in immediate (or working) memory, and it is tools for noninvasive monitoring of neural activity (in the
not involved in a wide range of nondeclarative (or form of local blood ow or oxygenation) throughout the
implicit) long-term memory tasks (e.g., delay conditioning, brain. When data analysis techniques are applied that respect
perceptual repetition priming, habit learning). These ideas and account for the morphological variability of MTL struc-
amplify and develop Milners suppositions in several ways. tures across individuals, fMRI data in particular can provide
As a component of this interconnected system, the hip- highly valuable data for understanding normal human hip-
pocampal region participates in declarative memory in some pocampal function (see Section 12.3.5).
way. However, the exact nature or particular memory func- What follows is rst an overview of these methods, with
tions in which it participates cannot be unambiguously deter- particular attention to their strengths and limitations for the
mined from patients such as H.M. or other severely amnesic study of the human hippocampus. This serves as the back-
patients with extensive damage to the medial temporal lobes ground for the rest of the chapter, which focuses on the func-
or from animal models with analogous lesions. Thus, knowing tion of the human hippocampus. Particular attention is paid
that H.M. is impaired regarding tasks of recognition memory to the role of the hippocampus in declarative memory and
(e.g., Milner et al., 1968) or that a similar patient (E.P.) is how it might be functionally dissociated from the adjacent
entirely at chance during tests of recognition memory cortical structures of the parahippocampal gyrus.
(Hamann and Squire, 1997; Stark and Squire, 2000b;
Stefanacci et al., 2000) does not denitively implicate the hip-
pocampal region in recognition memory (or any other indi-
vidual MTL region, as the entire MTL is extensively 12.3 Methods for Studying Human
damaged). Indeed, many researchers have proposed func- Hippocampal Function
tional dissociations within the MTL such that extrahippocam-
pal structures (e.g., perirhinal cortex) support recognition Each of the three sources of data discussed in the following
memory (or at least one form of recognition memory) and sections (hippocampal lesions, electrophysiology via depth
that the hippocampal region is not involved in recognition electrodes, functional neuroimaging) has its own particular
memory tasks (see Section 12.4.5). Of course, the study of strengths and limitations in providing data to help us under-
what is not impaired in patients with more extensive damage stand the function of the human hippocampus. Moreover,
(e.g., H.M. or E.P.) can be informative as well in letting us none of them can produce meaningful insight into under-
determine what forms of memory do not require the hip- standing hippocampal function without an understanding of
pocampal region (see Section 12.4.1). the behavioral tasks used to assess memory performance. As
To move beyond an understanding of what is and what is such, the cognitive and behavioral tasks used to assess mem-
not impaired overall in the amnesia that follows damage to the ory in humans are discussed rst. In addition, none of the
medial temporal lobe to an understanding of the specic con- three methods alone is sufficiently free of limitations to be
tributions of the components of the MTL (e.g., the hip- able to answer the fundamental questions surrounding the
pocampus), we must shift our focus. We must turn our study hippocampus. In Chapter 13, general theoretical concerns
to patients (and animals) with lesions to specic structures of from these lines of data are discussed (e.g., the interpretation
the medial temporal lobe and to recording or imaging tech- of correlational datasetssee also Henson, 2005), but several
niques that allow us to measure signals condently from these specic concerns exist for each. These strengths and weak-
structures. Three sources of such data exist for studying the nesses are discussed below (see Sections 12.3.312.3.5).
human hippocampus. First, anoxia or ischemia can lead to
fairly selective bilateral damage to the hippocampal region. 12.3.1 Behavioral Tasks and Terms
Although this damage is incomplete and although anoxia is by
no means certain to damage only the hippocampus (see Unlike assessing memory in nonverbal animals, assessing
Section 12.3.3), when it does the study of the memory impair- memory performance in humans (or at least adult humans)
ments in such patients can inform our understanding of the can be exceptionally straightforward. If we want to know
role of the hippocampus itself. Second, advances in diagnostic whether a participant remembers a previously seen list of
techniques for localization of epileptic seizure foci have led to items, we can simply ask him or her to recall as many items as
several cases in which depth electrodes have been placed into possible. The simplicity of this free recall task and the fact that
the hippocampal region and recordings made during various it is an obvious test of human memory but not of animal
memory tasks. The study of such patients provides a clear memory should not be overlooked. No training on the task is
bridge between electrophysiology in animal models (e.g., rats required, as it is perhaps the archetypal everyday memory
or nonhuman primates) and studies of the human hippocam- task. Despite the prevalence of free recall, or its cousin cued
pus. Finally, recent advances in noninvasive neuroimaging recall (in which some portion of the to-be-remembered item
techniques have led to several methods by which correlates of is provided) in everyday life, neither task is easily captured in
neural activity can be localized with sufficient precision to dif- animal studies of memory.
ferentiate signals from the components of the MTL. In partic- A second style of memory task, called a recognition memory
ular, positron emission tomography (PET) and functional task, is more easily captured in animal models. Like tests of
magnetic resonance imaging (fMRI) have become popular recall, recognition memory tasks begin with a study phase that
 Functional Role of the Human Hippocampus 553

can either be intentional (participants are explicitly instructed involved in recollective-based memory and the structures of
to learn the item for a later test) or incidental (no instructions the adjacent parahippocampal gyrus are particularly involved
are given to learn the items and no mention is made of a later in familiarity-based memory. This possibility is discussed in
test, often using an entirely unrelated task to serve as a dis- Section 12.4.5.
guise). During the test phase, unlike recall tasks, entire items Many variations on these basic tasks exist, each designed to
from the study list (target or old items) and entire items assess different kinds of memory. For example, for a paired-
that were not on the study list (foil or new items) are pre- associates task two stimuli are presented and the participant is
sented. In the simplest version of a recognition memory task, instructed to learn the association between the two items. For
yes/no recognition, single items (either targets or foils) are the test, a single item can be presented and the participant
presented to the participant with the instructions to indi- asked to recall the target member of the pair (cued recall) or
cate whether each item was on the study list. In a second to pick out the target item from a list of choices that include
typical version, two-alternative forced-choice recognition, a tar- one or more foil items (forced-choice recognition). Further-
get item and a foil item are presented, and the participant more, pairs of items can be presented at the test that are either
is instructed to indicate which of the two items was on the intact or recombined versions of the studied pairs. As such,
study list. each trial presents only familiar components, and perform-
In making their responses to recognition memory probes, ance must be governed by memory for the associations
participants are thought to use two types of memory or between the individual items. Finally, just as the study phase
sources of information: recollection and familiarity (for review can be conducted in an incidental way, the test phase can be
see Yonelinas, 2002). Recollection provides information about an incidental test of memory. For example, participants may
a specic prior event or episode and gives one the sense of be asked to complete the word-beginning win____ with the
being placed back in that moment in time. Certainly, if this is rst word that comes to mind (stem-completion task), making
the reaction one has to an item presented during a recognition no reference to the prior study episode. Such implicit, or non-
memory task, the item is endorsed as having been on the study declarative, tests of memory are forms of long-term memory
list. In contrast, one can still have a strong sense that the item that do not appear to rely upon the MTL (see Section 12.4.1).
is familiar without having a recollective experience. Such a
feeling that the item is familiar, albeit with no specic infor- 12.3.2 Behavioral Measures
mation about the study episode, can still lead to endorsing the
item in a recognition memory test. Thus, two sources of infor- In any of these tasks, performance can be quantied in several
mation can be used when endorsing an item at time of recog- ways. The most straightforward is to calculate an overall per-
nition. A variant on the yes/no recognition memory task, cent correctsimply the percentage of correct trials out of the
known as the Remember/Know task (Tulving, 1985), asks par- total number of trials. In a yes/no recognition task with 20 tar-
ticipants to split their endorsements into recollective get items and 20 foil items, 16 correct yes responses to the
(Remember) and familiarity-based (Know) responses in targets and 16 correct no responses to foil items would result
an attempt to isolate the two sources of information. in 80% correct (Table 121). Although quite straightforward
A key question that arises when discussing recollection and and intuitive, percent correct has signicant difficulty in accu-
familiarity is whether these two types of information repre- rately capturing the underlying strength of the memory when
sent the operation of two separate neural systems. If they do, participants exhibit a strong bias toward one response. An
a second question that arises is whether this functional disso- uneven distribution of yes and no responses can arise quite
ciation maps onto a division of labor in the medial temporal easily from either natural individual participant biases or task
lobe: for example, that the hippocampus is particularly instruction and result in an underestimation of the true

Table 121.
Comparison of Various Measures of Recognition Memory Performance

Memory Hits False %

capability (rate) alarms (rate) Correct Pr Br d

Normal 16 (0.8) 4 (0.2) 80 0.6 0.5 1.7

Normal, no bias 10 (0.5) 1 (0.05) 72 0.45 0.1 1.6
Normal, yes bias 19 (0.95) 10 (0.5) 72 0.45 0.9 1.6
Moderate 14 (0.7) 6 (0.3) 70 0.4 0.5 1.0
Chance 10 (0.5) 10 (0.5) 50 0 0.5 0
Chance, yes bias 16 (0.8) 16 (0.8) 50 0 0.5 0

Typical behavioral measureshit rate, false alarm rate, percent correct, corrected rejection (Pr), bias
(Br), and discriminability (d)are shown for three levels of underlying memory capability (Normal,
unimpaired memory; Moderate, mildly impaired memory; Chance, no actual memory); and under
conditions of response biases.
554 The Hippocampus Book

strength of the memory. For example, if one were to reward the two distributions and, as such, does not depend on where
participants with a small sum of money for each correct yes the participant decides to set the criterion threshold. Values of
response but administer a strong shock for each incorrect zero represent no memory, and positive values represent accu-
yes response, participants would be generally less likely to rate memory (negative values indicate a propensity to say
respond yes and do so only when they are quite condent yes to foil items and no to target items and, when reliable,
they are accurate. In our above example, this might reduce the indicate reliably below-chance performance). Different
correct yes responses to 10 and increase the correct no thresholds result in different biases (different propensities to
responses to 19, with a resulting percent correct of 72%. There respond yes or no) but merely represent different values of
is good reason to believe, however, that the actual memory is this single parameter that are used to threshold both distribu-
the same in these two cases, and this change in response rates tions. As such, d can index memory in a manner that is inde-
is the result of a shift in the participants response bias. pendent of individual response biases.
Several measures are commonly used to measure or Finally, one more means of quantication must be dis-
remove the detrimental effect of response biases. Response cussed because simply knowing the overall percent correct or
biases can be easily seen by dividing the recognition responses d in a memory task only tells a part of the story. For example,
into four categories: (1) hits, dened as correct yes responses if we measured recognition memory performance in a task
to studied target items; (2) misses, dened as incorrect no and obtained an average of 85% correct for a group of healthy
responses to studied target items; (3) correct rejections, dened control participants and 75% correct for a patient with dam-
as correct no responses to unstudied foil items; and (4) false age to the hippocampus, it is still unclear whether this
alarms, dened as incorrect yes responses to unstudied foil patients performance is impaired. If the control participants
items. Comparing the hit rate and correct rejection rate is one scores ranged from 80% to 90%, the patient is likely to be truly
method for determining whether a bias is present. A second impaired, but if the control scores ranged from 60% to 100%,
method uses the corrected recognition score, often denoted the participant is likely to be performing at a normal level.
Pr. Pr is simply the probability of a correct yes response (a In addition to the ubiquitous t-tests and analyses of vari-
hit) minus the probability of an incorrect yes response (a ance (ANOVAs) used throughout behavioral testing, analysis
false alarm). Pr scores are in a range between 0 and 1, with by z-scores is prevalent in the memory literature and in partic-
0 indicating chance performance. By themselves, Pr scores do ular in the analysis of small numbers of memory-impaired
not fully remove the effects of response biases; however, a patients covered in Section 12.4. A z-score is simply the num-
complementary measure of bias, Br, accurately estimates any ber of standard deviations away from the mean a given obser-
response bias that exists. vation lies. As such, it gives a way to determine how aberrant
a single observation is (e.g., a single amnesic patients recog-
Br  false alarm rate  (1
Pr)
nition memory score). Critically, this analysis relies on an
The most common method to correct for any bias that accurate estimate of both the average performance of the con-
exists is to calculate d, the discriminability measure from the trol population and an accurate estimate of the variance in
signal detection theory (Green and Swets, 1966). Here, it is control performance. A z-score of 0 represents performance
assumed that there are two normal distributions of stimuli (a equal to the mean of the control population. A z-score of 1.0
distribution of targets and a distribution of foils) that differ corresponds to a likelihood that the patients score was, in fact,
along a single parameter (e.g., some representation of per- drawn from the normal population (alpha level, two-tailed) of
ceived memory strength or familiarity at time of retrieval). It 0.32. A z-score of 1.96 has this probability drop to the tradi-
is further assumed that participants set a threshold or crite- tional threshold of signicance (0.05).
rion value for this single parameter and respond yes if the
stimulus presented for recognition exceeds this threshold and 12.3.3 Anoxia and Bilateral
no if the stimulus does not exceed this threshold. Thus, Hippocampal Lesions
there are four categories of responses corresponding to the
hits, misses, correct rejections, and false alarms dened above. Neuropsychologystudy of the relation between brain and
With respect to memory tasks, the amount of memory behaviorhas a long history of studying the cognitive and
present is viewed as the distance between the target and the behavioral results of brain lesions and using this information
foil distributions along this axis of familiarity (e.g., Morrell et to help understand the function of specic brain regions.
al., 2002). That is, the effect of studying the target stimuli is to Many common etiologies, and even many etiologies that pro-
shift the distribution away from the unstudied distribution. duce memory impairments, do not provide data that cleanly
The amount it has been shifted corresponds to the amount of isolate hippocampal function. For example, although a com-
memory the participant has for the studied items. If the dis- mon and valuable source of neuropsychological data, patients
tribution has been shifted a lot, it is easy to discriminate the who have suffered damage due to stroke are not common
target from the foil distributions, whereas if the distribution sources of restricted bilateral hippocampal damage, as the vas-
has been shifted very little, it is more difficult to discriminate culature does not provide a mechanism for selective bilateral
whether a given item presented at test was drawn from the tar- hippocampal damage. For example, although rupture or
get or foil distribution. The measure d is the distance between blockage of the posterior communicating artery often results
 Functional Role of the Human Hippocampus 555

Figure 123. Patient R.B.s performance on the Rey-Osterreith this delay). Postmortem histological analyses revealed his anoxic
gure-drawing task demonstrated intact ability to copy the drawing episode resulted in severe cell loss limited to the CA1 elds of the left
when it was placed in front of him but clearly impaired ability to and right hippocampus (left shown). Location and extent of damage
copy the drawing from memory 10 to 20 minutes later (healthy con- are indicated by the asterisk and arrows. (Source: Zola-Morgan et al.,
trols typically indicate the gure with only very minor distortions at 1986, with permission. 1986 by the Society for Neuroscience.)

in MTL damage (and amnesia), the damage is not restricted to patients have been reported to show quantiable bilateral
the hippocampus and is unilateral. Thus, although stroke or damage to the hippocampal region, little or no damage out-
aneurysm can lead to memory impairment, it does not give us side the hippocampal region, and behavior consistent with
the most powerful means of assessing hippocampal function. the amnesic syndrome after anoxic or ischemic episodes
Similarly, although physical infarct can lead to amnesia (Cummings et al., 1984; Zola-Morgan et al., 1986; Rempel-
(e.g., the case of N.A. who suffered mammillary body damage Clower et al., 1996; Spiers et al., 2001; Hopkins et al., 2004). In
following an accident involving a miniature fencing foil), it is one such case, that of R.B. (Zola-Morgan et al., 1986) (Fig.
hard to imagine how a physical infarct would selectively lesion 123), postmortem histological analysis conrmed that the
the hippocampus bilaterally given the physical location of the damage was almost entirely limited to the CA1 eld of the
hippocampus buried within the MTL. Alzheimers disease hippocampus with only minor damage observed elsewhere
does result in bilateral damage to the hippocampus, but it also (and located outside the MTL).
results in damage to the entorhinal cortex and damage outside The presence of these cases does not indicate that anoxia or
the MTL (particularly as it progresses). The thiamine de- ischemia necessarily result in selective damage to the hip-
ciency associated with chronic alcohol abuse and Korsakoff s pocampus. Although histological analyses such as R.B.s are
disease results in diencephalic damage and amnesia. Many rarely available (and only postmortem), even relatively basic
highly informative and inuential studies of amnesia have MRI techniques have been shown to correlate with post-
been conducted with patients suffering from Korsakoff s dis- mortem analyses of hippocampal damage, albeit at a coarser
ease (e.g., the seminal work of Warrington and Weiskrantz, resolution and potentially with less sensitivity (Rempel-
1968). However, as with patient N.A., it is clear that the dam- Clower et al., 1996). Using current quantitative MRI tech-
age suffered in such patients does not allow us to isolate the niques, a study of anoxia resulting from carbon monoxide
function of the hippocampus itself. poisoning and obstructive sleep apnea (Gale and Hopkins,
In contrast, anoxia (reduction in oxygen) and ischemia 2004) reported that only 30% to 36% of the patients were
(reduction in blood ow) have the potential to damage the observed to have hippocampal damage. Cortical atrophy (in
hippocampal region selectively and bilaterally. Such cases, the form of the ventricle/brain ratio) was present in 35% of
even if quite rare, can prove to be highly informative. Several the patients with damage resulting from carbon monoxide
556 The Hippocampus Book

poisoning. Furthermore, in a review of 43 patients with dam- 12.3.4 Depth Electrode Recordings
age following anoxia (Caine and Watson, 2000), 32 showed
evidence of cortical damage and 31 showed evidence of dam- Implanting electrodes and recording from neurons in the hip-
age to the pallidum or striatum. The hippocampal region was pocampal region and adjacent cortical structures has been a
only the third most common site of damage (30 cases). common technique for studying rodents and nonhuman pri-
Damage was also reported commonly in the basal ganglia and mates (see Chapter 13), but electrophysiology has not been
the thalamus. Most strikingly, in only 18% of these anoxic used extensively to study the human hippocampal region for
cases (8/43) was the damage present in and limited to the hip- obvious reasons. There have been a number of cases, however,
pocampal region. in which depth electrodes have been placed in the human
Thus, it should be quite clear that the study of patients with MTL to help identify the topography of epileptic seizures
damage resulting from anoxia or ischemia does not necessar- when other techniques for neurosurgical planning have not
ily imply the study of selective hippocampal damage or even proved sufficient. In a limited number of these cases, single-
hippocampal damage at all. To use anoxic patients as a route unit recordings from the depth electrodes have been made
to studying the consequences of hippocampal damage, during various explicit memory tasks (Heit et al., 1988, 1990;
detailed structural neuroimaging assessment of the damage Fried et al., 1997, 2002; Fernandez et al., 1999b, 2002; Kreiman
(whenever possible) and detailed neuropsychological assess- et al., 2000; Cameron et al., 2001; Paller and McCarthy, 2002;
ment of their behavior is certainly required. Improved tech- Fell et al., 2003) (Fig. 124).
niques for quantifying damage and for addressing the Such data are certainly highly important, but they are also
possibility of covert damage not resolved by current tech- scarce. Unfortunately, the scarcity of the data extends not only
niques are certainly warranted as well. In particular, detailed to the small number of studies but to the small number of
comparison of postmortem histological analyses and premor- recordings per participant in these studies. In electrophysio-
bid neuroimaging analyses are required to improve and fur- logical studies of the monkey hippocampus, one can expect to
ther validate the neuroimaging analyses. However, even if see data collected from 50 to 100 hippocampal neurons in
current neuroimaging techniques do not provide a denitive each monkey (e.g., Wirth et al., 2003). In depth electrode
assessment of the damage, they are innitely better than recordings on humans, one may see data from this same num-
assessing damage merely by etiology, as the above review has ber of neurons (e.g., Fried et al., 2002), but they are spread
shown. Unfortunately, although assessments to the best of across a large number of participants, yielding single-digit
current standards have been done for some of studies of numbers per participant. This is a limitation of the methodol-
memory, they have certainly not been done for all. ogy, not of the experimenters, as multiple electrode drops and

Figure 124. Electrophysiological data in the human. Left panel. the cue item in a paired-associate task. (Source: Left and middle
Trajectory of a depth electrode with microwires placed in the hip- panels: Cameron et al., 2001, with permission. 2001 Elsevier. Right
pocampus (arrow indicates where wires protrude from electrode). panel: Paller and McCarthy, 2002, with permission. 2002 Wiley-
Both peri-event spike-train histograms (middle) and eld potentials Liss, a subsidiary of John Wiley & Sons.) Right panel. Field poten-
(right) can be collected from such electrodes. Middle panel. A hip- tials from electrodes in the posterior portion of the hippocampus
pocampal neuron recorded from the electrode indicated on the left are shown for three patients demonstrating differences in activity
demonstrates an increase in activity at retrieval upon presentation of for the sample and match phases of a delayed match-to-sample task.
 Functional Role of the Human Hippocampus 557

continual adjustment of the electrode location or even initial have constraints and challenges that place limits on what one
placement of the electrode to isolate as many hippocampal can conclude from their results. If the challenges are well met
neurons as possible is not an option. The electrode placement and experiments designed and interpreted with the con-
must be motivated by the neurosurgical planning of which straints clearly kept in mind, neuroimaging studies can be
these data are a by-product. highly informative (Henson, 2005). If they are not, they can
Furthermore, we must remember that the recordings are present a confusing or conicting state of affairs (much as was
coming from brains that are decidedly not normal. The present at the time of the early studies).
recordings are being made from patients with epilepsy severe The rst constraint is that, on the whole, neuroimaging
enough to warrant neurosurgical intervention that is most and electrical recording techniques provide correlational data
likely going to target the medial temporal lobes. Whether and and cannot provide evidence as to the necessity (or even
how this affects the data cannot be known at present. To actual use of) a structure in a particular function. For exam-
address this concern, data are often analyzed only from sites in ple, although single-unit recording shows strong hippocampal
the unoperated hemisphere or otherwise distant from the activity during delay eyeblink conditioning in the rabbit (e.g.,
epileptic foci. Berger et al., 1980) and activity has similarly been observed in
humans using PET (e.g., Blaxton et al., 1996), complete bilat-
12.3.5 Neuroimaging eral lesions of the hippocampus do not impair acquisition or
expression of the response (e.g., Mauk and Thompson, 1987;
The initial set of neuroimaging studies exploring memory and Clark and Squire, 1998). Therefore, the activity observed in a
the medial temporal lobe using either PET or fMRI presented region may be incidental to the task at hand or may be the
a mixed set of results. There were a number of initial PET result of processing in another region that has projections into
studies with positive reports of MTL activity during memory the observed area. Whereas the use of parametric designs or
tasks (e.g., Squire et al., 1992; Grasby et al., 1993; Nyberg et al., other clever experimental manipulations that attempt to link
1995; Schacter et al., 1995). However, at the time there were imaging data to specic components of behavior (e.g., the
also a sizable number of studies that observed memory- recent attempts to link MTL activity to aspects of implicit
related activity outside the MTL but no memory-related activ- tasks reported by Rose et al., 2002 and Schendan et al., 2003)
ity in the MTL itself (e.g., Kapur et al., 1994; Shallice et al., can provide stronger evidence for a causal link than the use of
1994; Tulving et al., 1994; Buckner et al., 1995). For example, simpler designs, neuroimaging data cannot be resolute in this
Shallice et al. (1994) contrasted activity during word-pair regard.
encoding and cued recall with activity during a low-level base- A second problem faced by neuroimaging techniques is the
line control task (e.g., hearing the pair one thousand, two lack of a baseline. Although particularly problematic for
thousand repeatedly). Although the contrasts revealed BOLD (blood oxygenation level-dependent) fMRI (Fig. 125),
numerous regions of brain activity (e.g., prefrontal cortex), no which has a signal with an arbitrary offset and arbitrary units
MTL activity was present. As both contrast activity during of measurement, the lack of a clearly dened level of activity
what are clearly mnemonic tasks that are impaired in amnesic associated with a region not being involved in a task is
patients with activity during tasks that have no mnemonic endemic to all neuroimaging techniques. (How many spikes
demands and that are not impaired in amnesic patients, the per second constitute no activity in the neuron, and what is
negative results were unsettling. Many hypotheses for the an animal actually doing when this is assessed?) Neuroimaging
unreliability of ndings were put forth, such as the idea that techniques are contrastive in nature. Our data take the form of
the MTL might always be active (thus reducing the contrast a higher BOLD fMRI signal, greater regional cerebral blood
between states) or that MTL signals may be too weak to ow (rCBF), more spikes per second, a steeper excitatory post-
resolve (resulting from either poor imaging signals from the synaptic potential (EPSP) slope, or a sharper tuning function
MTL or inherently weak activity). It was even suggested that during task A than during task B. These numbers can often be
the MTL might not be involved in long-term memory after all. quantied and varied parametrically, but it is frequently diffi-
Developments in neuroimaging techniques and further cult to interpret a result zero activity. This contrasts with
studies have now led to many positive results, supporting the behavioral data. For example, in two alternative forced choice
former two hypotheses (and several new ones as well) but not recognition, scores necessarily vary between chance and 100%;
the latter. In fact, as of this writing, more than 1300 articles and for free recall, performance necessarily varies between 0%
indexed by PubMed during the last 5 years contained the key and 100%. Where there may be oor and ceiling effects to con-
words fMRI and hippocampus or medial temporal lobe. sider, one can, in principle, identify both perfect memory and
The numbers would certainly be signicantly higher if other absent memory.
neuroimaging techniques were included. The popularity of As noted, this lack of a standard of comparison may be
neuroimaging attests to the fact that it can provide useful data especially problematic for BOLD fMRI with its arbitrary units
to further our understanding of the neural mechanisms that and its complete lack of an estimate of what measurement
underlie memory. zero activity in a region should be. For example, Stark and
However, although useful and compelling data can be Squire (2001b) have shown that when randomly interspersed
obtained using neuroimaging techniques, these techniques 3-second periods of rest were used as a baseline to assess zero
558 The Hippocampus Book

Figure 125. Left. A current model of the chain of events that leads of signals during no visual stimulation is plotted. Note how the
to the blood oxygenation level-dependent (BOLD) effect measured BOLD effect is a temporally low-pass ltered response of the
on functional MRI (fMRI). Underlying neural activity results in underlying neural activity. One second of neural activity was
local increases in the cerebral metabolic rate of oxygen extraction recorded as a protracted response that peaked approximately
(CMRO2), cerebral blood ow (CBF), and cerebral blood volume 6 seconds after onset and did not return to baseline until approxi-
(CBF). Local increases in CBF affect the other two, and all combine mately 12 seconds after offset of the stimulus. Note also, though,
to change the ratio of oxygenated relative to deoxygenated hemo- that the response to two trials is a roughly linear summation of the
globin in a local area that is measured on fMRI. This signal is response to two individual trials. (Sources: Left: Buxton, 2001,
called the BOLD effect. Right. Typical BOLD effects. Here, visual p. 419, with permission of Cambridge University Press. Right: Dale
activity was recorded in response to either a single 1-second visual and Buckner, 1997, with permission. 2002, Wiley-Liss, a sub-
stimulus (ickering checkerboard) or to two 1-second trials, spaced sidiary of John Wiley & Sons.)
5 seconds apart. In both cases, the percent change from a baseline

activity, viewing novel or familiar pictures failed to elicit any the observation that the choice of baseline task determined
apparent activity in the hippocampal region. However, when whether activity during the mnemonic trials was above or
an active but menial task was used as a baseline (deciding below zero. When a trivially easy nonmnemonic perceptual
whether digits were odd or even), robust activity was task was used, activity increased with memory strength but
observed. was all negative. When a more difficult version of the same
Similarly, Law et al. (2005) collected fMRI images as par- task was used as the baseline, activity again increased with
ticipants learned a concurrent set of arbitrarily paired associ- memory strength but was now all positive. Contrasting the
ates gradually over multiple trials (Fig. 126). Each trial two baseline tasks revealed substantially greater activity in the
contained both encoding and cued-recall components as par- MTL for the easy task than for the difficult task, presumably
ticipants learned through trial and error which abstract geo- the result of participants minds wandering during the triv-
metric shapes were associated with which response options. ially easy trials. Such mind wandering is likely to include inci-
Notably, a number of MTL regions exhibited activity that dental encoding and retrieval of information, the hallmark of
increased in conjunction with the strength of the participants MTL function. This result draws into sharp focus the diffi-
memory for a particular region. Equally notable, however, was culty of not having an estimate of zero activity in a region.
 Functional Role of the Human Hippocampus 559

Figure 126. Activity in a region of the left hippocampus during a These data demonstrate the effect of activity during the baseline
paired associate task is shown for both strong, highly accurate task. Substantial activity in the hippocampus during the Easy
memories and for memories that are above chance levels of per- baseline task (presumably resulting from incidental encoding
formance but still only moderately accurate. Two perceptual tasks and retrieval unrelated to this rather boring task) served to deect
were used to estimate zero activity, neither of which was overtly activity during the memory task below zero. Note, however,
mnemonic. In both, the task was to identify the brightest square. that the relative differences between the curves are maintained,
The task was trivial in the Easy condition (98% correct) and chal- irrespective of baseline choice. fMRI has no baseline, and all data
lenging in the Difficult condition (54% correct). When the Easy must be interpreted in a relative manner. All we actually know
condition was used as an estimate of zero, strong memories from these data is that Difficult baseline  Strong memories 
yielded small negative activity, and moderate strength memories Moderate memories  Easy baseline. This property is often ignored
yielded large negative activity. When the Difficult condition yet is a clear source of several failures to observe activity. (Source:
was used as an estimate of zero, the same exact memory task Data are from Law et al., 2005, with permission. Society for
trials yielded large and small positive activities, respectively. Neuroscience.)

fMRI (and most other imaging techniques) yield purely rela- 12.3.6 Technical Challenge: Alignment
tive measures, and no claims can be made about a regions of MTL Regions Across Participants
absolute level of activity in any task. The only measures avail-
able are relative measures between tasks or conditions. Even if the above challenges are met and the constraints
Furthermore, simply because a task does not require a cogni- respected, one signicant challenge remains if neuroimaging
tive component (memory, in this case) does not mean the techniques are to be able to help answer questions concerning
component is not actually being used and that the region is the functional role played by the various structures in the
not active during the task. MTL. For neuroimaging to do so, it must be possible to local-
Finally, if one considers BOLD fMRI, the most popular ize signals of the specic subregions of the MTL. Therefore,
neuroimaging technique used in humans, we still have limited the data must be of sufficient resolution to allow condence in
temporal (~ 1 second) and spatial (~ 3 mm3 or ~10,000 neu- localization; and if group analyses are desired, it must be pos-
rons) resolution; and we still do not have a complete under- sible to transform the data from multiple participants in such
standing of the relation between neural events and the BOLD a way that cross-participant tests respect anatomical divisions
fMRI signal (or the measured by PET). Progress along these in the MTL. Techniques such as PET and magnetoencepha-
dimensions is being made, with some evidence from experi- lography (MEG) have clear strengths. PET can directly quan-
ments that BOLD is more closely linked to synaptic activity tify CBF and can be used to tag specic chemicals (e.g.,
than to spiking activity (Logothetis et al., 2001). One conse- neurotransmitters) in ways no other technique can; and MEG
quence of this is that inhibitory inputs may paradoxically has millisecond temporal resolution. The spatial resolution
serve to increase the BOLD effect rather than decrease it. and localization accuracy of both techniques have improved
Strong inhibition in a region results in substantial synaptic over recent years, but they still cannot approach the resolution
activity (and metabolic activity), even if spiking is reduced. possible with fMRI. The resolution in typical fMRI studies is
Direct tests of this hypothesis in the rat have yielded increases typically 3 to 4 mm3 but can be pushed into the cubic mil-
in the CBF, a precursor of the BOLD effect (Caesar et al., limeter range (Hyde et al., 2001; Kirwan et al., in press), mak-
2003), indicating that inhibition would result in an increase ing fMRI a leading candidate for imaging the small structures
rather than a decrease in BOLD. Thus, our current under- of the MTL.
standing of how to relate electrophysiological ndings to neu- However, even if the spatial resolution is theoretically suf-
roimaging ndings is certainly incomplete, although it does cient to have some condence in localization, it is still vital
appear that BOLD fMRI measures offer a relatively linear that any cross-participant analyses respect the structural
assessment of neural activity (Boynton et al., 1996; Rees et al., boundaries in the MTL. It would be pointless to attempt to
2000; Logothetis et al., 2001). discern what factors affect activity in the perirhinal cortex if
560 The Hippocampus Book

Figure 127. Coronal structural MRI sections through the hip- pant was initially manually segmented and then transformed along
pocampus of 20 participants, averaged following Talairach align- with the structural image using both techniques. In this group over-
ment (a) and region of interest alignment (ROI-AL) (b). White lay, white is ideal (no voxels in a and 23 voxels in b) and indicates
arrows indicate the location of the collateral sulcus used to identify that all 20 participants aligned segmentations identied the voxel as
the parahippocampal gyrus. Overlay on the left hippocampus indi- part of the left hippocampus. Black indicates 1 to 10 aligned seg-
cates the amount of cross-participant overlap of manual segmenta- mentations identied the voxel as part of the left hippocampus.
tions of each participants left hippocampus when brains were Light, medium, and dark gray indicate that 19, 18, and 16 segmen-
aligned using each technique. The left hippocampus of each partici- tations overlap, respectively.

our alignment techniques normalized brains in such a way from one hippocampal region are combined with another
that a given voxel in a group analysis was located in the participants entorhinal cortex and a third participants ven-
perirhinal cortex of one participant, the entorhinal cortex of tricle. If one extends this analysis to segmentations of all sub-
another, the hippocampal region of a third, and outside, in the regions of the MTL, the picture is worse still. Between any two
ambient cistern, of a fourth. Unfortunately, human brains are participants MTLs, only about half of the voxels are typically
sufficiently dissimilar from each other that alignment with identied as belonging to the same structure in both partici-
popular techniques (e.g., alignment to the atlas of Talairach pants (Stark and Okado, 2003; Kirwan et al., in press).
and Tournoux, 1988) often leaves us in this situation. Imperfect cross-participant alignment results in a blur of
An example of this problem is shown in Figure 127a. the data that has two detrimental effects. First, the blur
Here, 20 structural MRI scans were rst individually aligned to reduces the localization accuracy for any observed activity, as
the Talairach atlas before an average of the 20 scans was cre- signals from multiple regions may be combined. Second, the
ated (shown as a coronal image cropped around the MTL). blur reduces statistical power. If separate regions (or subre-
The white arrows indicate where the collateral sulcus (a den- gions) are behaving differently, the spatial blur smears activity
ing structure for the parahippocampal gyrus) is most likely to across regions, introducing noise that prevents a consistent
be. As one can see, it is far from clearly dened, indicating that pattern of activity from being observed. For example, the
the averaging across participants blurred this feature (which is main result of a study by Stark and Okado (2003) demon-
plainly visible in each individual scan) into an undifferentiated strating activity associated with encoding during a retrieval
mass. This poor level of cross-participant alignment arises task was not observed when the data were aligned with tradi-
from both global variability across participants (overall shifts tional Talairach techniques but was observed when the data
in the location, orientation, and size of structures) and from were aligned using more sophisticated techniques.
differences in the shape of structures (Preuessner et al., 2002). Recently, there have been three approaches taken to address
In addition to averaging the structural MRI images, Figure this issue: simple anatomical region of interest (ROI) analyses,
127a shows the result of averaging segmentations of the left cortical unfolding applied to the MTL, and ROI-AL (region of
hippocampal region. Each participants left hippocampal interest alignment). All three approaches can address the issue
region was manually segmented, and the same Talairach of improving cross-participant alignment and, in so doing,
transformation was applied to each segmentation prior to open up the possibility of using fMRI to help differentiate the
averaging the segmentations across participants. The result of role of individual structures in the MTL. Figure 127b shows
this averaging is a grayscale overlay that indicates how well the the result of aligning the medial temporal lobe structures
Talairach transformation was able to align all 20 hippocampi. from the same 20 participants from Figure 127a but using
In this slice, there are no voxels in which there was overlap one of these approaches (the ROI-AL method of Stark and
across all 20 manual segmentations of the left hippocampal Okado, 2003). Here, one can clearly resolve the collateral sul-
region, and there are only three voxels in which 19 of the 20 cus (white arrow) and differentiate it from the hippocampal
segmentations overlap. Even if our only goal is to assess activ- region immediately above. The 20 segmentations aligned with
ity in the hippocampal region overall (not in subregions of the this technique have perfect overlap in 23 voxels; and if one
hippocampus), this level of alignment is insufficient as signals extends this to near-perfect (19/20) levels, the count rises to 34
 Functional Role of the Human Hippocampus 561

voxels (versus 0 and 3 in the case of Talairach alignment). to align two brains with typical techniques (e.g., rotation,
These three approaches give us hope that fMRI may be able to translation, scaling, shearing) are focused on aligning a single
isolate hippocampal function from that of adjacent regions region (e.g., left hippocampal region). The net result is vastly
and perhaps even differentiate contributions of hippocampal improved cross-participant alignment, bringing with it
subelds. improved statistical power. Furthermore, by using anatomi-
With the most straightforward technique, several articles cally dened regions, ROI-AL (and unfolding) can localize
have collapsed activity across all voxels within a set of anatom- results from cross-participant analyses to specic anatomi-
ically dened ROIs (e.g., Stark and Squire, 2000a; Small et al., cally dened regions of interest (e.g., perirhinal versus
2001; Reber et al., 2002, 2003). Thus, there may be a single parahippocampal cortices). The exact location of any region
measure that represents activity for each participants anterior of activity in a group analysis can be compared with the seg-
left hippocampal region in a given condition. Although this mentations from each participant (or with a composite
technique has the potential for perfect alignment across par- anatomical model based on the individual participants
ticipants, it suffers from two drawbacks. First, by combining anatomically dened regions of interest). Thus, one can proj-
all voxels in an anatomically dened region into a single meas- ect backward from the group result to the individual partici-
ure (a single, large, irregularly shaped voxel), any functional pants anatomy and determine with some precision where a
variability in that region is lost. For example, if all data from signal was generated.
the right hippocampal region were treated as if it were one
large voxel, and if two opposing patterns of activity were pres-
ent in different subregions of the right hippocampal region,
this functional variability would be lost. A second, related 12.4 Dissociating Hippocampal Function
drawback is that if only a small subregion in the anatomically
dened ROI is active, noise from other included voxels distort In the following sections, ve aspects or potential divisions
the observed activity. in long-term memory are discussed with particular atten-
A second approach has been to adapt cortical unfolding tion paid to isolating the function of the hippocampus. For
techniques to the problem of unfolding the cortical MTL each of these aspects, the relevant data from patients with
structures and the spiral structure of the hippocampal region bilateral damage limited to the hippocampal region, from
(Zeineh et al., 2000, 2003). With this approach, anatomically depth electrode recordings, and from neuroimaging studies
localized boundaries are dened and used to map the three- are discussed.
dimensional data onto a common two-dimensional at map
and to align individual participants at maps. Unlike collaps- 12.4.1 Explicit Versus Implicit
ing all voxels in the anatomically dened ROIs, this technique
has the advantage of preserving the topography in each Perhaps the clearest example of functional dissociation in
region. In addition, because of the requirement for very high long-term memory is that between explicit, or declarative,
resolution of the functional data (1.6  1.6  4 mm) and the memory and implicit, or nondeclarative, memory. Explicit
techniques ability potentially to unfold the spiral structure of memory refers to intentional or conscious recollection of
the hippocampal region, this technique holds the promise of prior experiences, as assessed in the laboratory by traditional
differentiating signal from regions in the hippocampus itself. tests of recall or recognition, whereas implicit memory refers
However, the unwarping process is not entirely invertible, as to changes in performance or behavior, produced by prior
functional voxels that lie within a fold can be associated with experiences, on tests that do not require any intentional or
two different regions of the unfolded map. conscious recollection of those experiences (Schacter, 1999,
A third approach has been to use anatomically dened p. 233). These descriptive terms relate to specic task demands
ROIs to guide alignment directly (Stark and Squire, 2001a; rather than distinctions between memory systems or spe-
Stark and Okado, 2003; Miller et al., 2005; Kirwan et al., in cic brain structures. Yet, a wealth of data shown has shown
press). This technique (dubbed ROI-AL by Stark and Okado, that this descriptive distinction is strongly correlated with
2003) shares with the unfolding technique the advantage of hippocampal function. The terms declarative and nondeclar-
preserving topography in regions with the unfolding tech- ative are dened in similar terms when applied to the human.
nique but does not require very high-resolution functional The declarative/nondeclarative distinction goes a bit further,
voxels. ROI-AL takes a direct approach to aligning regions however, to embrace the apparent functional dissociation
across participants. Instead of using structural MRI to align between multiple memory systems. Declarative memory
gray matter, white matter, and cerebrospinal uid (CSF) identies a biologically real category of memory abilities
across participants, ROI-AL attempts to align anatomically (Squire, 1992, p. 232) that require structures in the medial
dened regions of interest based on rough segmentations of temporal lobes, whereas nondeclarative memory identies a
the regions. Furthermore, instead of attempting to arrive at heterogeneous collection of memory systems that bear resem-
the best tting alignment across the entire brain, ROI-AL blance to each other behaviorally (all are observed with
focuses only on alignment of a particular structure (or set of implicit memory tasks) but appear to rely on many brain
structures). Thus, all of the transformation parameters used structures.
562 The Hippocampus Book

Despite obvious memory impairments, it may come as a ceptual identication priming is normal even in the case of
surprise to some that patients with damage limited to the hip- E.P., who performs at chance when given a recognition mem-
pocampal region and those with extensive damage to the ory test under similar circumstances (Hamann and Squire,
medial temporal lobes demonstrate normal levels of perform- 1997).
ance on a wide range of long-term memory tasks. In an early In perhaps the most extreme example of this dissociation
report (Milner, 1962), H.M. was found to acquire a perceptual between intact priming and impaired recognition (Stark and
motor skill (learning to trace the outline of a shape when Squire, 2000b), E.P. studied a list of words (e.g., window)
viewed in a mirror) over several days at normal rates despite and after a 10-minute delay was presented with a test of repe-
not remembering the prior days training (see Fig. 122). tition priming (What is the rst word that comes to mind
Subsequent studies have shown that even severely amnesic that begins win____) and a test of recognition memory
patients exhibit normal rates of delay eyeblink conditioning (Which word did you see on the list 10 minutes ago: window
(e.g., Weiskrantz and Warrington, 1979; Clark and Squire, or winter?) on each trial. E.P. showed a normal priming effect
1998). Furthermore, the phenomenon of categorization in the form of a 26% increased likelihood of generating the
appears to be intact. In one task, a random dot pattern studied word (relative to baseline completion rates of generat-
(resembling an imaginary constellation of stars) is created, ing this word). However, his recognition memory perform-
and numerous distorted versions of this prototype are cre- ance was at chance, averaging 48% correct. Therefore, within
ated as well by moving the dots by random amounts (Posner seconds of each other, E.P. showed intact implicit memory for
and Keele, 1968). After studying only highly distorted ver- a word (in the form of stem completion priming) and no
sions, even severely amnesic patients such as E.P. incidentally detectible explicit memory for that same word. Given E.P.s
learn to abstract features of the prototype (or the category) so complete hippocampal loss (Stefanacci et al., 2000), it is clear
they can later correctly classify new patterns as either mem- that the hippocampus cannot be vital for this form of implicit
bers or nonmembers of the studied category (e.g., Knowlton memory.
and Squire, 1993; Squire and Knowlton, 1995). They do so, We should note that when comparing an amnesic patients
however, without any knowledge that they have ever per- level of performance on repetition priming tasks (or indeed
formed the task and without any ability consciously to recog- on many implicit memory tasks) relative to healthy controls,
nize any of the previously studied dot patterns (even if the there is always the possibility that the amnesic patients per-
study phase consisted of seeing the same dot pattern 40 formance may be impaired relative to that of the healthy con-
times). Thus, there appears to be a dissociation between the trols because the latter group may make use of covert explicit
ability to learn in these implicit tasks gradually and the ability memory. Although not the case in the above-mentioned stud-
to remember the study episodes or contents of those episodes ies overall, the repetition priming task is susceptible to
consciously or explicitly. explicit contaminationthe use of explicit memory on an
The term priming refers to another class of implicit implicit memory task. For example, if at this point in the
memory tasks that are not affected by MTL damage. In per- chapter the reader were asked to complete the word stem
ceptual priming tasks, exposure to a word or a picture (often win_____ with the rst word that comes to mind, you might
incidentally, with no instruction to study the item) improves follow these instructions faithfully and respond with the rst
the ability to perceive the item later, usually taking the form of word that simply pops into your mind. However, you might
increased accuracy or decreased reaction time during also realize that in the preceding paragraph the word window
degraded presentation. There is a long history in cognitive was used as a sample study item (especially if this were done
psychology dissociating perceptual priming and recognition in the context of not one implicit probe but an entire list of
memory in healthy individuals (for review see Schacter, 1994). them). Therefore, you might respond with window not
This behavioral dissociation is relevant here in that long-term because it was the rst word that freely popped into your
memory for an item in the form of a repetition priming effect mind, but because you tried to remember what study word
is normal following hippocampal damage, whereas explicit began with the letters win_____. Thus, despite the task
memory for the same item is clearly impaired. instructions, participants may treat an implicit task as a thinly
Warrington and Weiskrantz (1968, 1974) were the rst to disguised explicit task and show enhanced levels of per-
show that despite poor recognition memory after studying formance. For example, in the above-mentioned study with
lists of words or pictures, amnesic patients had an intact form E.P. that combined a repetition priming task with a forced-
of perceptual memory for these items. When tested with a choice recognition memory task during each trial (Stark
fragmented or degraded version of either type of stimulus and and Squire, 2000b, experiment 4), all but one of the controls
asked to complete this partial cue, studied items were com- had priming effects ranging from 0.21 to 0.29. The other con-
pleted more accurately or at greater levels of degradation than trol had a priming effect of 0.53, approximately 8 standard
nonstudied items. This perceptual priming effect was normal deviations (SD) away from the mean priming effect. Although
in the amnesic patients. Likewise, if words are presented rap- we do not know for certain that this particular control
idly at a test so correct identication is below ceiling levels, adopted an explicit strategy, it almost perfectly matches
previously studied words are identied more accurately than the mean hit rate in the recognition task, leaving this as
nonstudied words (e.g., Jacoby and Dallas, 1981). This per- the most reasonable hypothesis. Therefore, simply knowing
 Functional Role of the Human Hippocampus 563

Figure 128. Visual-paired comparison task. Participants view two copies of the same
image and after a variable delay are shown a new image and the old image. A bias exists to
spend more time looking at the new image. This is an implicit memory task, yet the bias
to look at the novel picture is dependent on the medial temporal lobes. (Source: Data are
from Manns et al., 2000.)

that the task requires no more than implicit memory does 128). During this time, the participant is free to examine
not guarantee that the participants will treat it as an implicit either copy of the object at will, and no task instructions are
task when there are multiple routes to solve the prob- given. After some delay (in which intervening items may be
lem. Where implicit tasks or any other tasks are impaired presented), a copy of the previously exposed object or scene is
in persons with amnesia, this possibility must certainly be presented along with a novel object or scene; and again the
considered. participant is free to examine either stimulus at will. The task,
A second form of priming, known as conceptual priming, is quite implicit in instruction and behavior. Participants have
is also apparently intact in amnesia. Here, the test presents an a tendency to spend a greater amount of time looking at the
item or a category that is semantically or conceptually related novel stimulus than the familiar stimulus. A small bias or
to the studied item. For example, if the word peach had been change is made in behavior as a result of experience on a task
presented at study (again, the study task is often incidental), that makes no reference to the prior study episodehall-
participants might be asked to generate exemplars of a cate- marks of implicit memory tasks. In fact, the task is commonly
gory (e.g., generate examples of fruits), verify category mem- used to assess memory in infants who would certainly not
bership (e.g., how long it takes to verify that peach is a fruit), understand any explicit instructions even if given (Fagan,
or perform a free-association task (e.g., free-associate to the 1970). Yet, the task is dependent on the hippocampus in that
word pear). In each of these, amnesic patients with damage no such bias is seen following hippocampal damage in
including but not limited to the hippocampus have shown humans (McKee and Squire, 1993; Manns et al., 2000; Pascalis
normal levels of conceptual priming (Graf et al., 1984; et al., 2004), the monkey (Bachevalier et al., 1993; Pascalis and
Shimamura and Squire, 1984; Vaidya et al., 1995; Keane et al., Bachevalier, 1999; Zola et al., 2000), or the rat (Clark et al.,
1997; Levy et al., 2004) while being impaired in explicit tests 2000). Furthermore, the degree of bias shown in the visual
for the same type of information. paired comparison task is predictive of subsequent recogni-
A task called visual-paired comparison appears to be the tion memory for the items shown, whereas the amount of
one exception to the rule that implicit memory tasks are non- repetition priming exhibited is not (Manns et al., 2000). Thus,
declarative in nature and are not impaired by damage to the even though the task requirements are implicit in nature (no
hippocampal region. In this task, two copies of an object or a reference is made to the study episode) and even though the
scene are presented to the participant for several seconds (Fig. memory is observed in the form of a change in a behavioral
564 The Hippocampus Book

bias, the task relies on the same underlying mechanisms that tasks (at least when the task is solved in an implicit way). The
are responsible for declarative memory. visual-paired comparison task is the one apparent exception
One nal task deserves consideration when discussing the to this rule, highlighting the importance of extending or ren-
role of the hippocampus in relation to explicit and implicit ing the descriptive distinctions created to understand human
memory. Chun and Phelps (1999) embedded an implicit behavior (e.g., implicit/explicit) to more fully encompass
memory task within a visual search task (locate a rotated T data from other sources. The simple description of the task
among numerous rotated L distractors). Although the displays requirements does an admirable job of dissociating two
appeared to be random, a set of 12 displays were repeatedly classes of tasks, yet there is clearly something about the visual-
intermixed with random displays throughout the experiment. paired comparison task that, despite its implicit nature,
As one might expect, the reaction time to locate the target makes it dependent on the hippocampal region and the adja-
item in the random displays gradually decreased over the cent medial temporal lobe cortices.
course of the task. This basic practice effect or increased skill
in performing the task was similar in a control group and an 12.4.2 Encoding Versus Retrieval
amnesic group (two anoxic patients and two patients with
encephalitic damage that included, but extended beyond, the Whereas studies of patients with damage to the hippocampal
hippocampal region), consistent with other implicit or non- region have been able to demonstrate a dissociation between
declarative tasks. Furthermore, although approximately half declarative memory tasks that involve the hippocampus and
of the healthy controls reported noticing repetition of several nondeclarative memory tasks that do not, studies of patients
displays when later asked, not even they could identify which are not particularly well suited to isolating the role of the hip-
they were (54% correct). Critically, however, the controls pocampus in encoding (or storage) versus retrieval processes.
reaction times to the repeated items were faster than their The inability to make accurate judgments in a recognition
reaction times to the random items in the later epochs of memory task or to recall items from a study list could be the
training. Although they lacked explicit knowledge of the result of failure to encode the items initially or failure to
repeated items, they had implicit knowledge of this repetition, retrieve a well encoded item. The observation that amnesic
evidenced by their reaction time. In marked contrast, the patients can retrieve fact or event-type memory that was
amnesic patients did not show this effect. The amnesic learned well prior to the onset of their amnesia (see Section
patients exhibited identical reductions in reaction time for 12.4.3) can be taken as evidence that the hippocampus is not
repeated and random displays. continually required for declarative memory retrieval; how-
Thus, a second task exists in which the task demands and ever, it still leaves open the possibility that when anterograde
memory are implicit in nature, yet performance requires and retrograde amnesia are observed the central impairment
structures in the medial temporal lobes. This said, perform- is one of retrieval rather than encoding or storage (e.g.,
ance may not require the hippocampus itself. A follow-up Warrington and Weiskrantz, 1970).
study by Manns and Squire (2001) solidly replicated the basic Thus, in humans, determining whether the hippocam-
ndings of Chun and Phelps (1999). Here, however, the pus plays a differential role in encoding or retrieval relies
amnesic patient group was large enough to separate into two heavily on electrophysiological and neuroimaging data.
groups based on size and extent of the lesion. Five patients had Complementing this, studies of animals offer the opportunity
damage limited to the hippocampal region or only mildly of employing reversible lesions, blockade of long-term poten-
extending into the parahippocampal gyrus. A separate group tiation (LTP), and other manipulations (discussed in Chapters
of three were encephalitic patients who had extensive, near- 8 and 13). One of the challenges that faces all such attempts is
complete damage to the MTL and mild damage that extended the observation that during memory retrieval tasks partici-
into the lateral temporal cortex (consistent with this etiology). pants unwittingly encode the test items and can often accu-
Notably, the only group that did not show reduced reaction rately remember whether an item was present during the test
times specic to repeated displays was the group with exten- (for review see Glover, 1989). fMRI activity associated with
sive MTL damage. These patients performed similarly to the this incidental encoding has been observed in numerous
mixed etiology group in Chun and Phelps (1999) study. In frontal and parietal regions (Buckner et al., 2001) and bilater-
contrast, the patients with damage more restricted to the hip- ally in the hippocampal region, perirhinal cortex, and
pocampal region performed much like the healthy control parahippocampal cortex (Stark and Okado, 2003).
volunteers in both studies, demonstrating improved reaction This challenge aside, hippocampal activity has been
times for repeated displays. Therefore, our best understanding observed relating to both encoding and retrieval success. A
of this task is that implicit memory for the repeated displays is number of fMRI studies (Small et al., 2001; Davachi and
not derived from the hippocampal region itself but, rather, is Wagner, 2002; Reber et al., 2002; Strange et al., 2002; Davachi
derived from some other temporal lobe structure. et al., 2003; Stark and Okado, 2003; Kirwan and Stark, 2004)
In summary, we can conclude that although the hip- have observed hippocampal activity correlated with encoding
pocampus plays a role in explicit or declarative memory tasks, success. In this so-called Dm (differences due to memory)
it does not play a role in implicit or nondeclarative memory effect (Paller and Wagner, 2002), there is greater activity dur-
 Functional Role of the Human Hippocampus 565

ing the successful encoding of items that are later remembered also observed a combination of these locations (and several
than during the unsuccessful encoding of items that are later others) during the encoding of facename pairs. At retrieval,
not remembered. Similar Dm effects have been observed in when participants were cued with a face and asked to recall the
several studies employing recording from the hippocampal name, the pattern of activity in the hippocampal region was
region using depth electrodes as well (Fernandez et al., 1999b; quite similar to the pattern of activity observed when encod-
Cameron et al., 2001; Fernandez et al., 2002; Fell et al., 2003). ing the facename pair. Similarly, using an anatomical ROI
We should note that these Dm effects have been frequently analysis, Reber et al. (2002) observed encoding effects for both
observed in the parahippocampal cortex (e.g., Brewer et al., words and pictures throughout the longitudinal axis of the
1998; Wagner et al., 1998; Fernandez et al., 1999b; Otten et al., hippocampal region and the adjacent cortical structures (dif-
2001; Davachi and Wagner, 2002; Fernandez et al., 2002; ferences were observed between picture and word encoding,
Strange et al., 2002; Davachi et al., 2003) and the entorhinal or however).
perirhinal cortices (Fernandez et al., 1999b; Cameron et al., Furthermore, in a study that attempted to reduce the
2001; Davachi and Wagner, 2002; Fernandez et al., 2002; effects of incidental encoding during retrieval (Stark and
Strange et al., 2002; Davachi et al., 2003; Fell et al., 2003; Okado, 2003), encoding and retrieval each resulted in activity
Kirwan and Stark, 2004) as well. in the hippocampal region and in the perirhinal and parahip-
Similarly, a number of fMRI studies have demonstrated pocampal cortices. It should be noted, however, that in this
activity in the hippocampal region related to retrieval success study the exact voxels associated with intentional encoding,
(e.g., Gabrieli et al., 1997; Eldridge et al., 2000; Stark and incidental encoding, and retrieval were not always the same.
Squire, 2000a, 2000c, 2001a; Stark and Okado, 2003; Kirwan Likewise, Pihlajamaki et al. (2003) found evidence for greater
and Stark, 2004). Here, successful retrieval (e.g., responding activity during retrieval than encoding and greater activity
old to previously studied targets) usually elicits greater during encoding than retrieval in areas of the perirhinal and
activity than unsuccessful retrieval (e.g., responding new to parahippocampal cortices and the hippocampal region. Thus,
unstudied foil items). Similar effects have been observed using it is apparent that although there may be differentiation
depth electrode recordings (Paller and McCarthy, 2002) and between encoding- and retrieval-related functions on a ne
magnetic source imaging (Papanicolaou et al., 2002). scale, encoding and retrieval processes appear to be present
Although apparently engaged in both successful encoding throughout the MTL.
and retrieval processes, a dissociation between the two is still
possible; and several studies have observed differences in the 12.4.3 Time-limited Role in Declarative Memory
hippocampal region between memory encoding processes
and retrieval processes. For example, Zeineh et al. (2003) used At the beginning of the chapter, we noted Milners observa-
the above-mentioned unfolding techniques to isolate encod- tion that patient H.M.s remote memory appeared to be intact
ing and retrieval-related activity for facename pairs. They in the face of both his profoundly impaired ability to learn
reported above-baseline (visual xation on a cross) activity new information and a profound loss of information that he
during encoding but not retrieval in CA2, CA3, and the den- had been exposed to for some amount of time prior to his
tate gyrus, with this activity decreasing across repeated pre- operation. Given his intact childhood memories, she con-
sentations (evidence of encoding-related activity was seen in cluded that the medial temporal lobes were not the ultimate
the parahippocampal cortex as well). In contrast, retrieval storage site for what we now refer to as declarative memory.
(and to some degree encoding) was associated with activity in Therefore, some form of systems-level consolidation occurs
the subiculum, also showing a decrease in activity with by which memories that initially rely on structures in the
repeated presentations. Thus, some differentiation between medial temporal lobe become independent of the medial tem-
encoding and retrieval was observed in components of the poral lobe over time. (This is not to be confused with an alter-
hippocampus. native use of the term consolidation to refer to the xation
In addition to differentiating encoding from retrieval of a memory over the course of seconds to hours.) This phe-
across these subregions (see Chapter 3), several attempts nomenon, termed temporally graded retrograde amnesia,
have been made to differentiate activity along the anterior- has been observed frequently in amnesic patients since rst
posterior (or longitudinal) axis (Gabrieli et al., 1997; LePage et being described by Ribot more than 100 years ago (Ribot,
al., 1998; Fernandez et al., 1999a). However, it appears as if the 1887).
initial support for this hypothesis has not been conrmed in This observation has been frequent but not entirely consis-
the further analysis of these and subsequent studies (Schacter tent. Although some have described it as temporally graded
and Wagner, 1999). For example Small et al. (2001) used an (e.g., Squire and Alvarez, 1995), others have described it as
anatomical ROI analysis to examine encoding- and retrieval- constant (e.g., Warrington and McCarthy, 1988). Some have
related activity along the longitudinal axis of the hippocampal found evidence that memory for both facts (semantic mem-
region (eight ROIs for each participant along this axis). Here, ory) and events (episodic memory) acquired before the onset
they observed activity in different locations along the longitu- of amnesia are similarly impaired (e.g., Verfaellie et al., 1995;
dinal axis for encoding faces and for encoding names. They Reed and Squire, 1998), whereas others have found more
566 The Hippocampus Book

selective (and constant) impairment in episodic memory actually engaged in an undetermined and uncontrolled task
(e.g., Nadel and Moscovitch, 1997). One potential source of (e.g., reecting upon the experiment). When activity during
variability in the data stems from the fact that all studies of episodic recollection (e.g., Reect upon your visit to Paris as
retrograde memory in human amnesic patients are retrospec- a child) is contrasted with rest, it is quite plausible that this
tive and quasi-experimental in nature. When testing a reection is carried into the rest periods. Such activity during
patients memory for knowledge acquired before the experi- rest could reduce the magnitude of an effect such that it might
ment (and potentially years or decades before the experi- not be observed or, if the amount of reection during rest or
ment), we cannot know how well the information had been the actual task performed during rest varied with condition
learned prior to the onset of amnesia or even if it had been (e.g., participants might be more likely to reminisce or
learned at all. Furthermore, we cannot know whether the attempt to further elaborate a memory following retrieval of a
information had been retrieved, and therefore reencoded, at very old than a recent memory), one could create or even
some time following the initial learning. Such retrieval- invert a gradient purely as an artifact (Stark and Squire,
induced encoding could easily affect the presence or absence unpublished data).
of a temporal gradient. Likewise, it is a matter of debate Given these challenges, it is not entirely surprising that in
whether such retrospective tests should employ information neuroimaging studies of consolidation the results have been
that shows a forgetting gradient in healthy controls or they mixed. For example, Ryan et al. (2001) asked participants to
should employ information that can be retrieved at a constant generate very remote ( 20 years) and relatively recent ( 4
level of performance across the supposed delay. years) episodic memories while outside the scanner. Inside the
Additionally, the exact location and extent of the hip- scanner, they were asked to recall these memories for 20 sec-
pocampal lesion is often not known nor whether it extends onds. In bilateral hippocampal regions that showed an overall
beyond the medial temporal lobes. There are a relatively small difference in activity between recollection and rest, there was
number of cases in which damage appears to be limited to no difference in activity between activity for very remote and
the hippocampal region and retrograde amnesia has been relatively recent memories (an alternate baseline of sentence
assessed (Zola-Morgan et al., 1986; Reed and Squire, 1998; completion was also used and did not differ from rest, but as
Kapur and Brooks, 1999; Holdstock et al., 2002a). In all but this task required one sentence completion in 5 seconds it may
one patient (Y.R. in Holdstock et al., 2002a), a temporally be open to the same difficulties as explicit rest). Furthermore,
graded retrograde amnesia spanning several years was no signicant activity was observed in any of the subjects in a
observed for both episodic and semantic memory. In the case remote-versus-recent contrast.
of Y.R., no retrograde amnesia was detected at all in this study. Similarly, Maguire et al. (2001) collected data as partici-
However, follow-up testing on Y.R. has revealed evidence for pants performed a verication task on both public event
at least some retrograde amnesia. Its precise nature with information and autobiographical information from periods
respect to temporal gradients and with respect to selectivity ranging from several weeks to more than 20 years prior to
for any kind of information or task is currently not known, scanning (e.g., Yes or no: You were at Tims wedding in
however (unpublished data, Andrew Mayes, personal commu- London). Although hippocampal activity during both tasks
nication, February 22, 2005). was greater than baseline (listening to a set of function
Using neuroimaging to study the time-limited nature of words), and it was greater for autobiographical than public-
hippocampal function is also associated with a number of event verication, it did not vary parametrically with memory
challenges. One clear challenge is the problem of activity asso- age. However, the participants knowledge of this information
ciated with incidental encoding. It is quite plausible, for exam- had been assessed several weeks prior to scanning, raising the
ple, that if memories have been consolidated and no longer possibility that they were retrieving the information from the
require structures in the MTL the retrieval of such a memory recent reencoded episode.
(e.g., a childhood memory that has not been thought about Finally, Stark and Squire (2000a) attempted a prospective
for some time) induces encoding in the MTL. If participants study of consolidation by having participants study pictures
can tell you what memories they had retrieved while inside of nameable objects at varying delays prior to scanning. At
the scanner, activity associated with this incidental encoding test, the names of the objects were presented. In so doing,
could mask (or even reverse) any retrograde gradient (Stark much of the incidental encoding-related activity appears to be
and Okado, 2003). conned to the left hemisphere, leaving retrieval-related activ-
A second challenge that perhaps faces studies of consolida- ity for the nameable objects relatively uncontaminated in the
tion more strongly than most is the difficulty posed by the right hemisphere (Stark and Squire, 2001a). Activity in
choice of baseline tasks. Several studies have shown that there anatomically dened ROIs in the MTL did not differ as a
is signicant activity throughout the brain (including the function of study-test interval. It should be noted, however,
MTL) during rest or other baseline tasks that do not actively that the delays employed were all relatively short, ranging
engage the participant (Binder et al., 1999; Gusnard et al., from an hour to a week.
2001; Gusnard and Raichle, 2001; Newman et al., 2001; Stark In contrast to these negative ndings, there have been sev-
and Squire, 2001b). One explanation for such activity is that eral studies that have shown gradients in MTL activity as a
during rest and other low-level baseline tasks participants are function of memory age. In the rst such study, Haist et al.
 Functional Role of the Human Hippocampus 567

(2001) used the famous faces task (Marslen-Wilson and that our best evidence for any time-limited role of MTL struc-
Teuber, 1975) in which participants were presented with pho- tures will be gained from animal models of amnesia.
tographs of faces from various decades (of both famous and Controlled prospective studies have been conducted in that
nonfamous people) and attempted to recall the names of arena that, by and large, reveal temporal gradients following
each. An analysis identifying regions whose activity varied lin- hippocampal damage (for review see Squire et al., 2001).
early by decade (activity determined by a contrast between
famous faces from each decade and rest) isolated activity in a 12.4.4 Spatial Memory
region that appeared to be the right entorhinal cortex. Here,
activity was greatest when participants attempted to recall The notion of a hippocampal place cell whose activity codes
names from the 1990s and 1980s and lowest when partici- for the current location in a spatial environment is covered
pants attempted to recall names from the 1940s. No other extensively in Chapter 11. The hypothesis that the hippocam-
region showed this linear trend. pus is primarily or even uniquely involved in spatial process-
In a second study, Niki and Luo (2002) used a version of ing and spatial storage is laid out there, drawing extensively on
the episodic reection task in which participants attempted electrophysiological data from the rat. This hypothesis is con-
mentally to revisit places they had visited either recently or sidered again in Chapter 13, where a wide array of data from
more than 7 years prior to scanning. In the direct comparison rats, monkeys, and humans are considered. Before this more
between recent and remote memories, activity in the left comprehensive treatment, the data concerning the role of the
parahippocampal gyrus was observed, with greater activity for human hippocampus in spatial memory and spatial process-
recent than remote memories. Although a number of other ing are considered. In so doing, the principal question is not
regions throughout the brain also showed this pattern, a sub- whether damage to the hippocampus impairs performance on
stantial number showed the reverse, with greater activity for spatial memory tasks or whether place cells are observed in
remote than recent memories. the human hippocampus. In fact, neuropsychological (e.g.,
Finally, Maguire and Frith (2003) again examined autobio- Holdstock et al., 2000) and electrophysiological data (Ekstrom
graphical and public event knowledge as a function of et al., 2003) have demonstrated both, implicating the human
remoteness using a task similar to that in their previous work, hippocampus in spatial memory tasks.
described above (Maguire et al., 2001). Although they again For example, Ekstrom et al. (2003) recorded data from
observed greater hippocampal activity for autobiographical depth electrodes implanted in the hippocampal region,
than public event memory, they did observe a gradient in parahippocampal gyrus, amygdala, and several frontal sites as
right hippocampal activity as a function of remoteness of the participants navigated in a virtual town. Their task was to pre-
autobiographical memory (no gradient was observed in the tend to be a taxi driver, picking up and dropping off passen-
left hippocampal region). gers. Several codes were observed in the data. In particular,
Thus, with respect to the expectation that a time-limited 11% of the cells (31/279) could be condently classied as
role of the MTL in declarative memories produces a gradient place cells, with their distribution skewed toward being found
in the BOLD fMRI signal, the current state of affairs is mixed. in the hippocampal region. Approximately 24% of hippocam-
There are several null results in which no gradient was pal neurons could be condently classied as place cells,
observed. There are, however, several positive results as well. whereas signicantly fewer neurons in the parahippocampal
None of the studies has been able to address fully the issue gyrus, approximately 8% of those sampled, could be con-
of incidental encoding (and how it might vary with remote- dently classied as place cells. A second code was observed pri-
ness), and many suffer from the difficulty that it is unclear marily in the parahippocampal gyrus. These cells, termed
whether participants are recalling events from several decades location-independent view cells, were observed to code for
ago or from their retrieval during prescreening sessions. particular objects or landmarks irrespective of their location.
Furthermore, all studies have ignored the possibility that over A total of 14 of the 279 neurons (5%) were classied as such
the course of decades the representation of the memory may and were observed largely in the parahippocampal gyrus.
change in such a way that the BOLD fMRI signal is affected Approximately 15% of neurons in the parahippocampal gyrus
without altering the functional role of a particular region were location-independent view cells (7 total) whereas only 1
(e.g., recently consolidated memories may have qualitatively of 55 neurons in the hippocampal region coded for this infor-
different representations in the tens of thousands of neurons mation. Although this dissociation was observed, one should
in a typical fMRI voxel than very long-standing memories). note that the numbers of cells and their proportion to the
One might consider this a clear criticism of the current stud- total number sampled is quite small. For example, of the 55
ies, and the solution to these problems is not clear; nor is it hippocampal cells recorded, 43 (78%) showed main effects of
clear how this question can be better approached with neu- factors other than place or coded for combinations of several
roimaging. The study of retrograde memory in humans has factors.
always been exceedingly difficult with most tasks being retro- Thus, the question at hand is not whether the hippocam-
spective rather than prospective in nature. When combined pus plays a role in spatial memory. Rather, the question is
with the above-mentioned difficulties presented by the limita- whether the function of the human hippocampus can be tied
tions of current neuroimaging techniques, it seems obvious strongly to spatial processing or the human hippocampus is
568 The Hippocampus Book

better viewed as playing a mnemonic role, with spatial mem- became amnesic). His performance dropped from 83% cor-
ory being only one example of hippocampal function. rect in his childhood environment to 0% correct in his current
One approach to this question that has been explored environment on the navigation tasks. Despite living on a hill
extensively in the rodent (see Chapter 11) (Eichenbaum et al., that overlooks the Pacic Ocean a mere 2 miles away, E.P. was
1999) is to ask whether damage limited to the hippocampal not even able to point in its direction. Thus, E.P. can retrieve
region impairs memory on nonspatial tasks. If such impair- spatial information and navigate in a spatial environment
ment exists and if spatial contributions to the task can be learned long before the onset of his amnesia but apparently
eliminated, one could posit a role for the human hippocam- has not learned any spatial information about his environ-
pus outside of spatial memory or processing. Although there ment after the onset of his amnesia. With complete hip-
are many reports of such impairment (see Spiers et al., 2001 pocampal loss, spatial processing therefore appears normal
for review), it is almost impossible to rule out the option that despite complete inability to acquire new spatial information.
healthy human control participants engage in a spatial strat- Neuroimaging studies have also begun to provide useful
egy on such nonspatial tasks. For example, even on tests of data on the role of MTL structures in spatial processing and,
verbal recognition memory, control participants might imag- consistent with lesion evidence (e.g., Epstein et al., 2001), have
ine the stimuli, imagine relations between stimuli, or use other clearly implicated the posterior portions of the parahip-
mnemonic techniques such as the method of loci to improve pocampal gyrus in topographical memory tasks. For example,
their performance. Were such spatial strategies not available to in an early study, Aguirre et al. (1996) observed MTL activity
patients with hippocampal lesions, their performance might that was conned to the posterior parahippocampal gyrus
be impaired for spatial, rather than mnemonic, reasons. As (likely parahippocampal cortex) as participants learned to
tests of recognition memory in humans have not aimed to navigate in a three-dimensional maze viewed from a rst-
address the spatial hypothesis directly, they have not con- person perspective (activity relative to a low-level control
trolled for these potential spatial aspects of the tests as care- task). Using a similar task, Maguire et al. (1998) also reported
fully as in much of the work in the rodent literature. parahippocampal gyrus activity associated with exploring and
An alternative approach to addressing the question is to learning the topography of a virtual three-dimensional envi-
assess the spatial abilities of patients who have lesions that ronment (at least when the environment contained objects
include the entirety of the hippocampus bilaterally. If spatial that could serve as landmarks). Whether the activity is specif-
processing appears normal despite a complete hippocampal ically related to topographical processing or is representative
lesion, it would be difficult to conclude that the hippocampus of more general mnemonic function cannot be determined
is required for spatial processing in humans. The case of from these data. However, it is of interest that they implicate
patient E.P. has provided compelling evidence that spatial the posterior portions of the parahippocampal gyrus and not
memory and processing may be intact despite hippocampal the hippocampal region in these topographical learning tasks.
loss (Teng and Squire, 1999). Unlike patient H.M., patient E.P. Consistent with these ndings, Shelton and Gabrieli (2002)
has what is most likely complete loss of the hippocampal also observed activity in the posterior parahippocampal gyrus
region bilaterally (Stefanacci et al., 2000), with only a small (with this region of activity extending into the posterior hip-
tag of tissue (~ 10% of the volume) remaining. With the pocampal region) as participants encoded a virtual three-
complete lack of entorhinal cortex in E.P., it is doubtful (even dimensional environment when viewed from a rst-person,
if this tissue contains healthy neurons, which it may well not) or route, perspective. Of interest, the pattern clearly differed
that it would be functional in any way. Yet, despite this com- when participants encoded environments from an aerial, or
plete loss of the hippocampal region, E.P. can perform a num- survey, perspective. Despite similar subsequent memory for
ber of complex spatial tasks at normal levels. When asked to environments learned from the two perspectives, activity in
navigate in the town in which he grew up, E.P. not only can the posterior parahippocampal gyrus was greater when par-
mentally navigate along familiar routes (from his childhood ticipants encoded the environment from a route perspective
house to other local landmarks), he can mentally navigate than from a survey perspective. The authors suggest that this
along novel routes (paths between randomly chosen land- difference may be the result of qualitatively different mnemo-
marks) and along novel routes with key paths blocked (e.g., nic demands placed on participants in the two conditions. For
imagine that a main road is closed). He also can perform a route encoding, participants must continually bind together a
dead reckoning task (imagine being at one landmark and representation of their present position to their past and
point in the direction of another landmark) at a level indis- future positions to create a layout of the entire environment.
tinguishable from that of healthy control volunteers who grew Although still spatial in nature, encoding the environment
up with E.P. and who also left during their young adulthood from a survey perspective did not place this same sort of
(Teng and Squire, 1999). The matched control volunteers had demand on participants. In a subsequent task that asked par-
an easier time mentally navigating in their current environ- ticipants to draw maps of the environment, the sequential
ments, scoring at the ceiling (100% correct). E.P., however, order of presentation was preserved following route learning
emphatically did not have an easier time navigating in the area and was not preserved following survey learning, further
he had moved into 6 years prior to testing (and 1 year after he suggesting a difference in the way the two environments were
 Functional Role of the Human Hippocampus 569

encoded that may have led to the observed difference in activ- pocampus is involved in spatial memory tasks but that its
ity in the parahippocampal gyrus. function is more generally mnemonic and not limited to spa-
Hartley et al. (2003) contrasted activity associated with tial memory. In humans, spatial memory is an excellent exam-
navigating in a large-scale virtual environment either by way- ple of complex, declarative memory.
nding (via route-based spatial knowledge gained by free
exploration) or by route-following (traversing the same route 12.4.5 Associations, Recollections,
that had been well learned during study). Unlike the simple Episodes, or Sources
route-following task, which required only recapitulation of a
well learned route, the way-nding task required knowledge A large amount of the research on the human hippocampus
of the global spatial relations in the environment and naviga- has been aimed at functionally dissociating the role of the hip-
tion along a novel path. Within the MTL, Hartley et al. (2003) pocampus from the role of adjacent cortical structures. As
reported greater activity in the posterior parahippocampal noted at the beginning of the chapter, a popular idea draws
gyrus overall during the way-nding task than during the on the anatomy to suggest that the hippocampus integrates
route-following task. Perhaps of more interest, with the way- information from and combines the processing of the adja-
nding condition, activity in the hippocampal region was cor- cent cortical structures that feed into the hippocampus. Two
related with accuracy of performance (ceiling effects made fundamental hypotheses that share this basic idea and that are
this test impossible with the route-following task). Further- both driven by data from human and nonhuman studies have
more, in an individual differences analysis that examined cor- been proposed and explored. One hypothesis states that there
relations between the participants overall performance in is a clear dissociation of function between the hippocampus
way-nding and neural activity differences between way-nd- and the adjacent structures. For example, the hippocampus
ing and route-nding, Hartley et al. (2003) reported a signi- has been described as being involved in memory that is
cant positive correlation in what is likely the perirhinal cortex associational, multi-item, spatial, episodic, and recollective,
and a signicant negative correlation in the head of the right whereas the perirhinal cortex (and by extension at times the
caudate. A positive correlation in the hippocampal region fell parahippocampal cortex) is involved in memory that is auto-
just short of signicance. Thus, participants who were better matic, noneffortful, single-item, and familiarity, or recency-
at navigation showed greater MTL recruitment during the based (in contrast to recollective), with this distinction being
way-nding task relative to the route-nding task than partic- qualitative rather than quantitative (Brown and Aggleton,
ipants who performed more poorly. Conversely, these same 2001).
good navigators showed greater caudate recruitment during The other hypothesis states that the dissociation of func-
the route-nding task relative to the way-nding task than tion is more quantitative than qualitative in nature and that
participants who performed more poorly. the hippocampus and the adjacent structures in the parahip-
One logical interpretation of these data (although not the pocampal gyrus are all broadly involved in declarative mem-
only one, given the contrastive, or relative, nature of fMRI) is ory. This is not to say that the medial temporal lobe is
that the better navigators recruited structures in the MTL equipotential in nature. By virtue of being farther up in the
more for the way-nding task and that they recruited the hierarchical structure of the medial temporal lobe (see
caudate more for the route-nding task. Furthermore, these Chapter 3), the hippocampus is proposed to combine and
results make the tantalizing suggestion that this differential extend the processing carried out by the entorhinal, perirhi-
recruitment might be causally related to their better perform- nal, and parahippocampal cortices (Squire and Zola-Morgan,
ance. Hartley et al. (2003) interpreted their results as being 1991). By virtue of receiving input from both perirhinal and
supportive of a role of the hippocampus in the use of a cogni- parahippocampal cortices (via the entorhinal cortex), the
tive map, but McNamara and Shelton (2003) suggested that hippocampus is in a position to be able to integrate informa-
the correlations observed in the hippocampal region are also tion across these structures and sources of information. Thus,
consistent with the view that the hippocampus processes associative, or conjunctive, processing can be attributed to
memory in a way that allows for its exible use in guiding the hippocampus. However, the structures in the parahip-
behavior (e.g., Eichenbaum, 2000). Moreover, in the study of pocampal gyrus also receive input from a wide range of
Hartley et al. (2003), the most reliable positive correlation in sources, putting them in a position to perform associative or
the individual differences analysis was observed in the perirhi- conjunctive processing as well. As the input to the hip-
nal cortex, not the hippocampal region. (One reliable correla- pocampus consists of more rened and further processed
tion was observed in the left hippocampal region, but this was information (its major input arrives from the entorhinal cor-
using a mnemonically laden contrast.) That the contrast tex, which receives approximately two-thirds of its input from
between way-nding and route-nding in the perirhinal cor- the perirhinal and parahippocampal cortices), the hippocam-
tex would strongly correlate with navigation skill provides a pus is in a position to perform different, potentially more
novel datapoint to help us understand the differentiation of abstract or complex associative or conjunctive processing.
function across structures in the MTL. This is not to say, however, that there is binary dissociation of
Thus, the evidence from humans suggests that the hip- function between the hippocampus and the adjacent cortical
570 The Hippocampus Book

structures according to associative or conjunctive versus sin- iarity signal to be used during recognition. Thus, we would
gle-item, episodic versus semantic, recollection versus famil- have clear functional dissociation between the hippocampal
iarity, and so on. region and the adjacent cortex.
The hypothesis that the hippocampus combines and The difficulty is that this pattern is not consistently
extends the processing of the adjacent cortex is underspeci- observed. For example, Manns et al. (2003) tested a group of
ed, as the current theories that take this view do not detail seven patients with bilateral damage limited to the hippocam-
how this process takes place (nor, often, do the theories that pal region (as determined by MRI) using the same tasks
propose a more binary dissociation). Data from amnesic employed by Yonelinas et al. (2002). Consistent with the
patients with damage limited to the hippocampus and data results of Yonelinas et al. (2002), the patients were impaired
from neuroimaging studies are presented in the following two on both tests of recognition and recall. However, there was no
sections. Overall, the data are not consistent with a clean, evidence of disproportional impairment. When recall per-
binary division of labor between the hippocampus and the formance was measured against recognition performance, the
cortical components of the medial temporal lobes. They are patients recall score was in the 30.5th percentile of their
more consistent with either the graded division of labor recognition distribution and the controls recall score was in
implied by the concept of combine and extend or with a the 30.6th percentile of their recognition distribution. When
division of labor along lines that have yet to be explored sig- the same z-score analysis was performed, the amnesics recog-
nicantly using these two techniques. nition performance was worse than their recall performance,
indicating greater impairment in recognition than recall.
Amnesic Patients Furthermore, the amnesics performance on the Remember-
Know task showed similar impairments for both Remember
The data available from amnesic patients with bilateral dam- (60% reduction) and Know (6071% reduction, depending
age thought to be limited to the hippocampal region are far on the scoring method) responses. Thus, although a null
from allowing consensus in support of either hypothesis. For result, the data of Manns et al. (2003) leave little room to sup-
example, Yonelinas et al. (2002) reported data from 56 port a clear functional dissociation between the hippocampal
hypoxic patients with presumed (although not conrmed region and the adjacent MTL structures with respect to recol-
see Section 12.3.3) bilateral damage to the hippocampal lective processes.
region. When compared to a similar-sized group of healthy Unfortunately, neither nding is entirely atypical of the
controls, the patients were impaired on both recall and recog- results found in the literature. With regard to a specialized role
nition memory tasks. When transformed into z-scores (so the for the hippocampus in this form of memory, the results are
patients performance is expressed relative to the mean and decidedly mixed, indicating a critical gap in our theories of
standard deviation of the controls performancesee Section hippocampal function or a critical methodological problem
12.3.2), the impairment in recall was larger than the impair- (such as the often-suggested possibility that our assessment of
ment in recognition. Thus, if recall tasks place greater the damage in these patients is incompletesee Section
demands on episodic or recollective processing than recogni- 12.3.3). In the particular case of these two studies, there is a
tion memory tasks, these data point to a differential role for potential solution to the disparity in the ndings in the
the hippocampus in this kind of memory. methodology. The potential solution serves to highlight how
In the same study, four of the hypoxic patients were tested difficult this research can be and how tenuous the observa-
using the Remember-Know procedure to assess recollective tions of differential impairments often are.
and familiarity components of recognition (Tulving, 1985). In In a detailed reanalysis of the individual participant data
this task, participants indicate at the time of retrieval whether from both data sets, Wixted and Squire (2004) noted that a
their memory is best described as recollective, containing single control participant of the 55 tested in the Yonelinas et
episodic components and clear knowledge of the source of the al. (2002) study had a recognition score that was a marked
memory (a Remember response), or best described as a feel- outlier in the distribution of scores. The effect of this data-
ing of familiarity (a Know response). Whereas patients with point was sufficient to skew the z-score analysis (see Section
MTL damage known to extend beyond the hippocampal 12.3.2) and mask the apparent recognition memory impair-
region were impaired in both recollective and familiarity com- ment.
ponents, the hypoxic patients were signicantly impaired only Results such as these led to a confused state of affairs in the
in the recollective component. Thus, barring any concern literature, as it was often unclear what to make of the conict-
about the lesion locations in the hypoxic patients, these data ing data and what factors have inuenced the differing results.
appear to support the conclusion that the hippocampus plays Looking at the effects on recall versus recognition, recollection
a key role in recall and in recollective processing, and that versus familiarity, associations versus single-items, or episodic
these processes may not be able to be supported by the adja- versus semantic memory, conicting results abound. For
cent cortex. Conversely, if the lack of familiarity impairment example, Jon, a developmental-amnesic patient (neonatal
is true (in the four participants the 23% impairment was hypoxia) with damage limited to the hippocampal region (as
unreliable, but this may be the result of insufficient statistical assessed by MRI), has been shown to exhibit clearly impaired
power), the hippocampal region may not provide any famil- episodic memory but has done relatively well in school and
 Functional Role of the Human Hippocampus 571

obtained a solid vocabulary despite his amnesia (Vargha- region (location or extent), the stimuli, the tasks, or the man-
Khadem et al., 1997). Jon has also been shown to demonstrate ner in which the patients approach the tasks is currently
relatively normal levels of recognition memory (except on unknown. Perhaps most critically, the disparate results could
cross-modal tasks), relatively intact familiarity, and impaired arise from attempting to nd evidence for the wrong dissoci-
recollection (Baddeley et al., 2001; Vargha-Khadem et al., ation. If the functional role of the hippocampus is correlated
2001). Although Jons amnesia was neonatal, it appears that with episodic, recollective, and recall aspects of declarative
this is not the source of the dissociation. While examining memory, but in fact none of these aspects best describes its
early-onset (perinatal to 3 months) versus late-onset (614 role, we could be left with a pattern of positive and negative
years) amnesia in children with MRI-conrmed hippocampal results such as we currently face. For each positive result, some
damage, Vargha-Khadem et al. (2003) noted substantial aspect of the task or stimuli correlated better with the
impairments on several standardized tests of episodic mem- unknown true functional role of the hippocampus than for
ory in both groups. In contrast, semantic memory, as assessed each negative result.
by measures of vocabulary acquired outside the laboratory,
was in the low-average range in both groups, perhaps indicat- Neuroimaging
ing only mild impairment (unfortunately, it is difficult to
know if the amount of vocabulary training in the school or at A signicant number of recent neuroimaging studies have
home was similar to that of normals in these patients. (See sought to complement the amnesic patient data on the func-
Vargha-Khadem et al., 2001 for a similar nding.) tional dissociation between the hippocampal region and the
Furthermore, Y.R., an adult-onset amnesic patient adjacent cortical structures with respect to the associative, rec-
(Holdstock et al., 2002a,b; Mayes et al., 2002) showed a simi- ollective, or source components of declarative memory.
lar impairment pattern. Across a series of 34 recall tests, Although several patterns are beginning to emerge, the recent
Y.R. was substantially impaired relative to a group of healthy neuroimaging data also do not support a clean distinction
controls, obtaining an average z-score of 3.6. However, across between the role of the hippocampal region and the role of the
a series of 43 recognition memory tests, Y.R. was only mildly adjacent cortical structures along these lines. Clearly, most
impaired, averaging a z-score of 0.5 (Mayes et al., 2002). studies to date have observed activity in the hippocampal
(Interestingly, by not showing a bias to view the novel item, region associated with associative, recollective, or source com-
Y.R. exhibits impaired performance in the visual paired com- ponents of memory. However, in almost all of these studies,
parison task at 5- to 10-second delays). Thus, even in a case activity in the (presumed) parahippocampal cortex mirrors
of adult-onset amnesia, selective impairment in recall relative that in the hippocampal region. Furthermore, several have
to recognition has been observed following hippocampal reported associative, recollective, or source components in
damage. (presumed) perirhinal or entorhinal cortices. In contrast,
In contrast with the reports on Y.R. and the developmental activity correlated with single-item memory or familiarity
amnesic patients, several studies other than that of Manns et may not be limited to the parahippocampal gyrus, or even the
al. (2003) have shown substantial impairments in item-recog- perirhinal cortex specically (see Section 12.4.5). For exam-
nition memory tasks in patients with hippocampal damage ple, Henson and colleagues (2003) reported activity associated
(Hopkins et al., 1995; Reed and Squire, 1997; Stark and Squire, with familiarity in several studies within anterior portions of
2000b, 2003; Stark et al., 2002). Thus, in some patients, hip- the MTL that appear to include entorhinal/perirhinal cortices
pocampal damage leads to substantial recognition memory and the hippocampal region as well. Thus, the conclusion best
impairments that are similar to their recall impairments. drawn from the following review is that the existing data are
Furthermore, in two of these studies (Stark et al., 2002; Stark certainly not supportive of a clean functional distinction
and Squire, 2003), single-item recognition (e.g., Did you see between the hippocampal region and the adjacent structures
this object before?) was compared with associative recogni- according to the recollective, associative, or source compo-
tion (e.g., Youve seen both these objects before, but were they nents of declarative memory.
paired together?) in patients with damage limited to the hip- Several studies (e.g., Henson et al., 1999; Eldridge et al.,
pocampal region. In both, single-item and associative recogni- 2000) have examined the dissociation between recollection
tion patients with hippocampal damage were impairedand and familiarity using the Remember-Know (Tulving, 1985)
to the same degree (e.g., 15% impairment in single-item task. In this task, Remember responses are used to index rec-
recognition and 13% impairment in associative recognition in ollective memory, and Know responses are used to index
Stark et al., 2002, experiment 1). recognition based on familiarity. Thus, a contrast between the
Thus, we have a number of studies that report a dissocia- two might be used to assess whether a region such as the hip-
tion showing selective impairment in recall, recollection, or pocampus is particularly involved in recollective forms of
associative memory following hippocampal damage and a memory retrieval. Henson and colleagues (1999) reported lit-
number of studies that clearly show no such dissociation. tle differentiation between Remember and Know responses
Currently, there is no clear means of explaining the varying in the MTL, with no differences observed at the time of the
patterns of results. Whether this is due to differences in unde- test and only a small region in the parahippocampal gyrus
tected extrahippocampal damage, damage in the hippocampal showing less activity for subsequent Remember responses
572 The Hippocampus Book

than Know responses at the time of the study (signicant dif- restricted its analysis to the hippocampal region (Small et al.,
ferences were observed in frontal and parietal regions). 2001), activity was observed not only for the encoding and
In contrast, Eldridge and colleagues (2000) observed retrieval of facename associations but also for the encoding
greater activity for Remember judgments than Know judg- and retrieval of individual faces and names (although hip-
ments, correct rejections, and misses in the hippocampal pocampal activity for facename associations was more wide-
region. These data are consistent with a greater role for the spread than the combination of activity for faces and activity
hippocampal region in Remember responses than in Know for names alone). In both, a blocked design was used to con-
responses. However, there are some limitations to the trast activity associated with viewing novel facename pairs
Remember-Know task. First, it is difficult to perform in the with activity associated with viewing repeated facename
fMRI scanner because it requires two steps to avoid becoming pairs. Unfortunately, the blocked nature of the design
a simple condence rating (Hicks and Marsh, 1999). If two impeded the ability to isolate individual encoding trials based
steps are used (as was done in Eldridge et al., 2000), two cog- on the quality or amount of information subsequently
nitive processes are being engaged (rst deciding yes/no retrieved. Thus, in this study, it is difficult to know whether
recognition and then classifying the retrieval as Remember or the greater activity associated with viewing novel facename
Know), and it is impossible in fMRI data to separate activity pairs was the result of encoding the face, the name, or the
from two events that always immediately follow each other association or whether the activity resulted from a novelty
(see Section 12.3.5). Additionally, it is difficult to ascribe the detection (e.g., Strange et al., 1999).
enhanced activity for Remember responses entirely to a func- Other studies contrasting associative and nonassociative
tional dissociation favoring the recollective component of memory have observed more widespread activity. For exam-
recognition. It is quite plausible that any more detail-rich ple Henke and colleagues (1997) used PET to measure activity
retrieval would result in more activity than a detail-poor during the encoding and retrieval of facehouse pairs. More
retrieval. As such, Remember responses might yield more recently, Henke et al. (1999) used fMRI to measure activity
activity than Know responses in a region not particularly during the encoding of semantic associations between words.
involved in the recollective component itself. Finally, even if In both studies, greater activity for associative relative to
one reasonably assumes that some portion of the enhanced nonassociative memory was observed in both the hippocam-
activity for Remember responses over Know responses can be pal region and the parahippocampal gyrus. Similarly, in an
attributed to recollective or associative aspects of processing, fMRI study of recognition memory, Yonelinas et al. (2001)
Eldridge et al. (2000) observed greater activity for Remember reported greater activity during associative than nonassocia-
than Know responses not only in the hippocampal region but tive recognition in both the hippocampal region and the pos-
also in the parahippocampal gyrus (likely to be in the parahip- terior parahippocampal gyrus. Likewise, Pihlajamaki et al.
pocampal cortex). Thus, this recollective signal was not lim- (2003) observed their most consistent activity during the
ited to the hippocampal region. encoding of picture pairs not in the hippocampal region itself
Activity tied to familiarity must be considered as well. (where there was some evidence) but in the perirhinal cortex;
Henson and colleagues (2003) reported activity associated and Sperling et al. (2003) reported activity associated with
with familiarity in several studies in anterior portions of the high-condence encoding of facename pairs in anterior por-
MTL that appear to include entorhinal/perirhinal cortices and tions of the MTL that covered the hippocampal region and the
the hippocampal region as well. Likewise, Stark and Squire parahippocampal gyrus bilaterally (but whose most active
(2000c, 2001a) reported activity in the hippocampal region voxels were in the hippocampal region bilaterally and in the
during simple recognition memory tasks that can rely purely right entorhinal cortex).
on familiarity, with no apparent increase in activity when the In addition to studying arbitrary associations, a number of
task became more associative or recollective (Stark and attempts have been made to examine the automatic associa-
Squire, 2001a). Thus, the data do not support a clean func- tions that are formed between memory for an item itself and
tional dissociation between recollective processing in the hip- the source memory or episodic knowledge of when and
pocampal region and familiarity processing in the structures where that item was encountered (e.g., Did you see this item?
of the parahippocampal gyrus. If so, which study task was it in?). In an elegant example that
Several other studies have examined activity during the looked at encoding activity that predicted subsequent mem-
formation or retrieval of paired associates that may be used to ory for an item along with the source relative to encoding
determine whether the hippocampal region might be more activity that predicted memory for the item alone, Davachi et
involved in associative components of declarative memory al. (2003) reported a functional dissociation. Activity predict-
than in nonassociative or single-item components. In one case ing subsequent source memory was somewhat widespread,
(Sperling et al., 2001), associative encoding-related activity with this associative activity observed bilaterally in the hip-
was observed in the hippocampal formation (dened as the pocampal region and in the left parahippocampal cortex. In
hippocampus proper, subiculum, and entorhinal cortex) contrast, item-only activity was observed solely in the left
without observing significant activity elsewhere in the perirhinal cortex. Here, activity at encoding predicted subse-
parahippocampal gyrus, supporting the notion that the hip- quent memory in general (using as contrast greater activity at
pocampus is particularly involved in associative aspects of the time of encoding for items subsequently remembered
memory. However, in a similarly designed study, but one that relative to items subsequently forgotten), but the activity did
 Functional Role of the Human Hippocampus 573

not differ as a function of whether the source aspect was sub- whereas the hippocampal and posterior parahippocampal
sequently remembered. Thus, the hippocampal region and the gyrus activity could be easily explained as indexing source
parahippocampal cortex appeared to play a role in encoding retrieval, the entorhinal/perirhinal activity was a more com-
the source component, whereas the perirhinal cortex appeared plex combination of source and recency-related activity.
to play a role only in encoding the item component of the Thus, the current neuroimaging evidence has clearly impli-
memory (see Ranganath et al., 2003 for a similar nding). cated both the hippocampal region (Henke et al., 1997, 1999;
In a related study, Kirwan and Stark (2004) examined Eldridge et al., 2000; Small et al., 2001; Sperling et al., 2001,
activity during both the encoding and the retrieval of 2003; Stark and Squire, 2001a; Yonelinas et al., 2001; Dobbins
facename pairs. Like Davachi et al. (2003), greater activity et al., 2002; Davachi et al., 2003; Ranganath et al., 2003;
during the successful encoding of associations relative to their Kirwan and Stark, 2004) and the parahippocampal cortex
unsuccessful encoding (an associative pattern) was observed (Henke et al., 1997, 1999; Eldridge et al., 2000; Yonelinas et al.,
in the hippocampal region and the parahippocampal cortex 2001; Dobbins et al., 2002; Davachi et al., 2003; Ranganath et
(right unilateral in this case). Also like Davachi et al. (2003), a al., 2003; Kirwan and Stark, 2004) in recollective and associa-
portion of perirhinal cortex showed a general subsequent tive memory encoding and retrieval. Furthermore, there has
memory effect that did not vary with the associative compo- been some support for the perirhinal cortex and the entorhi-
nent (a single item pattern). Thus, associative signals were nal cortex in this regard as well (Dobbins et al., 2002;
observed in the hippocampal region and the parahippocam- Pihlajamaki et al., 2003; Sperling et al., 2003; Kirwan and
pal cortex whereas single-item signals were observed in the Stark, 2004) (see also the discussion of Hartley et al., 2003, in
perirhinal cortex. Section 12.4.4). Thus, it would be an oversimplication of the
However, unlike Davachi et al. (2003), reliable signals that data to conclude that the fMRI data support a specic or
ran counter to this simple dissociation were observed as well. unique role for the hippocampal region in associative or rec-
First, the associative pattern was observed in one portion ollective aspects of declarative memory, as the same patterns
of the right parahippocampal cortex during encoding, of activity clearly extend into the adjacent cortical structures
whereas the single-item signal was observed in a different por- (most often the parahippocampal cortex). Likewise, it would
tion of the right parahippocampal cortex. Thus, at the time of be an oversimplication to conclude that the fMRI data sup-
encoding, there was no clear differentiation observed between port a specic or unique role for any of the adjacent cortical
an associative hippocampal region and parahippocampal cor- structures in nonassociative forms of declarative memory.
tex and a nonassociative perirhinal cortex (nor, of course, was Although there is evidence for nonassociative or familiarity-
there clear differentiation between the hippocampal region based activity in the entorhinal and perirhinal cortices
and the parahippocampal gyrus as a whole). Furthermore, (Dobbins et al., 2002; Davachi et al., 2003; Henson et al., 2003;
when examining activity during recognition memory testing, Ranganath et al., 2003; Kirwan and Stark, 2004), there is evi-
the left perirhinal cortex, right entorhinal cortex, right hip- dence for such activity in the parahippocampal cortex
pocampal region, and right parahippocampal cortex all (Kirwan and Stark, 2004) and in the hippocampal region
showed an associative pattern of activity (greater activity dur- (Stark and Squire, 2000c; Small et al., 2001; Stark and Squire,
ing retrieval of the face(name pair and their association rela- 2001a; Henson et al., 2003). In addition, each of these struc-
tive to the retrieval of the face and name without their tures has been shown to respond according to the associative
association). Thus, at time of retrieval, greater activity in the or recollective nature of the task as well. Accordingly, although
associative than in the nonassociative condition was observed these neuroimaging data cannot be resolute with respect to
throughout the regions of the MTL, with similar amounts of function (all are correlational data sets), they certainly do not
associative activity (contrast between associative and nonasso- support clean dissociation of function in the MTL according
ciative retrieval) across MTL structures. to recollective, associative, or source components of declara-
Dobbins et al. (2002) reported a similar dissociation dur- tive memory.
ing retrieval of source information (what task was performed
at the study?) and recency information (which word appeared
more recently on the study list?). When examining activity in
the source retrieval task, activity in the hippocampal region, 12.5 Conclusions
the posterior parahippocampal gyrus (likely the parahip-
pocampal cortex), and activity in the entorhinal/perirhinal We began this chapter by asking the question, What does the
cortex all were associated with greater activity during correct human hippocampus do? The signicant amount of research
source retrieval than incorrect source retrieval. In the hip- that has followed Milners initial hypotheses has allowed us to
pocampal region and the posterior parahippocampal gyrus, reach a number of clear conclusions. First, along with the adja-
an overall task effect was observed, with both regions showing cent cortical structures in the parahippocampal gyrus, the
greater overall activity during source than recencyjudgments. human hippocampal region is critically involved in memory
In contrast, although activity in the entorhinal/perirhinal cor- for facts and events (explicit or declarative memory). Second,
tex indexed accuracy in the source retrieval task (it was this involvement is time-limited. Third, the hippocampal
affected by accuracy of source judgments), its activity was region and the adjacent cortex are not involved in immediate
similar for source and recency judgments overall. Therefore, or working memory process and are not involved in a wide
574 The Hippocampus Book

range of implicit or nondeclarative long-term memory process not entirely novel or arbitrary. If it were not, wed have a hip-
(although they may be used in working memory or implicit pocampus that is rarely doing anything at all, as few experi-
tasks that evoke declarative processes). Fourth, the hippocam- ences would satisfy the extreme end points of the dimensions
pal region is not involved in nonmnemonic aspects of cogni- laid out here.
tion including spatial processing (although spatial memory is Furthermore, it might well be that if one or more of these
a clear example of its time-limited mnemonic role). constraints were dropped or weakened, adjacent cortical
However, despite these strides, the differentiation of func- regions might be able to perform the task to some degree.
tion between the hippocampal region and the adjacent corti- With the bias toward visual information found in perirhinal
cal structures of the parahippocampal gyrus is not currently cortex, perhaps associative, complex, relational, novel, and
apparent. That the hippocampus is involved in associative, arbitrary information could be learned to at least some degree
recollective, or source components of declarative memory is if it were not heavily multimodal and if the test were not all-
quite clear. Lesions to the hippocampal region yield decits on or-none but sensitive to somewhat more gradual learning. By
tasks that assess these forms of memory, and neuroimaging still being somewhat multimodal, perirhinal cortex may sup-
studies have observed activity correlated with them as well. port learning of multimodal information but only when other
However, it is equally clear that the hippocampus can be constraints are weakened.
involved in single-item, nonassociative, and familiarity-based In this situation (which parallels the data at least in spirit),
components of declarative memory as well. Hippocampal what would we claim to be the function of the hippocampus?
lesions have resulted in decits on these tasks, and hippocam- Are there tasks that only the hippocampus can perform? Yes,
pal activity has been observed during them. Furthermore, the but, is that all that the hippocampus does? No. Is there a sin-
adjacent cortical structures (most notably the parahippocam- gle feature of a task that allows us to determine that the hip-
pal cortex) may also serve to support associative, recollective, pocampus and only the hippocampus can perform the task?
or source components of declarative memory. Thus, the avail- Not really. There is a combination of features, each of which
able data do not support a clean division of labor along any of points toward strengths of the hippocampus that, when taken
the lines proposed. together, isolate it as the only structure capable of performing
Some differentiation exists. Not only does the anatomy sug- the required task. By virtue of the anatomy, the hippocampus
gest functional differentiation, the research to date has often receives the most highly processed, multimodal information
yielded solid evidence of functional dissociations. However, and is well designed to learn new arbitrary information rap-
pushing our interpretations of these dissociations and the idly. Structures such as the perirhinal cortex receive less
theories that fall out of them has often led to clear disconnects processed, more modal-specic information and are less well
between the theoretical predictions and additional data. designed to learn new arbitrary information rapidly.
There is a difficulty faced when one tries to assign function Therefore, despite nding what might be a task clearly to
to a region based on the impairment observed following dam- isolate the hippocampus, we may very well be right back in the
age or based on a set of signals recorded from the region. If a opposite class of theory. Even with a task that dissociates the
hippocampal lesion impairs task X, does it mean that the hip- hippocampus from the rest of the brain, the hippocampus
pocampus does task X? Difficulties with dissociations aside may be combining and extending the processing of the adja-
(which are also discussed in Chapter 13), even if we are certain cent cortex in the medial temporal lobe. It is not that the hip-
that the ability to perform task X critically depends on pro- pocampus is doing anything all that different from the
cessing that occurs in the hippocampus, it is still a large leap adjacent cortex, it is merely that it has access to more com-
to the conclusion that the purpose of the hippocampus is to plete, more rened information and can learn a bit more rap-
give us the processing required by task X. Yet, this is the kind idly. Thus, what initially appears as a clear qualitative
of conclusion the dissociation approach can easily encourage dissociation may, in truth, be far more quantitative in nature.
unless one is extremely cautious about the interpretation of This is not meant to discourage the quest for dissociations.
the data. By nding such dissociations, we get to know in what tasks the
A more concrete example might make this problem clear. hippocampus (or any other structure) plays a critical role. In
Suppose there is a task that requires one-trial learning of asso- so doing, and when looking across data sets, we can attempt to
ciative, complex, cross-modal, relational, novel, arbitrary discern just what aspects make a task fully dependent on the
information. Further suppose that a complete bilateral lesion hippocampus and what aspects make a task partially depend-
to the hippocampus entirely removes the ability to perform ent on the hippocampus; we therefore can then dene just
this task at above-chance levels. One might conclude quite what the hippocampus may be doing for us. What this is
reasonably that the hippocampus allows us to, or is necessary meant to argue against is the simple extension from dissocia-
to, perform one-trial learning of associative, complex, cross- tion to functional interpretation. The shorthand of labeling a
modal, relational, novel, and arbitrary information. However, structure such as the hippocampus as performing task X
it would not be reasonable to conclude that this is the only memory or being responsible for a particular kind of mem-
thing the hippocampus does. The hippocampus might well be ory implies that it does not perform task Y memory, and other
involved in one-trial associative, complex, cross-modal, and structures cannot be responsible for this kind of memory. It is
relational memories between pieces of information that are not the dissociation in the data that is problematical but its
 Functional Role of the Human Hippocampus 575

interpretation as a binary functional dissociation that does REFERENCES

injustice to the data and oversimplies the operation of a
complex, highly interconnected and dynamic system. Aguirre G, Detre J, Alsop D, DEsposito MD (1996) The parahip-
What we are then currently left with is incomplete under- pocampus subserves topographical learning in man. Cereb
standing of the medial temporal lobe and the role the hip- Cortex 6:823829.
Bachevalier J, Brickson M, Hagger C (1993) Limbic-dependent
pocampus plays in its function. As several have suggested, the
recognition memory in monkeys develops early in infancy.
division of labor is likely to be graded rather than absolute
Learn Mem 4:7780.
(e.g., Lavenex and Amaral, 2000; Suzuki and Eichenbaum, Baddeley A, Vargha-Khadem F, Mishkin M (2001) Preserved recogni-
2000; Stark et al., 2002; Norman and OReilly, 2003). Although tion in a case of developmental amnesia: implications for the
it may best be thought of currently in terms of combining and acquisition of semantic memory? J Cogn Neurosci 13:357369.
extending the processing of the adjacent cortex, this is still an Berger T, Laham R, Thompson R (1980) Hippocampal unit-behavior
unsatisfying denition of the role of the hippocampus. We correlations during classical conditioning. Brain Res 193:
still do not know the exact nature of this division of labor, 229248.
what exactly can be combined and extended, how it is done, Binder J, Frost J, Hammeke T, Bellgowan P, Rao S, Cox RW (1999)
and how the collection of structures interact to perform this Conceptual processing during the conscious resting state: a
mnemonic function. We also do not know the role of the functional MRI study. J Cogn Neurosci 11:8093.
Blaxton T, Zeffiro T, Gabrieli JD, Bookheimer SY, Carrillo M,
perirhinal, entorhinal, and parahippocampal cortices; and we
Theodore W, Disterhoft J (1996) Functional mapping of human
do not know how they all interact as a dynamic system. These
learning: a positron emission tomography activation study of
issues represent currently active areas of experimentation and eyeblink conditioning. J Neurosci 16:40324040.
theoretical development, and all can benet from the recent Boynton G, Engel SA, Glover GH, Heeger D (1996) Linear systems
methodological advances laid out earlier in the chapter. In analysis of functional magnetic resonance imaging in human
short, although our present understanding of human hip- V1. J Neurosci 16:420742212.
pocampal function is far from complete, we have certainly Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD (1998)
made great strides. This is an exciting time for human hip- Making memories: brain activity that predicts how well visual
pocampal research. experience will be remembered. Science 281:11851187.
In conclusion, one nal point needs to be addressed. Brown MW, Aggleton JP (2001) Recognition memory: what are the
Throughout this chapter, the hippocampus and the rest of the roles of the perirhinal cortex and hippocampus? Nat Rev
Neurosci 2:5161.
medial temporal lobes have been discussed as providing a crit-
Buckner RL, Petersen SE, Ojemann JG, Miezin FM, Squire LR,
ical role in declarative or explicit long-term memory. The
Raichle ME (1995) Functional anatomical studies of explicit
reader should not take this to mean that the structures of the and implicit memory retrieval tasks. J Neurosci 15:1229.
medial temporal lobe are the only ones responsible for long- Buckner RL, Wheeler ME, Sheridan MA (2001) Encoding processes
term memory performance. Their role is essential, as they during retrieval tasks. J Cogn Neurosci 13:406415.
appear to act as a repository for this kind of information or at Buxton RB (2001) Introduction to functional magnetic resonance
least for some compressed version of this information (i.e., a imaging: principles and techniques (translated). Cambridge:
pointer) for some period of time. However, encoding and Cambridge University Press.
retrieval tasks engage structures well beyond the medial tem- Caesar K, Gold L, Lauritzen M (2003) Context sensitivity of activity-
poral lobes. For example, in cued recall, the cue must be dependent increases in cerebral blood ow. Proc Natl Acad Sci
assembled in some format suitable for probing memorys con- USA 100:42394244.
Caine D, Watson J (2000) Neuropsychological and neuropathological
tents, and the results of the retrieval must be interpreted and
sequelae of cerebral anoxia: a critical review. J Int Neuropsychol
evaluated. Quite likely, the results of the initial retrieval
Soc 6:8699.
attempt leads to an extended or revised cue that is again used Cameron K, Yashar S, Wilson C, Fried I (2001) Human hippocampal
in a retrieval attempt. Much of this appears to be the purview neurons predict how well word pairs will be remembered.
of structures outside the medial temporal lobe (of frontal Neuron 30:289298.
structures in particular), making memory, and even long- Chun MM, Phelps EA (1999) Memory decits for implicit contextual
term declarative memory, the operation of a large system of information in amnesic subjects with hippocampal damage. Nat
structures. The medial temporal lobe and the hippocampus in Neurosci 2:844847.
particular are but one component, albeit a vital component, of Clark RE, Squire LR (1998) Classical conditioning and brain systems:
this larger system. the role of awareness. Science 280:7781.
Clark RE, Zola SM, Squire LR (2000) Impaired recognition memory
in rats after damage to the hippocampus. J Neurosci 20:
88538860.
ACKNOWLEDGMENTS Cohen NJ, Poldrack RA, Eichenbaum H (1997) Memory for items
and memory for relations in the procedural/declarative mem-
The author thanks Shauna Stark, Brock Kirwan, Yoko Okado, Marci ory framework. Memory 5:131178.
Flanery, and Richard Morris for their comments on drafts of the Corkin S, Amaral DG, Gonzalez RG, Johnson KA, Hyman BT(1997)
chapter; Susan Corkin for providing previously unpublished data on H.M.s medial temporal lobe lesion: ndings from magnetic res-
H.M.; and the National Science Foundation for support. onance imaging. J Neurosci 17:39643979.
576 The Hippocampus Book

Cummings JL, Tomiyasu U, Read S, Benson DF (1984) Amnesia with Glover JA (1989) The testing phenomenon: not gone, but nearly
hippocampal lesions after cardiopulmonary arrest. Neurology forgotten. J Educ Psychol 81:392399.
34:679681. Graf P, Squire LR, Mandler G (1984) The information that amnesic
Dale AM, Buckner RL (1997) Selective averaging of rapidly presented patients do not forget. J Exp Psychol Learn Mem Cogn 10:
individual trials using fMRI. Hum Brain Mapp 5:329340. 164178.
Davachi L, Wagner AD (2002) Hippocampal contributions to Grasby P, Frith C, Friston K, Bench C, Frackowiak R, Dolan RJ (1993)
episodic encoding: insights from relational and item-based Functional mapping of brain areas implicated in auditory-ver-
learning. J Neurophysiol 88:982990. bal memory function. Brain 116:120.
Davachi L, Mitchell J, Wagner AD (2003) Multiple routes to memory: Green D, Swets J (1966) Signal detection theory and psychophysics
distinct medial temporal lobe processes build up item and (translated). New York: Wiley.
source memories. Proc Natl Acad Sci USA 100:21572162. Gusnard D, Raichle ME (2001) Searching for a baseline: functional
Dobbins IG, Rice HJ, Wagner AD, Schacter DL (2002) Memory ori- imaging and the resting human brain. Nat Rev Neurosci
entation and success: separable neurocognitive components 2:685694.
underlying episodic recognition. Neuropsychologia 41:318333. Gusnard D, Akbudak E, Shulman G, Raichle ME (2001) Medial pre-
Eichenbaum H (2000) A cortical-hippocampal system for declarative frontal cortex and self-referential mental activity: relation to a
memory. Nat Rev Neurosci 1:4150. default mode of brain activation. Proc Natl Acad Sci USA
Eichenbaum H, Dudchenko P, Wood E, Shapiro M, Tanila H (1999) 98:42594264.
The hippocampus, place cells, and memory: is it spatial memory Haist F, Gore J, Mao H (2001) Consolidation of human memory over
or a memory space? Neuron 23:209226. decades revealed by functional magnetic resonance imaging.
Ekstrom A, Kahana M, Caplan J, Fields T, Isham E, Newman E, Fried Nat Neurosci 4:11391145.
I (2003) Cellular networks underlying human spatial naviga- Hamann SB, Squire LR (1997) Intact perceptual memory in the
tion. Nature 425:184187. absence of conscious memory. Behav Neurosci 111:850854.
Eldridge LL, Knowlton BJ, Furmanski CS, Bookheimer SY, Engel SA Hartley T, Maguire E, Spiers H, Burgess N (2003) The well-worn
(2000) Remembering episodes: a selective role for the hip- route and the path less traveled: distinct neural bases of route
pocampus during retrieval. Nat Neurosci 3:11491152. following and waynding in humans. Neuron 2003.
Epstein R, DeYoe E, Press D, Rosen A, Kanwisher N (2001) Heit G, Smith M, Halgren E (1988) Neural encoding of individual
Neuropsychological evidence for a topographical learning words and faces by the human hippocampus and amygdala.
mechanism in parahippocampal cortex. Cogn Neuropsychol Nature 333:773775.
18:481508. Heit G, Smith M, Halgren E (1990) Neuronal activity in the human
Fagan J (1970) Memory in the infant. J Exp Child Psychol 9:217226. medial temporal lobe during recognition memory. Brain
Fell J, Klaver P, Elfadil H, Schaller C, Elger C, Fernandez G (2003) 113:10931112.
Rhinal-hippocampal theta coherence during declarative mem- Henke K, Buck A, Weber B, Wieser HG (1997) Human hippocam-
ory formation: interaction with gamma synchronization? Eur J pus establishes associations in memory. Hippocampus 7:
Neurosci 17:10821088. 249256.
Fernandez G, Brewer JB, Zhao Z, Glover GH, Gabrieli JD (1999a) Henke K, Weber B, Kneifel S, Wieser HG, Buck A (1999) Human hip-
Level of sustained entorhinal activity at study correlates with pocampus associates information in memory. Proc Natl Acad Sci
subsequent cued-recall performance: a functional magnetic res- USA 96:58845889.
onance imaging study with high acquisition rate. Hippocampus Henson R (2005) What can functional neuroimaging tell the experi-
9:3544. mental psychologist? Q J Exp Physiol 58:193233.
Fernandez G, Effern A, Grunwald T, Pezer N, Lehnertz K, Henson RNA, Rugg MD, Shallice T, Josephs O, Dolan RJ (1999)
Dumpelmann M, Van Roost D, Elger CE (1999b) Real-time Recollection and familiarity in recognition memory: an event-
tracking of memory formation in the human rhinal cortex and related functional magnetic resonance imaging study. J Neurosci
hippocampus. Science 285:15821585. 19:3962972.
Fernandez G, Klaver P, Fell J, Grunwald T, Elger CE (2002) Human Henson RNA, Cansino S, Herron J, Robb W, Rugg MD (2003) A
declarative memory formation: segregating rhinal and hip- familiarity signal in human anterior medial temporal cortex?
pocampal contributions. Hippocampus 12:514519. Hippocampus 13:301304.
Fried I, MacDonald K, Wilson C (1997) Single neuron activity in Hicks JL, Marsh RL (1999) Remember-know judgments can depend
human hippocampus and amygdala during recognition of faces on how memory is tested. Psychonom Bull Rev 6:117122.
and objects. Neuron 18:753765. Holdstock JS, Mayes AR, Cezayirli E, Isaac C, Aggleton JP, Roberts N
Fried I, Cameron K, Yashar S, Fong R, Morrow J (2002) Inhibitory (2000) A comparison of egocentric and allocentric spatial mem-
and excitatory responses of single neurons in the human medial ory in a patient with selective hippocampal damage. Neu-
temporal lobe during recognition of faces and objects. Cereb ropsychologia 38:410425.
Cortex 12:575584. Holdstock JS, Mayes AR, Isaac C, Gong Q, Roberts N (2002a)
Gabrieli JD, Brewer JB, Desmond JE, Glover GH (1997) Separate neu- Differential involvement of the hippocampus and temporal lobe
ral bases of two fundamental memory processes in the human cortices in rapid and slow learning of new semantic informa-
medial temporal lobe. Science 276:264266. tion. Neuropsychologia 40:748768.
Gale S, Hopkins R (2004) Effects of hypoxia on the brain: neu- Holdstock JS, Mayes AR, Roberts N, Cezayirli E, Isaac C, OReilly RC,
roimaging and neuropsychological ndings following carbon Norman KA (2002b) How recall and recognition are affected by
monoxide poisoning and obstructive sleep apnea. J Int Neu- focal damage to the human hippocampus. Hippocampus 12:
ropsychol Soc 10:6071. 341351.
 Functional Role of the Human Hippocampus 577

Hopkins R, Kesner R, Goldstein M (1995) Item and order recognition Manns J, Stark C, Squire L (2000) The visual paired-comparison task
memory in subjects with hypoxic brain injury. Brain Cogn as a measure of declarative memory. Proc Natl Acad Sci USA
27:180201. 97:1237512379.
Hopkins R, Myers C, Shohamy D, Grossman S, Gluck MA (2004) Manns JR, Hopkins R, Reed JM, Kitchener EG, Squire LR (2003)
Impaired probabilistic category learning in hypoxic subjects Recognition memory and the human hippocampus. Neuron
with hippocampal damage. Neuropsychologia 42. 37:171180.
Hyde JS, Biswal B, Jesmanowicz A (2001) High-resolution fMRI Marslen-Wilson W, Teuber H (1975) Memory for remote events in
using multislice partial k-space GR-EPI with cubic voxels. Magn anterograde amnesia: Recognition of public gures from news
Reson Med 46:114125. photographs. Neuropsychologia 13:353364.
Jacoby LL, Dallas M (1981) On the relationship between autobio- Mauk M, Thompson R (1987) Retention of classically conditioned
graphical memory and perceptual learning. J Exp Psychol Gen eyelid responses following acute decerebration. Brain Res
110:306240. 403:8995.
Kapur N, Brooks D (1999) Temporally-specic retrograde amnesia in Mayes AR, Holdstock JS, Isaac C, Hunkin N, Roberts N (2002)
two cases of discrete bilateral hippocampal pathology. Relative sparing of item recognition memory in a patient with
Hippocampus 9:247254. adult-onset damage limited to the hippocampus. Hippocampus
Kapur N, Craik FIM, Tulving E, Wilson A, Houle S, Brown G (1994) 12:325340.
Neuroanatomical correlates of encoding in episodic memory: McKee R, Squire LR (1993) On the development of declarative mem-
levels of processing effect. Proc Natl Acad Sci USA 91:20082011. ory. J Exp Psychol Learn Mem Cogn 19:397404.
Keane M, Gabrieli JD, Monti L, Fleischman DA, Cantor J, Noland J McNamara T, Shelton A (2003) Cognititve maps and the hippocam-
(1997) Intact and impaired conceptual memory processes in pus. Trends Cogn Sci 7:333335.
amnesia. Neuropsychology 11:5969. Miller MB, Beg M, Ceritogulu C, Stark CEL (2005) Improving statis-
Kirwan C, Stark CEL (2004) Medial temporal lobe activation during tical power and fMRI data localization using large deformation
encoding and retrieval of novel face-name pairs. Hippocampus metric mapping. Proc Natl Acad Sci USA 102:96859690.
14:910930. Milner B (1962) Les troubles de la mmoire accompagnant des lsions
Kirwan CB, Jones C, Miller MI, Stark CEL (in press) High-resolution hippocampiques bilatrales. In: Physiologie de lhippocampe, pp
f MRI investigation of the medial temporal lobe. Hum Brain 257272. Paris: Centre National de la Recherche Scientique.
Mapp. Milner B (1972) Disorders of learning and memory after temporal-
Knowlton BJ, Squire LR (1993) The learning of categories: parallel lobe lesions in man. Clin Neurosurg 19:421446.
brain systems for item memory and category knowledge. Science Milner B, Corkin S, Teuber H (1968) Further analysis of the hip-
262:17471749. pocampal amnesic syndrome: 14-year follow-up study of H.M.
Kreiman G, Koch C, Fried I (2000) Category-specic visual responses Neuropsychologia 6:215234.
of single neurons in the human medial temporal lobe. Nat Milner B, Squire LR, Kandel ER (1998) The cognitive neuroscience of
Neurosci 3:946953. memory. Neuron 20:445468.
Lavenex P, Amaral DG (2000) Hippocampal-neocortical interaction: Morrell H, Gitman S, Wixted J (2002) On the nature of the decision
a heirarchy of associativity. Hippocampus 10:420430. axis in signal-detection-based models of recognition memory. J
Law JR, Flanery MA, Wirth S, Yanike M, Smith AC, Frank LM, Suzuki Exp Psychol Learn Mem Cogn 28:10951110.
WA, Brown EN, Stark CEL (2005) fMRI activity during the Nadel L, Moscovitch M (1997) Memory consolidation, retrograde
gradual acquisition and expression of paired-associate memory. amnesia, and the hippocampal complex. Curr Opin Neurobiol
J Neurosci 25:57205729. 7:217227.
LePage M, Habib R, Tulving E (1998) Hippocampal PET activation Newman S, Tweig D, Carpenter P (2001) Baseline conditions
of memory encoding and retrieval: the HIPER model. and subtractive logic in neuroimaging. Hum Brain Mapp 14:
Hippocampus 8:313322. 228235.
Levy D, Stark CEL, Squire LR (2004) Intact conceptual priming in the Niki K, Luo J (2002) An fMRI study on the time-limited role of the
absence of declarative memory. Psychol Sci 15:680686. medial temporal lobe in long-term topographical autiobio-
Logothetis N, Pauls J, Augath M, Trinath T, Oeltermann A (2001) graphic memory. J Cogn Neurosci 14:500507.
Neurophysiological investigation of the basis of the fMRI signal. Norman KA, OReilly RC (2003) Modeling hippocampal and neo-
Nature 412:150157. cortical contributions to recognition memory: a complemen-
Maguire E, Frith CD (2003) Lateral asymmetry in the hippocampal tary learning systems approach. Psychol Rev 110:611646.
response to the remoteness of autobiographical memories. J Nyberg L, Tulving E, Habib R, Nilsson L, Kapur S, Houle S, Cabeza R,
Neurosci 23:53025307. McIntosh A (1995) Functional brain maps of retrieval mode
Maguire E, Frith CD, Burgess N, Donnett J, OKeefe J (1998) and recovery of episodic information. Neuroreport 7:249252.
Knowing where things are: parahippocampal involvement in OReilly RC, Rudy JW (2000) Computational principles of learning
encoding objection locations in a virtual large-scale space. J in the neocortex and hippocampus. Hippocampus 10:389397.
Cogn Neurosci 10:6176. Otten LJ, Henson RNA, Rugg MD (2001) Depth of processing effects
Maguire E, Henson RNA, Mummery C, Frith CD (2001) Activity on neural correlates of memory encoding: relationship between
in prefrontal cortex, not hippocampus, varies parametrically ndings across- and within-task comparisons. Brain 124:
with the increasing remoteness of memories. Neuroreport 399412.
12:441444. Paller KA, McCarthy G (2002) Field potentials in the human hip-
Manns JR, Squire LR (2001) Perceptual learning, awareness, and the pocampus during the encoding and recognition of visual stim-
hippocampus. Hippocampus 11:776782. uli. Hippocampus 12:415420.
578 The Hippocampus Book

Paller KA, Wagner AD (2002) Observing the transformation of expe- Schacter DL, Wagner AD (1999) Medial temporal lobe activations in
rience into memory. Trends Cogn Sci 6:93102. fMRI and PET studies of episodic encoding and retrieval.
Papanicolaou AC, Panagiotis GS, Castillo EM, Breier JI, Katz JS, Hippocampus 9:724.
Wright AA (2002) The hippocampus and memory of verbal and Schacter DL, Reiman E, Uecker A, Polster M, Yun LS, Cooper L
pictorial material. Learn Mem 9:99104. (1995) Brain regions associated with retrieval of structurally
Pascalis O, Bachevalier J (1999) Neonatal aspiration lesions of the coherent visual information. Nature 376:587590.
hippocampal-formation impair visual recognition memory Schendan H, Searl M, Melrose R, Stern CE (2003) An fMRI study off
when assessed by the paired-comparison task but not by delayed the role of the medial temporal lobe in implicit and explicit
nonmatching-to-sample task. Hippocampus 9:609616. sequence learning. Neuron 37:10131025.
Pascalis O, Hunkin N, Holdstock JS, Isaac C, Mayes AR (2004) Visual Scoville WB (1954) The limbic lobe in man. J Neurosurg 11:6466.
paired comparison performance is impaired in a patient with Scoville WB, Milner B (1957) Loss of recent memory after bilateral
selective hippocampal lesions and relatively intact item recogni- hippocampal lesions. J Neurol Neurosurg Psychiatry 20:1121.
tion. Neuropsychologia 24:12931300. Shallice T, Fletcher PC, Frith CD, Grasby P, Frackowiak R, Dolan RJ
Pihlajamaki M, Tanila H, Hanninen T, Kononen M, Mikkonen M, (1994) Brain regions associated with acquisition and retrieval of
Jalkanen V, Partanen K, Aronen H, Soininen H (2003) Encoding verbal episodic memory. Nature 368:633635.
of novel picture pairs activates the perirhinal cortex: an fMRI Shelton A, Gabrieli JD (2002) Neural correlates of encoding space
study. Hippocampus 13:6780. from route and survey perspectives. J Neurosci 22:27112717.
Posner MI, Keele SW (1968) On the genesis of abstract ideas. J Exp Shimamura AP, Squire LR (1984) Paired-associate learning and
Psychol 77:353363. priming effects in amnesia: a neuropsychological study. J Exp
Preuessner J, Kohler S, Crane J, Pruessner M, Lord C, Byrne A, Kabani Psychol Gen 113:556570.
N, Collins DL, Evans AC (2002) Volumetry of temporopolar, Small S, Arun A, Perera G, De La Paz R, Mayeaux R, Stern Y (2001)
perirhinal, entorhinal, and parahippocampal cortex from high- Circuit mechanisms underlying memory encoding and retrieval
resolution MR images: considering the variability of the collat- in the long axis of the hippocampal formation. Nat Neurosci
eral sulcus. Cereb Cortex 12:13421353. 4:442449.
Ranganath C, Yonelinas AP, Cohen MX, Dy CJ, Tom SM, DEsposito Sperling RA, Bates JF, Cocchiarella AJ, Schacter DL, Rosen BR, Albert
MD (2003) Dissociable correlates of recollection and famil- MS (2001) Encoding novel face-name associations: a functional
iarity within the medial temporal lobes. Neuropsychologia MRI study. Hum Brain Mapp 14:129139.
42:213. Sperling R, Chua E, Cocchiarella A, Rand-Giovannetti E, Poldrack
Reber PJ, Wong EC, Buxton RB (2002) Encoding activity in the RA, Schacter DL, Albert M (2003) Putting names to faces: suc-
medial temporal lobe examined with anatomically constrainted cessful encoding of associative memories activates the anterior
fMRI analysis. Hippocampus 12:363376. hippocampal formation. Neuroimage 20:14001410.
Reber PJ, Gitelman D, Parrish T, Mesulam M (2003) Dissociating Spiers H, Maguire E, Burgess N (2001) Hippocampal amensia.
explicit and implicit category knowledge with fMRI. J Cogn Neurocase 7:357382.
Neurosci 15:574583. Squire LR (1992) Declarative and nondeclarative memory: multiple
Reed JM, Squire LR (1997) Impaired recognition memory in patients brain systems supporting learning and memory. J Cogn Neurosci
with lesions limited to the hippocampal formation. Behav 4:232243.
Neurosci 111:667675. Squire LR, Alvarez P (1995) Retrograde amnesia and memory con-
Reed JM, Squire LR (1998) Retrograde amnesia for facts and events: solidation: a neurobiological perspective. Curr Opin Neurobiol
ndings from four new cases. J Neurosci 18:39433954. 5:169177.
Rees G, Friston K, Koch C (2000) A direct quantitative relationship Squire LR, Knowlton BJ (1995) Learning about categories in the
between the functional properties of human and macaque V5. absence of memory. Proc Natl Acad Sci USA 92:1247012474.
Nat Neurosci 3:716723. Squire LR, Zola-Morgan S (1991) The medial temporal lobe memory
Rempel-Clower NL, Zola SM, Squire LR, Amaral DG (1996) Three system. Science 253:13801386.
cases of enduring memory impairment after bilateral damage Squire LR, Ojemann JG, Miezin FM, Petersen SE, Videen TO, Raichle
limited to the hippocampal formation. J Neurosci 16:52335255. ME (1992) Activation of the hippocampus in normal humans: a
Ribot T (1887) Diseases of memory: an essay in the positive psychology functional anatomical study of memory. Proc Natl Acad Sci USA
(Smith WH, translator). New York: Appleton. 89:18371841.
Rose M, Haider H, Weiller C, Buchel C (2002) The role of medial Squire LR, Clark RE, Knowlton BJ (2001) Retrograde amnesia.
temporal lobe structures in implicit learning: an event-related Hippocampus 11:5055.
fMRI study. Neuron 36:12211231. Stark CEL, Okado Y (2003) Making memories without trying: medial
Ryan L, Nadel L, Keil K, Putnam K, Schnyer D, Trouard T, Moscovitch temporal lobe activity associated with incidental memory for-
M (2001) Hippocampal complex and retrieval of recent and mation during recognition. J Neurosci 23:67486753.
very remote autobiographical memories: evidence from func- Stark CEL, Squire LR (2000a) fMRI activity in the medial temporal
tional magnetic resonance imaging in neurologically intact lobe during recognition memory as a function of study-test
people. Hippocampus 11:707714. interval. Hippocampus 10:329337.
Schacter DL (1994) Priming and multiple memory systems: percep- Stark CEL, Squire LR (2000b) Recognition memory and familiarity
tual mechanisms of implicit memory. In: Memory systems 1994 judgments in severe amnesia: no evidence for a contribution of
(Schacter DL, Tulving E, eds), pp 233268. Cambridge, MA: repetition priming. Behav Neurosci 114:459467.
MIT Press. Stark CEL, Squire LR (2000c) Functional magnetic resonance imag-
Schacter DL (1999) The seven sins of memory. Am Psychol ing (fMRI) activity in the hippocampal region during recogni-
54:182203. tion memory. J Neurosci 20:77767781.
 Functional Role of the Human Hippocampus 579

Stark CEL, Squire LR (2001a) Simple and associative recognition Verfaellie M, Reiss A, Roth H (1995) Knowledge of new English
memory in the hippocampal region. Learn Mem 8:190197. vocabulary in amnesia: an examination of premorbidly
Stark CEL, Squire LR (2001b) When zero is not zero: the problem of acquired semantic memory. J Int Neuropsychol Soc 1:443
ambiguous baseline conditions in fMRI. Proc Natl Acad Sci USA 452.
98:1276012766. Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Dale AM,
Stark CEL, Squire LR (2003) Hippocampal damage equally impairs Rosen BR, Buckner RL (1998) Building memories: remember-
memory for single-items and memory for conjunctions. ing and forgetting of verbal experiences as predicted by brain
Hippocampus 13:281292. activity. Science 281:11881191.
Stark CEL, Bayley PJ, Squire LR (2002) Recognition memory for sin- Warrington E, McCarthy G (1988) The fractionation of retrograde
gle items and for associations is similarly impaired following amnesia. Brain Cogn 7:184200.
damage limited to the hippocampal region. Learn Mem Warrington E, Weiskrantz L (1968) New method of testing long-term
9:238242. retention with special reference to amnesic patients. Nature
Stefanacci L, Buffalo EA, Schmolck H, Squire LR (2000) Profound 217:972974.
amnesia after damage to the medial temporal lobe: a neu- Warrington E, Weiskrantz L (1970) The amnesic syndrome: consoli-
roanatomical and neuropsychological prole of patient E.P. J dation or retrieval? Nature 228:628630.
Neurosci 20:70247036. Warrington E, Weiskrantz L (1974) The effect of prior learning on
Strange BA, Fletcher PC, Henson RNA, Friston KJ, Dolan RJ (1999) subsequent retention in amensic patients. Neuropsychologia
Segregating the functions of human hippocampus. Proc Natl 12419428.
Acad Sci USA 96:40344039. Weiskrantz L, Warrington E (1979) Conditioning in amnesic
Strange BA, Otten LJ, Josephs O, Rugg MD, Dolan RJ (2002) patients. Neuropsychologia 17:187194.
Dissociable human perirhinal, hippocampal, and parahip- Wirth S, Yanike M, Frank L, Smith A, Brown E, Suzuki WA (2003)
pocampal roles during verbal encoding. J Neurosci 22:523528. Single neurons in the monkey hippocampus and learning of
Sutherland RJ, Rudy JW (1989) Congural association theory: the new associations. Science 300:15781581.
role of the hippocampal formation in learning, memory, and Wixted J, Squire LR (2004) Recall and recognition are equally
amnesia. Psychobiology 17:129144. impaired in patients with selective hippocampal damage. Cogn
Suzuki WA, Eichenbaum H (2000) The neurophysiology of memory. Affect Behav Neurosci 4:5866.
Ann N Y Acad Sci 911:175191. Yonelinas AP (2002) The nature of recollection and familiarity: a
Talairach J, Tournoux P (1988) A co-planar stereotaxic atlas of the review of 30 years of research. J Mem Lang 46:441517.
human brain (translated). New York: Thieme Medical. Yonelinas A, Hopnger J, Buonocore M, Kroll N, Baynes K (2001)
Teng E, Squire LR (1999) Memory for places learned long ago is Hippocampal, parahippocampal and occipital-temporal contri-
intact after hippocampal damage. Nature 400:675677. butions to associative and item recognition memory: an fMRI
Tulving E (1985) Memory and consciousness. Can J Psychol 26:112. study. Neuroreport 12:359363.
Tulving E, Kapur N, Markowitsch H, Craik FIM, Habib R, Houle S Yonelinas AP, Kroll NEA, Quamme J, Lazzara MM, Suave M,
(1994) Neuroanatomical correlates of retrieval in episoidc Widaman K, Knight RT (2002) Effects of extensive temporal
memory: auditory sentence recognition. Proc Natl Acad Sci USA lobe damage or mild hypoxia on recollection and familiarity.
91:20122015. Nat Neurosci 5:12361241.
Vaidya C, Gabrieli JD, Keane M, Monti L (1995) Perceptual and con- Zeineh MM, Engel SA, Bookheimer SY (2000) Application of cortical
ceptual memory processes in global amnesia. Neuropsychology unfolding techniques to functional MRI of the human hip-
9:580591. pocampal region. Neuroimage 11:668683.
Vargha-Khadem F, Gadian DG, Watkins KE, Connelly A, Van Zeineh MM, Engel SA, Thompson PM, Bookheimer SY (2003)
Praesschen W, Mishkin M (1997) Differential effects of early Dynamics of the hippocampus during encoding and retrieval of
hippocampal pathology on episodic and semantic memory. face-name pairs. Science 299:577580.
Science 277:376380. Zola SM, Squire LR, Teng E, Stefanacci L, Buffalo EA, Clark RE (2000)
Vargha-Khadem F, Gadian DG, Mishkin M (2001) Dissociaitons in Impaired recognition memory in monkeys after damage limited
cognitive memory: the syndrome of developmental amnesia. to the hippocampal region. J Neurosci 20:451463.
Philos Trans R Soc Lond B Biol Sci 356:14351440. Zola-Morgan S, Squire LR, Amaral DG (1986) Human amnesia and
Vargha-Khadem F, Salamond C, Watkins KE, Friston K, Gadian DG, the medial temporal lobe region: enduring memory impair-
Mishkin M (2003) Developmental amnesia: effect of age at ment following a bilateral lesion limited to eld CA1 of the hip-
injury. Proc Natl Acad Sci USA 100:1005510060. pocampus. J Neurosci 6:29502967.
This page intentionally left blank
13 Richard Morris

Theories of Hippocampal Function

stability over time, and the subsequent retrieval and reactiva-

13.1 Overview tion of hippocampus-dependent memories during recall.
Newer research techniques, such as functional magnetic reso-
What does the hippocampal formation do? Does it have one nance imaging (fMRI) in humans and invasive techniques
function or many? Can we understand its functions in isola- with animals such as ensemble single-cell recording and selec-
tion from that of other areas of the brain? The theories we tive neuropharmacological perturbations, all point to a role
review in this chapter all focus on the idea that it is a special for specic aspects of hippocampal neural activity in succes-
kind of memory machine. The link between the hippocam- sive phases of memory processing.
pus and memory is an old idea, having its roots in clinical and Two theories have dominated research on hippocampal
experimental observations on patients who sustained brain function over the past quarter century. The rst, discussed in
damage that included the hippocampus but all too often Section 13.3, is that it is involved in the formation of memories
extended into other areas of the medial temporal lobe or mid- for everyday facts and events that can be consciously recalled
brain (see Chapter 2). Careful study of the small number of collectively called declarative memory (Squire, 1992; Squire et
patients with more selective damage (see Chapter 12) and that al., 2004). Complementing the human studies of the preceding
of animals given experimental lesions or other interventions chapter, a particular focus has been the development of a pri-
(see later in this chapter) has provided further grounds for mate model of amnesia built largely on studies of recognition
believing that the integrity of the hippocampal formation is memory. Other behavioral tasks have also been examined, but
causally involved in implementing some but not all types of they have been secondary to testing a key issue: whether the
memory. Given that multiple types of memory have now hippocampus is involved in recognizing stimuli that have been
been proposedsuch as declarative, spatial, and episodic presented before. The other major theory (see Section 13.4),
memorythe focus of this chapter is on identifying which of emerging from observations rst made during the recording of
these or other putative types of memory are mediated by the single-cell activity in freely moving rodents (see Chapter 11), is
hippocampal formation. Success in this task requires that we the idea that it is involved in spatial memory and, more specif-
recognize that the hippocampus operates in conjunction with ically, the formation of cognitive maps and their use in naviga-
other networks and brain structures, including neuromodula- tion through space (OKeefe and Nadel, 1978; Burgess et al.,
tory afferent systems emanating from the midbrain and 2002). This theory makes testable predictions about the cell
diverse regions of the neocortex. Any claim that the integrity activity that occurs during spatial navigation and the impact of
of the hippocampus is required for one or another type of hippocampal dysfunction on spatial memory. Although enor-
memory does not, therefore, mean that other brain areas are mously inuential, and often at loggerheads with each other, it
not involved as well. Discussion of types of memory and their will become clear that neither of these two major theories
anatomical substrate carries us forward to consider the logi- offers a fully satisfactory account of the available data.
cally distinct issue of the various phases of memory process- Section 13.5 discusses a range of alternative theories, par-
ing in which the hippocampus may be engagedthe ticularly those built around how memory systems handle
encoding of information into memory, the storage of memory ambiguity, associative-relations, and context. In 1984,
traces as biochemical and structural changes in the nervous Schmajuk identied more than 20 psychological theories of
system, the consolidation processes responsible for their hippocampal function (Schmajuk, 1984). There have been yet

581
582 The Hippocampus Book

others since. Here we focus on certain particularly inuential so difficult to capture (Hampton and Schwartz, 2004), this
ideas of the past 25 years, namely that the hippocampus is now often referred to as episodic-like memory (Clayton
implements a learning process involving the formation of and Dickinson, 1998). Although also applied elsewhere, it is
stimulus congurations (Sutherland and Rudy, 1989a; OReilly in this section that we see the greatest implementation of
and Rudy, 2001), that it is involved in the relational processing new research technologies, such as novel behavioral tasks,
of stimuli in a manner that enables exible retrieval (Cohen reversible pharmacological manipulations, and region-specic
and Eichenbaum, 1993; Eichenbaum and Cohen, 2001), or gene targeting. This new way of thinking about hippocampal
that it provides the substrate for encoding and retrieval of function also draws more explicitly on the anatomical, physi-
information in relation to the spatiotemporal context in which ological, and cell biological ideas discussed in earlier chapters
that information occurs. Contexts, relations, and congura- of this book.
tions may each help disambiguate conicting associative rela- In each section, the main concepts of each theory are out-
tions in which a specic stimulus occurs (Hirsh, 1974; Good lined followed by an examination of exemplar experimental
and Honey, 1991). Although perhaps less prominent than the work that supports or conicts with it and then a general cri-
two major theories, these ideas have been important in tique. Given the scale of the literature, no attempt has been
expanding our concepts of what memory is all about and its made to be denitively inclusive, and the choice of studies
mediation by diverse structures in the brain. cited is inevitably subjective. When possible, an attempt is
The present focus on memory is, arguably, too exclusive. made to refer to ideas about hippocampal function that have
The hippocampus has also been implicated in a range of other emerged from the principles discussed in earlier chapters.
brain functions, including behavioural inhibition and anxiety Understanding the hippocampal memory machine is an ambi-
(Kimble, 1968; Gray, 1982; Davidson and Jarrard, 2004), sen- tious exercise that requires integrating top-down psycholog-
sorimotor function (Vanderwolf and Cain, 1994), and acting ical concepts about the distinct types and processing phases of
as a comparator to detect novelty (Gray, 2000). The presence memory with bottom-up physiological and cell biological
in the hippocampus of a high density of receptors for adreno- ideas about the local networks, cells, and signal transduction
cortical hormones such as cortisol and corticosterone has also pathways that constitute this beautiful brain structure.
implicated the hippocampus in the cognitive regulation of
stress and the hypothalamic-pituitary axis (Sapolsky, 1985;
McEwen and Sapolsky, 1995; de Kloet et al., 1999). For clarity,
these issues are considered separately in Chapter 15. 13.2 Cognitive and Behavioral
Certain concepts relevant to identifying hippocampal Neuroscience, Interventional
function in relation to memory are excluded from detailed Techniques, and the Hippocampus
discussion in this chapter because it seemed most appropriate
to outline them elsewhere. They include the widely discussed 13.2.1 Value of Interventional
idea that activity-dependent synaptic plasticity, such as long- Studies to Identify Function
term potentiation (LTP) and long-term depression (LTD),
might play a role in learning and memory (Martin et al., 2000; If you want to learn how a machine works, it is a good idea to
Bliss et al., 2004). This important topic is discussed at the end watch it working. An appropriate next step is to look at selec-
of Chapter 10. Hebbs concepts of the cell assembly and tive interventions that disrupt one part of the machine or
phase sequence (Hebb, 1949) and Marrs proposal that dis- another and study the various changes in the machines oper-
tributed associative memories could be implemented by hip- ation that then occur. In the case of a modular organ such as
pocampal local circuitry (Marr, 1971) were both touched on the brain, it is essential to complement studies that involve its
in our historical overview (see Chapter 2) and are developed cells being cut up, put under microscopes, poked with elec-
more fully in Chapter 14. Neural network modeling is a eld trodes, or bathed in drugs (see Chapters 311) with studies of
in its own right that has matured to a level of conceptual and selective interventions that may affect its behavior. This is part
mathematical precision that is beyond the scope of this chap- of what cognitive and behavioral neuroscientists do to try to
ter (Willshaw and Buckingham, 1990; OReilly, 1998; Rolls understand the functions of the distinct but interacting
and Treves, 1998; Lisman and Zhabotinsky, 2001; OReilly and regions of the brain. Lesion studies in animals, on which we
Norman, 2002). largely focus in this chapter, have been the classic interven-
The chapter concludes by zeroing in on the idea that neu- tionist technique for testing theories of function. They have
ral activity in the hippocampal formation contributes to their methodological and interpretative problems, as do phar-
episodic memory (Tulving, 1983; Aggleton and Brown, 1999; macological and genetic techniques. Intervention is, however,
Redish, 1999). This contribution has been variously charac- the only way to secure denitive information about whether
terized as remembering the scenes in which events take place the integrity of a particular anatomical region of the brain or
(Gaffan, 1994a), or as processing the automatic components specic physiological and cell biological processes in it is nec-
of recording and recalling experience that are downstream of essary for a particular function.
neocortical attentional systems (Morris and Frey, 1997). In In humans, damage to the brain may arise because of a
animals, where the phenomenological aspects of memory are tumor, a stroke, or a viral infection, or it may be incurred
 Theories of Hippocampal Function 583

through neurosurgery undertaken to prevent further brain to the neuroscientists efforts to interpret lesion effects in
damage (e.g., the surgical management of epilepsy). A few relation to the functional theories they are intended to test.
striking and analytically valuable exceptions aside, lesions that Notwithstanding these obstacles along the path, interven-
arise from these causes are rarely restricted to anatomically tional techniques can be used successfully, provided a number
circumscribed areas of the brain; nor when they are is the of conceptual issues are also considered.
damage incurred ever complete (see Chapter 12). In animal
studies, in contrast, specic brain regions can be subject to 13.2.2 Lesions, Functional
selective lesions to evaluate whether one or more functions Hypotheses, and Behavioral Tasks
can still be carried out normally. Postmortem histological
techniques complement the behavioral observations, provid- Dissociations, Double
ing a vital, relatively immediate check on an experimentalists Dissociations, and Hypotheses
intentions with respect to the site and size of the lesion. It is
also now possible to use histologically calibrated structural Experimental lesions are widely used to establish functional
MRI techniques in vivo to establish the accuracy and extent of dissociations between tasks that are thought to depend on dis-
damage before extensive behavioral testing is undertaken tinct forms of information processing. Thus, if behavior A is
(Malkova et al., 2001). impaired by a lesion made in area X and, simultaneously,
The lesion method is, of course, used widely in biology, the behavior B is unaffected by lesion X, there are grounds for sus-
modern variants of it being pharmacological techniques pecting that behaviors A and B may, at least in part, depend on
directed at specic receptors or targeted molecular engineer- different brain circuits. The condence in such assertions
ing to investigate gene function (gene knockouts). The com- grows when a so-called double dissociation is obtained
mon analytical principle underlying such approaches is that when a different lesion Y impairs behavior B but not behavior
the study of dysfunction offers insights into normal function. A (Teuber, 1955). However, discovering a list of behaviors that
Creating dysfunction does not necessarily require invasive are affected or unaffected by a particular set of lesions (or
interventionsthe study of visual illusions being a case in other intervention) is only the rst step toward such a theory.
pointbut the neurological approach, as twentieth century An essential parallel step is the development of hypotheses
neurologist Sir Henry Head put it so well, is one in which it about information processing in the brain (Shallice, 1988).
should not be the injury that captures our attention but how, Numerous hypotheses have been developed by cognitive
through injury or disease, normal function is laid bare. neuroscientists to account for the cerebral organization
There are, however, problems strewn along the path of lay- of perceptual processes, attention, language processing, exec-
ing things bare. In some cases it is the pitfalls of the approach utive function. and memory (Shallice, 1988; McCarthy and
that lead an investigator astray. First, a brain reacts to perma- Warrington, 1990; Gazzaniga, 2002). With respect to the hip-
nent damage with repair mechanismscerebrovascular, neu- pocampal formation, numerous hypotheses have been devel-
ral, glialthat collectively result in dynamic changes from the oped and are discussed in this chapter. As we saw in the
moment the damage occurs. It may be some while before overview, these theories differ in their claims about the char-
brain function restabilizes after surgically induced diaschisis acter and timing of the information processing that is taking
(Stein et al., 1983; Finger et al., 2004). Second, even when it place in the hippocampal formation during learning and
does restabilize, we should not assume that adjacent undam- memory.
aged brain areas necessarily function normally, as the neural
activity carried on input pathways may have been affected by Learning, Performance, and Hippocampus-dependent
the nearby damage. Lesions can also induce structural Learning and Memory Tasks
changes, such as axonal growth and synaptogenesis (see
Chapter 9). This sprouting can be compensatory and thus One important theme in studies of learning and memory is
homeostatic, or it may give rise to abnormal circuitry not the distinction between learning and performance. In a behav-
usually seen in the normal brain (Lynch et al., 1973; Steward ing animal, performance is what we can see and measure,
et al., 1974). Third, even if postlesion reorganization achieves whereas learning reects a variety of covert processes that we
relatively little in the way of compensation for lost tissue, presume are taking place in the brain. It is these covert
the person or animal may still alter his behavior in a manner processes that are responsible for the acquisition of knowl-
that enables him to compensate for lost function through edge. Performance is the expression of this knowledge. Of
the use of undamaged systems. Behavioral compensation can course, performance may also itself be learned, as occurs with
help mask a neurological decit. With respect to damage sensorimotor skills, but it is often a reection of learning
to the medial temporal lobe and the ensuing amnesia, for rather than being a participant in such a process. Obstructing
example, the use of memory aids such as electronic devices to performance in some way (e.g., by putting a barrier in a maze
remind people to do things is one example of such compen- that an animal has been trained to run through) may prevent
satory behavior (Wilson, 1999). Desirable as these aids the usual expression of learned performance, but it does not
are therapeutically, the possibility that behavioral compensa- affect what the animal knows. The animal generally tries to
tion occurs naturally over the course of time adds difficulties nd an alternative way in which to express this knowledge,
584 The Hippocampus Book

such as its memory of where some desired food is located afferent stimuli or to move around effectively (Vanderwolf
(Tolman, 1948). and Cain, 1994).
The concept of learning itself must also be fractionated as
the brain has evolved to have a number of qualitatively dis- What, How, and When: Different
tinct forms of learning. Most theories of hippocampal func- Aspects of Hippocampus-dependent Memory
tion are statements about the type (or types) of learning and
memory in which this group of structures is engaged. There is The overall aim of the functional enterprise is to nd out what
passionate debate about how to characterize them. Over the the hippocampus does and how it does it. Identifying which
years, a range of learning tasks have been developed that are tasks are hippocampus-dependent and which are not is an
now widely used in studies with rodents, primates, and birds important part of this endeavor, particularly when they shed
to try to identify which of them involve hippocampus- analytical light on the relative merits of different theories.
dependent processes. Widely used tasks for rats and mice There are, however, a number of ways in which learning
include alternation protocols in T mazes and cross mazes, for- and memory tasks can be theoretically ambiguous. The rst
aging for food in radial mazes, and other spatial tasks such as has to do with what an animal learns. In a simple T maze, a rat
nding the way home in an open arena or swimming to safety might be taught to run from a start box in the stem and then
in a watermaze. Recognition memory can be studied by pre- turn left to nd food at the end of the armbut has it learned
senting novel and familiar objects or by taking advantage the motor response of turning left? Or that food is to be found
of rats superb olfactory sense and then asking which object near landmarks positioned on the left? One way to nd out is
or smell is familiar. One widely used protocol is the condi- to rotate the maze through 180 and now start the animal
tioning of fear to the context where testing takes place, and from the stem of the maze. If it has learned the motor
another examines the social transmission of food preferences. response of turning left, it should move toward what had been
Certain protocols of some of these tasks seem to be hip- the right-hand side of the maze during training. Conversely, if
pocampus-dependent (i.e., affected by hippocampal lesions), it has learned to approach landmarks on the left-hand side of
whereas others are not. The community that conducts the room, it should now turn right at the choice point. This
research on temporal lobe function in primates has its family manipulation nicely dissociates control and hippocampus-
of tried and trusted tasks for investigating memory ranging lesioned rats early in training; but, interestingly, if training
from manual protocols with three-dimensional objects to continues too long, normal rats develop a turning habit that
automated touch-screen technology, some of which has now is inexible to manipulations of the starting location (Packard
seen application as neuropsychological tests for humans and McGaugh, 1996). Another example of ambiguity about
(Robbins et al., 1994). Birds are also used extensively, with due what has been learned concerns recognition memory.
consideration to their natural behavioral ecology. Innovation Recognition can be dened as the memory of past occurrence
has gone into the design and development of rigorous labora- (Brown and Xiang, 1998), but there are several ways it can be
tory-based tasks such as food caching and recovery (Suzuki achieved. For example, recognition of a past stimulus by a
and Clayton, 2000). In using this range of tasks, there is ongo- monkey may be mediated by a feeling of familiarity evoked by
ing debate about the relative value of standardization on a the stimulus itself. Alternatively, the stimulus may evoke a
subset of tasks that are then used in a common way within covert mental recollection of past experience. The monkey
and across laboratories and of innovation with respect to cannot tell us about either of these possibilities in the way that
procedure in the search for better ways of characterizing a person would through language, but its sense of familiarity
hippocampus-dependent learning and memory. or its sense of recollection is expressed as appropriate choice
One complication is that there is rarely, if ever, a simple behavior in a memory test (performance). A current challenge
one-to-one mapping between a specic task and a single puta- is trying to nd a way of distinguishing these two possibilities
tive type of learning. All of them rely on effective sensory/per- of what the animal has learned.
ceptual processing in upstream brain structures and on The second kind of ambiguity arises if two or more theories
effective motor processing downstream. Researchers have offer different accounts of how an animal has learned a partic-
then to be careful that any alteration in task performance ular task. In this case, there is agreement about what the
caused by interventions is really due to an interruption of a mouse, monkey, or human has learned but not how it has been
neural process mediated by the hippocampus or to some sec- achieved. Allocentric spatial learning is a relevant case in point,
ondary effect on upstream or downstream processes. The con- as there is widespread agreement that animals can learn where
ceptual boundary between input, cognition, and output is things are and that the hippocampus is involved but disagree-
hazy, particularly as it becomes ever more apparent that bot- ment about how it is done. Questions include whether there is
tom-up processes can be modulated by top-down mecha- anything special about spatial learning that might require spe-
nisms. Indeed some researchers of a behaviorist persuasion cialized neural circuitry, or is it just an instance of a more gen-
even question whether the conceptual boundary between eral form of declarative memory? Are different learning rules
learning and performance exists, in part because an animal involved for spatial learning than other types of learning? Do
may appear to be unable to learn when, in practice, a treat- different brain areas have to interact for spatial learning than
ment has been given that has affected only its ability to sense for other forms of declarative or relational learning?
 Theories of Hippocampal Function 585

A third ambiguity arises from the distinction between types reasons why behavioral inexibility could arise from other
of learning and memory processes. The classic concept of than excision of neural machinery that had (somehow)
a hippocampus-dependent learning task is one that is evolved to inhibit inexibility. For example, if the lesioned
impaired by permanent hippocampal lesions. An unambigu- animal has a decit in spatial memory and so cannot learn or
ous hippocampus-dependent learning task is one that is remember that one end of a runway (A) is in a different
affected by such lesions but unaffected by others. However, place than the other end of a runway (B), it cannot acquire
this categorization does not capture the subtlety that neural knowledge about their relative location. It learns only to run.
activity in the hippocampus may be required at one stage of Changing a spatial performance strategy is much easier for
learning but not at another. It might, for example, be required normal animals than the slamming on the brakes to con-
at encoding but not at retrieval or during the initial stages of strain a habitual running response (Nadel et al., 1975). The
memory storage but not during consolidation. The logically debate about the relative merits of these specic views has
orthogonal concept of a memory process refers to impor- largely subsided, although the inhibitory view still has its fol-
tant distinctions that must be made between the distinct lowers (Chan et al., 2001; Davidson and Jarrard, 2004), but the
memory processes associated with encoding, storage, consoli- logical point remains that the overt phenotype of response
dation, and retrieval. It is orthogonal, as each of these stereotypy need not imply the existence of an inhibition mod-
processes applies to many types of learning and memory. ule localized in the hippocampus or, indeed, anywhere else in
They are distinguished experimentally through the various the brain.
hemodynamic, physiological, or cell biological correlates that Gregorys (1961) critique should, however, not be over-
are seen in association with these various phases, by the stated. If one were to lesion a cars exhaust system, it would be
impact of lesions made at different stages of training, and by right to infer that its function is to make the engine quieter.
the effects of reversible interventions (such as drugs) used to Damaging the earphones of an iPOD limits its ability make a
inactivate brain areas temporarily or to disrupt specic neural sound. By the same token, in a modular structure such as the
mechanisms. brain, inferences about localization may often be correct. If
Thus, the overarching task of dissociating theories of the overarching aim of the enterprise is to develop a theory of
regional brain function involves zeroing in on what, how, what different parts of the brain do and how they do it, we
and when. Many subtle variations in behavioral protocol must also consider how brain areas interact to realize the
are designed to explore the hypotheses. These variations seamless control of all aspects of behavior. It is an article of
of protocol are guided by predictions, better still by the coun- faith in the eld that different brain areas or neuromodulatory
terintuitive predictions, of functional hypotheses about transmitter systems must have different and therefore, at least
structurefunction relations (Crabbe and Morris, 2004). The in principle, dissociable functions.
declarative memory, cognitive mapping, and predictive ambi- Equally, however, it is also understood that no brain area is
guity theories outlined below often make common predic- an island; different brain areas work together in networks. In
tions about standardized versions of specic tasks but distinct recent years, there has been growing interest in the issue of
predictions about their variants. how memory systems interactwhen they compete, when
they complement each other, when the interruption of one
Cerebral Localization or Dissociation enables another to learn faster or slower, and so on (White and
Between Mental Processes? McDonald, 2002; McDonald et al., 2004; Voermans et al.,
2004). This change of emphasis is refreshing, as it counterbal-
A frequent non sequitur in the interpretation of an interven- ances the near-exclusive stress on dissociations of earlier
tional study is the assertion, following a positive experimental years. A desire to understand how these systems interact is tak-
nding, that a function can somehow be localized to the tar- ing its place as the brain sciences mature. In any event,
geted brain structure or pathway. A classic and amusing cri- whether the immediate aim is functional dissociation or the
tique of this localizational way of thinking was presented by understanding of competition and/or complementarity, data
Gregory (1961), who pointed out that if removing a part of an secured using lesions and other interventions are a powerful
old valve radio causes the radio to howl it need not mean that way of testing theories.
the function of this bit is to suppress howling. This overly
symptomatic way of thinking pervaded the hippocampal 13.2.3 Contemporary Lesion Techniques:
eld early on and indeed much of the brain sciences since the Pharmacological and Genetic Interventions
early days of experimental neurology. One early claim about
hippocampal function was that it is involved in behavioral Neurotoxic Lesions
inhibition (Kimble, 1968). This claim rested, in part, on the
observation that hippocampus-lesioned animals are very The technique by which a lesion is made can be important
inexible in their behavior; they often persist much longer in because certain techniques, such as the use of excitotoxins,
carrying out well learned habits in inappropriate situations rst introduced to study hippocampal function by Jarrard
than do control subjects. Analysis of behavior in runways, (1989), have the effect of damaging cells while leaving bers of
mazes, and operant tasks has revealed, however, a number of passage intact (Fig. 131). Thus, if a brain function critically
586 The Hippocampus Book

A. Topological arrangement of HPC Formation B. Selective neurotoxic and knife-cut

lesions in rats

1 2

3 4

C. Quantitative assessment of partial lesion volume

sham partial complete

lesion hippocampal hippocampal
50 lesion lesion
% of hippocampus spared

40 5
30

20

10

0
al

al

al

l
ra

ra

ra
pt

pt

pt
po

po

po
se

se

se
m

m
te

te

te

Figure 131. Excitotoxic and knife-cut lesions of the hippocampal establish the quantitative extent of the cell loss associated with par-
formation in rats. A. Topological arrangement of the structures tial lesions placed at different points along the septotemporal axis.
comprising the hippocampal formation (from Chapter 3, Figure (Source: Courtesy of Livia de Hoz.) D. Use of a knife cut through
31). B. Photomicrographs of horizontal sections from a represen- the longitudinal-association pathway in area CA3. This cuts bers
tative control rat (B1) and rats with neurotoxin lesions of hip- (not shown) and damages a small number of cells in CA3 at one
pocampus (2), subiculum (3), and presubiculum/parasubiculum particular point along the axis. Animals to which the same amount
(4). EC, entorhinal cortex; HIP, hippocampus; PaS, parasubiculum; of neurotoxic damage was deliberately applied without damaging
PrS, presubiculum; Sub, subiculum. (Source: Courtesy of Len the longitudinal bers are used as controls in such a study. (Source:
Jarrard.) C. Use of three-dimensional volumetric techniques to Courtesy of Hill-Aina Steffenach.)

depends on communication from area X to area Z, with bers ergic cells afferent to the hippocampus while leaving nearby
passing through area Y, old-fashioned lesions might lead to -aminobutyric acid (GABA)ergic neurons intact (Gallagher
the false conclusion that area Y is essential for this function and Rapp, 1997). It has been claimed that lesions restricted to
whereas the modern approach would not. By the same token, area CA1 but not CA3, or vice versa, can be achieved with pre-
a positive outcome of a lesion study using an excitotoxin gives cise stereotaxic techniques (Lee and Kesner, 2004). This preci-
comfort to the notion that it is the targeted neurons that are sion can certainly be achieved along a small length of the
functionally signicant for the task under examination. longitudinal axis of the hippocampus, but it remains unclear
Conversely, lesions that cut bers while leaving cells intact can how complete and yet still selective such lesions can be.
also be valuable, especially when the axons of a population of Lesions that cut bers while leaving cells intact can also be
neurons have axon collaterals. Neurotoxic lesions can be valuable, especially when the axons of a population of neu-
highly localized (e.g., to distinct components of the hip- rons have axon collaterals, but it is extremely difficult to do
pocampal formation such as the dentate gyrus, subiculum, or this in a manner that interferes with one and only one ber
entorhinal cortex). They may even extend to killing specic pathway.
cell types in the region to which the toxin is delivered, such as Techniques for making permanent lesions are accompa-
the use of immunoglobulin G (IgG) saporin to ablate cholin- nied, when testing is complete, by histological analysis of the
 Theories of Hippocampal Function 587

damage caused. Was the target brain area damaged as as the mbria-fornix, or by selective knife cuts of intrinsic cir-
expected? Was the lesion complete? Did it encroach on other cuitry such as the longitudinal-associational pathway of area
brain areas? Did it damage cells and bers or only cells? If the CA3 (Steffenach et al., 2002).
latter, was conrmation obtained through tract tracing tech-
niques that the apparently intact bers are capable of axonal Pharmacological and Genetic Manipulations
transport? Doing this is part of the point of using animals
rather than just studying people. Nissl stains, silver stains that Drugs and genetic manipulations are alternative ways of inter-
reveal degenerating bers, and other histochemical techniques vening in brain function. The turn of the twenty-rst century
can all be used to good effect. More ambitiously, one might is something of a methodological tipping point (Gladwell,
enquire whether an apparently intact ber tract at or near to a 2002) as the major behavioral neuroscience laboratories move
lesion site is electrophysiologically functional. Likewise, the on from deploying classic lesions on their own to using them
check might be on whether the lesion has selectively affected in conjunction with single-unit recording and tract-tracing
one but not another neuromodulatory system by using in vivo techniques, together with reversible pharmacological and cell
neurochemistry to monitor changes of particular neurotrans- biological manipulations. With respect to the opportunities
mitters. The use of all of these techniques for postexperiment afforded by targeted molecular engineering Wulff and Wisden
histology is the gold standard to which most laboratories quoted Francis Cricks view that the lesion approach needs
aspire, but a difficulty shared by all is that it is not always prac- new methods, especially as the usual ways of ablating the parts
tical to dot every i and cross every t in every experiment. of the brain are so crude. For example, it would be useful to
Much has to be taken on trust so research can proceed at a rea- inactivate, preferably reversibly, a single type of neuron in a
sonable pacebut this trust is sometimes misplaced, which single area of the brain (Wulff and Wisden, 2005, p. 44). As it
sows seeds of confusion. As will become clear, the size of a happens, this would not be very useful for behavioral studies,
lesion can be a major factor to consider when interpreting as a certain minimum amount of tissue must be affected, but
experiments on hippocampal function in relation to recogni- the point is taken nonetheless.
tion memory or spatial memory. The use of drugs in conjunction with behavioral studies
Permanent lesions are often used to realize network dis- has a long scientic history (McGaugh, 2000). One aspect
connections (Aggleton and Pearce, 2001). Most lesion studies of this approach has been the analytical use of intracerebral
in animals involve bilateral damage to a brain structure, but, infusions of microiter and nanoliter quantities of active
in an ingenious twist, a unilateral lesion of one structure can compounds targeted to specic regions, often (although
be made in conjunction with contralateral damage to another. not always) given with posttraining to avoid the impact of
In right-handed people language is largely mediated by struc- known side effects. In relation to hippocampal function, typi-
tures in the left hemisphere such that a stroke in the left brain cal drugs used include excitatory amino acid antagonists
can interfere with speech, whereas one on the right does acting at -amino-3-hydroxy-5-methyl-4-isoxazolepropi-
so rarely (McCarthy and Warrington, 1990). Conversely, onate (AMPA), N-methyl-D-aspartate (NMDA), and
right-hemisphere damage to the parietal cortex is associated metabotropic glutamate (mGluR) receptors, and ligands at
with the fascinating syndrome of spatial neglect (Behrmann et inhibitory synapses such as GABAergic agonists (as described
al., 2004; Farah, 2004). Laterality is barely present in animals in Chapter 6). These compounds interfere with (or potenti-
(one exception being song birds), such that unilateral lesions ate) normal excitatory or inhibitory neurotransmission for as
in animals are often behaviorally benign. Thinking is shifting long as the drug remains present and active at its site of
from a focus on localization toward a crossed-lesion administration. Drugs interacting with the synthesis, trans-
approach to reveal the importance of functional connectivity. port, or degradation of neuromodulatory transmitters
For example, a unilateral lesion might be made to the entorhi- acetylcholine (ACh), norepinephrine (NE), dopamine (DA),
nal cortex (on the left) and a second unilateral lesion to the and serotonin (5-HT), for exampleor antagonists of their
mbria-fornix (on the right) with, perhaps, section of the hip- target receptors are also analytically powerful. There is
pocampal commissures to prevent cross-talk of information growing interest in drugs acting intracellularly on signal-
within the hippocampus via residual interhemispheric con- transduction pathways downstream of postsynaptic receptors,
nectivity. Such a lesion might produce a syndrome similar to such as CAMKII and MAPkinase (see Chapters 6, 7, and 10).
that of an entire hippocampal lesion (even though the hip- Their use requires compounds capable of penetrating cell
pocampus and dentate gyrus are actually intact, and neither membranes to reach their site of action rather than acting
brain area that is lesioned has a bilateral lesion) because it has extracellularly at membrane-bound receptors (Sweatt, 2003).
successfully disconnected cortical and subcortical structures The exploitation of drug actions in the hippocampal for-
that normally interact via the hippocampus. The disconnec- mation as a tool to study learning and memory function is a
tion approach is increasingly popular (Warburton et al., relatively new departure. One aspect capitalizes on the devel-
2001; Gaffan, 2002; Miyashita, 2004). Disconnection of the opments in our understanding of synaptic function discussed
hippocampus need not always be achieved by damaging affer- earlier in the book. An AMPA receptor antagonist, such as
ent or efferent structures in opposite hemispheres. It can also CNQX, shuts down normal fast synaptic transmission for a
be done by sectioning major interconnecting pathways, such period of time (see Chapter 6). In principle, its effects should
588 The Hippocampus Book

be similar to that of a neurotoxic lesion that spares bers of (Morris and Kennedy, 1992). Knockouts are now (almost)
passagewith the analytical advantage of reversibility. In routine, with many a PhD student in well-founded molecular-
contrast, an NMDA antagonist such as D-AP5 has the distinct genetic laboratories saying that they would like to make a
effect of blocking NMDA receptor-dependent mechanisms, mouse during the course of their studies. The development
such as activity-dependent synaptic plasticity (LTP and LTD), of these techniques has been accompanied by databases and
while leaving fast synaptic transmission intact. This should repositories of mouse lines, such as those at the Jackson
and does have very different effects on cognition than any per- Laboratory in the United States (http://www.jax.org/lab), the
manent lesion could have, allowing previously acquired spa- development of testing facilities such as those at the Mary
tial memory to be displayed in performance but preventing Lyon Centre in the United Kingdom (http://www.mlc.har.
new learning (Morris et al., 1986b). Interestingly, the effects of mrc.ac.uk/aboutUs.html), and meetings at which ideas and
an NMDA antagonist such as D-AP5 are regionally depend- recommendations about breeding different lines of mice are
ent because NMDA receptors mediate different physiological debated in an open, constructive manner (Conference, 1997).
functions in different brain areas. In the spinal cord, NMDA Neuroinformatics expertise is having a considerable impact
receptors help mediate suprasegmental connectivity, partly by on the eld (http://www.neuroinf.de/Members/stefan/mbl).
enabling the rhythmical activation of Ca2-dependent K This molecular-genetic approach has now advanced from
currents (Dale and Roberts, 1985; Grillner et al., 1998), homologous recombination to regionally specic manipula-
whereas in the hippocampus they serve as a trigger for certain tions, using the cre-LOX procedure, enabling studies to focus
forms of neuronal plasticity (see Chapter 10). on the role of genes in specic brain areas such as the hip-
What virtually all drugs have in common and what makes pocampus (Tsien et al., 1996). The introduction of inducible
them so useful compared to lesions is that their actions are, at constructs, such as tetracycline, has taken the mouse knockout
least in principle, reversible. Reversibility has two main rami- enterprise into a third phase of sophistication, potentially
cations. First, the intended dysfunction lasts only a short allowing the inactivation of specic genes at specic times in
time, and thus the chances of brain damage or compensatory specic cells of specic regions of the hippocampal formation
changes are reduced. Second, reversibility makes it possible to or other areas (Mayford et al., 1996). For all this enviable
dissect different phases of memoryencoding, storage, con- specicity, however, there remain formidable technical prob-
solidation, retrievalin a more exacting way. However, no less lems to be solvednot least the discovery of region-specic
than lesions, the impact of drugs must be studied with care. promoters and the apparent leakiness of inducible manipu-
Drugs diffuse, raising questions about variation in concentra- lations. Another problem is that most studies have to be con-
tion across a target brain region. This variation may itself vary ducted on mice. Many molecular behaviorists look on mice
with time, particularly with acute injections. Osmotic as miniature rats, but this is clearly absurd. Rats are different.
minipumps enable drugs to be infused at a steady rate over A rarely tested assumption in the eld is that hippocam-
periods ranging from 1 day to 1 month. A steady-state con- pus-dependent tasks for rats may be used similarly in mice.
centration is then achieved when the rates of infusion and dif- Other technical issues have to do with running appropriate
fusion of a drug match preciselya great advance on acute controls for knockout mice made using homologous recom-
infusions. Last and by no means least, it is essential to establish bination (Gerlai, 1996), pleiotropic developmental abnor-
precisely what and where a drug has had its effect. It is aston- malities associated with deletion of a gene throughout
ishing how rarely this is done independently of the effects on embryogenesis and postnatal life (Tonegawa et al., 1995), and
behavior that the drug may have. The use of autoradiography issues associated with the genetic or pharmacological rescue
to establish the local spread of action of a radiolabeled version of a normal phenotype by transgenic expression of a gene in a
of a drug is valuable (Steele and Morris, 1999; Attwell et al., mouse with a knockout background (Ohno, 2001). Others are
2001), but this is a surrogate for an independent marker of its longstanding conceptual problems such as the need, so rarely
functional effect. If an AMPA antagonist is used, physiology addressed, for the double dissociations that are common in
should be undertaken to establish the extent to which synap- conventional lesion studies. Fortunately, these problems are
tic transmission has been compromised. Monitoring glucose being recognized, and the solutions are forthcoming. The
metabolism is one way of looking at this and direct physiolog- technological innovations are remarkable; they are continu-
ical recording another (Riedel et al., 1999). ing, and the molecular-genetic approach has a great deal to
The last 15 years have witnessed considerable excitement offer.
regarding the introduction of transgenic and knockout mice
as in vivo methods of genetic manipulation. Their application 13.2.4 Biological Continuity
to study the systems-level and cellular mechanisms of learning of Hippocampal Function
and memory has been particularly imaginative (Tonegawa et
al., 1995). Beginning with pioneering studies that targeted Virtues of a Comparative Approach
molecules such as CAMKII and fyn tyrosine kinase, which
were implicated in LTP (Grant et al., 1992; Silva et al., 1992), All mammals possess a brain with the same basic pattern
an industrial production line of transgenic and mutant mice of organization: spinal cord, hindbrain, midbrain, forebrain,
soon became available in virtually all elds of neuroscience and cerebral hemispheres (Swanson, 2000). The simplest view
 Theories of Hippocampal Function 589

The genetic similarity of mammals

Platypus
Opossum
Elephant

Anteater
Armadillo
Hedgehog

Cow
Dog
Figure 132. Genomic sequence similarity in Rabbit
mammals. Publication of the genomic sequence of
the rat brain, following those of the mouse and of
Rat
humans, provided an opportunity to assess the Mouse
degree of sequence similarity across species. The Macaque
outcome places rodents on a line less distant from
Chimpanzee
primates and humans than that of many other
mammalian species. (Source: Consortium, 2004.) Human

of this common pattern is that although there has been take ight with birds, but unlike rodents they are visual ani-
substantial encephalization of function in primates and mals and have a number of other advantages than make them
humans relative to rodents the core functions of various useful for certain studies.
areas of the central nervous system (CNS)cerebellum, When using animals, it is essential to be respectful of their
midbrain, neocortexare most likely conserved across all biological needs, to use the minimum number necessary to
mammalian species. Even where anatomical similarities realize the required level of statistical power, and to think care-
may seem cryptic, there are striking genetic similarities fully about how to motivate and reward them in an ethically
that justify the use of primates and other mammals (Fig. acceptable manner. Reduction and renement are laudable
132), and there may even have been convergent evolution of goals in animal experimentation. In general, we cannot ask
certain aspects of anatomy and intelligence in avian species animals to do something through language, but mild food
(Jarvis et al., 2005), as in the case of corvids (Emery and deprivation creates the opportunity to reinforce an animal
Clayton, 2004). with rewards for performing a task. This is more charitable
If this comparative assumption also applies to the hip- than it sounds because, as it happens, food deprivation actu-
pocampus, which it surely should, the key advantage of using ally lengthens the life of laboratory animals (Nelson, 1988).
animals is that invasive anatomical, physiological, and behav- The use of mild electric shock to motivate animals in studies
ioral studies can be carried out. Although not everyone shares of learning and memory may seem unfortunate to some, but
the view that such work is ethically acceptablea concern the neurobiological study of fear and stress is important; and
exacerbated by growing evidence of the complex emotional, there are circumstances in which the rapid learning achieved
cognitive, and cultural life of higher primatesmost neuro- with shock cannot easily be achieved in other ways. Electric
scientists adopt the utilitarian perspective that such work is shock has the advantage of being variable and can be motivat-
justied on its scientic merits and by its potential to relieve ing even when quite mild, but it is articial. It is probably no
both human and animal suffering. The macaque monkey is a more stressful than the rst day of swimming in a watermaze,
particularly suitable subject because Old World monkeys are although the stress of the latter habituates quickly, and escap-
closely related to humans, although the last common ancestor ing from water may be a more biologically meaningful form of
was about 30 million years ago (Kay et al., 1997). Its brain is motivation.
remarkably similar to that of humans (see Chapter 3), and
functional imaging studies are revealing striking similarities Limitations in the Use of Laboratory
in, for example, the dynamic organization and responsiveness Animals to Investigate Brain Function
of the cerebral cortex (Van Essen and Drury, 1997). Rodents
may seem less well justied on this account, although the Animal work does have its disadvantages. In comparison with
topological organization of their brain is also quite similar human studies, animal studies include problems associated
(Whishaw, 2004b). The comparative similarity argument may with what the animals may be said to know, with language
590 The Hippocampus Book

and communication, and the articiality of the physical and engaging their curiosity and the use of reward and punish-
social environments in which animals are tested. ment. Such techniques are more painstaking than merely ask-
Can we be condent that animals possess knowledge in ing a person to remember something; and they run the risk of
the same way humans do? As certain kinds of human knowl- missing important subtleties. The answers you get from
edge come to us via language (such as semantic, or fact, humans who are doing memory tasks depend critically on the
memory), some observers have questioned whether such way the questions are put and thus the subjects awareness of
forms of knowledge exist in animals (Macphail, 1998). An the problem at hand (Graf et al., 1984; Squire, 2004). This
American knows that a slam dunk is a shot in basketball, issue was discussed in detail in Chapter 12: recognizing the
and he can explain it to a puzzled foreigner; but how do distinction between explicit and implicit memory process-
we teach knowledge to animals? Do they have semantic- ing. Although this distinction may also apply to animalsand
like knowledge or merely conditioned responses? Should we the declarative memory theory argues that it doesit is diffi-
assume that a monkey knows that a particular object cult to get a handle on it analytically. It seems unlikely that we
means reward simply because it has been consistently paired shall ever know whether animals are conscious, although
with reward, or that one stimulus leads to another because here again experimenters have come forward with ingenious
they are associated? The presence of pair-asssociate neurons in ideas to try tackle the problem.
the temporal regions of the neocortex of monkeys is sugges- Finally, behavioral tasks in the laboratory are clearly arti-
tive (Miyashita, 2004); and at a strictly behavioral level some- cial, but then so are cyclotrons and supercolliders and, for that
thing akin to word learning has been demonstrated in dogs matter, test tubes. Science often has to be articial to proceed.
(Kaminski et al., 2004; Markman and Abelev, 2004). The pri- Biomedical laboratory work generally requires that animals
mary reason for believing that animals know things is are kept in somewhat articial conditions, trained when they
because they behave in relation to stimulifood, water, shel- are experimentally nave rather than after they have accu-
ter, dangeras if they know what these stimuli are. A monkey mulated knowledge, skill, or experience, and often trained on
that has learned to use a rake to collect food that is beyond protocols that seem to bear little relation to normal behav-
arms length probably does know what a rake is and what it ior in the wild. Many primatologists and evolutionary psy-
can be used for. Receptive elds in appropriate somatosensory chologists see this as a major problem (Tooby and Cosmides,
areas of the cortex expand in association with learning 2000), particularly with respect to the evolution of social
(Maravita and Iriki, 2004). behavior. In contrast, others look upon ecological factors (and
The problem is that, although the possession of knowledge species differences) as somewhat tangential to the primary
does not logically require language, communication about it task of searching for general principles of brain function
to others usually does. For example, episodic and autobio- (Macphail, 1982; Bolhuis and Macphail, 2001). This disagree-
graphical memories are about specic events (what) that ment has provoked much debate (Healy and Braithwaite,
happen in particular places (where) at particular times 2000; Hampton et al., 2002). What is not always appreciated is
(when) (see Section 13.6). The memory that one person has the extent to which certain laboratory tasks are either explic-
of what, where, and when may be different from that of itly designed to be analogous to things that animals do in the
another person who experiences the same event; in fact, much wild (e.g., recognizing a previously seen stimulus) or unex-
discourse is concerned with debating such differences. It fol- pectedly turn out to be analogous to naturalistic behavior. An
lows that it is difficult to investigate episodic memory in ani- example is the ability of rhesus macaques to learn arbitrary
mals. The medium of communication is different and unlikely sensory-motor mappings. These are tasks that require
to be one that satisfactorily reects that animals personal a monkey to perform action A to stimulus A, action B to stim-
experience. Indeed, Tulving took the view that as far as we ulus B, and so on. In one laboratory investigation of this abil-
know, members of no other species possess quite the same ity, different visual cues were presented to rhesus monkeys,
ability to experience again now, in a different situation and with the monkeys then having to make different movements
perhaps a different form, happenings from the past, and know of a computer joystick (Brasted et al., 2002). Such sensorimo-
that the experience refers to an event that occurred in another tor mappings are learned quite slowly; and on the face of it,
time and in another place (Tulving, 1983, p. 1). Regardless of this may seem hardly surprising as monkey cognition did
whether this skepticism is justied, we should recognize that not evolve to play computer games. However, as pointed out
distinguishing both episodic-like and semantic-like mem- by the authors, the different actions of vervet monkeys to dif-
ory from mere changes in learned behavior is far from ferent alarm calls made by troop members (Seyfarth et al.,
straightforward. Several studies have suggested that animals 1980) are also examples of arbitrary sensory-motor map-
probably can remember events as events, and there is no a pri- pings. Monkeys climb trees in response to the barking call
ori reason why language is necessary for them to be able to do uttered when a leopard is nearby and look skyward followed
so at another time and in another place. by hiding in bushes to the cough call uttered when an eagle
The training (of animals) and the asking (of humans) is spotted. Brasted et al. (2002) summarized evidence indicat-
may not always yield the same answer in experiments on ing that these different actions to distinct stimuli seem to
learning and memory. The primary ways of teaching animals be learned, just as the computer joystick movements have
to perform particular behavioral tasks are, as we have seen, by to be. An apparently abstract laboratory task may be
 Theories of Hippocampal Function 591

somewhat closer to primate cognitive ecology than might be distinction was prompted by their nding that amnesic
appreciated. patients could acquire and retain the cognitive skill of mirror-
To conclude: interventional studies in animals are a power- reading novel words (in which words and all their letters were
ful way of investigating brain function. When using them in printed backward) despite showing no conscious memory of
conjunction with behavioral tasks, we must accept that the the training experience itself. Control subjects learned at an
way in which hippocampal function is studied in animals has identical rate and, of course, could also recall their training on
to be different from the approaches described in Chapter 12 the task (Cohen and Squire, 1980). Squire went on to develop
because animals cannot talk, they cannot tell us of what they and articulate the declarative memory theory of amnesia,
are aware (if anything), and they are unlikely to have a sense of implicating the hippocampal formation and other structures
their past or their future or to have, to quote Tulving (1983) of the medial-temporal lobe (Squire and Cohen, 1984; Squire
again, quite the same ability to experience the past that and Zola-Morgan, 1991; Squire, 1992; Squire et al., 2004). Four
humans do. It does not follow, however, that these differences key propositions of this theory (Box 131) are as follows.
of approach mean that the hippocampal formation in animals The rst propositionthat the hippocampus is primarily
is carrying out a different memory function than exists in involved in memoryis now relatively noncontroversial. It is
humans. strongly supported by work on amnesic patients and func-
The enlightened possibility, reecting the commonalities of tional imaging studies (see Chapter 12). Observations on peo-
anatomy, physiology, and cell biology, is that it carries out ple have been extensively conrmed and elaborated in studies
essentially the same information-processing algorithms but of the effects of experimental lesions in animals. In keeping
that they might seem to be different because the human hip- with the overriding theme of this chapter, we accept this rst
pocampus and the animal hippocampus are operating on dif- proposition of the theory without further comment.
ferent inputs and so provide different outputs to downstream The second proposition denes declarative memory as
targets. A metaphor may help to explain the gist of this idea. representing both facts and events (Fig. 133). What fact
Imagine being able to transplant the hippocampus of a mon- (semantic memory) and events (episodic memory) have in
key into the brain of a human. Suppose the neurosurgeon is common is that both are propositional and can be brought to
the wizard we know all neurosurgeons to be, and that she suc- conscious awareness. Information about facts and events can
cessfully connects up all the nerves and blood vessels to their be declared (Paris is the capital of France or I ate an
appropriate targets. Our perspective is that the result of such almond croissant for breakfast). and such declarations can be
an operation is that all should be wellour monkey hip- either true or false. By virtue of this shared property of fact
pocampus should work normally inside the human brain. and event memory, such memories can be combined inferen-
However, because it would now be receiving different kinds of tially with other information to yield new propositions (a
information from its cortical and subcortical inputs, the end point developed further in the relational-processing version
result would be appropriate to the human, not the monkey. of this theorysee Section 13.5). This inferential property is
Laterality of function in the human brain is a less fanciful
illustration of the same argument because it seems unlikely
that the left and right hippocampi of humans are anatomi-
cally, physiologically, or biochemically very different; yet the Box 131
outcomes of their processing are clearly distinct. With this Declarative Memory Theory
optimistic perspective of the enterprise at hand, we now pro-
ceed to discuss the major theories of hippocampal function. 1. Memory: The primary function of the hippocampal for-
mation is in memory.
2. Selectivity: The role of the hippocampal formation in
memory is selective. It mediates the memory of facts and
events, called declarative memory. This is the type of
13.3 Declarative Memory Theory
memory that, in humans, can be consciously recalled.
3. Memory systems: The hippocampal formation is one of a
13.3.1 Outline of the Theory number of structures that comprise a medial temporal
lobe memory system. Although the components of this
In 1980, Cohen and Squire suggested that the distinction be- system may have distinct subfunctions, it operates collec-
tween declarative and procedural information processing tively to mediate the formation and initial storage of
could be relevant to the pattern of memory decits seen with declarative memories.
amnesia. Drawing upon Gilbert Ryles distinction between 4. Time-limited: The role of the hippocampus in memory is
knowing that and knowing how (Ryle, 1949), they pro- time-limited. Memory is gradually reorganized as time
passes after learning. The hippocampus contributes to a
posed that amnesic patients are impaired in forming conscious
time-dependent systems-level consolidation process such
memories of facts and events (remembering that) but have
that, once completed, long-term memory traces are stored
relatively normal learning motor and cognitive skills (remem- in the cortex and neural activity in the hippocampus is no
bering how). Following Milners observations of successful longer required for or involved in recall.
motor learning in patient H.M. (see Chapters 2 and 12), this
592 The Hippocampus Book

Declarative memory theory: a taxonomy of mammalian memory systems

Declarative Nondeclarative
memory memory

Episodic Semantic Procedural Classical Non-associative Figure 133. Taxonomy of mammalian memory sys-
Priming
memory memory memory conditioning learning tems. The declarative memory theory recognizes many
types of memory system and their mediation by dis-
tinct neural brain areas. The scheme was rst intro-
duced by Squire (1987) and has been updated many
Striatum; times since (e.g., Squire, 2004). The relative independ-
Hippocampus- Amygdala; Reflex
motor cortex; Neocortex ence and interdependence of these separate systems
medial temporal lobe cerebellum pathway
cerebellum has not been clearly delineated.

an important point of similarity between information about Tulving has long argued, the input to episodic memory must
facts and events, but controversy surrounds the theorys asser- be partly via semantic memory; equally, we add to that knowl-
tion that there is a common neural substrate for the formation edge structure through our experience and memory of events,
of both episodic and semantic types of memory. so the development of new semantic memories generally
A key psychological difference is that unique events occur depends on episodic memory. The two forms of memory are
only once and are specic to particular contexts and moments interdependent. However, they can also act in parallel and
in time, whereas factual knowledge is often gradually accu- independently with, for example, direct input of information
mulated and (generally) nonspecic with respect to the con- from perceptual-representational systems to both semantic
text of learning. Certain other neuropsychological theories of and episodic memory (Graham et al., 2000). Investigations
human memory argue that the nervous system encodes, of developmental amnesiafocusing on a set of young people
stores, and retrieves episodic information in a way that is referred to a neurological clinic who had grown up being
qualitatively different from that for semantic information very forgetful about everyday life but able to do reasonably
(Shallice, 1988; McCarthy and Warrington, 1990; Schacter and well in schoolraised the alternative possibility that the
Tulving, 1994; Aggleton and Brown, 1999). Such theories gen- hippocampus plays an exclusive role in the formation of
erally embrace or, at least, start off from the concept of episodic memory (Vargha-Khadem et al., 1997). However, rig-
episodic memory as a distinct entity (Tulving, 1983) and orous assessment of this idea is hampered by uncertainties
question whether it is helpful to subsume it with semantic about the extent of hippocampal damage in these cases and
memory into the unitary category of declarative memory. In theoretical uncertainties about the relative degree of inde-
contrast, declarative memory theorists see the distinction pendence and interdependence of episodic and semantic
between facts and events as descriptive rather than funda- memory. We discuss these cases in the critique below.
mental, both being initially encoded by structures in the Notwithstanding this qualication, the second proposition
medial temporal lobe (MTL) that collectively comprise what of the declarative memory theory implies that a central focus
they believe to be a unitary memory system. Anatomically, of the theory has to do with the information processing
the MTL is positioned downstream of structures in the ventral and neural representation of what we ordinarily think of
what pathway of the visual system in primates (Ungerleider as memorythe things we bring to mind as conscious rec-
and Mishkin, 1982) and so is well placed to receive attended, ollections. For some, the term explicit memory is preferred
perceptually processed information. Damage to this down- to describe certain types of memory task in which the subject
stream system is therefore held as likely to cause a propor- is consciously aware of the earlier study episodein contrast
tionate impairment in both fact and event processing, at least to implicit memory tasks, which lack the requirement
in the anterograde domain.. for recollective awareness (Graf and Schacter, 1985). The
Whether viewed as a merely descriptive or fundamental differences between the declarative/nondeclarative and
distinction, these two forms of memory are not thought (by explicit/implicit distinctions are important at some levels
anyone) to operate in isolation. There is a dialectic here, such of analysis but are arguably little more than terminological
that our ability to form new event memories depends on both at another. In practice, the declarative memory theory treats
our general and our personal knowledge of the world (seman- explicit and declarative memory as interchangeable
tic knowledge) and on accumulating information from our terms (see Chapter 12). In the same vein, no strong philo-
experience of events in the world (episodic memory). As sophical position about consciousness is implied by the
 Theories of Hippocampal Function 593

theorys supposition that declarative memories entail aware- Perceptual and cognitive skills (e.g., the mirror-reading
ness or consciousness. The theory is not tied to any partic- task of Cohen and Squire, 1980); priming phenomena
ular theory of this problematic but fascinating concept (e.g., word-stem completion) (Graf et al., 1984), percep-
(Zeman, 2004); it quite reasonably takes a folk psychology tual priming (Hamann and Squire, 1997), and concep-
view of the matter. tual priming (Levy et al., 2004) (see discussion in
Another issue about the second proposition of the theory Chapter 12)
is its skepticism that a categorical subdivision of memory Simple forms of nonassociative learning such as habitu-
retrieval with respect to familiarity and recollection (Yoneli- ation (waning of responsiveness following repetition)
nas, 2001) can be mapped onto distinct MTL brain structures. and sensitization (augmented responsiveness) (e.g.,
Both forms of remembering involve conscious awareness, the Kandel et al., 2000)
dening attribute of declarative memory. Familiarity appears
The collective attributes of this heterogeneous group of
to be a phenomenological attribute that is acquired by a novel
skills is that they are not propositional (and so can be neither
stimulus, within several sensory systems, after its rst presen-
true nor false), are generally learned gradually (although are
tation. A stimulus that has been presented before may then be
sometimes rapidly acquired), do not require conscious aware-
judged familiar, but this can happen without the subject
ness of the knowledge implicit in their execution, and are
having any recollection of when or where that stimulus was
simply things that people or animals learn to do. For some
seen, heard, or smelled previously. Recollection is more com-
(Macphail, 1998), they are the sum-total of what animals can
plicated, as it also involves source memory pertaining to
learn to dosuch as the proverbial rat that presses the lever in
where or when the stimulus was previously presented.
an operant chamber because doing so in the past he has been
Subjects may even engage in mental time travelthe act
rewarded but not because it has any expectation of securing
of moving in their minds eye back to the time or place of
reward. Although not what we would ordinarily regard in
prior occurrence while simultaneously remaining in the tem-
everyday discourse as remembering, nondeclarative learning
poral present and responsive to their perceived surroundings
is important in constituting the dispositions, habits, atti-
(such as in conversations about the past with someone while
tudes, and preferences that are inaccessible to conscious recol-
simultaneously driving a car). Certain theorists look upon
lection, yet are shaped by past events, inuence our behavior
these two forms of memory retrieval as independent
and our mental life, and are a fundamental part of who we
(Baddeley et al., 2002), with only recollection or relational
are (Kandel et al., 2000, p. 1119).
memories critically dependent on processing in the hip-
The third proposition of the theory concerns the existence
pocampal formation and diencephalon (Cohen and
of distinct brain systems for memory. Systems are distinguished
Eichenbaum, 1993; Aggleton and Brown, 1999; Eichenbaum
with respect to structure as well as function. Different forms
and Cohen, 2001). Yonelinas (2001) has outlined quantitative
of memory are held to depend on anatomically distinct,
techniques derived from signal detection theory to identify
though partially overlapping, brain regions. Different brain
circumstances in which people may be using recollection
systems for memory probably evolved to mediate functionally
rather than familiarity, arguing that recollection is a thresh-
incompatible purposes (Sherry and Schacter, 1987a),
old process, whereas familiarity can be better understood in
although the lack of a fossil record for memory makes it
signal-detection terms; they then applied this to the analysis of
extremely difficult to draw rm conclusions about their evo-
selected amnesic patients (Yonelinas et al., 2002). However,
lution. However, that different memory systems might be
this approach has been the subject of erce criticism by
mediated by anatomically different brain substrates nicely
Manns et al. (2003) and Wixted and Squire (2004). As some of
reects Francois Jacobs idea that evolution is a tinkerer,
these ideas constitute major changes to the declarative mem-
adding qualitatively distinct circuits rather than bringing
ory theory as it was originally conceived, they are presented in
about wholescale changes to the brain.
separate sections (see Sections 13.5 and 13.6). For the present,
The putative medial temporal lobe memory system
we accept declarative memory as a conceptual category on
(Squire and Zola-Morgan, 1991) is held to be one of these
its own merits.
brain systems for memory. It comprises the hippocampal for-
How are other types of long-term memory considered in
mation (as dened in Chapter 3) and both the perirhinal and
the theory? Collectively called nondeclarative memory, they
parahippocampal cortices of the medial temporal lobe (Fig.
are thought to reect acquired information, habits, and
134). Damage to these structures in humans and nonhuman
learned dispositions that are expressed in behavior but cannot
primates causes disturbances of declarative memory without
be declared. Initially called procedural (Squire and Cohen,
necessarily affecting other types of learning or memory.
1984), a term rst used in the eld of articial intelligence by
Damage to homologous structures in rats causes apparently
Winograd (1975), these types of learning are held to include
similar dissociations. In an evocative phrase, Squire wrote of
the following.
the distinction between declarative and nondeclarative mem-
Motor skills and learned dispositions (e.g., the stimu- ory as being honored by the nervous system. He suggested
lusstimulus associations and stimulusresponse habits that, in addition to the medial temporal lobe mediating declar-
acquired during simple conditioning tasks) (Corkin, ative memory, the striatum is important for stimulus-response
1984; Cavaco et al., 2004) habits, the neocortex is the substrate for simple perceptual
594 The Hippocampus Book

The medial-temporal lobe memory system even outside the medial temporal lobe itself, are also involved
in declarative memory. For example, Shimamura and Squire
(1987) claimed that the frontal lobes contribute memory of
S
Hippocampal the context in which information is acquired (i.e., source
regions CA1 memory) on the basis of what they observed in amnesic
patients with additional frontal damage. As recalling the con-
CA3 text where an event happened is a fact that one can declare and
DG of which one is conscious, it must (by denition) be a type of
declarative memory. Similarly, Squire has written of a lim-
bic/diencephalic brain system that incorporates structures
Other direct Entorhinal such as the anterior thalamic nucleus, the mediodorsal
projections cortex
nucleus and connections to and from the medial thalamus
that lie within the internal medullary lamina (Squire et al.,
Perirhinal Parahippocampal 1993, p. 462). Squire and Zola asserted that declarative mem-
cortex cortex
ory is dependent on the integrity of medial temporal lobe
and midline diencephalic structures (Squire and Zola, 1998,
Unimodal and polymodal association areas p. 205). This anatomical liberalism wisely recognizes that
(Frontal, temporal, and parietal lobe)
medial temporal structures do not operate in isolation from
the rest of the brain (hence the reference to direct projections
Figure 134. Medial temporal lobe memory system. Major compo-
nents of Squire and Zola-Morgans (1991) medial-temporal lobe
in Fig. 134), but it leaves skeptics wondering where the
memory system (structures with gray shading), which is believed to anatomical boundaries of the medial temporal lobe memory
be the neural substrate for the formation of declarative memories system begin and end. In any event, the key point of this third
(as updated in 2004). Information enters the system from neocorti- proposition is clear: The declarative memory theory places the
cal association areas. It is projected via the adjacent perirhinal and hippocampal formation at the apex of declarative memory
parahippocampal cortices to the entorhinal cortex, which has the processing supported by a network of other brain structures.
pivotal role of being both the point of entry of information into the The fourth major postulate of the theory concerns memory
hippocampal formation and a point of exit. Information passes consolidation. This is a particularly important aspect of the
within the hippocampal formation via a cascade of unidirectional theory and a focus of much current research. The basic idea
projections before signals relevant to the encoding of new memo- is that the hippocampal formation has a time-limited role
ries, or their consolidation, are fed back to the neocortex. CA1,
in memory for an individual fact or event. Some interaction
CA3, and SUB represent areas CA1 and CA3 and the subiculum, as
dened in Chapter 3; DG, dentate gyrus.
is held to take place between it and presumed long-term stor-
age sites in the neocortex. Gradually, through some time-
dependent consolidation process in which the hippocampus
representations and priming (see Chapter 12), the amygdala and neocortex interact, initially labile memory traces in the
for emotional and social learning, and the cerebellum for the cortex become stabilized. This renders them resistant to later
learning of skeletal components of classic conditioning. This brain damage to the medial temporal lobe itself, although not,
anatomically based taxonomy is seen as a crucial advance in of course, to brain areas that are the eventual storage sites of
understanding by those neuroscientists who hold a relatively lasting memory traces (Fig. 135). Implicit in this fourth
localizational view of brain systems. proposition of the theory is the supposition that memory for
However, as outlined in Section 13.2, current ideas about remote events depends on the strength of memory traces (i.e.,
information processing in the CNS refer to networks of inter- information encoded within ensembles of spatially dispersed
acting brain areas whose normal operation can be disrupted neocortical neurons), where trace strength can be roughly
by disconnections rather than lesions of specic brain areas thought of in reductionist terms as alterations in membrane
(McCarthy and Warrington, 1990; Aggleton and Brown, 1999; excitability, the strength of synaptic connections, or other
Gaffan, 2002; Murray and Wise, 2004). Human brain imaging activity-dependent properties of cellular physiology.
work has also moved rapidly from a localizationist perspec- Memories with high trace strength can be recalled easily; those
tive to one that stresses the importance of effective connec- with lower trace strength cannot. Consolidation is thought to
tivity between brain areas (Friston, 1995, 2004) and the be a process that gradually, sometimes through the interleav-
dynamic interaction of medial temporal and frontal regions ing of new traces with existing traces, brings trace strength
during intentional aspects of memory retrieval (Miyashita, from low to high levels.
2004). Accordingly, the one-to-one mapping of memory A key issue when thinking about consolidation is exactly
type to brain area of the original statements of declarative what is temporarily stored in the hippocampal formation
memory theory is now somewhat dated. itself. There are several possibilities. One view is that detailed
Another complication associated with the mapping of sensory/perceptual information is stored there for an inter-
types of memory onto brain areas, explicit in Figure 133, is mediate period of time and then literally shuttled to the neo-
that even the proponents of declarative memory theory recog- cortex. At the start of the consolidation period, the memories
nize that brain areas outside the hippocampal formation, and would be in the hippocampus; by the end, they would be
 Theories of Hippocampal Function 595

Systems-level memory consolidation

encoding and cellular consolidation processes during systems-level consolidation systems-level consolidation complete
completed, soon after learning

passage of time

Initial memory traces

New memory traces forming during Neocortex

systems-level consolidation Hippocampus
Consolidated memory traces

Figure 135. Systems-level memory consolidation. Pathways areas and the hippocampal formation. Over time, a consolidation
between neocortical areas representing recent events or recently signal emanating in the hippocampus is thought to strengthen neo-
acquired facts are held to be weak initially, even after the protein cortical connections (middle) to the point where hippocampal
synthesis-dependent cellular consolidation that sometimes follows activity is no longer necessary (right). The time this consolidation
immediately after initial encoding (left). Memory soon after learn- process takes may be quite longweeks, months, or even longer
ing relies on rapidly formed connections between these neocortical (Squire and Alvarez, 1995).

in the cortex. No one has much condence in this possibil- though systems-level memory consolidation is a process
ityas if the brain were some kind of railway shunting yard; revealed through experiments in which the hippocampus is
and it was never considered seriously by Squire. Another idea, rendered dysfunctional, it is unlikely that the evolutionary
albeit a somewhat metaphorical one, is that hippocampal pressure to develop a consolidation mechanism was to avoid
storage consists of pointers or indices (Teyler and the deleterious effects on memory of brain damage. So what is
DiScenna, 1986). These representations do not contain the function of consolidation and why the need for a brain
detailed information; they are more like cartoons and are area with a time-limited role in memory? One answer, from
thought to do two things. First, they help activateby point- McNaughton et al. (2003), is that the numerous neocortical
ing at themthe relevant but dispersed neocortical neurons interconnections that exist for creating an associational
at which traces representing detailed sensory information are framework of acquired knowledge are weak, slowly formed,
located (presumably through alterations in synaptic strength and/or liable to rapid decay. It is essential to protect against
between interconnecting neuronssee Chapter 10). This is a everything-becoming-connected-to-everything during the
lovely metaphorneurons pointing at each otherbut long process of learning throughout development and adult
exactly how such an addressing mechanism would work is life. More generally, consolidation is held to be a gradual,
unclear. How does a neuron in the hippocampus point at selective process precisely because it has to beensuring that
one or more neurons in the cortex or, even more difficult, a only the relevant connections are made across ensembles in
subset of their synaptic interconnections? The answer, if this and between cortical networks to represent information about
metaphor is on the right lines, is likely to involve diffuse acti- facts and events accurately (OReilly and Norman, 2002). We
vation coupled to local signals, in much the same way that in revisit the issue of whether consolidation always has to be
the dark a ashlight (diffuse) lights up a prowling cats eyes gradual in Section 13.6.
(local). It is the conjunction of light and the cat that enables The second thing that hippocampus indices and pointers
spatial specicity, it not being a property of either feature are supposed to do is guide the consolidation process over
alone. More generally, specicity is one of the big issues of time (Squire and Alvarez, 1995). It may take place over hours,
modern neuroscience at every level of analysis applying as days, weeks, months, or even yearsthe virtue of such slow
much to intracellular synaptic stabilization mechanisms processing being that it helps avoid catastrophic interference
(Goelet et al., 1986; Frey and Morris, 1997) as to the systems- in distributed associative memory traces stored in the cortex
level memory consolidation processes that involve several (McClelland et al., 1995). This is the teenagers bedroom
brain structures (Dudai and Morris, 2001). type of interference that occurs when one set of new memo-
Given that areas in the neocortex are the likely sites of per- ries (new junk) overlays earlier formed traces (old junk) in
ceptual processing and the eventual sites of storage, we might such a manner as to interfere with their subsequent retrieval
reasonably wonder why the hippocampus is ever needed. Why (Has anyone seen my iPOD?). Cautious interleaving, guided
bother with such apparent duplication? Moreover, even by hippocampal pointers, prevents this from happening. As
596 The Hippocampus Book

this consolidation process eventually comes to an end for a memory. Recognition memory refers, at least operationally,
particular set of information, the pointers have done their job. to the ability to discriminate between stimulus items that have
There is therefore no need for extensive long-term memory been seen recently and others that have not. A judgment of
retention in hippocampus. Like forgotten movie stars, hip- prior occurrence (Brown, 1996), it is generally impaired in
pocampal pointers just fade away. global amnesia, as shown by a standard test such as
Declarative memory theory therefore asserts that after con- Warringtons Recognition Memory Test; however, it is not
solidation is complete damage to the hippocampus should always severely impaired in patients with more restricted hip-
have no effect on memory. However, as consolidation takes pocampal damage (Aggleton and Shaw, 1996). Associative
time, the theory makes the important prediction that tempo- memory, on the other hand, refers to the ability to learn that
ral gradients of retrograde amnesia should be obtained in two stimulus items go together, such as a stimulus and a
both humans and animals after damage to the medial tempo- reward or a response and a reward, without regard to their
ral lobe. This prediction is attractively counterintuitive: It familiarity. This is a habit type of learning that Gaffan
states that if the hippocampus is damaged at time t an event argued was preserved in the presence of amnesia.
experienced several weeks before time t would be remembered To model recognition memory, he collected a set of 300
better than an event experienced only a few days earlier junk objects and presented 60 of them to the monkeys each
because the former, despite being older, would have enjoyed a day. Each trial consisted of a sample phase, during which the
longer period of consolidation. Testing this prediction has experimenter presented one object on its own, and a choice
become a cottage industry in its own right within the eld. phase in which two objects were presented: the one that had
This completes our presentation of the major features of just been shown as a sample together with another, novel
the theory. In the discussion that follows, we rst present rel- object. These objects were presented on trays inside a
evant interventional studies on animals (primarily primates Wisconsin General Testing Apparatus (WGTA) (Fig. 136A).
but with some mention of other species) and then a critique. To motivate the monkeys to perform the task, a sugar puff
As noted earlier, the relevant human studies have been dis- reward could be retrieved when they displaced the object pre-
cussed in Chapter 12, and an important extension of the sented during the sample phase and again when they chose
declarative memory theory called the relational processing that same object during the choice phase. The only way the
theory (Cohen and Eichenbaum, 1993; Eichenbaum and monkeys could perform correctly was if they: (1) learned and
Cohen, 2001) is considered separately (see Section 13.5). It has then applied the rule that a reward was available for displac-
inspired a body of experimentation somewhat separate from ing the familiar object rather than the novel one; and (2)
Squires version of declarative memory theory. Similarly, work remembered which object was familiar during the choice
on the consolidation of spatial memory is reserved for discus- phase of each trial after the memory interval following sample
sion in Section 13.4. presentation. This technique of probing recognition memory
was therefore called delayed matching to sample with trial-
13.3.2 Development of a Primate unique cues, where matching-to-sample refers to the rule
Model of Amnesia for solution, trial-unique to the fact that an absolute judg-
ment of familiarity of the objects is required, and delayed
In the historical survey of Chapter 2, we noted that the initial because of the memory interval between the sample and
efforts to model human amnesia in animals were largely choice phases of each trial. It is usually abbreviated as DMTS.
unsuccessful. Monkeys given lesions of the medial temporal After learning using the short memory delay of 5 to 8 seconds
lobe showed decits in certain tasks. but overall they were (at which the monkeys performed extremely well), two per-
capable of learning tasks that one might not expect an formance challenges were added. One was to increase the
amnesic monkey to learn. Similarly, rats given hippocampal memory delay up to several minutes; the other was to present
lesions showed diminished exploration and striking inexibil- a list of several sample objects, one after the other, and then
ity in their learned behavior but were quite capable of learn- do a seriesof choice trials. Delays and lists are widely used
ing many quite complex tasks. The apparent discrepancy in modern variants of this task. Finally, to model associa-
between the human and animal data did not go unnoticed, tive memory, Gaffan used a small subset of the same junk
and various attempts were made to explain it (Iversen, 1976; objects but a different protocol. The monkeys had to learn
Weiskrantz, 1982). which objects were consistently rewarded and which consis-
tently unrewarded, presenting the objects time and again
Development of New Tasks for Monkeys until they were all completely familiar. The animals indi-
to Parallel the Types of Memory Lost in Amnesia cated their knowledge of the object-reward relations correctly
by reaching for rewarded objects when they were presented
An important insight by Gaffan was that the tasks then used and toward a small brass disk placed alongside when the
in animal studies were quite unlike those on which amnesics nonrewarded objects were presented. Correct choices were
were seen to fail (Gaffan, 1974). He suggested that new tasks rewarded; incorrect choices were not (Box 132).
be developed to capture what he then saw as an important dis- The monkeys given fornix lesions showed delay-dependent
tinction between recognition-memory and associative- impairment in the recognition task (Fig. 136B) and list
 Theories of Hippocampal Function 597

A Recognition memory in a WGTA B Delayed matching to sample (DMTS)

100

90

% correct
80 NC
F
70

60

50
10 70 130 delays (s)

C Delayed non-matching to sample (DNMTS)

100 100
C C
90 90
NC NC

% correct
% correct

80 80 A
A
70 H 70 H
A+H A+H
60 60

50 50
10s 30s 60s 120s delays (s) 1 3 5 10 lists of objects

Figure 136. Recognition memory in the monkey. A. Rhesus maca- of the hippocampus and amygdala lesions, including neocortical
que in a Wisconsin General Testing Apparatus (WGTA) reaching structures of the medial temporal lobe, cause delay-dependent
for and displacing objects that hide a food reward. (Source: Courtesy (left) and list-dependent (right) impairment of delayed nonmatch-
Christopher Coe, Harlow Primate Laboratory.) B. Fornix lesions ing to sample (DNMS) after the animals reach the initial criterion
cause delay-dependent impairment of delayed matching to sample of 90% correct with single-sample objects. (Source: After Mishkin,
(DMS). (Source: After Gaffan, 1974). C. Conjoint aspiration lesions 1978.)

length-dependent impairment, but their associative learning memory or (2) intellectual function (they had learned the rule
was unimpaired. As correct performance at short delays preoperatively and could still execute it postoperatively), but
depends on successful execution of the procedural rule (3) they prevent the formation of memories lasting much
learned preoperatively (matching-to-sample), the deficit longer than about 2 minutes. This conjunction of three
caused by the fornix lesion at longer delays is unlikely to have important characteristics of human amnesia in a single exper-
been due to the monkeys not remembering the rule. Rather, it iment represented the rst apparently successful attempt to
suggested that fornix lesions have no effect on (1) short-term model key features of the syndrome in animals. Ironically,
although explicitly introduced on the grounds that it would
cause the least extrahippocampal damage (Gaffan, 1974, p.
Box 132
1100), fornix lesions are associated with only modest amnesia
Procedures for Testing Learning
and Memory in Primates
in humans (see Chapter 12). This represented something of a
puzzle but one that was soon solved.
Apparatus: Wisconsin general testing apparatus
Computerized touch screens Emergence of Delayed Nonmatching to Sample as the
Procedures: Pattern discrimination Benchmark Test of Recognition Memory in Primates
Spatial discrimination and reversal
Object discrimination and reversal DMTS tasks had been used previously with animals, but
Concurrent object discrimination Gaffans innovation of trial-unique cues during each week of
Delayed response
testing appeared to ensure that the familiarity discrimination
Delayed matching-to-sample (DMTS)
demanded of the monkeys was effectively an absolute judg-
Delayed nonmatching-to-sample (DNMTS)
Visual paired comparison
ment (Have I ever seen this object before?) rather than, as
Object-in-place memory happens with pairs of cues that are repeatedly given across tri-
Scene memory als, a judgment of relative recency (Which of these objects
have I seen most recently?). Relative recency may still be a
598 The Hippocampus Book

factor, even with 300 junk objects, as testing requires that they subjected to hippocampal lesioning, three to amygdala lesion-
be reused within a week or so (Charles et al., 2004a). In any ing, and three to lesioning of both structures. Postoperatively,
event, this subtle procedural change successfully engaged a the animals that had lesions of the hippocampus (but includ-
brain system that operated in the domain of long-term, rather ing the caudal perirhinal cortex and parts of the parahip-
than short-term, memory. Additionally, the use of lists of pocampal cortex because an aspiration technique was used)
items appeared to make the task analogous to the list-learning and those with lesions of the amygdala (including parts of the
tasks so widely used in human studies. It is, perhaps, a slightly rostral perirhinal cortex) all relearned the task quickly. Those
unrealistic model of human list learning because the stimulus with the combined lesiona lesion analogous to the medial
materials usually presented to people (e.g., words or word- temporal lobectomy in patient H.M.were severely impaired.
pairs) are items with which people are already familiar. The Mishkin (1978) also found that once criterion levels of per-
human test is not whether the word cat has ever been seen formance had been reached by all groups with a short mem-
before, but if this word appeared in the list presented by the ory delay the combined lesion group showed pronounced
experimenter several minutes earlier. Thus, in the human list-length and delay-dependent impairments in recognition
experiment, word lists are context-specic, a distinction not memory (Fig. 136C). The other groups performed well
thought to matter at the outset but, as we shall see later, indistinguishably from the nonoperated controls. Thus, the
arguably critical to the anatomical mediation of different group that had been subjected to lesioning similar to that
types of memory. For the monkeys, whether a judgment of applied to H.M. showed strikingly similar memory impair-
absolute familiarity or relative recency is required may not ment. This was a great step forward, and the paper has
matter provided a sufficiently large number of objects are been justly celebrated as a classic paper of twentieth-century
used. However, it most certainly does matter if only two neuroscience (Aggleton, 1999).
objects are used, as Owen and Butler (1981) were to show, Both lesions involved removing cortical tissue adjacent to
because fornix lesions have no effect on remembering which the target structures of the amygdala and hippocampus. At the
of two objects has been seen most recently. This is not sur- time, this was not thought to be of particular signicance by
prising because normal monkeys can do this well only over a most observers (Horel, 1978), and the results were long
few seconds, and they use short-term memory to do so; the described by Mishkins group and by almost everyone else in
task is impaired by a combined orbitofrontal and temporal the eld as lesions of the hippocampus and of the amyg-
stem lesion (Cirillo et al., 1989). Thus, fornix lesions do not dala, respectively. The pattern in the ndings suggested that
interfere with short-term memory but may affect long-term these two structures played a parallel role in memory (which
memory. might yet be dissociated by other tasks), the dual circuit idea
Another, more important difficulty with DMTS relates to emerging as part of a theory of memory proposed a few years
the matching rule. With such a rule, the monkey is rewarded later (Mishkin, 1982).
during the sample trial for displacing a novel object but Delayed nonmatching-to-sample (DNMTS) rapidly
rewarded on the choice trial for displacing what is then the emerged as, in Zolas phrase, the benchmark test of recogni-
familiar object. This potential source of confusion can be tion memory. It was used in numerous studies, notably by the
avoided with several variants of the delayed-matching proto- National Institutes of Health (NIH) group led by Mishkin and
col, but the simplest and most inuential change in procedure colleagues Bachevalier, Murray, Saunders, and others and sep-
was the shift to tasks employing a delayed nonmatching-to- arately by the San Diego group of Squire, Zola-Morgan
sample rule. With nonmatching, the monkey is rewarded for (Zola), Amaral, Suzuki, and others. A distinctive and laudable
reaching for a novel object each time he reaches out: on sam- feature of their work over the 20-year period from 1978 to
ple trials for the single sample object and on choice trials 1998 was partial standardization of the manual testing proto-
when presented with a pair of objects. Mishkin and Delacour cols such that comparisons could be made across studies,
(1975) developed such a task and found that normal monkeys at least within each laboratory. With the sole but impor-
could learn nonmatching more easily than matching, proba- tant exception that the NIH experiments typically involved
bly because of their natural inquisitiveness about novelty. extensive pretraining prior to the lesion whereas the San
Being easier and being later shown to be sensitive to large Diego series always involved surgery prior to training (Zola-
lesions of the medial temporal lobe, delayed nonmatching to Morgan et al., 1982), this standardization was extremely valu-
sample became the task of choice for many years. able. In the case of the San Diego experiments, it made
In an important study, Mishkin (1978) exploited both possible a comparison of lesion groups with rst one and later
characteristics of delayed nonmatching to sample to investi- a second standardized group of unoperated control animals.
gate the effects of hippocampal, amygdala, and combined hip- Research using primates is not undertaken lightly because of
pocampal/amygdala lesions on recognition memory. He the ethical, conservation, and nancial reasons referred to ear-
trained monkeys on the delayed nonmatching to sample task lier (see Section 13.2). Thus, establishing a set of repeatable
until they were performing at 90% correct or better. They were protocols for a test that models several aspects of amnesia was
then divided into four groups, with most undergoing bilateral an important step forward. If there was a weakness of this
surgery to the medial temporal lobe; three monkeys were left standardization, it was that different experimenters were
as normal controls. Of the nine operated monkeys, three were inevitably involved in testing the two control groups and the
 Theories of Hippocampal Function 599

various lesion groups over the years, a practice that would not Suzuki et al., 1993). However, the inference about other
ordinarily be accepted in research on rodents. The experi- modalities is insecure. The available data on auditory DNMTS
menter was typically not blinded with respect to the lesion indicates no impact of perirhinal lesions (Fritz et al., 2005), a
created in the animal, whereas this also is common practice in nding also observed in dogs (Kowalska et al., 2001). The data
the best laboratories working with rodents. However, the of Fritz et al. (2005) led to the intriguing speculation that lan-
robustness of the task and the tight variability of the forget- guage may be unique to humans not only because it depends
ting functions with memory delays make it unlikely that these on speech but because it requires long-term auditory mem-
weaknesses seriously limit its validity, even if it slightly injures ory, although, as they pointed out, this appears to be contra-
their elegance. dicted by eld studies showing their ability to recognize alarm
The outcome of studies conducted throughout the 1980s calls that signal predators (Seyfarth et al., 1980). To our
and early 1990s indicate that after the combined lesion knowledge, the impact of medial temporal lesions on olfac-
(lesions that include the underlying cortical structures of the tory recognition has not been reported in primates (T. Otto
posterior entorhinal cortex, perirhinal cortex, and parahip- and M. Munoz, personal communication), partly owing to the
pocampal gyrus), a DNMTS decit is observed whose main difficulty of making olfactory cues salient for monkeys (M.
characteristics can be summarized as follows (Box 133): Baxter, personal communication). Fourth, the effects of dis-
Working through these numbered points in turn, the fol- traction in DNMTS are analogous to what happens with
lowing additional comments and qualications should be amnesic patients. They can retain small amounts of informa-
noted. First, despite the use of different monkey species and tion for extended periods when not distracted but forget more
different surgeons, experimenters, and WGTA designs, bilat- rapidly when their attention is diverted (Zola-Morgan and
eral damage to the medial temporal lobe similar to that pro- Squire, 1985)although interaction between distraction and
duced in the original study of Mishkin (1978) always causes the effects of a lesion is not always present (Zola-Morgan et
a memory decit in DNMTS (Squire et al., 2004). It is, how- al., 1989). Fifth, there is no recovery of function; a decit is
ever, important to stress that this does not mean that the still seen when memory testing is conducted up to 4 years after
hippocampus alone is involved in recognition memory, as surgery as well as when it is carried out shortly after surgery
the lesions that were used initially typically encompassed (Zola-Morgan et al., 1986). These ve characteristics of
extrahippocampal structures. Second, performance is good at DNMTS performance after large medial temporal lobe lesions
short delays but appears to decline monotonically as the time map neatly onto several features of the amnesic syndrome. It
between sample and choice is lengthened (Mishkin, 1978; is not without reason that object recognition memory, as
Zola-Morgan and Squire, 1985; Overman et al., 1990; Alvarez measured in DNMTS, was championed as an animal model
et al., 1994a; but see Ringo, 1991, 1993). Securing this claim of amnesia (Squire and Zola-Morgan, 1991) and even
involved the development of computer-automated systems for declared a millennial achievement in neuroscience (Kandel
presenting stimulus material to get around the difficulty that et al., 2000).
there was always a small delay in the WGTA when the screen
separating the monkey and the experimenter was raised and Other Tasks Used in a Comprehensive Test Battery
lowered. Notwithstanding this technological improvement in to Explore Declarative Memory in Primates
testing procedure, there was an intense if somewhat technical
debate about delay-dependence, an issue we shall come to Although DNMTS became the most widely used task, it was
shortly. Third, a decit occurs in the visual and tactile modal- complemented by other tasks designed to probe different
ities, and by inference it is probably multimodal. For example, types of declarative memory. During the early years, Squire
monkeys tested in the dark so they can only feel the sample and Zola-Morgan often used a standardized test battery
and choice objects also show a memory decit following consisting of several of the tasks shown in Box 132. Large
medial temporal lobe lesions (Murray and Mishkin, 1983; medial temporal lobe lesions cause a learning impairment in
the concurrent object discrimination task involving the inter-
leaved presentation of eight pairs of objects of which one
Box 133 member of each pair is consistently rewarded, and the delayed
Delayed Nonmatching to Sample Reveals retention of object discriminations (Zola-Morgan and Squire,
an Important Role for the Medial Temporal 1985). Decits in the delayed response task are sometimes seen,
Lobe in Recognition Memory but they are capricious. This is because lesioned monkeys have
no decit in short-term memory and even normal monkeys
1. The decit in DNMTS after large medial temporal lobe can tolerate only short memory delays before they become
lesions is robust and repeatable. confused about whether they are remembering the present or
2. The decit is apparently delay-dependent.
a previous trial. This is never a problem for DNMTS because
3. The decit may be independent of modality.
of the use of trial-unique cues.
4. The decit is exacerbated by distracting stimuli presented
during the delay interval. Each of these and other tasks in the battery is held to be
5. The decit is enduring. declarative in the sense that, with only slight suspension of
disbelief, they can be seen as analogous to tasks given to
600 The Hippocampus Book

Test battery and different MTL lesions formance) using a linear scale. Ringo argued that statistical
considerations require an arcsine or other transformation for
such data, particularly if scores come close to 90% correct as
1.0
they often do at short memory delays. His reanalysis of several
0.5 published DNMTS studies using such a transformation led
mean z score

him to conclude that medial temporal lesions may not, in fact,

0.0 cause delay-dependent forgetting. He suggested that apparent
delay dependence in several studies could arise because lesion-
-0.5
induced decits are concealed at short memory-delay inter-
-1.0 vals through use of the linear scale. For example, controls may
be scoring 92% correct and lesioned monkeys 90%per-
-1.5 formance levels that do not differ statistically when the vari-
NC H H H PR H ability is computed using the entire data set. However, an
(RF) (ISC) + + +
EC PH EC arcsine transformation (which stretches the scale as values
+ + approach either 0 or 100%) may show a different picture, as
type of lesion PH PR Ringos reanalysis of several data sets indicated. This is a
+
PH
cogent argument, although it did play down the overlap of
performance by lesioned and control monkeys at short delays
Figure 137. Pooling data across the test battery. The results in some studies, as pointed out by Alvarez-Royo et al. (1992).
of several tasks of the San Diego test battery were pooled using Its importance was nevertheless taken on board. It spurred the
z-scores to reveal a graded decit proportional to the extent of
development of, or at least the better appreciation of, auto-
the medial temporal lobe lesion and the way it was made. EC,
mated versions of the WGTA that permit the presentation of
entorhinal cortex; H, hippocampal lesions; ISC, ischemic; NC,
normal control; PH, parahippocampal cortex; PR, perirhinal very short sample-choice intervals. Under these conditions,
cortex; RF, radiofrequency. (Source: Zola-Morgan et al., 1994.) no lesion-induced decit is observed in the rate of acquisition
of the DNMTS task at zero delay, with control and lesioned
animals reaching the same performance asymptote. An unam-
people. For example, the concurrent object discrimination biguous delay-dependent decit still pops out at long delays
task can be thought of as analogous to learning a list of words, (Overman et al., 1990; Alvarez et al., 1994). Ringos concern,
the delayed retention of object discriminations to remember- however, should not be forgotten, as ndings obtained more
ing a passage of prose, and so on. Moreover, when the data recently with neurotoxic lesions of structures within the hip-
from all of tasks used in the San Diego test battery are com- pocampal formation add a further note of caution. To claim a
bined across experiments by averaging z-scores (reecting rel- differential effect on memory rather than perception, it is
ative statistical variability rather than absolute values), a clear essential to secure the statistical interaction between group
decit is observed after lesions of the medial temporal lobe and memory delay. This is because a lesion that impaired per-
that, with some exceptions (Baxter and Murray, 2001b), is ception would be expected to have as large an effect at short
nicely graded with respect to extent of damage (Zola-Morgan memory delays as at long onesthe animals being just as
et al., 1994) (Fig. 137). However, the test battery has gradu- poor at seeing the objects in front of them at both intervals.
ally fallen out of favor, partly because components of it are This is the gold standard for establishing that the performance
now thought to depend on other brain structures. For exam- of groups differs across the domain of long-term memory.
ple, the decit in the concurrent object discrimination task Many studies have aspired to this standard; the truth is that
after large medial temporal lobe lesions is probably due to few primate DNMTS studies have reached it.
damage to area TE (Buffalo et al., 1998b), and this same task A separate concern about the test battery is that the psy-
involves habit learning in humans rather than declarative chological justication for regarding tasks in the test battery as
memory (Bayley et al., 2005a). declarative is not always clear, and in some cases the argu-
ment borders on the circular. We have just noted that concur-
Adequacy of These Tasks rent discrimination learning might be thought akin to a
as Models of Declarative Memory person learning a list of words. The trouble is that there are
lists and there are lists. For example, Squire (1992) claimed
One qualication of this ostensibly tidy picture was an argu- that the concurrent discrimination task in which eight pairs of
ment rst posed by Ringo to the effect that the delay depend- objects are presented a number of times during a single test-
ence of the DNMTS decit may depend on the choice of ing session (a task developed originally by Moss et al., 1981) is
numerical scale with which performance is scored (Ringo, declarative, whereas a 20-pair concurrent task in which each
1991, 1993). This is an important technical point. The out- pair of objects is presented only once per day is not (Malamut
come of a single trial in DNMTS is the monkey choosing the et al., 1984). The basis for these assertions was that the former
correct or incorrect object. It is therefore usual to plot per- was impaired by medial temporal lobe lesions, whereas the
formance on a scale from 50% (chance) to 100% (perfect per- latter was not. Clearly this circularity, if that is all there is to it,
 Theories of Hippocampal Function 601

is unsatisfactory. In fairness, there is more to it, namely that trained to do, not because it has in any sense remembered
many behavioral tasks can be learned using different strate- that doing so actually leads to anything. Murray and her col-
gieseven those of ostensibly the same logical structure (task leagues at the NIH laboratory have used this reinforcer-deval-
ambiguity being one of the key conceptual issues discussed in uation procedure to dissociate fact and habit learning in
Section 13.2). Varying the pattern and temporal spacing of the studies that have revealed that excitotoxic damage to the
trials may bias animals toward or away from using a declara- amygdala disrupts the ability of animals to change their inter-
tive strategyan example of a quantitative change having a nal representations of reinforcement value (Malkova et al.,
qualitative effect. In the case of concurrent object discrimina- 1997). Stimulusstimulus associations can therefore be repre-
tion tasks, normal monkeys might be able to remember the sented in a manner that might be thought of as a fact.
fact that a particular object was paired with reward for a However, neither the hippocampus nor the rhinal cortex
period of time despite the interference produced by the other seems to be a brain structure mediating reinforcer devaluation
eight interposed problems with different objects. Provided (Thornton et al., 1998).
another trial of the rst problem comes along soon enough, As an aside, exclusion of the amygdala from a revised
they could then use their fact-based declarative memory medial temporal lobe memory system may also be mis-
to solve the problem. However, if the intertrial interval is long, taken because there are aspects of recognition memory that
as with the 24-hour delay task, even the normal monkey do seem to involve stimulus value. Seeing old friends
may nd a declarative strategy less reliable. It might then fall again, such as colleagues at an annual scientic gathering,
back on a stimulus-response strategy, by which is meant the invokes great pleasure. We all witness and experience this
gradual strengthening of a disposition to reach for the on the rst day of such a meetinga curious pleasure as it is
rewarded object and not for the nonrewarded object, rather all too often followed by heated debate in the lecture theater
than explicit memory of a prior object and reward episode. with these self same people. Although this emotional charac-
Because this nondeclarative strategy, according to the theory, teristic of stimulus familiarity may be mediated by the hip-
is also available to the lesioned monkey, no difference in per- pocampus and/or the perirhinal cortex, it seems unlikely. It is
formance would be expected between groups. Interestingly, tempting to speculate that there may be role for the amygdala
people generally use a declarative strategy to learn lists, irre- as well.
spective of how long the list and how many trials are presented
per day, although, as already noted, people also learn the con- 13.3.3 Domains of Preserved Learning Following
current object discrimination task in a habit-like manner Medial Temporal Lobe Lesions in Primates
(Bayley et al., 2005a). Debate about the interpretation of tasks
rst published 20 years ago may seem like a historical sideline, When constructing an animal model of amnesia, it is impor-
but it is not. For the issue of strategy used versus logical tant to model spared memory as well as impaired memory, of
structure is a key one and, as we shall see, has reemerged as an which Gaffans (1974) use of an associative task was an early
issue with respect to the memory strategies mediating example. Historically, systematic analysis of the issue of mul-
DNMTS. tiple types of learning and memory and their differential sen-
The way to avoid circularity is to come up with independ- sitivity to different lesions began in the rodent literature (see
ent psychological evidence that a particular strategy is being Sections 13.4 and 13.5). It was, however, not long before the
used or that performance on one task correlates with that idea was taken up in studies with primates beginning with
on another and allow this evidence to help guide denitive tasks involving motor skill. For example, Zola-Morgan and
predictions from the theory. In the case of concurrent object Squire (1984) found that control and medial temporal lobe
discrimination learning, one way might be to use the reward- (MTL)-lesioned monkeys could learn to thread a lifesaver
devaluation procedure as originally developed in experi- candy (which contains a central hole) along and then off a
ments on rats (Adams and Dickinson, 1981). If the association thin wire. Eating the candy was the monkeys reward. With
between reaching for an object and securing reward is learned experience, normal and lesioned monkeys got steadily faster,
as a fact, pairing the reward with a toxin such as lithium and they learned at an equivalent rate. Presumably, the
chloride should render it less palatable and so make the ani- lesioned monkeys did not remember doing the task from one
mal less inclined to reach for the object. This is because day to the next or recognize the apparatus when it was shown
the animal would, in some sense, infer from his propositional to them again, but they threaded the lifesaver nonetheless. If
knowledge that reaching for the object leads to the reward, only we could ask them about their mental experience in
and thus seeing the object should retrieve a memory repre- such a situation!
sentation of the reward and that it is now no longer worth Perceptual skills have also been studied in monkeys using
having. However, if the animal had merely learned the habit pattern discrimination tasks (Squire and Zola-Morgan, 1983).
of reaching for the object as a disposition (and learned These are slowly learned discriminations in which two similar
it because it was repeatedly paired with reward during train- patterns (such as alpha-numeric characters) are differentially
ing), reward devaluation would not have any immediate paired with reward. Learning typically takes place slowly, over
effects on performance. The animal would reach for the hundreds of trials; and with one exception no decit in learn-
reward because reaching for rewards is what it had been ing is seen following even large MTL lesions or after
602 The Hippocampus Book

hippocampus-specific lesions. The interesting exception Mishkins dual-circuit theory relating to the putative role
noted by Squire is that performance is sometimes poorer over of the hippocampus and amygdala in recognition memory
the rst few trials of the day. He suggested that this difference prompted the question of whether different types of informa-
between groups is due to controls remembering the rst few tion processing occurred in the two routes of his dual pathway
trials in an explicit or declarative way prior to the build-up of (Mishkin, 1982). One prescient idea, based in part on theoriz-
within-session interference and the reliance on learned habits. ing by Mandler (1980) was that recognition memory could
However, more recent experiments, discussed in the critique be based on either a memory of the polymodal features of
later on, challenge the idea that components of the MTL an object or on memory of where the object had been seen
memory system are not involved in learning stimulus-reward before. We may think of this as, on the one hand, distinguish-
associations (Murray and Wise, 2004). ing between features that are intrinsically part of the object
With respect to cognitive skills, that lesioned monkeys can and, on the other, between features that reect the context in
learn tasks such as DNMTS at a normal rate at zero delay which an object is seen. Whereas the former are part of (and so
(Alvarez-Royo et al., 1992) seems to imply that they are able would move with) an object when it is displaced, the latter
to abstract, from the sequence of trial events, the appropriate might not.
rule for performance. This nding is inconsistent with their Two studies began the systematic exploration of this issue.
having any major decit in intelligence. However, although One examined cross-modal transfer of information, and the
the investigation of primate cognitive skills, knowledge, and other examined memory for the place where an object had
metaknowledge (i.e., what they know they know) has been presented. Murray and Mishkin (1985) found that
been considered by those working in a more ecological context amygdala lesions (which included damage to the perirhinal
(Hauser, 2003), it has only recently been considered seriously cortex) caused impaired cross-modal transfer, whereas hip-
by behavioral neuroscientists (Hampton, 2001). Monkeys pocampal lesions (which also damaged the parahippocampal
have not yet been taught the subtle tasks of learning a nite cortex) were with without effect. Monkeys were trained to
grammar or to make accurate weather forecaststasks that sample one of a restricted set of 40 objects in complete dark-
have imaginatively extended the domain of preserved learning ness and then make a choice between the sample and another
in amnesics (Knowlton and Squire, 1993), but the range member of the set in the light. Information acquired in
of laboratory tasks of nondeclarative memory is steadily the tactile modality was sufficient to guide visually directed
expanding. choices accurately, but performance, at even short memory
delays, was selectively impaired by the amygdala lesions.
13.3.4 Selective Lesions of Distinct Components In contrast, Parkinson et al. (1988) found that their hip-
of the Medial-temporal Lobe Reveal pocampal lesions selectively impaired both a place task and
Heterogeneity of Function an object-in-place task, whereas the amygdala lesions had
no effect. In a concurrent place and object-in-place tasks,
Squire and Zola-Morgans development of a test battery was two objects were presented during the sample trial in two
motivated by the reasonable ambition of developing tasks that of three possible locations on the tray in front of the monkey.
are analogues of the verbal and visual memory protocols On the choice trial, these objects were again put in front
on which MTL amnesics are impaired. Combining data from of the monkey. In the place choice trials, the monkey had
these tasks in the form of z-scores to create a single quantita- only to choose on the basis of the locations occupied by the
tive index (Fig. 137) revealed a monotonic effect of lesions objects in the sample trial; whereas for the object-in-place
within the MTL: the larger the lesion, the greater the decit choice trials, the animals had to choose on the basis of
(Zola-Morgan et al., 1994). However, mindful of the uncer- remembering the particular places occupied by particular
tainties surrounding the classic but misleading concepts of objects. It was later established that monkeys with hip-
mass action and equipotentiality (Lashley, 1950), it is not sur- pocampal lesions could remember one place but not two
prising that critics of declarative memory theory are wary of (Angeli et al., 1993).
the notion that the entire MTL functions as a single homoge- The sufficient lesions for seeing these decits are now
neous unit. Indeed, further research has revealed that known to be purely neocortical, but the historically still
restricted lesions of MTL structures cause little or no impair- important point to emerge from these studies is that recogni-
ment in some declarative tasks but do affect others. This is not tion of novelty might be mediated by either intrinsic or
just a matter of graded task difficulty, as double dissociations contextual components that are not disambiguated in the
are seen (as we shall see shortly). Particularly problematic are standard DNMTS test. The distinction echoes the concept of
data suggesting that neither restricted hippocampal lesions recognition being mediated either by a sense of familiarity
nor damage conned to the entorhinal cortex necessarily (intrinsic) or by recollection (contextual recall) (Mandler,
cause an enduring decit in DNMTS. 1980). Recognition by familiarity would involve the monkey
Absolute condence in the benchmark test of DNMTS making its choice because of the two objects confronting him
began to unravel for a number of reasons. One problem has to in the choice test one evokes a sense of familiarity. It could do
do with understanding what the test is really measuring psy- that even if the monkey was unable to recollect the occasion
chologically; another relates to identifying the necessary and when or where it had seen it before. Recognition by contextual
sufficient lesion that impairs it. These two issues are connected. recall, on the other hand, would involve the monkey utilizing
 Theories of Hippocampal Function 603

the cues of the WGTA surrounding him to bring back to tions, Gaffans laboratory in Oxford being one of them, these
mind that of the two objects before him one had been pre- studies remained focused on using the old warhorse DNMTS.
sented before in this context. Although we cannot talk to the Lesions described as being in the rhinal cortex (part of the
monkey about it, we can nonetheless imagine the animal hav- anterior perirhinal cortex) were sufficient to cause severe,
ing a private recollective experience in much the same way enduring delay-dependent impairment of DNMTS irrespec-
that we would do in a similar situation. tive of whether the hippocampus had also been damaged
Unfortunately, the important line of psychological think- (Meunier et al., 1993; Suzuki et al., 1993; Mishkin and Murray,
ing embedded in these ingenious studies was largely over- 1994). Following these important discoveries, Murray and
shadowed by preoccupation with the anatomical implications Mishkin (1998) reinvestigated the role of the hippocampus
of the aspiration lesion technique. Damage to tissue lying and amygdala using excitotoxic lesions. Not only does this
in the entorhinal, perirhinal, and/or parahippocampal cor- type of lesion leave the surrounding cortex unaffected, excito-
tices might have been contributing to poor performance in toxins should not damage bers from the anterior perirhinal
the amygdala- and hippocampus-lesioned animals rather cortex passing through the ventral amygdalo-fugal pathway to
than damage to these target areas. Later studies using more the medial thalamus or the posterior perirhinal efferents pro-
selective lesions have borne this out. For example, we now jecting through the mbria-fornix and posterior thalamus to
know that cross-modal memory is affected principally by the the same nucleus (see Chapter 3). The key nding was that
perirhinal but not the amygdala component of the original monkeys with average lesion sizes of 88% damage to the
lesions (Murray and Bussey, 1999). Similarly, the spatial task amygdala and 73% damage to the hippocampus showed no
appears to be unaffected by neurotoxic hippocampal lesions impairment in DNMTS (Fig. 138A). This comprehensive
(Malkova and Mishkin, 2003). study included varying list lengths and memory delays and,
Studies during the late 1980s revealed, somewhat surpris- following the protocols set by Alvarez et al. (1995), included
ingly, that neurons in the perirhinal cortex were sensitive to delays of up to 40 minutes in a subset of four monkeys. As no
familiarity and relative recency (Brown et al., 1987; Brown and decit was observed, Murray and Mishkin concluded that, in
Xiang, 1998). In parallel, careful anatomical studies began to the MTL, the rhinal cortex is not only necessary but also suf-
focus on the neocortical regions neighboring the hippocam- cient to sustain visual recognition ability (Murray and
pus, such as the entorhinal cortex (Amaral et al., 1987; Mishkin, 1998, p. 6579). This conclusion is entirely compati-
Insausti et al., 1987a,b) and both the perirhinal and parahip- ble with the available electrophysiological and lesion data.
pocampal cortex (Suzuki and Amaral, 1994a, b). This work In passing it should be noted that these ndings rescue part
was accompanied by further lesion studies. With few excep- of the memory circuit for memory originally proposed by

Figure 138. Conicting ndings in studies A Delayed nonmatching to sample (DNMTS; NIH Lab)
of selective lesions of the hippocampus upon
100
delayed nonmatching to sample (DNMTS).
A. Murray and Mishkin (1998) found that con- 90
joint damage of both amygdala and hippocam- C
pus without major damage to surrounding 80 A+H
% correct

perirhinal and parahippocampal cortex had no

effect on DNMTS with memory delays as long 70
as 40 minutes. B. Zola et al. (2000) reported
60
that damage restricted to the hippocampus
made using ischemia, radiofrequency heating, 50
or excitotoxins caused a delay-dependent 10 30 60 120 LL3 LL5 LL10
decit in DNMTS over the same time delays.
delays (s) list lengths
C. Zola et al. (2000) observed a similar decit
in another test of recognition memoryvisual
paired comparison. B DNMTS; San Diego Lab C Visual paired comparison (VPC)

100 70
% time viewing novel stimulus

90 65
% correct

80 C
H 60
70
55
60

50 50
8s 15s 1m 10m 40m 1s 10s 1m 10m
delays delays
604 The Hippocampus Book

Mishkin (1982) by showing that the critical pathway for a cedural differences. One of these was the use of extensive
judgment of familiarity emanates from the perirhinal cortex pretraining by the NIH laboratory. Zola et al. commented
to the medial nucleus of the thalamus, bypassing the hip- that: Training on the rule provides the monkeys with
pocampus itself. extended practice at holding novel objects in memory across
However, in contrast and over memory delays of as short as short delays, which might then make it easier to hold novel
2 and 10 minutes, Beason-Held et al. (1999) found hippocam- objects in memory across the longer delays from which the
pus lesion-induced impairment. Zola et al. (2000) also performance scores are derived (Zola et al., 2000, p. 459). The
reported that restricted hippocampal damage can be sufficient implication seems to be that avoiding preoperative training
to cause impairments in both DNMTS and another test of makes for a more sensitive behavioral assay, a view consistent
recognition memory called the visual paired-comparison with the apparently greater sensitivity to hippocampal lesions
(VPC) task (Fig. 138B,C). The latter task, like certain tasks of the VPC task in which there is no formal training at all. This
developed by Gaffan much earlier (e.g., Gaffan et al., 1984), may be true empirically but is unsatisfying intellectually, as
involves no formal trainingmerely exposing the monkey to there is nothing in declarative memory theory to explain this
pairs of black and white line drawings on a computer screen dependence. Indeed, we should be wary of a logical inconsis-
(Bachevalier et al., 1993). After looking at these drawings for tency with respect to mission of this enterprise, namely, mod-
about 30 seconds, monkeys spontaneously direct their eye eling amnesia. Amnesic patients have extensive experience
movements to a new drawing presented on the screen beside recognizing things prior to becoming amnesic but, according
one of the old ones (a phenomenon rst observed experimen- to Squire, have a recognition decit even if their brain damage
tally in studies of human infant perception (Fantz, 1964). is restricted to the hippocampus. Why should the animal
Many pairs of such stimuli can be presented, one after the model be any different? Furthermore, why did the San Diego
other, to build up a prole of the animals ability to detect laboratory insist on using a protocol so different from what
novelty spontaneously. they were doing with their patients?
Drawing together data from experiments over 10 years A major of focus of attention has been on the extent
using a variety of lesion techniques and a large number of of damage to hippocampal and extrahippocampal regions.
monkeys, Zola et al. (2000) claimed that ischemic, radiofre- Growing awareness of the importance of the extent of lesion
quency, and excitotoxic lesions of the hippocampal region all damage in individual animals led to the development of MRI-
cause a modest but signicant decit in recognition memory based evaluations of the locus and extent of damage (Malkova
in both DNMTS and VPC. The decit is apparent at delays as et al., 2001). Suitably calibrated, these evaluations offer the
short as 15 seconds, with larger effects seen (at least on opportunity of providing accurate, noninvasive estimates of
DNMS) at longer delays of 10 and 40 minutes. Consistent brain damage in advance of extensive postoperative testing
with the declarative memory theory, they concluded that the (an ethically desirable development apart from anything else).
integrity of the hippocampal region is essential for recogni- The more usual postmortem histology has oftenno doubt
tion memory. The extent of hippocampal damage varied sub- to the dismay of experimenters (after years of training an
stantially among these three studies, ranging from 33% to individual animal)revealed wide variations in locus and
62% in the San Diego animals but, paradoxically, averaging size after lesions made by various techniques. There was,
73% in Murray and Mishkins NIH study that found no for example, an average of 18% extrahippocampal damage (to
decit. However, Nemanic et al. (2004) came to a somewhat the parahippocampal cortex) in the monkeys trained by Zola
different conclusion after a similar comparison of DNMTS et al. (2000), who claimed a specic hippocampus lesion-
and VPC. Their data point to a substantial decit in both tasks induced decit in DNMS; but there was only 4% extrahip-
after perirhinal lesions but no decit in VPC in hippocampus- pocampal damage in a study by Murray and Mishkin (1998),
lesioned monkeys until longer delays are interposed and only who claimed the opposite. To be fair, despite incursion into
a slight decit in DNMTS at a memory delay interval of 10 this neighboring brain area, there was no evidence of any
minutes. Bachevalier raised the important qualication about within-group correlation between performance and the
her own study (Nemanic et al., 2004) that the average lesion extent of parahippocampal damage in the Zola et al. (2000)
size in her hippocampal group was only 43.5% (this being study (damage ranged from 0% to 46%). Not much comfort
made with ibotenic acid), but this does not appear to be sub- should be drawn from this, however, as the study did not
stantially different from the mean lesion size prevailing in the report any correlation between the extent of hippocampal
San Diego studies that also used excitotoxins. damage and the recognition performance (ranging from 13%
to 76% across all 14 animals tested). In contrast, Murray and
Comparison of Conicting Studies Reveals Subtle Mishkins ndings indicated, paradoxically, a positive correla-
Differences in Lesion Size and Methodology tion between the extent of hippocampal damage and per-
formance at the longer delays: the greater the hippocampal
How is the discrepancy between these experiments to be damage, the better the recognition performance.
explained? Some reports have indicated that apparently Baxter and Murray (2001b) took this curious inverse rela-
restricted hippocampal lesions did impair recognition tion further with a meta-analysis of these three studies of
memory and others that they did not. There were several pro- DNMTS in monkeys with restricted hippocampal lesions.
 Theories of Hippocampal Function 605

Using an optimum d statistic (which takes into account dif- most effectively mediated by the hippocampus, animals with
ferences in the performance of control monkeys across stud- large hippocampal lesions would most likely use a familiarity
ies), their analysis revealed an inverse correlation between the strategy and so rely on their intact perirhinal cortex to make
loss in d and percent damage to the hippocampus (Fig. 139). judgments of prior occurrence using the neuronal ensembles
Zola and Squire (2001) argued that this meta-analysis was identied in the pioneering work of Brown and colleagues
invalid because it failed to partial out the potential inuence (Brown et al., 1987; Brown, 1996). Conversely, if recollection
of various factors that differed across studies, other than remains feasible and is ordinarily preferred, animals with
lesion size, such as whether pretraining had been given, the small hippocampal lesions may automatically continue to use
way the lesions were made, and the delays used to assess mem- their damaged hippocampus, attempt a recollection method
orya caution that ironically did not prevent Zola himself of solving the problem, but perform less well than control ani-
from generating a z-score statistic to characterize a wide range mals precisely because this structure is damaged. A negative
of tasks (as in Fig. 137, p. 600). Baxter and Murray (2001a) correlation between performance and size of lesion would
conceded the weakness of pooling data across studies that then emerge. It is not unreasonable to suppose that extensive
used slightly different training protocols but asserted that pretraining may also predispose animals to use a less effortful
even when factors relating to lack of task identity are partialed familiarity strategy preferentially, thereby accounting, at least
out a nonparametric analysis of the inverse relation between in part, for the persistent failure to see restricted hippocampal
loss of d and lesion size remains signicant. As in Ringos ear- lesion effects on DNMTS in the NIH laboratory but their
lier analysis of delay dependence, the use of a d statistic was presence in the San Diego laboratory (that generally did not
helpful, though by no means critical, as the same result per- use preoperative training). The situation is laden with irony, as
tains when raw percent correct difference scores are used Squire and colleagues have been uncertain of any simple map-
(M.G. Baxter, personal communication). Moreover, each of ping of the recollection/familiarity distinction onto structures
the pooled studies secured a trend or signicant inverse rela- in the MTL (Squire et al., 2004) (a view endorsed by Stark in
tion on their own. Chapter 12), yet it could be precisely because the unoperated
Baxter and Murrays meta-analysis is an empirical observa- monkeys in San Diego used recollection to perform DNMTS
tion, not an explanation about why there might be an inverse that hippocampal decits were seen.
correlation. One possibility is that residual hippocampal tissue The available data suggests that the VPC test may be a more
adjacent to the lesion produces aberrant neural activity that hippocampally sensitive test of recognition memory than
disrupts neighboring brain regions. There is, however, another DNMTS (Zola et al., 2000; Bachevalier et al., 2002; Nemanic
possible reason for this paradoxical correlation that deserves et al., 2004). Unfortunately, we do not yet have a princi-
careful discussion. Once again, like the debate about concur- pled understanding of why. Manns et al. (2000b) have shown
rent object discrimination learning (above), it is the old prob- in humans that performance on VPC is predictive of subse-
lem of task ambiguity. We saw earlier that DNMTS is quent recognition memory performance in a standard two-
amenable to two distinct strategies: a familiarity strategy and alternative forced-choice test, whereas performance in
a recollection strategy. If recollection is either exclusively or perceptual priming (a measure of perceptual uency) is unre-
lated to recognition performance. This is helpful because it
suggests that, in humans, VPC really is in the declarative
Figure 139. Meta-analysis of DNMTS. Systematic comparison of domain. Subjects do have some awareness of having seen
data from several laboratories reveal the paradoxical inverse relation one of two stimuli before. However, this analytical study does
between DNMTS performance and hippocampal lesion size (Baxter not directly address the familiarity versus recollection issue.
and Murray 2001). The various symbols represent data from differ- The problem we face is that monkeys may be using explicit
ent studies: squares, Murray and Mishkin (1998); triangles, Beason- recollection to direct their gaze, or they may be using stimulus
Held et al. (1999); circles, Zola et al. (2000). familiarity (or perhaps both). If VPC is a pure familiarity
Paradoxical relationship between DNMTS and lesion size task uncontaminated by recollection, it is unlikely to prove to
be of continuing value for understanding the role of the pri-
mate hippocampus in more complex aspects of event memory
-0.5 good
in which recollection is denitively engaged (e.g., cued recall
in which one stimulus brings to mind another with which it
has been associated). Bachevalier suspected that it entails little
0.0
performance
loss in d

more than familiarity but went on to suggest that its greater

sensitivity may reect the greater role that novelty plays with
0.5 respect to all the stimuli in the task.

In VPC, animals are passively exploring two-dimensio-

1.0 poor
nal back/white novel stimuli, not actually memorizing
0 20 40 60 80 100 the sample to select a future response (i.e., incidental
% damage learning). It is presumably more ecological for
606 The Hippocampus Book

monkeys (and humans) passively witnessing a new not judgments of absolute familiarity. However, such a task
event to keep a trace (however weak it is) of the whole might only work in incidental mode. Deliberate training
event, because anything can later prove to be behav- with multiple objects in the different contexts, or objects over
iorally relevant (i.e., the stimulus, its elements, and its multiple trials, may engage conditioning processes that utilize
spatial and temporal contexts). This incidental encod- congural cues mediated by neocortical circuitry (see Section
ing could favor the formation of conjunctive represen- 13.5). However, if this or another appropriate protocol could
tations not only of the different elements of the sample be developed, disagreement about whether the hippocampus
but also of its location and contexts. (Nemanic et al., is or is not involved in recognition might then move forward
2004, p. 2025) toward discussion about qualitatively different types of recog-
nition memory. Such experiments would start the process
We come back to this thoughtful comment in Sections 13.5 of fractionating declarative into different kinds of proposi-
and 13.6, where the argument is presented that incidental tional knowledge, just as the visual system has been frac-
encoding of stimuli does not necessarily commit an animal to tionated into different streams of processing. Precisely such
being able to make only familiarity judgments about them context-specic recognition experiments are already under-
and that the formation of conjunctive representations, even if way using rodents (Dix and Aggleton, 1999; Eacott and Nor-
encoded automatically, is an essential building block of recol- man, 2004). In primates, context-specic discrimination tasks
lection. Whatever the psychological basis of the VPC task, pro- have been explored (Dore et al., 1998), showing decits after
ponents of declarative memory theory do, nonetheless, like it. neurotoxic hippocampal lesions but not yet recognition
However, there is a lurking suspicion of circularity in this memory.
attraction; unlike DNMS, the task is reliably impaired by hip- The era of primate lesion experiments on recognition
pocampal lesions at reasonably short memory delays. memory using DNMTS has probably drawn to a close. This is
Again, what is needed is an information-processing analy- partly because of the ambiguities discussed above but also
sis of specic tasks that is independent of any tests of their because the individuals particularly interested in these issues
sensitivity to brain damage. DNMTS and VPC both suffer have moved on and a new generation of primate researchers is
from the problem that there are at least two ways in which a tackling other issues. Some primate lesion experiments are
monkey might perform the test. We need either new, less underway using new tasks, and others are focusing on single-
ambiguous tests of memory or new behavioral assays to estab- unit and multiple single-unit recording during memory tasks
lish when an animal is performing these ambiguous tasks one (Suzuki and Eichenbaum, 2000; Squire et al., 2004). One com-
way or the other. Assuming that these could be developed, we mon feature of these experiments is abandoning the notion
could then return to the main task of mapping the cognitive that recognition and association are likely to be fundamentally
process onto underlying brain systems and networks. Given different; both may be associative processes. Buckmaster et al.
that stimulus familiarity suffices to solve the DNMTS task, (2004) have developed tests of paired-associate learning that
instances where excitotoxic hippocampal lesions have no are more sophisticated than merely pairing an object with
effect (Murray and Mishkin, 1998) may be because the task a reward. The animal must learn that object A goes with object
has been set up to encourage no more than a judgment of B. Tests of transitive inference and delayed spatial recall are
familiarity (extensive pretraining?). By the same token, also being added to the arsenal of tests with which the multi-
instances where a decit is seen (Zola et al., 2000) may occur ple types of memory in the primate brain will eventually be
because the control animals enjoy the benet, at least on some uncovered. Similarly, Suzuki and her colleagues have exam-
trials, of explicit recollection. In such cases, the hippocampus ined the hippocampal single-unit correlates of paired-associ-
may provide the processing necessary for remembering the ate learning in monkeys, for new pairs and for well-established
object or its image on a computer screen in its spatiotemporal pairs (Wirth et al., 2003; Yanike et al., 2004). We consider the
context. application of new behavioral tests of declarative memory
If this analysis is correct, a novel prediction is that context- later in our discussion of relational-processing theory (see
specic recognition memory is impaired by discrete hip- Section 13.5).
pocampus damage in monkeys. A possible experiment might
be one in which the monkey would be shown object A once in 13.3.5 Remote Memory, Retrograde
context 1 and a short while later object B in context 2. After a Amnesia, and the Time Course of
memory delay, it would then be presented with different types Memory Consolidation in Primates
of recognition judgment. Numerous novel objects would be
used across of a series of trial triads. Some tests would require Given the uncertainties of testing remote memory in humans
no more than an absolute familiarity judgment: Has either of (see Chapter 12), there has been interest in examining
the two objects been presented before? Others would involve retrograde amnesia in animals. The value of doing this is
making a context-specic judgment: Can the monkey indicate that studies can be done prospectively. The training experience
that object B in context 1 (or A in context 2) is a novel condi- of laboratory animals can be accurately controlled, with no
tion? On the analysis just presented, discrete hippocampal ambiguity about precisely what eventsand when or
lesions might affect context-specic recognition memory but wherehave occurred prior to a lesion. The experimenter
 Theories of Hippocampal Function 607

knows and does not have to rely on the uncertain testimony suggests that a memory consolidation process must have been
of relatives. There are, however, looming difficulties in the taking place in normal animals that was interrupted by the
use of animals for such studies, one being the critical episodic/ hippocampal/parahippocampal lesion.
semantic distinction. Other primate studies failed to obtain positive evidence for
Zola-Morgan and Squire (1990) sought evidence for grad- consolidation (Dean and Weiskrantz, 1974; Salmon et al.,
ual memory consolidation by rst teaching monkeys a series 1987; Gaffan, 1993), but there are several reasons why no
of 100 object discrimination problems. They were divided temporal gradient may have been observed. Inferotemporal
into ve sets of 20 problems scheduled at intervals of 2, 4, 8, lesions, as studied by Dean and Weiskrantz, could have dam-
12, and 16 weeks prior to creating aspiration lesions in the aged the actual site of memory storage in the neocortex, mak-
animals hippocampus and surrounding parahippocampal ing it impossible to see the effect on remote memory of any
cortex. Two weeks later they were retested on each of the 100 putative process of consolidation orchestrated by the MTL.
problems by presenting pairs of discriminanda just once (to The study by Salmon et al. (1987), also using large MTL
examine retention uncontaminated by new learning). The lesions, showed little forgetting in control animals and very
lesioned monkeys were impaired relative to controls on the poor performance in the lesioned animalsconcerns about
problems learned shortly before surgery, but the two groups ceiling and oor effects that led directly to the design of the
performed at a comparable, above-chance level for problems later experiment with more restricted lesions. Gaffan (1993)
learned 12 or 16 weeks earlier (Fig. 1310). examined picture memory with various retention intervals
As just described, the results do not necessarily require ref- prior to the administration of fornix lesions, but it had several
erence to a concept such as memory consolidation they curious features: First, a different number of pictures were
could reect no more than damage to a storage site and dif- presented in the set just before surgery than at the earlier time
ferential rates of forgetting in control and lesioned animals point; and second, a retention test was given just before
(Fig. 1310, pattern in panel B1). Of greater signicance are surgery that could have reminded the animals of the pictures
the within-subject comparisons. The controls did best on the and so altered the extent to which they can be considered
most recent problems and worst on the problems learned ini- as exclusively belonging to the recent or remote set.
tially, a pattern that reects gradual forgetting over time. In Indeed, demonstrable improvement in the performance of the
contrast, the lesioned monkeys did not fail at both training- control group between the two retention tests strongly sug-
lesion intervals (Fig. 1310, data pattern in panel B2), but gested that reminding altered the effective age of the memo-
actually did worse on problems learned 2 weeks before surgery ries. Although this weakens the force of Gaffans study, the
than those learned 12 weeks earlier (Fig. 1310, the true pat- possibility that recall can induce re-storage of information
tern, in panel B3). This dual pattern of performanceforget- and so alter trace strength should not to be ignored.
ting in controls but gradual improvement in the lesion The study by Zola-Morgan and Squire (1990) is therefore
groupis critical to the interpretation of the data. It strongly widely considered the denitive study of gradual time-

Figure 1310. Retrograde amnesia and declarative memory theory. Performance by the lesion group is always poorer than that of con-
A. Average performance during a single postsurgery probe trial for trols, although they may display a shallower gradient of forgetting.
each of 100 object discrimination problems learned earlier. The data B2. The lesion causes disruption of retrieval. Performance is poor at
are plotted as a function of the time interval between training and all time intervals. B3. The lesion causes selective disruption of long-
the subsequent lesion. Problems learned 12 weeks before surgery are term consolidation. Performance shows an inverted-U shape, being
remembered better than those only 2 weeks beforehand. (Source: better for problems when there has been time for consolidation
Zola-Morgan and Squire, 1990.) B. Data to be expected on various prior to the lesion but, like normal controls, also forgetting over
models of how lesions might affect storage sites or memory consoli- time. The exact form of the memory gradients is critical to the
dation. B1. The lesion causes partial disruption of the site of storage. interpretation.

A Memory for object discriminations B Theoretical models

B1 B2 B3
90
partial
storage retrieval disruption of
80 failure consolidation
damage
performance
% correct

70
C
H
60 chance
recent remote
50
2 4 8 12 16
e
re nt
ot
ce
m

learning-surgery interval (weeks)

re
608 The Hippocampus Book

dependent memory consolidation in primates, although there dependent changes in 2-deoxyglucose utilization in the hip-
are aspects of the results that are troubling. The use of lesions pocampus after learning are also consistent with the idea that
encompassing both the hippocampus and the parahippocam- increases in hippocampal activity occur for a limited period
pal cortex raises the question of whether damage restricted to time after learning (Bontempi et al., 1999; Frankland et al.,
the hippocampus would cause a retrograde effect. This is 2004; Maviel et al., 2004b). These changes may constitute a
important given the lack of much retrograde amnesia in physiological and molecular signature of consolidation.
patients R.B. and G.D. (see Chapter 12). A study using mon- There are numerous outstanding issues concerning mem-
keys in whom ischemic and/or neurotoxic lesions were created ory consolidation as envisaged within the declarative memory
could offer a more exact model of the etiology and damage in theory. A key issue is what determines whether new memories
these patients. To our knowledge, this has not been done. The are consolidated or allowed to fade. Most event memories that
use of a within-subjects design is a strength (with respect to humans make automatically during a typical day are lost;
mimicking the human syndrome and allowing modest use of some passive or active selection process must come into play
animals), but it raises the question of whether the repeated act that determines what is retained or discarded (Morris et al.,
of reminding the animal of the testing situation preopera- 2003). Second, once set in motion, is consolidation a gradual,
tively could inuence the memory strength of earlier trained inexorable, largely time-dependent process? Or is it a more
items. Putting a monkey back into the WGTA apparatus to quantal process in which short consolidation episodes occur
learn a new set of problems 4 weeks after it has learned an ear- repeatedly over a longer period of time, perhaps triggered by
lier set could re-activate memory representations of the rst contextual retrieval (Dudai and Morris, 2001)? Both gradual
trained problems even though the specic stimulus items are and quantal ways of thinking about consolidation could result
not presented again. This could happen because the training in gradual temporal functions of retrograde amnesia when
context may be sufficient to activate hippocampus-specic average data are considered, but the underlying mechanism by
pointers. Put simply, being in a place you have been in pre- which individual traces are strengthened might be very differ-
viously reminds you of things that happened there and may ent. Third, how does consolidated information become inte-
make these remembered events more memorable in the grated or interleaved with information already stored in the
future, a process that is more akin to a cyclical reactivation neocortex (McClelland et al., 1995)? Fourth, should a distinc-
and re-storage process than an inexorable time-dependent tion be made between consolidation and semanticization?
consolidation process. Insofar as this did happen in the The former would allow the persistent memory of discrete
within-subjects design, it would particularly affect the prob- events and enable the recovered consciousness and mental
lems learned earliest before surgerythe very ones remem- time travel characteristic of episodic memory (Moscovitch,
bered best by the lesioned animals (a different problem from 1995). The latter would involve an interleaving process
that in the Gaffan study where reminding occurred just before through which the regularities that emerge from successive
surgery). This argument would be more convincing had there similar episodes are abstracted and so be able to add to the
been an improvement in memory for these same early prob- subjects semantic knowledge. The later recall of such infor-
lems in the control group, but the controls exhibited forget- mation need not entail access to spatiotemporal tags
ting over time. here is also the further possibility of all animals (Winocur et al., 2005). Fifth, is there just one type of memory
developing a learning-set over the succession of discrimina- consolidation, or is there a family of distinct processes? Some
tion problems (Murray and Bussey, 2001). What is needed to have argued for a distinction between cellular consolidation
settle this issue is a systematic comparison of training condi- that putatively operates rapidly within single neurons in a sin-
tions that involve repeated reexposure to the context and of gle brain area, and systems-level consolidation involving the
training conditions that do not. Such a study is now most interaction of different brain areas (Dudai and Morris, 2001).
unlikely to be conducted in primates, but fortunately this Finally, what are the indices or pointers, what do they index or
issue has been followed up using rodents. point at, and how is neuronal and synaptic specicity in neo-
Studies of retrograde amnesia in nonprimate species have cortex realized by the signals emanating from hippocampus?
been conducted using social transmission of food preferences, These and related issues constitute an important next step in
cue and context fear conditioning, trace-eyeblink condition- the development of declarative memory theory.
ing, spatial learning, and object discrimination learning tasks
(e.g., Winocur, 1990; Kim and Fanselow, 1992; Bolhuis et al., 13.3.6 Critique
1994; Anagnostaras et al., 1999; see Squire et al., 2001 for
review). Some studies suffer from ceiling or oor effects in The declarative memory theory of hippocampal function has
the data (i.e., performance being so good or so bad across a been immensely inuential and remains the most cited theory
range of intervals that not all studies have both a forgetting of memory in neuroscience textbooks (Kandel et al., 2000;
and a consolidation gradient). However, the general pattern is Bear et al., 2001). Yet, although simplicity is desirable, one
of temporal gradients of retrograde amnesia consistent with a should always remember Einsteins dictum that scientists
gradual process of memory consolidation, with the sole should keep things as simple as possible, but no simpler.
exception being certain spatial tasks that are discussed in Several critics empathize with this view, two describing the
detail later (see Section 13.4). Data showing gradual time- theory bluntly as follows: It is attractive, it is parsimonious, it
 Theories of Hippocampal Function 609

is extraordinarily popular, and it is wrong (Murray and Wise, Tulvingss alternative SPI framework for propositional memory
2004, p. 194). Others, notably Gaffan (2002), go further in
opposing the whole concept of memory systems, particularly Sensory inputs
that of a unitary declarative memory system encapsulated
entirely within the MTL. Whether these and other critiques are Flow of Perceptual representation system
well founded or merely idiosyncratic is a matter of opinion. activated
subsets of
What is clear is that several problems with declarative information
memory theory are now widely recognized. They range Working
Semantic memory
from empirical issues to do with how well the data really do t memory
the theory, many of which have already been discussed, to
conceptual issues such as the status of the taxonomy of mem-
ory systems, the role of conscious awareness in the encoding Episodic memory
and retrieval of declarative information in humans and ani-
mals, and problems surrounding the concept and time scale of Figure 1311. Tulvings SPI model of propositional memory. An
systems-level memory consolidation. alternative framework for the ow of information into and between
memory systems. Information ows into memory systems that
What Is the Status of the Memory Taxonomy? operate in a serial, parallel, independent (SPI) manner (Schacter
and Tulving, 1994). The ow is from a perceptual-representation
The taxonomy depicted in Figure 133 is easy to grasp and system into semantic memory and then into episodic memory.
Activated subsets of these systems interact with working memory
easy to remember. It is a joy to teach, and many a lecture in
(Baddeley, 2001).
research seminars begins with this diagrambut what does it
really mean? Is it meant to be an accurate depiction of evolu-
tionary or logical distinctions between different forms of ing rules. Do they encode, store, and retrieve information
learning and their mapping onto specic structures in the differently and so provide outputs different from those of
human brain? Or is it actually no more than an aide memoire? the rest of the brain? Do their storage mechanisms express
Taking issue with the taxonomy may seem no more than a memory traces differently, with differing patterns of persist-
semantic side show; but if the taxonomy is conceptually con- ence or susceptibility to consolidation? Others are concerned
fused, it is important to reect on this. The term taxonomy that certain types of memory storage may start off in one way
is used in evolutionary biology to denote relatedness, such as (e.g., declarative) but then become another (e.g., nondeclara-
between species, orders, and groups. Clearly, the use of the tive) through repetition in multiple contexts and/or the pas-
term here is not intended to imply that the evolution of mem- sage of time. That is, the ostensibly sharp boundaries of any
ory proceeded rst through some ancient differentation memory taxonomy may neither be immutable nor adequately
between short-term and long-term memory and then on reect dynamic changes in memory representation that occur
through all the binary divisions of the hierarchy. To the con- during the course of learning. However, to be fair to the archi-
trary, it is generally assumed that nondeclarative memory tects of declarative memory theory, these are in part issues for
evolved earliest (Sherry and Schacter, 1987). The term taxon- the future and precisely the sort of topics they have been
omy is being used in a different senseone expressing osten- attempting to address, notably in the many studies character-
sibly qualitative distinctions between types of memory. izing the nature of preserved learning in the presence of
Nonetheless, some critics of declarative memory theory feel amnesia (Squire et al., 1993; Cavaco et al., 2004).
the need to move beyond mere taxonomy to more precise, These computational- and algorithmic-level questions
noncircular statements of what is different (in information- have to be addressed before we can proceed securely toward
processing terms) about the identied forms of memory any mapping onto the neural substrate in which they are
depicted within the framework. Do the distinct nodes of the expressedat the circuit, cellular, and even intracellular levels
taxonomy have different inputs such that they operate on of analysis. As currently drawn, the taxonomy implies that
different types of information? Schacter and Tulving (1994), types of learning and memory at the bottom of the hierarchy
for example, argued for the ow of information into memory can be mapped onto specic brain areas, ranging from regions
occurring in a serial, parallel, independent mannerthe in the MTL to such structures as the striatum and cerebellum.
information passing rst via perceptual representational sys- Matters, however, are unlikely to be so simple. One controver-
tems, then through semantic memory, and on into episodic sial claim, made by Murray and Wise (2004), is that radically
memory (Fig. 1311). Activated subsets of semantic and new concepts of the embryology and anatomical organization
episodic memory constitute the inputs to working memory, of major parts of the primate and rodent brain need to be
thereby creating a set of input-output relations between mem- taken on board by memory researchers, notably ideas most
ory systems quite different from those envisaged in the declar- closely associated with those of Swanson (2000, 2004). One
ative memory theory. Continuing in this vein, we may ask of aspect of this is a shift away from thinking of brain functions
the standard taxonomy (and of Schacter and Tulvings alter- as localized to discrete brain regions or to closely connected
native), whether the memory systems in it use different learn- brain areas such as MTL, toward thinking in terms of recur-
610 The Hippocampus Book

rent loops. It is too early to assess this idea securely. However, which rats traverse runways to displace goal objects that are
one lesson of the last decade of functional imaging research in either familiar or unfamiliar (Rothblat and Hayes, 1987;
humans is that distributed cerebral networks for memory Mumby et al., 1990; Kesner et al., 1993), and even procedures
involving top-down interactions between the medial tempo- for rats using images on computer screens (Gaffan and Eacott,
ral and frontal lobes need to be considered and analyzsed 1995). A continuous delayed nonmatching paradigm, teas-
(Fletcher and Henson, 2001; Miyashita, 2004). ingly called cDNM, uses a go/no-go digging response for the
A last point about the taxonomy is the growing concern is recognition of odors (Otto and Eichenbaum, 1992).
that it does not readily capture the sense that the seamless Spontaneous novel object recognition (NOR), analogous to
control of behavior is almost certainly a matter of the coordi- the VPC protocol for primates, is an increasingly popular par-
nated regulation of numerous brain networks. When learning adigm for recognition memory rst developed by Ennaceur
to drive a car for example, there is perceptual learning, motor and Delacour (1988). After prior habituation to the context of
control, knowledge of facts about road signs, and memory for testing, the animal is confronted by two objects that it investi-
previous similar traffic situations that one may have encoun- gates and explores, though never formally rewarded for doing
tered. The taxonomy is suspiciously silent about how the out- so. The objects are generally identical in a sample phase but
puts of the ostensibly independent forms of memory interact, differ in the memory test phase. Trial-unique goal boxes,
compete, or are coordinated in the brain. Experimental work objects, or smells are used for rodent DNMS, cDNM, and
to date has been devoted largely to devising tasks that dissoci- NOR, with the relative probability of the choice or the extent
ate the psychological and anatomical components of memory of spontaneous investigation of the novel cue serving as the
processing rather than investigating how these entities com- index of memory of prior occurrence. One issue of concern
pete or cooperate. As noted in Section 13.2, new developments has been that not all of the rodent studies have routinely var-
in memory research are turning toward this more synthetic ied the memory delay within the protocol as has been the cus-
goal (Poldrack et al., 2001; White and McDonald, 2002). None tom in the primate work; this is important, as use of only
of these comments and criticisms on their own denitively short memory delays can yield inappropriate conclusions
indicate that the standard memory taxonomy and its mapping (Clark and Martin, 2005). There are also subtle but important
onto brain areas is wrong, but they are grounds for caution. differences of protocol across ostensibly similar paradigms.
For example, some studies of NOR require all subjects to
Inconsistencies Between the Animal Lesion accumulate some minimum period of object exploration dur-
Data and the Declarative Memory Theory ing each trial (e.g., 30 seconds), these trials of necessity then
being of indeterminate duration; others have a set trial dura-
Several lines of inquiry have turned up data that appear incon- tion but then average the normalized exploration scores across
sistent with the declarative memory theory or, at best, are han- subjects. Clark et al. (2000) presented data that favor the for-
dled awkwardly. Much of this is considered later in relation to mer approach, and they thereby set new standards for these
other theories of hippocampal function, notably spatial, con- types of experiment in rats (a move by the San Diego group
gural, and episodic theories. Studies of the neural basis of into rodent work that may have been prompted by a certain
recognition memory in rats and of other forms representa- frustration about the way so many rodent experiments were
tional memory in primates have been conducted as explicit being conducted). Mumby (2001) provided a perceptive dis-
tests of the declarative theory and so are considered here. cussion of many other differences between these and the pri-
Hardly unique in science, there is conict about the data. mate tasks discussed earlier.
Like the proverbial housewives living in medieval tenements Most of the data reported on rats indicates that restricted
who always bickered at each other across the narrow alleyways hippocampal dysfunction (lesions of various kinds, fornix
(they were always arguing from different premises), there is section, intrahippocampal drug infusions) has minimal effect
a persisting conict about the way studies of recognition on these tasks (see Table 1 in Mumby, 2001). In contrast,
memory in rats are perceived. Mumby summarized the situa- perirhinal and postrhinal cortex lesions, including reversible
tion fairly as one in which most investigators are looking out- inactivation using glutamatergic ligands and disruption of
side the hippocampus (Mumby, 2001, p. 159) to explain the cholinergic neuromodulation, cause clear decits (Bussey et
neurobiological basis of recognition memory. In contrast, al., 1999; Warburton et al., 2003; Winters et al., 2004; Winters
Squire et al. (2004) continued to defend the view that damage and Bussey, 2005). There are exceptions to this pattern, but
to the hippocampus does impair recognition memory at long when taken as a whole it is inconsistent with the prediction
memory delays. from declarative memory theory that damage anywhere
A range of techniques have been developed to study recog- within this group of MTL structures, including the hip-
nition memory in rodents that complement the DNMTS and pocampus, should cause a proportionate decit in recognition
VPC paradigms for monkeys. There are several DNMTS pro- memory. Starting with a study by Aggleton et al. (1986) using
tocols in which rats are explicitly trained to learn a matching novel and familiar goal-boxes in a Y maze, a series of analyti-
or nonmatching rule and selectively rewarded on correct cal studies led by Rawlins resolved why hippocampal dysfunc-
choice trials. They include a Y maze nonmatching goal-boxes tion impaired nonspatial working memory in a radial maze
procedure (Aggleton, 1985; Aggleton et al., 1986), tasks in but appeared to have no reliable effect on DNMTS (Rawlins et
 Theories of Hippocampal Function 611

al., 1993; Steele and Rawlins, 1993; Cassaday and Rawlins, Aggleton, 1994; Winters et al., 2004; Forwood et al., 2005)?
1995, 1997). It turned out that the size of the goal box is crit- Lesion size or, more generally, the degree of hippocampal dys-
ical, with small compact goal boxes being treated by rats as function is one possibility. For example, Ennaceur and
discrete objects, whereas larger boxes are treated as spaces Aggleton (1994) found no effect of fornix lesions on NOR at
and so engage spatial and relational encoding (see Sections delays varying from 1 to 15 minutes but dropping to chance
13.4 and 13.5). Hippocampal lesions only affect memory for levels over 4 hours. Other studies using fornix lesions reported
large goal boxes. These ndings have been complemented by similar results, but it should be remembered that fornix
concern that certain protocols for testing nonspatial recogni- lesions in monkeys have little impact on DNMTS and may
tion memory may have cryptic spatial or contextual compo- leave many aspects of MTL function intact. Perirhinal lesions
nents. For example, Nadel (1995) was the rst to point out have consistently been observed to impair NOR (Bussey et al.,
that the decit in DNMTS at long memory delays in certain 1999; Murray and Bussey, 1999). Consistent with these nd-
primate studies was confounded by the monkeys being, for ings but drawing a very different conclusion, Clark et al.
practical reasons, removed from and then later returned to the (2000) conducted a comprehensive study of NOR using
WGTA testing apparatus at the longer but not the shorter groups of rats subjected to sham surgery, fornix lesioning, or
delays. Control but not hippocampus-lesioned monkeys radiofrequency and excitotoxic hippocampal lesioning.
might then benet from contextual cues aiding recall. Importantly, they required the absolute accumulation of 30
However, cryptic spatial processing is unlikely to be an seconds of total object exploration time by all subjects.
issue for a different DNMTS protocol developed by Mumby et Fornix-lesioned subjects did not show a decit, consistent
al. (1992a). They trained rats with memory delays varying with earlier ndings. However, rats with either type of hip-
from 4 to 600 seconds with the sample and choice compo- pocampal lesion showed impairments at a 1-hour memory
nents at opposite ends of a runway during a trial but sched- delay. In a follow-up study by Broadbent et al. (2004), a decit
uled equally often at both ends across trials. Subsequent to in NOR was observed as a function of hippocampal lesion
surgery, no impairment was seen in rats with aspiration size. Small lesions caused no impairment, whereas larger
lesions of the hippocampus, or of hippocampus and the lesions did. Given the parallel, distributed nature of hip-
amygdala, at any delay except the longest (10 minutes) at pocampal processing (see Chapter 14), it is reasonable to sup-
which the hippocampal lesion groups then performed more pose that many functions of the MTL that require
poorly than controls. Although sometimes cited as support for hippocampal processing could continue relatively normally
the idea that hippocampal lesions can affect recognition after partial lesions.
memory in rats, there is a caveat here also. Specically, the Finally, turning back to primates, Murray and Wises
within-subject comparisons showed that the control group (2004) critique of the declarative memory theory offers sev-
got paradoxically better at the 10-minute delay rather than the eral other examples of instances where the two sides in the
lesion group getting worse. This improvement by controls was debate draw different conclusions from a common set of data.
not seen in several later studies, and thus Mumbys (2001) rea- One example has to do with whether the perirhinal cortex has
sonable conclusion was that this single statistical difference only memory functions (Buffalo et al., 1998a; Teng et al.,
was the exception to the rule that hippocampal lesions have 2000) or also participates in aspects of perceptual processing.
minimal if any effect on rodent DNMTS. However, using a For example, Teng et al. (2000) reported that rapid learning of
large test battery in the manner of Zola-Morgan and Squire a discrimination between simple three-dimensional objects
(1985), Mumby et al. (1995) again found a small but signi- (e.g., a red versus a green peanut shell) is only slightly
cant hippocampal lesion decit in DNMTS at the longest impaired by hippocampal lesions, notably over the rst
delay tested (120 seconds) relative to both no surgery and par- few trials of the day (which were described as more declara-
tial parietal cortex lesion control groups. Clark et al. (2001) tive in nature), whereas the much slower learning of two-
later also secured a decit in rats with excitotoxic hippocam- dimensional pattern discrimination (e.g., N versus W) is
pal lesions at the longest delay tested (also 120 seconds). Thus, unaffected by hippocampal lesions unless they extend to
another equally reasonable way of summarizing the data is include damage to the tail of the caudate nucleus. In contrast,
that when hippocampal lesion-associated decits have been Bussey et al. (2002) reported that discrimination of com-
seen they have tended to be at long memory delays or with pound visual stimuli with the ambiguity between the features
long list lengths (Mumby et al., 1992b; Steele and Rawlins, maximized is strikingly impaired by perirhinal lesions,
1993; Mumby et al., 1995; Wiig and Bilkey, 1995; Clark et al., whereas the discrimination of those with minimal common
2001). This would be consistent with the declarative memory features is unaffected. A follow-up study by Bussey et al.
theory were it not for the much greater sensitivity of these (2003) used morphed images to produce a single-pair dis-
tasks to perirhinal lesions of comparable or even smaller size. crimination task in which the two stimuli to be discriminated
Either the hippocampus is at the apex of declarative mem- had either very high or very low feature ambiguity (Fig.
ory processing or it is not. 1312). The results show a clear perirhinal decit on the
What might be the key difference between studies that nd slowly learned, maximal feature ambiguity task. These nd-
a deficit after hippocampal lesions (Clark et al., 2000; ings are problematic for the declarative memory theory for
Broadbent et al., 2004) and those that do not (Ennaceur and two reasons. First, the theory asserts that the MTL receives
612 The Hippocampus Book

A Low feature ambiguity

100 100
90 90

mean % correct
80 80
% correct

C C
70 PRh 70 PRh

60 60
50 50
40 40
1 2 3 4 5 6 7 8 9 10 11 12

B High feature ambiguity

100 100 **
90 ** ** ** 90
** ** *

mean % correct
80 ** 80
C
% correct

* C
70 PRh 70 PRh

60 60

50 50

40 40
1 2 3 4 5 6 7 8 9 10 11 12
Sessions
Figure 1312. Perirhinal cortex and perception. Lesions of the perirhinal cortex, ostensibly
part of a medial temporal lobe memory system, cause a decit in the perceptual discrimination
of stimuli with high-feature ambiguity (B) but not low-feature ambiguity (A). The symbols in
the bar graphs on the right indicate scores of individual monkeys. (Source: After Bussey et al.,
2003.)

perceptually processed visual inputs from the inferotemporal laboratory tasks, animals have been shown to remember
cortex and is not itself involved in making perceptual dis- things that people would describe as facts (e.g., that a certain
criminations. Second, it asserts that rapidly learned visual dis- food is safe to eat) and that they can remember events (e.g.,
crimination tasks, other things being equal, are most likely to that an initially novel object has been seen before). However,
be learned in a declarative manner whereas slowly learned dis- will we ever know whether an animal is conscious of its mem-
criminations are acquired in a habit-like way, as in the Teng et ories? Moreover, even if we did, how could an animal ever
al. (2000) study. Yet here it is the slowly learned discrimination consciously declare that it knows or remembers something
that is most sensitive to lesions of the perirhinal cortex. from the past? Does it have a sense of its own life in the way
Indeed, a body of data from rats, monkeys, and humans as that humans doof the state of mental awareness that
well as relevant computational modeling support the notion Tulving (1983) called autonoetic consciousness? If the
that the perirhinal cortex has a role in object perception answers to these questions are negative, one is tempted to
(Murray et al., 2005). Data from amnesic patients with wonder whether a central feature of the theory lies more in the
restricted hippocampal damage or damage extending into realm of metaphysics than in empirical science. In fact, a com-
neocortical structures of the MTL (identied radiographi- prehensive theory of brain and mind should identify under
cally) indicate that the discrimination of faces and scenes of what circumstances animals and humans are conscious, what
ever greater similarity can pose a particular challenge for they are conscious of, and perhaps why consciousness is
patients (Lee et al., 2005). advantageous for some but not other forms of memory
(Zeman, 2002).
Certain Comparative Problems When Asserting This issue is usually nessed in what was referred to earlier
that Declarative Memory Must Be Conscious as the folk psychology approach to consciousness that the
declarative memory theory has taken to date. Nonetheless, in
The insistence on MTL structures mediating memory that can the context of work on blindsight, animal experiments are
be consciously declared is clearly problematic for a theory that making inroads into the issue of mental awareness. For exam-
seeks to encompass both humans and animals. In numerous ple, Cowey and Stoerig (1995) devised an ingenious procedure
 Theories of Hippocampal Function 613

in which monkeys that had been trained to reach accurately al., 2003) to examine whether rhesus monkeys know when
toward one of several targets were also required to report they remember. Using computer-controlled techniques, each
whether they were visually aware of the targets. Monkeys were of two monkeys was briey shown an image on a touchscreen
placed in front of a touch screen. Each trial began with them (Fig. 1314). The image then disappeared for a delay interval
looking at a xation point and ended with their reaching out during which the animals may have sometimes remembered it
to touch various images presented a little while later in the left and other times forgotten it. They were later tested for their
or right visual eld. Operated monkeys (with large unilateral memory of the image. What made this experiment different
striate cortex lesions) were observed to reach as accurately in from conventional delayed match-to-sample testing was that
their blind hemield as normal monkeys (Fig. 1313A). The the monkeys were allowed, for two-thirds of the trials, to
reporting of awareness was achieved in a separate training choose between progressing to the memory test or declining
condition in which the monkeys had to touch a specic area to do it. Declining resulted in a guaranteed but less preferred
of the screen if a target failed to occur at a time when they reward than could be obtained by accepting to do the memory
might have expected to see one (the possibility of a target was test and choosing correctly. For the remaining one-third of the
indicated by an auditory cue). The lesioned monkeys correctly trial sequences, the monkeys were not given a choice and were
reported that they could not see a target when targets were forced to take the memory test. The results showed that both
deliberately not presented on selected trials to their intact monkeys performed more accurately on memory tests they
hemield, but the monkeys incorrectly reported not seeing tar- had opted to take than on enforced tests. A control experiment
gets that were actually presented to their blind hemield (Fig. included trial sequences that began without an image to
1313B). These unseen targets were the very ones to which, remember during the delay interval; the monkeys routinely
in the rst training condition, the animals had reached accu- reacted to these sequences by declining to do the memory test
rately. To all intents and purposes, these monkeys lack phe- when given the option to do so. Moreover, as the memory
nomenal vision; that is, they lacked the ability to comment delay interval was extended (a procedure likely to promote
on seeing things to which they could reach accurately forgetting), the monkeys showed a temporally graded reluc-
(Weiskrantz, 1997). tance to take the memory tests. Taken together, these results
Analogous procedures might be developed for studies of suggest that the monkeys decision whether to take the mem-
memory in primates, the point being to devise a way of disso- ory test at the end of a trial sequence was likely to have been
ciating between the successful performance of a memory task based on their own self-generated awareness of whether they
and the quite separate display of awareness that one is (or is could remember the image. This sort of procedure might be
not) remembering. In an important step, Hampton (2001) adapted to establish whether lesions of the hippocampal for-
exploited techniques developed in parallel for pigeons (Sole et mation (or, better still, reversible inactivation) disrupt aware-

Figure 1313. Blindsight in primates. Cowey

and Stoerigs (1995) commentary procedure
for studying visual awareness in primates. A. The
animals watch the xation spot (center) and
then upon hearing a tone reach toward one of
four peripherally located and briey illuminated
targets. Monkeys with striate cortex lesions can
do this as well as control animals. B. The animals
again watch the xation spot and, upon hearing
the auditory cue, reach for one of the ve briey
illuminated lights on the left, the light on the
right or, when no ash occurs, for the panel at
the top left. Normal monkeys correctly identied
all trial conditions. Monkeys with unilateral stri-
ate cortex lesions identied lights ashed in
their good hemisphere but treated the right-
hand illumination as if it had not occurred.
Thus, monkeys display blindsight in being able
to reach toward objects they report being
unable to see. Can an analogous reporting pro-
cedure be developed in the domain of memory?
(Source: Cowey and Stoerig, 1995).
614 The Hippocampus Book

A Metamemory - experimental design B Recognition Memory

100
Study Choose


Proportion correct
phase Memory
Test

Delay 50
Delay
Forced
interval
p = 0.33 p = 0.67 Memory
Test

Choice 0
phase C.
 

Proportions
Test phase

or small 
reward  
Preferred Primate
Peanut Pellet
if Correct
Figure 1314. Metamemory. A. Monkeys were allowed, during two-thirds of the trials of a
recognition test, to choose whether to take a memory test on the basis of their covert judgment
of whether they could remember. B. Memory performance on trials when they chose to take
the recognition test was signicantly better than on the remaining one-third of trials for which
they were forced to take the test. (Source: Cowey and Stoerig, 2001).

ness of memory to the same extent that they disrupt accurate their conscious awareness of the past? Anecdotes about spared
performance. Might monkeys with hippocampal disruption unconscious memory function abound. Each new genera-
show above-chance familiarity for a previously presented tion of neuropsychologists add their own; an excellent exam-
stimulus even though they declare, in the optional choice ple is an endearingly personal article entitled Memories of
test, that they would prefer not to take the test? Might they H.M. (Ogden and Corkin, 1991). Patient E.P.one of the
know but not remember? A human imaging study of this issue most profoundly amnesic patients to have been subject to
(Kao et al., 2005) suggests that such a task, even if it could be detailed neuropsychological and neuroradiological examina-
developed for monkeys, would be likely to involve an interac- tion in the San Diego neuropsychological studiesis another
tion between the MTL, mediating successful memory, and the case in point.
ventrolateral prefrontal lobe, mediating memory awareness.
During the rst 2 or 3 years in which we visited his
house, he was wary and slow to accept the idea that we
Nondeclarative Memory: Are Spared Learning wished to talk with him and administer tests. After
Abilities Nonpropositional or Learned Tasks some conversation, and with encouragement from his
That Can Be Performed Without Awareness? wife, E.P. would after a number of minutes seat himself
at a table for testing. During the subsequent years, the
A critical claim of the theory is the independence of declara-
same tester has visited his house more than 150 times.
tive and nondeclarative memory. The two key differences
Now when she arrives he greets her in a friendly man-
claimed by Squire (1992) for these superordinate categories of
ner and moves readily and promptly to the table even
memory relate to the propositional status of the informa-
when his wife is not present. Yet, his pattern of greet-
tion being processed and the necessity of awareness during
ing and acceptance occurs without any recognition of
the encoding and recall of such information.
who the tester is, and he will repeatedly deny that he
The existence of spared learning in the presence of amne-
has seen her before. (Stefanacci et al., 2000, p. 731.)
sia has been the subject of comment for about a century
(Claparede, 1911; Weiskrantz, 1997). It is a particularly The declarative memory theory, perhaps more than any
intriguing feature of the syndrome to discuss with students or other theory of memory, has put forward this disconnection
present to a public audience. How can it be that H.M. could of consciousness as a centerpiece. It makes the distinctive
learn to draw in a mirror yet not remember doing so? How claim that such nondeclarative learning is not merely
could he and other patients have this curious disconnection in spared, as others had claimed previously, but that such learn-
 Theories of Hippocampal Function 615

ing occurs in amnesics at a normal rate and in a manner indis- tion being drawn to the matter of location. Later, without
tinguishable from that of controls. To emphasize this key prior warning, the experimenters asked subjects to recall
point, it is not only that amnesic patients can display classic where the objects had been placed. The subjects were never
conditioning, as Warrington and Weizkrantz (1968) were the asked to pay attention to and, in that sense, be attentively
rst to describe, it is that they do so at a normal rate despite a aware of the location of the objects until the retention test.
lack of awareness of the fact of learning (Gabrieli et al., 1995). That normal subjects could do this task suggests that they may
However, because certain ostensibly nondeclarative tasks automatically encode attributes of a stimulus to which their
may be subject to contamination of learning using a declara- attention is not directed despite paying attention to other
tive strategy, ndings from patients with mild amnesia who attributes of the stimulus, such as its value. In contrast, people
have residual declarative memory can be misleading. The with large right MTL lesions were impaired. In this case, the
importance of drawing rm theoretical conclusions only from experimental manipulation of awareness occurred at the
those rare individuals with severe amnesia was a theme of an point of encoding, but manifestations of this differential
early critique of the declarative memory theory (Weiskrantz, awareness only came to light at the point of recall. We return
1997) and, in a paradoxical twist, one now taken up by Squire to the issue of differentiating automatic and intentional
et al. (2004) in their careful comparisons of performance by encoding in Sections 13.5 and 13.6.
people and animals with partial versus complete damage to This discussion leads us to suggest that the nondeclarative
the MTL (e.g., patient R.B. versus patient E.P.). nature of nondeclarative memory may be less to do with
Nondeclarative learning does not, of course, usually take whether the skill being performed can ever be declared than
place outside of consciousness. Learning to ride a bicycle, for with whether subjects are consciously unaware of the prior
example, is a motor skill for which most of us are acutely con- occurrence of the stimulus at the time of recall. The impor-
scious of what we are doing as we attempt to learn. The declar- tance of awareness rather than propositional status is particu-
ative memory theory makes the radical claim, however, that larly well revealed in studies of eyeblink conditioning, an
although aware of the act of learning this awareness does not advantageous paradigm for the present purposes as it can be
play any causal role in the encoding or expression of learned studied in both humans and animals. In this form of condi-
motor acts. This is a rather interesting idea but one that is dif- tioning, an initially neutral stimulus such as a tone is pre-
cult to test rigorously. It cannot be tested by seeing what sented for a few hundred milliseconds before a puff of air to
learning capacity remains while we are unconsciousnot the eye. Through repeated pairings of these two stimuli, the
because we cannot then ride bicycles (though that too) but subject develops a conditioned response (CR) to the tone (the
because the lack of awareness to which the theory makes ref- conditioned stimulus, or CS). Eyeblink conditioning has been
erence is not a lack of consciousness as such but an absence of extensively studied in humans and animals and is known to
awareness of the information relevant for learning at the obey the basis principles of classic (i.e., Pavlovian) condition-
moment of motor recall. ing (see Section 13.5). Learning depends on the contiguity and
This absence of the referent of awareness is highly specic contingency between the CS and the puff of air (the uncondi-
in the nondeclarative domain. For word-stem completion tioned stimulus, or US). It is a form of conditioning for which
priming, for example, people are asked to think of a word the cerebellum was found to be essential (Clark et al., 1984).
any wordthat begins with the stem given as a cue. Subjects Following the observation that amnesics could acquire classic
are aware they are being tested, aware that the test involves conditioning (Warrington and Weizkrantz, 1968) and that it
words, and attentive to the task in hand. Amnesic subjects, occurs at a normal rate (Gabrieli et al., 1995), it had been
however, are unaware that the words they come up with are widely assumed that this is not a form of learning in which the
often words they were shown earlier. What they fail to MTL is ever engaged. This turns out to be not quite right.
declare is not the words themselves, but that they have the Two frequently studied forms of human eyeblink condi-
phenomenological attribute of being words they remember tioning are named (somewhat inappropriately) the delay
having seen earlier. This simultaneous sense of being tested in paradigm (in which, confusingly, the US follows the CS with
the present and bringing to mind events from the past is what no delay between the end of the CS and the onset of the US);
is meant by mental time travel (Schacter and Tulving, 1994; the trace paradigm in which CS offset occurs 300 to 1000 ms
Tulving and Markowitsch, 1998), and it is an attribute of mind prior to US onset (Fig. 1315). As the CS and US are not con-
that amnesics lack (Weiskrantz, 1997). tiguous in time, Pavlov suggested (and others since) that some
In a similar vein, studies of incidental learning suggest that trace of the CS must linger in memory if an association is to
unilateral MTL lesions can disrupt memory for information be formed with the US to allow effective conditioning (hence
to which a persons attention is not overtly directed until the the name). Although no decit is seen in amnesics using the
point of recall. In a famous study, Smith and Milner (1981) delay paradigm, McGlinchey-Berroth et al. (1997) have shown
gave subjects a series of small toys, one by one, asking them that such patients have impaired trace eyeblink conditioning
the likely price of the real thing (e.g., a car). Their attention at across a range of trace intervals. The contrast between these
encoding was directed to the value of the toys. One by one, the two protocols is instructive as what subjects learn to do in
toys were put down by the experimenter in various places on each caseto make automatic, appropriately timed CRscan
the table in front of the subjects, without the subjects atten- hardly be said to be nonpropositional in the former case but
616 The Hippocampus Book

Delay conditioning Trace conditioning

Figure 1315. Protocols for delay and
trace conditioning. Time lines showing the
CS onset and offset of the conditional stimu-
lus (CS) and unconditional stimulus (US)
US in two distinct forms of classic condition-
ing. Appropriately timed CRs develop in
300-1000 msec both protocols, yet only one is thought to
time be hippocampus-dependent.

propositional in the latter. It is therefore an important test for through some interaction between hippocampus and neocor-
the theory to identify what might be declarative about trace tex during the course of conditioning and is somehow fos-
conditioning. tered by using the trace conditoning paradigm. By virtue of
Clark and Squire (1998) used a differential trace condi- this, a suitably timed signal can be sent to the cerebellum
tioning paradigm in which the CS (a tone) was followed by where the eyeblink response component of the conditioning
the air-puff US, whereas a second stimulus, the CS- (a noise), takes place. The eyeblink CR, a learned but automatic defen-
was presented on its own. Sessions consisted of multiple pre- sive response, can then be executed at the appropriate time.
sentations of each stimulus to human subjects, some of whom To summarize, what is nondeclarative about nondeclara-
were subject to a delay paradigm and others to a trace para- tive learning may actually be the lack of awareness of what is
digm. All were required to watch, attend to, and try to remem- being remembered at the point of recall, rather than any
ber a silent movie (Charlie Chaplins The Gold Rush) while intrinsic inexibility in what has been learned or a lack of
these CS and US events were happening, thereby command- propositional status of the content of learning.
ing a great deal of their conscious awareness. At the end of the
experiment, in addition to being quizzed about the movie, the Are Fact-Memory and Event-Memory
subjects were asked a set of questions with true/false answers Processed by a Common Brain System?
designed to explore their awareness of the stimulus contin-
gencies. The key nding was that successful trace conditioning Supercially, amnesics present with a decit restricted to
was correlated with levels of awareness of these contingencies. episodic memory. They cannot remember events for any
A subset of control subjects who showed good awareness of length of time, but their factual knowledge about the world
the CS predicting the US after the trace delay conditioned and their knowledge of language are both intact. Semantic
well. Other control subjects, and all amnesic subjects tested, memory shows all the appearances of being preserved. The
were less aware or even unaware of the stimulus contingen- declarative memory theory holds that this dissociation is pro-
ciesthey conditioned poorly. Awareness was unrelated to foundly deceptive. It asserts that there is a unitary process
levels of conditioning in the delay paradigm, a result that is underlying the formation of both event and fact memories.
not universally obtained (e.g., Knuttinen et al., 2001). The reason amnesic patients display intact semantic memory
Furthermore, to explore whether the correlation between level is because so much of a persons factual knowledge was
of awareness and degree of trace conditioning was causal, acquired years earlier, extending from the years of childhood
Clark and Squire (1998) manipulated awareness directly. on through life. Consolidation of such memory traces would
Manipulations that facilitated awareness facilitated eyeblink be long completed. Conversely, a patients failure of event
conditioning; manipulations that reduced it blocked condi- memory often relates to relatively recent events such as a for-
tioning. A subsequent critique (LaBar and Disterhof, 1998) gotten conversation of the day before.
queried Clark and Squires use of a differential conditioning Although most amnesic subjects have some residual
protocol somewhat different from those used in many previ- episodic memory function, a few are claimed to have little or
ous studies of human eyeblink conditioning; they also won- none. In Chapter 12, we noted the striking case of patient E.P.,
dered whether a better online measure of awareness than a who has essentially no measurable recent episodic memory
postexperimental questionnaire might be developed. Manns but who, becoming amnesic when adult, displays good factual
et al. (2000a) provided both in a study of normal subjects who knowledge of the world acquired earlier in life and remarkably
showed a close predictive relation between developing aware- good memory for spatial and other episodic-like information
ness of the CS-US trace contingency and subsequent levels of acquired during childhood. However, if this interpretation is
conditioning. These investigators also showed that the critical correct, people who have sustained damage to the MTL at a
feature of this awareness is the subjects awareness of the CS- much younger age should also show impaired semantic mem-
US contingency rather than their awareness of the likelihood ory. Data concerning the effects of bilateral hippocampal
of blinking. Thus, it appears that what is declarative about pathology sustained early in life are relevant to this issue
trace conditioning is an awareness of the fact that the tone (Vargha-Khadem et al., 1997). Three young developmental
predicts that an air-puff will come shortlyan awareness that amnesics who sustained brain injuries at birth, 4 years, and 9
can be expressed propositionally. This awareness may arise years of age, respectively, were aged 14, 19, and 22 at the time
 Theories of Hippocampal Function 617

of rst testing. The anoxic-ischemic episodes they had are Does it make sense to suppose that the memory of facts
likely to have affected the hippocampus preferentially. MRI and events could depend on a single brain system? According
measurements revealed hippocampal volumes ranging from to the theory, both types of information require the integrity
43% to 61% of normal. T2-weighted relaxometry measure- of the hippocampal formation for their formation (encod-
ments also revealed hippocampal abnormalities, but MR spec- ing); and both result, after memory consolidation, in stable
troscopy values were within the normal range. Additional long-term traces in neocortex. Success in recalling either a fact
MRI measurements in other brain areas suggested that bilat- or an event depends only on the strength of the traces estab-
eral hippocampal changes are not only the primary pathology lished through consolidation. In contrast, Nadel and col-
but might be essentially the only pathology in all three cases leagures (Nadel and Moscovitch, 1997, 1998; Nadel et al.,
(Mishkin et al., 1997). 2000) have argued that memory for an event requires access to
The striking neuropsychological feature is that their amne- a hippocampally based contextual trace (where the event
sia is largely exclusive to episodic memory. They are tempo- happened) and a prefrontal temporal trace (when the event
rally and spatially disoriented: forgetting the date and happened). Retrieval of event memory is a reconstruction
appointments, getting lost, mislaying their belongings. based on both of these traces and other information about
Despite severe problems of episodic memory, these children the event itself stored elsewhere in the neocortex. Retrieval of
all attained levels of speech and language competence and of factual information, on the other hand, is not thought to
literacy and factual knowledge within the low average to aver- require either a contextual or a temporal trace; it is context-
age range. They all attended mainstream schools, and their independent.
acquisition of semantic information appears to be normal (or In conclusion, we still do not understand the precise role of
at least near-normal) as assessed by performance on the the hippocampus in episodic and semantic memory or, within
Wechsler Adult Intelligence Scale (WAIS) verbal intelligence the domain of episodic memory, in familiarity and recollec-
tests. Their performance on a Basic Reading and Reading tion. The immediately preceding discussion, complementing
Comprehension subtest was normal in all three cases, as was Chapter 12, points to the need to develop new animal models
their Spelling subtest (with the exception of one subject whose of these forms of memory and of retrieval to help resolve the
spelling is poor). issues. We return to these important ideas in Section 13.6 after
The pattern of memory preservation and memory loss we have discussed the other major theories of the hippocam-
shown by these children is in striking contrast to what would pus: cognitive mapping and predictable ambiguity.
have been expected on the basis of the declarative memory
theory. Vargha-Khadem and her colleagues argued that these
three patients raise the possibility that the basic sensory
memory functions of the perirhinal and entorhinal cortices 13.4 Hippocampus and Space: Cognitive
may be largely sufficient to support the formation of context- Map Theory of Hippocampal Function
free semantic memories but not context-rich episodic memo-
ries, which must therefore require the additional processing Space, wrote OKeefe and Nadel (1978, p. 5), plays a role in
provided by the hippocampal circuit (Vargha-Khadem et al., all our behavior. We live in it, move through it, explore it,
1997, p. 379). However, further research is required before defend it . . . yet we nd it extraordinarily difficult to come to
such conclusions can be drawn with any certainty as later grips with space. This strident rhetoric, together with discus-
commentaries on these cases and parallel primate studies have sion of the philosophical concept of space, introduced a book
recognized (Mishkin et al., 1998; Bachevalier and Vargha- that outlined the then new cognitive map theory of hip-
Khadem, 2005). For one thing, the children appear to have pocampal function. The theory has developed substantially
some residual episodic memory, and this may have been suffi- over the past quarter century.
cient for them to learn well in school, albeit more slowly than The primary stimulus for its development was the discov-
other children and perhaps only after frequent repetition. ery of place cells in the hippocampus of freely moving rats
Moreover, although Vargha-Khadem and her colleagues inter- (OKeefe and Dostrovsky, 1971). These cells are neurons that,
viewed the childrens parents, they did not report on any dis- as discussed in Chapter 11, are characterized by patterns of r-
cussions with the childrens schoolteachers. It is difficult to ing that increase from a generally low level to higher rates as a
believe that their teachers would have failed to notice their freely moving animal moves through a particular region of
severe episodic memory difficulties and thus compensated for space. As different place cells are responsive in different loca-
these problems with their teaching. The early damage to the tions and are found throughout the septo-temporal axis of
brain may also have resulted in morphological reorganization the hippocampus, it was natural to suppose that the collective
of MTL structures in a way that does not occur when adults ring of different place-cell ensembles in different environ-
become amnesic. Certainly the MRI measurements do not all ments might be the neural substrate of distinct maps of
point to a completely destroyed hippocampus. Taken together, space. Observations on place cells were quickly followed by
although fascinating cases, these concerns blunt the critical the discovery that lesions of a major ber tract into the hip-
impact on declarative memory theory that they might other- pocampal formation, the mbria/fornix, resulted in an appar-
wise have had. ently selective decit of spatial, but not cue, learning by rats
618 The Hippocampus Book

in a circular maze task (OKeefe et al., 1975). These two nd-

ings were the trigger for a comprehensive review of the litera- Box 134
Cognitive Map Theory
ture and one that offered an imaginative reassessment of
ndings hitherto explained in other ways (OKeefe and Nadel, 1. Representation: The vertebrate brain has a neural system
1978). the locale systemthat organizes the encoding and rep-
The cognitive map theory was the rst of several spatial resentation of perceived stimuli with respect to an
theories of hippocampal function. Its authors have since out- allocentric spatial framework, or cognitive map. The
lined modications of the theory, beginning with Nadels spatial locations of landmarks are stored in this map dur-
work on context (Nadel and Willner, 1980), through ideas ing exploration. This locale system is in the hippocampus.
about multiple memory traces stored in the hippocampus 2. Navigation: The locale system is used for spatial naviga-
(Nadel and Moscovitch, 1997), and now incorporating neu- tion. The various anatomical components of the hip-
roimaging ndings from humans performing virtual naviga- pocampal formation are involved in mediating different
aspects of spatial information processing, with distinct
tion tasks (Burgess et al., 2002). OKeefe has also explored the
classes of neurons responsive to an animals place, head-
relevance of the theory to the neural basis of language, in par-
direction, view and its movement through space.
ticular spatial prepositions (OKeefe, 1996). Other spatial 3. Evolution and the laws of learning: Spatial mapping
memory theories include the scene memory hypothesis of evolved as one of the multiple memory systems of the ver-
Gaffan (1991), the fragment assembly hypothesis (Worden, tebrate brain with its own distinctive learning rules. Other
1992), and the closely related idea of the hippocampal forma- types of learning and memory include simpler associative
tion being involved in path integration (McNaughton et al., mechanisms and certain geometrical tasks that can be
1996; Whishaw, 1998; Redish, 1999). There have also been acquired using taxon strategies. These other systems
ideas about the metrics in which spatial information is obey laws of learning that are different from those of the
processed (Poucet, 1993) and a dual, or parallel-map, the- locale system.
ory (Jacobs and Schenk, 2003). This latest development argues 4. Sites of storage: Spatial maps are stored in the hippocam-
pus. They are not consolidated or stored elsewhere in the
for a synergistic interaction of a directional bearing map and
brain, although information stored in the hippocampus
an allocentric sketch mapeach mediated by distinct but
does interact with information stored elsewhere for the
overlapping subregions of the hippocampal formation. Like purpose of guiding navigational behavior. An intrahip-
the ideas about path integration, it incorporates the discovery pocampal consolidation mechanism may nonetheless exist.
of another category of spatially tuned cells, the head-direction 5. Extension of the theory to humans: Whereas the cognitive
units (Taube and Schwartzkroin, 1987; Taube, 1998). These map is purely spatial in animals, it subserves the storage
are cells, described in Chapter 11, that re when a rats head is and recall of linguistic and episodic memories in humans.
facing a particular direction in a familiar environment irre- Lateralization enables the right hippocampus to maintain
spective of the animals location. In parallel with this neurobi- a primarily spatial function, whereas the left hippocampus
ological theorizing, there has been continuing interest in the incorporates a temporal sense and linguistic entities that
psychological literature concerning the concept of cognitive together provide the basis for mediating episodic memory.
mapping (Gallistel, 1980) that dates back to Tolmans classic
paper (Tolman, 1948). Although the term cognitive mapping
was used in a slightly different sense by Gallistel, this renewal As already noted, numerous developments of the original
of interest has prompted debate that, in an unexpected turn, theory have occurred since 1978, leading to slightly different
has cast doubt on the explanatory power of the concept ideas about spatial memory than those in the list in Box 134.
(Bennett, 1996; Wang and Spelke, 2002). Mackintosh has also In the discussion that follows, there are references to these
queried its adequacy as an account of how animals get from developments, as appropriate. Although distinct, they are in
one place to another but expressed the view that the 1978 the same genre as the original cognitive map theory.
book deserves its status as a modern classic (Mackintosh, The rst of OKeefe and Nadels (1978) ideasthe concept
2002. p. 181). Numerous reviews have been published of a locale, or mapping, systemimplies that when forming
(Poucet, 1993; Muller, 1996; Taube, 1999), as have two com- memories of the world neural activity in the brain automati-
prehensive books focusing on the neurobiology of spatial, cally encodes stimuli in a connected mental framework rather
context, and episodic memory (Redish, 1999; Jeffery, 2003). than as isolated stimulus traces or independent pairwise asso-
ciations. This encoding happens, for example, as laboratory
13.4.1 Outline of the Theory rats explore a novel environment (Fig. 1316A). It occurs
quickly, sometimes in one trial; and it relies on information
The ve main suppositions of the original (1978) cognitive from the place cells indicating the animals location, particu-
map theory relate to representation of the environment, nav- larly the distance from the boundaries of the test arena. Head-
igation through it, the evolution of an ostensibly specialized direction cells provide a representation of direction based on
spatial memory system, storage of spatial information, and what are described as innitely long vectors. This curiosity-
the relevance of a specically spatial system in animals to the driven exploration involves something more than perceptual
mediation of episodic and semantic memory in humans. learning about the identity of individual cues (objects, land-
 Theories of Hippocampal Function 619

Figure 1316. Place cells, boundary vectors, and the puzzle of place rat at place A wants to navigate to B but not to C. Place cells (gray
cell-guided navigation. A. As rats explore a simple environment in a circles and ellipses) have ring elds in these three areas. A subset of
test arena, they encode information about location, such as distance the cells with elds at place A re with a certain probability with the
from the side walls, into a map-like internal representation. Place animal there, but it is not clear how at place A the rat can activate
cells emerge as the interaction of two or more Gaussians (inset) the subset of place cells for place B but not those for place C to
whose centers are determined by the distance from prominent fea- direct its choice behavior appropriately. Additional goal cells seem
tures such as walls. (Source: After OKeefe and Burgess, 1996.) B. A to be required to achieve navigation. (Source: After Morris, 1991.)

marks) or the development of familiarity that later enables canonical local circuits of neocortex (Douglas and Martin,
recognition. The putative hippocampal mapping system also 2004), could be precisely what is needed to perform the geo-
encodes where landmarks are located in relation to each other metrical computations involved in nding the way around.
in some kind of geometrical framework. This might include computations for encoding the distance
The idea of an organized structure to memory is, of course, and direction between landmarks, rotating the map for appro-
hardly unique to the cognitive map theory, as many other the- priate alignment with perceived cues, identifying the location
ories of human memory embody the same idea, including and heading direction of the animal within it, and performing
articial intelligence-derived theories of semantic memory various other geometrical functions essential for effective allo-
(Collins and Quillian, 1969) and connectionist theories of centric navigation through space. In short: Where am I? Where
memory consolidation (McClelland et al., 1995; McClelland am I going? How do I get there?
and Goddard, 1996; OReilly and Norman, 2002). Some form OKeefe and Nadel (1978) offered several speculations
of memory organization would also be helpful for encoding about how such spatial operations could be implemented
one-time events in relation to static landmarks; and, partly in the hippocampal formation. Considering the original ideas
for this reason, the hippocampus in humans could be help- on representation and navigation together (Box 134, propo-
ful for encoding information into episodic memory. The recall sitions 1 and 2), they supposed that sensory information
of a specic event, such as what one did during a recent holi- entered the hippocampal formation via the entorhinal cortex,
day, generally involves rst remembering where one was from which it projected into the dentate gyrus and then to
before proceeding to remembering with whom one spent the succeeding stages of the trisynaptic circuit as it was then
holiday or what unusual things happened. Gaffan (1991) understood (but see the discussion of hippocampal connec-
was the rst to make this point explicitly, later building it into tivity in Chapter 3, Section 3.7). Different pyramidal cells were
a scene-specic theory of episodic memory (see Section thought to be driven by different proportions of excitatory or
13.6). inhibitory inputs. The group with mainly excitatory inputs
The second supposition is that, once stored, locale informa- (place cells) were dened as those that red when an animal
tion can be used for spatial navigation. Whereas the rst propo- occupied a particular position in relation to a constellation of
sition was that information is stored in a map-like framework, sensory cues (e.g., visual, auditory, olfactory). Because of the
the second relates to how that information is used. How does convergent nature of hippocampal circuitry, it was suggested
an animal navigate using its cognitive map? This question that the probability (or rate) of place-cell ring should decline
prompted the supposition that the unusual anatomical cir- in a monotonic but nonlinear fashion as individual cues
cuitry of the allocortex (see Chapters 3, 4, and 8), so unlike the become obscured or are removed but be relatively insensitive
620 The Hippocampus Book

to the loss of any one cue. This prediction about place cells Notwithstanding these developments, we still do not fully
was upheld (OKeefe and Conway, 1978). Other pyramidal understand how a map of space could be represented (Box
cells (misplace cells) were presumed to have, in addition, 134, proposition 1) or used (Box 134, proposition 2)
sensory-driven inputs relayed via feedforward inhibition such by means of hippocampal circuitry. The relation between
that they would increase their rate of ring when sensory cues location ring (in pyramidal cells of CA1 and CA3), head-
were removed. This increased ring was held to be the neural directional ring (in the presubiculum and anterior thala-
event that triggered exploratory behavior, perhaps via the mus), and movement ring (in the interneurons of the
subicular output through the fornix to the nucleus accumbens hippocampus and dentate gyrus) is also poorly understood.
(see Chapter 3, Section 3.4.2). The original theory also stated Those concerned with specic details might reasonably won-
that the ring of theta cells is coupled to the translational der, for example, whether the head-direction ring of pre-
movement of the animal from one position in the map to subiculum cells is a component of locale or taxon
another, an idea supported by studies in which the different processing (or both). Various logical puzzles about the link
frequencies of theta activity observed during jumping between mapping and navigation have also been pointed out.
through space correlated directly with the distance moved For example, as place cells are dened as cells that re only
rather than the force exerted (Morris, 1983). More recent neu- when an animal occupies a specic position in space, how can
ral modeling studies have modied these ideas substantially the mapping system access information pertaining to places
(see Chapter 14). The unfolding picture about the wide range other than the one presently occupied by the animal (Morris,
of distinct inhibitory interneurons and their axonal connec- 1991)? That is, if a rat is at place A and wants to navigate to
tions (see Chapter 8) has also raised the specter of yet deeper place B but not to place C, how does it at place A access infor-
subtlety at the level of local circuits. mation relevant to going to B rather than going to C (Fig.
Since 1978, there have been many technical and conceptual 1316B)? Place-cell ring on its own does not seem to help
developments in hippocampal single-unit recording in freely because the place cells corresponding to B and C cannot, by
moving rats. The development of multiple single-unit record- denition, re until the animal gets to B or C, respectively. This
ing (ensemble recording) with stereotrodes and tetrodes kind of logical conundrum is clearly problematic for the the-
has shed fresh light on a range of theoretical issues: multiple ory, but it may not be fatal if goal cells can be identied.
place elds, local views, gating by the potential for movement, Perhaps these are in other brain areas (Hok et al., 2005). Other
memory properties, multiple reference frames, pattern com- models of the neural implementation of goal representations
pletion, and pattern separation (as discussed in Chapter 11). are also being developed (Redish, 1999; Biegler, 2003). Several
Numerous uncertainties have been claried. It is now known, neural network and temporal difference learning models that
for example, that pyramidal cells acquire their place elds rap- grapple with these issues are outlined in Chapter 14.
idly, re in proportion to an animals speed of motion, tend to The third feature of the theory is that spatial mapping has
be directionally sensitive only in directionally constrained evolved in response to specic environmental demands, and that
environments (e.g., linear tracks), and show temporal preces- it constitutes one of the multiple memory systems of the verte-
sion in their phase relation to hippocampal theta as an animal brate brain (Fig. 13-17). The emphasis on evolution, lacking
moves through the cells place eld. Improved recording tech- prominence in the declarative memory theory, has been
niques have also established that single cells do not, in general, informed by studies of cache recovery in passerine birds,
have place elds in different parts of a single environment, homing in pigeons, territory size and mating systems in small
and that physically adjacent pyramidal cells may have place mammals, and other aspects of naturalistic spatial behavior
elds in quite different parts of the environment. A boundary (Sherry and Schacter, 1987b; Clayton and Krebs, 1995; Dyer,
vector model (BVC) has been developed (Hartley et al., 2000) 1998; Healy, 1998; Shettleworth, 1998; Jacobs and Schenk,
according to which places can be modeled as the sum of two 2003). The neuroethological component of the thinking
or more Gaussians, with the center of each Gaussian located behind the cognitive map theory has been explicit from the
by its distance from a specic environmental feature such as a outset (OKeefe and Nadel, 1979; Nadel, 1991), fostering stud-
wall (as in Fig. 1316A). With this view, place elds represent ies of sex and seasonal differences in spatial memory that
probability distributions and do not specify precise locations. might not have happened had all hippocampal research
Certain theoretically unanticipated properties of CA1 place remained anchored solely in a paramedical context. Work in
cells have also been discovered, such as their sensitivity to both vertebrate and invertebrate species has revealed both
physical restraint, their apparent insensitivity to lesions of the convergent and divergent evolution with respect to how to
dentate gyrus or CA3, and alterations in receptive eld size nd food, water or sexand then get safely back home.
and stability with aging (see Chapter 11). The discovery of Studies of animal navigation have, however, revealed mecha-
head-direction cells in the presubiculum and thalamus (Taube nisms that appear to have nothing to do with cognitive maps,
et al., 1990a, b) and of grid cells in the dorsomedial entorhi- such as ideothetic path integration by insects (Wehner et al.,
nal cortex (Fyhn et al., 2004; Hafting et al., 2005) are impor- 1996) and rodents (Etienne and Jeffery, 2004) and the snap-
tant additions to the idea that the hippocampal formation and shot processing of landmarks by bees (Collett, 1992). The rel-
interconnecting structures constitute, in animals, an essen- evance of the neuroethological approach to hippocampal
tially spatial system. function and mechanisms has been the subject of cogent crit-
 Theories of Hippocampal Function 621

Types of learning between spatial memory and nonspatial associative learning

(such as classical and instrumental conditioning). Whereas
allocentric locale learning was held to be motivated only by
curiosity and to be extremely fast, associative learning was
thought to be motivated by rewards related to basic motiva-
tional needs (such as food for a hungry animal) and to be
Spatial Associative slower, cumulative, and governed by quite different learning
rules. However, this spatial versus associative distinction has
come under close scrutiny (just as it did in connection with
recognition memory in the previous section of this chapter).
Locale Taxon Classical Instrumental Certain classic phenomena in the associative domain (such as
Conditioning Conditioning overshadowing and blocking, to be explained later) have
also been found to operate within the domain of spatial learn-
ing (Biegler and Morris, 1999; Chamizo, 2003) although sub-
Orientation Cue guidance
ject to certain constraints associated with the perception and
local processing of geometry (McGregor et al., 2004a).
Associative learning can also be very fast, and in phenomena
Hippocampal Formation such as sensory preconditioning, is not obviously motivated
Figure 1317. Multiple forms of memory in the cognitive map the-
by biological rewards. Although OKeefe and Nadels original
ory. Only the locale system, requiring cognitive mapping, is medi- dichotomy now feels incomplete, understanding the bound-
ated by the hippocampus. The theory has less to say about other aries between cognitive mapping and other forms of learning
forms of learning and memory. and memory is proving to be a fruitful area of research.
Also of interest is how spatial and nonspatial systems compete
or work synergistically to ensure the seamless control of
icism (Squire, 1993; Bolhuis and Macphail, 2001), but many behavior; and it remains a fundamental issue (White and
see it as an important adjunct to the development of cognitive McDonald, 2002).
mapping theory. The fourth proposition of the 1978 theory is that allocen-
However, the bulk of the work addressing the cogni- tric spatial information is stored in the hippocampus. Unlike the
tive map theory has been conducted with the laboratory rat time-limited role for the hippocampus in the declarative
and, more recently, various laboratory strains of mice. A veri- memory theory, the original version of the cognitive map the-
table hippocampal industry emerged, beginning with ory has both initial encoding and long-term storage in the one
lesion/behavior studies and single-unit recording work on brain structure. Nadel and Moscovitch (1997) extended this
animals, that now nds modern expression in a variety of idea in a multiple-trace theory of episodic memory in which
behavioral tasks used in conjunction with the full spectrum of memory retrieval episodes can initiate re-storage. The possi-
contemporary neuroscience techniques. Particularly exciting bility of multiple traceseven for single episodesleads to a
are the technical innovations that complement classic lesion different account of the temporal gradients of retrograde
techniques, with the new tasks including ensemble single-unit amnesia that are sometimes seen for information acquired
recording from large numbers of hippocampal neurons prior to brain damage. The key difference from the account
(Wilson and McNaughton, 1993), monitoring the expression offered by the declarative memory theory lies in the impact of
of immediate early genes (Guzowski et al., 1999; Vann et al., partial damage. Complete lesions at various time points after
2000), and reversible pharmacological interventions that are training should always result in temporally extensive, or at
sufficiently sustained in time to investigate memory processes, gradients, of remote memory loss. In contrast, partial lesions
such as systems-level consolidation (Riedel et al., 1999). Mice might result in a reverse gradient if older information, recol-
have nibbled their way onto the hippocampal stage with tar- lected previously, had laid down additional traces in areas of
geted gene deletions, hippocampus-specic deletions, and the hippocampus spared by the lesion. The declarative mem-
even inducible gene knockouts (Tonegawa et al., 1995; May- ory theory makes no differential prediction as a function of
ford et al., 1996; Tsien et al., 1996; Abel and Lattal, 2001). The lesion completeness.
hippocampal research industry, which has seen the number of Although the original cognitive map theory of hippocam-
papers on rodent spatial representation and navigation rise pal function was based largely on observations in rats, its
from around 10 per year in 1980 to several hundred per year extension and likely modication to account for primate and
(Sutherland and Hamilton, 2004), owes a great debt to the human hippocampal function was recognized at the outset.
cognitive map theory. Proposition 5 in Box 134 recognizes that later chapters of
With respect to multiple memory systems, the scope of OKeefe and Nadels monograph addressed the relevance
OKeefe and Nadels (1978) original analysis was narrower of their ideas to human semantic memory (e.g., Chapters 14
than that encompassed in the declarative framework dis- and 15) and, specically, the way in which spatial relations
cussed earlier (Figure 13.4). Their book distinguishes only are fundamental to certain linguistic prepositions that reect
622 The Hippocampus Book

the knowledge of relations (e.g., beside, near, above). forms of propositional knowledge, whether it comes to us via
Although not originally inuential in the human neuropsy- visual, tactile, olfactory, or other sensory modalities. Why,
chology community, these ideas have nonetheless been their concern continues, should space be any different? The
extended to such domains as reasoning and abstract thought no different perspective appears to look on memory pro-
and are now an active area of research. For example, Gattis cessing of space as involving multimodal sensory informa-
(2001) identied the sharp distinction between a generalist tion, with cues that may be familiar or unfamiliar and with the
school of thought, in which space is merely a special case of capacity to realize declarations of spatial knowledge such A
relational processing, and a Kantian perspective, in which is near B. Given this, why would encoding or storage be any
space (and time) are special. Nadel and Hardt (2004) strongly different from that of other object-orientated information
reasserted the latter position, whereas OKeefe (2005) has con- that is processed via the ventral stream of the perceptual
tinued his exploration of vector grammar and how it might processing pathways of the neocortex terminating in the
explain the essentially various meanings of spatial propo- inferotemporal, perirhinal, and parahippocampal cortices
sitions. (Ungerleider and Mishkin, 1982)? Indeed, so absurd may it
Cognitive science apart, one obstacle to pursuing these seem to think that space could be different, a major review of
lines of thought has been in securing denitive evidence for the functions of the MTL considered the adequacy of the spa-
causally linking hippocampal cell activity in humans to any tial theory of hippocampal function in little more than a cou-
such processes. Hippocampal damage in humans does not ple of paragraphs (Squire et al., 2004).
obviously disrupt a persons ability to use spatial prepositions. Not surprisingly, the opposing camp is resistant to being so
Invasive single-unit recording in humans is rarely feasible, but subsumed. For those prepared to contemplate a neural system
some intriguing examples of place cells have been recorded for cognitive mapping (including critics of OKeefe and
from the human hippocampus with depth electrodes placed Nadels theory, such as the present author), thinking about
to monitor epileptic seizure activity (Fried et al., 1998; space in this declarative way reects a conceptual misunder-
Ekstrom et al., 2003). Place cells have also been observed in standing: Space is not a sensory modality. We do not have sen-
the hippocampus of monkeys free to drive themselves around sory organs for space or a cranial nerve devoted to it. Space
slowly in a motorized cab (Matsumura et al., 1999). Spatial is a construct of mental processing. Moreover, space has
correlates are therefore by no means unique to rodents, but a structure that requires but is logically independent of the
securing single-unit evidence for vector grammar has not yet multimodal sensory information used to identify it. For
been achieved. example, many objects whose location we know are often hid-
Functional brain imaging has also been used to study den, whether it is acorns cached by a squirrel or food in a
hemodynamic activation during human navigation. The kitchen cupboard. It is precisely in this situation that spatial
problem here is that fMRI is a technique that does not obvi- memory becomes so usefulthe exact opposite of the situa-
ously lend itself to the study of a mental processnaviga- tion that applies to normal recognition memory. Similarly,
tionthat ostensibly requires subjects to be mobile. One the appearance of a specic area of space can change across
solution is to use virtual reality. Whether virtual reality in the seasons despite its containing inanimate or sessile living
subjects lying prone in a brain scanner really engages the same things that do not ordinarily move around. What does
neural mechanisms as true navigation can be debated not change when a cupboard door is closed, a tree loses its
place cells in rats being sensitive to whether animals are leaves, or a rock becomes covered in snow is the location of
restrained or allowed to move (Foster et al., 1989)but if one the remembered item. There is, in short, a geometric stabil-
accepts that a virtual reality approach might be viable, mod- ity about remembered space that is independent of
ern computing software has certainly made it possible. the appearance of the stationary objects and landmarks that,
Numerous studies have now examined whether the hip- paradoxically, dene that stability (Biegler and Morris, 1993).
pocampus and adjacent structures of humans are activated This paradox is at the heart of why pattern separation and
during apparent navigation through imaginary towns pattern completion are network operations at the heart
(Maguire et al., 1998), featureless arenas, and swimming pools of hippocampal processing (see Chapter 14). The closely
(Thomas et al., 2001; Parslow et al., 2004) as well as other tasks related concept of context is also not necessarily something
that distinguish mapping from merely following a sequential that is out there waiting for a sensory/perceptual system to
route (Morris and Mayes, 2004). Burgess et al. (2002) made a process it; it is a neural construct (Jeffery, 2003). These and
strong case that virtual reality is informative and specically other features of space could have been major driving forces
discussed patterns of activation in relation to spatial frame- in the evolution of the distinctive cognitive systems that medi-
works, the dimensionality of space, and both orientation and ate it.
self-motion. In contrast, a more cautious stance about the This completes the initial overview of the ve key features
ndings to have emerged from functional imaging of spatial of the cognitive map theory. A selective presentation of rele-
navigation is presented in Chapter 12. vant data follows, in four successive sections, followed by a
Finally, by way of introduction, is there anything special concluding critique. The relevance of the theory to humans is
about space? Declarative memory theorists think not. In their not considered further, and the reader referred to the very dif-
view, a memory system exists in the MTL that processes all ferent perspective offered in Chapter 12.
 Theories of Hippocampal Function 623

13.4.2 Representing Spatial Information, Locale contributes (Vanderwolf and Cain, 1994). There remains,
Processing, and the Hippocampal Formation however, the question of what information is gathered.
As the cognitive map theory asserts that this curiosity-
A key concept with respect to the representation and storage driven exploration is the basis for encoding a map, a critical
of spatial information is the distinction between egocentric experimental prediction is that renewed exploration would be
and allocentric space. Egocentric space refers to the locations triggered by alterations in the spatial arrangement of familiar
of objects relative to the viewerto the left or the right objects. Critically, hippocampus-lesioned subjects should dis-
directions that necessarily alter as the viewer moves around. play selective failure of renewed exploration to object dis-
Representing this kind of space is not thought to require placement but not to object novelty. Using a paradigm
a cognitive mapping system. Allocentric space, on the developed by Thinus-Blanc et al. (1987), Save et al. (1992a)
other hand, is a representation of spatial relations within some compared the impact of small dorsal hippocampal lesions,
kind of absolute framework. For a freely moving subject, anterior and posterior parietal lesions, and sham lesions on
an allocentric representation is required to identify whether exploration in a circular arena containing ve objects. In a sys-
an object or landmark has actually moved in relation to tematic series of short exploration sessions, the animals
others. In contrast, objects are constantly moving in egocen- became familiar with the initial set of objects and their spatial
tric representations. The distinction between egocentric locations, after which the critical manipulations of spatial dis-
and allocentric space has not always been appreciated by location and object novelty were introduced. The results
critics of the cognitive map theory, with some incorrectly showed renewed exploration by controls and anterior parietal
assuming that the theory predicts that lesions of the hip- cortex-lesioned animals in response to spatial dislocation;
pocampal formation must impair any and all spatial learning there was no detectable reaction by hippocampus- or poste-
tasks (e.g., Nadel and Hardt, 2004). Many spatial tasks given to rior parietal-lesioned subjects. In contrast, renewed explo-
animals, such as the simple T maze task described in Section ration was shown by all groups in response to object novelty.
13.2 and certain touchscreen tasks, are ambiguous because However, the effect was not large. In an ingenious follow-up of
they can be solved using either an allocentric or an egocentric this dissociation, Save et al. (1992b) examined exploration at
strategy. The study of exploratory behavior has provided a key a particular point in space that was triggered by the absence of
test of this feature of the cognitive map theory together with an expected object (Fig. 1318). Rats that had been subject
studies examining how animals localize objects in space. to preexperimental surgery rst explored an open circular
Lesion studies and observations of immediate early gene acti- arena that was slightly unusual in that it had a transparent
vation have contributed to a detailed examination of this oor. Another rat, in a small box, was sometimes placed
issue. underneath this oor in full view of the exploring animal.
In a sequential four-phase experiment, the experimental rats
Lesion Studies of Exploratory Behavior Reveal the rst explored the open but empty arena. They were then given
Importance of Allocentric Spatial Representation two successive opportunities to explore the arena again with
the stimulus rat visible at one zone of space, and the study
It has long been known that rats explore spontaneously, and ended with the nal phase that, like the rst, had no stimulus
the phenomenon has been systematically in the laboratory for rat available under the oor. A neat feature of this design
years (Berlyne, 1950, 1966; Halliday, 1968; Archer and Birke, is that the overt visual appearance of the arena was identical
1983; Renner, 1990). Placed in a novel test arena, the rats in sessions 1 and 4. As expected, introduction of the stimulus
remain motionless for a period of time and then start to move rat in session 2 triggered extensive investigation centered
around, rst at the periphery and then throughout the avail- on the zone above the stimulus rats location, and this habitu-
able space, where they explore objects and landmarks one by ated between sessions 2 and 3. The key test was session 4
one, often returning to an object just investigated until they with the stimulus rat now removed. In control rats with sham
nally come to rest in one corner. This is not mere movement; lesions, exploration was focused at the zone, but this did
it is exploration to gain information. For example, renewed not occur in animals with dorsal hippocampal or posterior
exploration can be triggered easily after habituation by intro- parietal lesions. The use of a recall paradigm may have
ducing a novel object into the arena, a phenomenon exploited made the magnitude of the effect much larger. Taken together,
in the visual-paired comparison task (see Section 13.3). these ndings imply three things: (1) exploration can be
Hippocampal lesions impair certain aspects of exploration. triggered in normal rats by object novelty, spatial novelty,
They tend to make animals hyperactive yet also disinclined to and the absence of an expected object; (2) damage to hip-
visit all regions of a space systematically and less likely to dis- pocampus and posterior parietal cortex selectively impairs
play habituation of activity over time (OKeefe and Nadel, exploration triggered by spatial novelty, including the absence
1978; Whishaw et al., 1983). OKeefe and Nadel (1978), Morris of an expected object; (3) when it occurs, this exploration is
(1983), and Renner (1990) have repeatedly made the point not merely activated but spatially directed in an appropriate
that the hippocampus-dependent information-gathering way. Such ndings are consistent with the construction of a
function of exploratory behavior should be clearly distin- map of space in the manner predicted by the cognitive map
guished from any motor functions to which the hippocampus theory.
624 The Hippocampus Book

A Response to spatial change in rats with HPC and parietal cortical (PPC) lesions

session S1 sessions S2 & S3 session S4

oriented exploration response to change
transparent floor
exploring rat stimulus-rat

z2
z1
NC
open field

z2 ???
H or PPC z1
lesions Figure 1318. Exploration guided by the
absence of an expected stimulus. A. In a
multiple phase study, rats explored an
arena with a transparent oor that some-
B Response to remembered places by C, HPC and PPC lesioned rats
times revealed the presence of another rat
200 under one zone (z1). Normal and both
C
hippocampus- and posterior parietal
mean % re-exploration

150 PPC lesions

H lesions cortex-lesioned rats explored differently
100 when this second rat was removed: only
control animals spent more time in z1
50
than z2, reecting their recall of the now
0 absent cue animal underneath. B. Data
-50 showing greater exploration of the area
where the now absent animal had been
-100 located. See text for more details. (Source:
zone1 zone2 After Save et al., 1992b.)

An outstanding puzzle is why similar disruption of explo- representation of space. This implies that steps in the compu-
ration induced by spatial novelty was seen in the posterior tation of allocentric space can occur upstream of the hip-
parietal cortex-lesioned group, a nding not predicted by the pocampus, although the multipeaked properties of entorhinal
theory if map storage is in the hippocampus. The absence of grid cells indicates that an individual entorhinal cortex
direct anatomical connections between this structure and the cell provides a metric rather than a specic indication of
hippocampus led Parron and Save (2004) to wonder whether location. Perhaps ensembles of entorhinal cells are able to
there may be some larger neuronal network involving the disambiguate on their own, or they may provide a Fourier-
interconnected regions of the retrosplenial and entorhinal like input to the dentate gyrus and hippocampus, where
cortices. In a follow-up study using the same behavioral pro- single-peaked cells are primarily observed. The discovery
cedures, they observed that lesions of these cortical structures of the entorhinal grid cells will have a big impact on the cog-
also caused disruption of renewed exploration in response to nitive map theory, but the full implications remain to be
spatial novelty. This nding adds to a growing body of evi- worked out.
dence from exploration and navigation tasks that the process-
ing of spatial information extends beyond the hippocampus Immediate Early Gene Studies Reveal
(Poucet and Benhamou, 1997; Kesner, 2000; Whishaw, 2004a; Anatomical Dissociations with Respect
Hafting et al., 2005). to Environmental Representations
For example, although some studies have suggested that
neurons in the entorhinal cortex are only weakly modulated Instead of using lesions, visualizing the expression of immedi-
by the position of the animal (Quirk et al., 1992; Frank et al., ate early genes (IEGs), such as c-fos, Arc, and zif-268 (Morgan
2000), the dorsomedial region of the entorhinal cortex has et al., 1987; Link et al., 1995; Lyford et al., 1995; Tischmeyer
grid cells whose ring shows multipeak elds that are small, and Grimm, 1999), has been used to examine the patterns
stable, and directionally insensitive (Fyhn et al., 2004; Hafting of activation in multiple brain areas during both spatial and
et al., 2005). These ring elds are unaffected by hippocampal nonspatial tasks. Some of these studies have focused on the
lesions, implying that their existence is not secondary to representation of the environment, others on spatial naviga-
upstream hippocampal processing that might be projected tion, and yet others on memory consolidation (Dragunow,
back to the entorhinal cortex. Instead, they suggest that at the 1996; Guzowski, 2002). Different IEGs are used in different
border between the entorhinal and postrhinal cortex (a region studies. The choice can take advantage of the fact that some
almost certainly lesioned in the Parron and Save, 2004, study IEGs are regulatory transcription factors that inuence the
of exploration) there is an allocentric, view-independent activity of other downstream genes (e.g., c-fos), whereas
 Theories of Hippocampal Function 625

other IEGs are effector genes that directly affect cellular phys-
computer monitor
iology (e.g., Arc).
There are several features of the IEG approach that are
noteworthy. First, from a low level of basal expression in ani- novel picture familiar picture
mals, experience-dependent activation of IEGs is correlated
with increases in neuronal activity (Herrera and Robertson,
1996). There is, for example, a broadly similar activation pro-
le for c-fos, zif-268, and Arc in association with spatial learn- 30cm
ing, with some indication of differential sensitivity to
manipulation of task demands (Guzowski et al., 2001). 50
50
Second, the imaging observations made after sacrice are juice tube perspex screen
45
from animals whose brains are normal at the time of testing;
thus, IEG mapping links behavior to physiology in a manner
Rat observing hole
that complements the single-unit recording studies of
Chapter 11. This does not preclude combining the lesion and
IEG approaches, and studies using such a combined approach Figure 1319. Bilateral viewing technique to image immediate
have revealed alterations in activity in areas remote from the early genes such as c-fos. Novel and familiar visual patterns are pre-
site of the lesion (Aggleton and Pearce, 2001). Third, although sented simultaneously to each eye. The largely crossed visual system
the analytical focus of a given study may be on activation of a of the rat enables the various c-fos patterns to be compared within
specic brain structure, the approach does not require a individual animals and then averaged. (Source: Courtesy of M.
prospective choice, as quantitative measures may be taken Brown and J. Aggleton.)
of a large number of brain structures. The drawback of the
IEG approach is that, unlike single-unit recording for which A limitation of the bilateral viewing technique is that there
the measured output is a signal that is well understood in is no behavioral output of the simultaneous processing of
information-processing terms, the meaning of an IEG signal novel and familiar stimulus combinations in the two hemi-
is unclear. There has been substantial discussion about the spheresindeed, no way of establishing whether the animals
possible functions of IEGs, ranging from metabolic replenish- even attend to the cues. Jenkins et al. (2004) therefore followed
ment to regulation of synaptic plasticity, memory consolida- up the Wan et al. (1999) study by a systematic analysis of nor-
tion, and even memory retrieval (Guzowski, 2002). malized c-fos patterns in animals running through and making
The technique was rst used in relation to behavioral stud- choices in a radial maze task (see below for a description of this
ies of exploration by Hess et al. (1995), who observed wide- classic navigation task). Following a procedure rst developed
spread hippocampal IEG activation in various subelds. This by Suzuki et al. (1980), they rst trained rats with eight dis-
was a valuable rst step, but it did not speak to whether the tinctive cue cards placed between the arms of the maze. Then,
hippocampus encodes information allocentrically. IEG activa- on the nal day of training, the cue cards were left in their
tion may merely occur in the hippocampus in response to standard arrangement for half the animals (Group Familiar)
novelty, stress, or the motor activity that accompanies explo- or shifted to a novel spatial arrangement for the others (Group
ration. An important step forward was made in a study using Novel) (Fig. 1320). Spatial novelty caused c-fos activation in
c-fos by Wan et al. (1999) who presented rats with computer- CA1, CA3, and the dentate gyrus, whereas no changes were
displayed images in which novelty was introduced as either observed in the surrounding cortical regions such as the
new stimuli or novel spatial arrangements of hitherto familiar perirhinal and parahippocampal cortex. This was not due to
stimuli, a comparison analogous to the original lesion studies differential sensitivity of the imaging method between hip-
just discussed (Save et al., 1992a,b). A within-subjects proto- pocampus and cortex because successful c-fos activation fol-
col was used in which the images were presented bilaterally lowing spatial novelty was seen in the posterior parietal cortex.
such that familiar stimuli were presented to one eye and novel In addition, animals in a companion study were shown to be
stimuli (or novel rearrangements of familiar stimuli) to the using the cue cards in their representation of the maze, as
other eye (Fig. 1319). As the projections of the rat visual sys- indexed by behavioral sensitivity to cue rotation. These nd-
tem are largely crossed, it was possible to compare normalized ings nicely complement the lesion observations of Save et al.
patterns of activation in the two hemispheres within individ- (1992a,b). That the hippocampal formation can detect alter-
ual animals. Any nonspecic c-fos activation associated with ations in the spatial arrangement of cues and respond by dif-
novelty, stress, or motor activity was effectively and elegantly ferential activation of an IEG strongly suggests that the
subtracted out. The key nding was that novel rearrange- information encoded about the familiar stimuli presented ear-
ments of familiar stimuli to one eye differentially activated the lier must have included some representation of their spatial
CA1 region of the hippocampus. Novel stimuli activated the layout in the image or in the room. They support the notion
perirhinal cortex, as previously shown by Zhu et al. (1995). that the hippocampus is both necessary for and involved in the
Brown and Aggleton (2001) summarized a body of work representation of space. However, they also point to unex-
using this approach. pected involvement of the posterior parietal cortex as well.
626 The Hippocampus Book

A Spatial arrangement of cues during training (left) B Hippocampal (left bars) and cortical (right) c-fos
and c-fos measurement period (right) patterns following spatial cue re-arrangement

Group Familiar

normalised counts of c-fos nuclei

c-fos test day 100
Familiar
Novel
75
* * *
50

Group Novel
25

0
CA1 CA3 DG Peri EntL EntM

Figure 1320. Immediate early gene activation reveals regional c-fos patterns show the differential sensitivity of structures in the
patterns of sensitivity to spatial novelty. A. The spatial arrangement hippocampus (CA1, CA3, DG) versus neocortical regions such as
of the cues placed between the ends of the arms of the maze was perirhinal and entorhinal cortex (medial and lateral). (Source: After
retained between training and the c-fos test session for Group Jenkins et al., 2004.)
Familiar but was changed for Group Novel. B. The normalized

Guzowski et al. (1999) introduced a highly illuminating technique would not have revealed that the activation
new technique in which it is possible to identify time-locked response of the different environments was specic, whereas
expression of IEGs using uorescent in situ hybridization Arc catFISH provides novel information about the informa-
(called catFISH after the cellular compartment in which time tion content of neuronal activation.
of expression is observed). It was rst applied to the imaging In an extension of this work, Vazdarjanova and Guzowski
of Arc on its own and later extended to enable simultaneous (2004) have compared the sequential IEG activation patterns
imaging of Arc and Homer. The catFISH technique enables a in CA3 and CA1 in an exploration task very like that used by
within-subject comparison of a different kind in which the the Marseille group: a box containing objects that can be
population of neurons activated by one experience (e.g., moved about. Exploiting a further technical innovation, con-
exploring a box) can be distinguished from a second popula- focal images of Homer reected neuronal activity that had
tion activated by a later experience (e.g., exploring a different occurred approximately 30 minutes before sacrice, whereas
box). Arc RNA rst appears as discrete small intranuclear sig- nuclear images of Arc reected more immediate neuronal
nals within 2 to 15 minutes of the cells being activated (e.g., activation. The study revealed that when confronted by two
the activation of a place cell as an animal occupies its place object arrangements successively (A then B), cells in area CA3
eld). After 20 to 45 minutes the nuclear signal disappears, showed discontinuous activation patterns that were highly
and a cytoplasmic mRNA signal develops that is translocated suggestive of pattern separation. However, when confronted
to dendrites. As the time course of the nuclear signal is distinct with either identical or very similar boxes (A then A or A), the
from the cytoplasmic signal, the activity history of individual patterns overlapped very well, suggestive of pattern comple-
neurons at two time points can be inferred (see this chapter, tion. The patterns in CA1 were more continuous across these
gures in color centerpiece). Two predictions of the cognitive environmental manipulations. Integration of distinct CA3
mapping theory were then tested. First, it was established that and CA1 ensemble representations may provide a rened rep-
a common population of place cells was activated when the resentation of spatial context reecting intrahippocampal
animals explored the same box (A) on two successive occa- processing of information projected from the entorhinal cor-
sions, with the same population of cells expressing Arc mRNA tex. The observations of Leutgeb et al. (2004) on the ring
in dendrites (signaling the rst visit to the box) and in the properties of CA3 and CA1 cells in various boxes and contexts
soma (more recent visit). Second, they observed Arc RNA are consistent with this conclusion.
expression in two distinct but overlapping populations of cells
on sequential visits to two different boxes (A then B). In boxes Distributed Representations, Pattern
approximately 70 cm in diameter (about the same as those Separation, and Pattern Completion
used in many studies of place cell ring elds), they found
that approximately 38% of the CA1 pyramidal cells showed The lesion and IEG studies of exploratory behavior and reac-
both the dendritic and the somatic IEG signals, whereas oth- tions to novelty clearly point to a role for the hippocampus in
ers displayed only one of the two signals. The standard c-fos representing spatial information, but on their own they say lit-
 Theories of Hippocampal Function 627

tle about how spatial information is represented. Single-cell D-aspartate (NMDA) receptor (see Chapters 6, 7, and 10) was
and ensemble recording is the way forward to understand and excised selectively in the CA3 region of the hippocampus.
quantify the representational structure of information (see Using the Cre-Lox technique to realize this selective ablation,
Chapter 11) guided by neural network models of how the sys- the resulting progeny (CA3-NR1 knockout mice) were shown
tem might work (see Chapter 14). There is also a role for to have normal physiology and behavior on a range of tests
behavioral studies in identifying neural network processes but certain highly specic impairments. For example, whereas
such as pattern separation and pattern completion that act on LTP was normal in the mossy ber input to CA3 and in the
such representations (Rolls and Treves, 1998). CA3 output to CA1, it was absent in the longitudinal associa-
With one approach to pattern separation, Gilbert et al. tion pathway of CA3. This pathway is likely to be important
(1998) trained hungry rats to forage for food on a cheese- for building associative spatial relations between cues in an
board apparatus with a series of food wells arranged in a row environment (McNaughton and Morris, 1987; Treves and
such that the separation between the wells varied from 15 to Rolls, 1994). When trained in a watermaze (see Fig. 1323,
105 cm. Training took the form of pairs of trials, a sample trial below, for a description of this task), the mice learned nor-
followed by a choice, with as many as 16 trial-pairs per day. mally with all cues present but showed selective decit in
During each sample trial, the rats ran from a start box at the recall when tested with a subset of the training cues.
periphery of the cheeseboard to displace a single object over Interestingly, place-specic ring of CA1 complex-spike cells
one of the wells to secure food. During the choice trial that fol- was also fuzzier in the mutants when recording sessions
lowed a few seconds later, two identical copies of this same took place using partial cues. Nakazawa et al. (2002) inter-
object were placed to cover two wells, with one object covering preted this observation as implying a role for NMDA recep-
the same baited well, the other an empty one farther down the tors in area CA3 of the hippocampus in pattern completion,
row (i.e., delayed matching to place). This task provided a way and they outlined a model of the way areas CA3 and CA1 may
to measure how easily the rats could discriminate spatial loca- interact during the storage of information (Fig. 1321). This
tions in memory as a function of their distance apart. Prior to study also nicely illustrated the power of adult-onset, cell
being lesioned, choice performance was very good, as high as type-restricted genetic manipulations in the integrative study
80% on even the smallest distance and rising to 90% at the of the molecular, cellular, and neuronal circuitry mechanisms
largest separation. Large electrolytic lesions of the hippocam- underlying spatial cognition, a considerable advance on rst-
pus caused performance at the smallest separation (15 cm) to generation transgenic techniques (Nakazawa et al., 2004).
fall to chance, but there was a graded recovery, as this distance
increased reaching nearly 90% at the largest distance (105 cm). 13.4.3 Using Spatial Information: Spatial
Performance was also unaffected by changing the start loca- Navigation and the Hippocampal Formation
tion by 180 between sample and choice trials. This pattern of
performance indicates that one function of the hippocampus The second key idea in cognitive map theory is that neural
may be to separate patterns of spatial information, this being activity in the hippocampal formation occurs during and is
possible without the hippocampus for widely separated cues required for spatial navigation. Allocentric navigation utilizes
but not for cues close together. Findings consistent with this the information stored as a map during earlier locale pro-
interpretation using a radial maze task had also been reported cessing, and reading information out of maps requires hip-
by McDonald and White (1995). In a follow-up study using pocampal place cells to re. Other cognitive processes related
the cheeseboard, Gilbert et al. (2001) showed that discrete to navigation involving the hippocampus include the ability to
neurotoxic lesions of the dentate gyrus (DG) caused a compa- mentally rotate a stored representation to align it with a cur-
rable pattern of impairment, whereas CA1 lesions had no rent view, to discover and use shortcuts, and so on. A key pre-
effect. This dissociation is intriguing, but it raises the puzzle of diction is that the integrity of the hippocampal formation is
how spatial information, orthogonalized at the level of the absolutely required for locale-based navigation. This predic-
granule cells in the DG, is able to inuence target structures of tion led to the development of a number of novel navigational
the hippocampus in animals that have been subjected to CA1 tasks for rats, the rst of which was the radial maze.
lesioning. The usual output for information, however effec-
tively it is processed in DG, would no longer be available. An Radial Maze
output route via the ventral hippocampus is one possibility as
the CA1 lesions were restricted to the dorsal hippocampus. The radial maze is an apparatus in which rats (or mice) are
There may also be the possibility of output directly from the given the opportunity to collect food hidden at the end of
CA3 to the lateral septal nucleus, but the route to the neocor- arms of the maze that radiate out from a central choice area
tex is then indirect. It would be interesting to explore this (Olton and Samuelson, 1976). In its simplest version, a hun-
route, as the nding hints at a radical new conception of infor- gry rat is placed on the maze and allowed to run around until
mation ow within the hippocampal formation. it has collected all the food. Once the food at the end of one
Another innovative approach to pattern completion has maze arm has been eaten, it makes sense to avoid that arm,
involved molecular-genetic techniques to address what is, in which normal rats gradually learn to do. Once mastered, the
effect, a systems-level issue. Nakazawa et al. (2002) engineered rats run to each arm without revisiting arms a second time
a mutant mouse in which the NR1 subunit of the N-methyl- (Fig. 1322A). They have, in effect, solved a special case of the
628 The Hippocampus Book

Control Mice CA3-NR1 mutant mice

A Full cue condition B Full cue condition

EC input DG/EC input EC input DG/EC input

SC SC

CA1 CA3 CA1 CA3

firing firing
rate RC rate
position

C Partial cue condition D Partial cue condition

SC SC

CA1 CA3 CA1 CA3

firing firing
rate RC rate
position

Figure 1321. Pattern completion in the storage and retrieval of the normal mice make associative changes in synaptic efficacy in the
spatial memory using areas CA3 and CA1. Note the interconnecting commissural-association pathway (small circles representing poten-
circuitry of CA3 and CA1, emphasizing pathways that form NR- tiated synapses), whereas CA3-NR1 KO mice do not. C, D. In later
dependent modiable synapses in CA3. A, B. Basic wiring of CA3 testing under partial cues (bottom row), a fraction of the original
and CA1 illustrates a possible encoding mechanism for pattern com- input is provided (fewer downward arrows) to activate any memory
pletion. In normal and CA3-NR1 knockout (KO) mice (left and traces formed during training. Only control mice are able to benet
right columns, respectively), a full cue input (top row) is provided to from the associative plasticity in CA3 that completes the pattern
CA3 from the dentate gyrus (DG) or the entorhinal cortex (EC) and and so provides an output to CA1 that can then be used for spatial
to CA1 from the EC (downward-facing arrows). During learning, recognition and navigation. (Source: After Nakazawa et al., 2002.)

traveling-salesmans problem of getting everywhere with the Olton and Papas, 1979) adopted a modication of the task,
least effort. Olton (1977) reported that rats with various originally suggested by Jarrard (1978), in which both sham-
lesions that deafferented or deefferented the hippocampus operated and fornix-lesioned rats were tested in a 17-arm
(e.g., fornix or entorhinal cortex lesions) were severely radial maze in which a xed set of arms were baited before
impaired in their initial learning. Whereas normal animals each trial and others were never baited (Fig. 1322B).
collected all available food from the ends of the maze arms Following pretraining as normal animals, these investigators
with little or no retracing, those with lesions moved around found that fornix-lesioned rats were impaired in remember-
the maze haphazardly, making numerous repeat entries into ing to avoid repeat entries into arms from which they had col-
arms from which they had already obtained food. The initial lected the available food but were as successful as the
interpretation of these data was that these surgical disconnec- sham-lesioned animals in avoiding the arms that were never
tions of the hippocampus disrupted spatial navigation and so baited. Essentially the same results were obtained whether the
supported the cognitive map theory. baited arms were arranged adjacent to each other or were
However, Olton began to doubt his interpretation follow- mixed haphazardly with the never-baited arms. This nding
ing a study suggesting that the decit was really in trial- conicts with the cognitive map theory, as neither a distur-
specic working memory. This term was introduced into the bance of spatial representation nor of navigation can explain
animal literature by Honig (1978) to refer to the memory of the dissociation between failure to avoid repeat entries into
information useful for only a single trial in a behavioral task.* the initially baited arms and successful avoidance of the never-
*Working memory has a different meaning in human memory research, baited arms.
where it refers to a limited capacity consisting of several subsystems such Olton et al. (1979) therefore argued for a distinction
as a phonological store and visuospatial sketch pad (Baddeley, 2001). between the anatomical mediation of working memory (i.e.,
 Theories of Hippocampal Function 629

A The 8 arm radial-maze B A 17 arm radial-maze with some maze-

arms baited and others unbaited.
reward to be collected
reward collected
no reward
rat path

Figure 1322. Radial maze. A. The eight-arm radial maze is a task might be that of a fornix-lesioned rat, initially trained preopera-
in which rats run from a central region to collect food reward from tively, that is selectively visiting the maze arms that are always
the ends of the maze arms. The path shown is of a rat that has suc- baited, making occasional repeat errors at these arms but success-
cessfully made ve choices without retracing its path and has three fully avoiding the maze arms that are never baited. (Source: After
arms remaining where food is available (notional path data). B. A Olton and Papas, 1979.)
17-arm radial maze with baited and unbaited arms. The path shown

memory required for only one trial) and reference memory even with doors, it is unclear whether a spatial map is guiding
(i.e., memory used for a series of trials). They proposed that in successive choices in this task (Brown, 1992; Brown et al.,
radial maze tasks in which some arms are baited and some 1993). Third, in the Olton et al. (1979) study, the animals were
unbaited, the animal uses working memory to keep track of initially trained as normal animals and were lesioned only
which arms it had entered as the trial proceeds and reference after asymptotic performance had been reached. This training
memory to avoid entries down arms that never contained arrangement leaves open the possibility that the integrity of
food. Oltons insistence on thinking about reference and the hippocampal formation is required for the initial learning
working memory as independent rather than interacting of which arms to avoid but perhaps not required for retention.
memory systems carries with it the implication that the ani- Using animals that had undergone surgery prior to training, a
mal is reading out from two memory systems simultaneously series of studies by Jarrard (1978, 1986) established that the
as it works its way from choice point to choice point a pre- initial learning of place reference-memory in an eight-arm
scient idea, as many behavioral tasks are likely to involve mul- radial maze task in which four arms were never baited is con-
tiple memory systems. In any event, if the hippocampus is sistently impaired by both complete hippocampal aspiration
involved exclusively in working memory, not only could this lesions and by more selective ibotenic acid lesions of the hip-
account for the data it also implies that there may be a subset pocampus and dentate gyrus. These studies were important
of allocentric spatial tasks for which the integrity of the hip- for introducing the most selective type of hippocampal lesion
pocampal formation is not requiredspatial reference mem- that has yet been developed. In the process, the cognitive map
ory tasks. theory was partially salvaged as the integrity of the hip-
The results of the Olton et al. (1979) study were, on the face pocampus is required for initial learning. However, the long-
of it, damaging to the cognitive map theory; but several prob- term storage of information about which arms are baited or
lems soon became apparent. First, Olton examined a variety of unbaited may be outside the hippocampal formation (Barnes,
lesions but, strangely, never looked at discrete hippocampal 1988). The theme that the hippocampal formation may be
lesions. Second, there is nothing to stop rats choosing adjacent required for the initial learning of a spatial task with eventual
arms in the standard radial maze, thereby performing per- storage in the cortex is one that recurs in many tasks, includ-
fectly without having to remember anything about the spatial ing the next one we consider.
layout of the maze. The path shown in Figure 1322A indi-
cates that our notional rat did not do this, but many radial Watermaze
maze studies ran into this problem. It was solved by placing
doors between the central area and the maze arms that pre- Further evidence that the integrity of the hippocampal for-
vented successive left or right turns by allowing the experi- mation is essential for allocentric spatial reference memory
menter to control the timing of the animals choices. However, came with the introduction of the watermaze (Morris, 1981).
 Box 135
Watermaze

The watermaze has become one of the most widely used tasks in behavioral neuroscience. It
was developed to study spatial learning and memory but is now often used as a general assay of
cognitive function for testing new drugs or other treatments of the nervous system.
The basic task is simple. Animals, usually rats and mice, are required to escape from water
onto a hidden platform whose location can normally be identied only using spatial memory.
There are no local cues indicating where the platform is located. Conceptually, the task derives
from place cells, as these cells also identify points in space that cannot be dened in relation
to local cues. This simple water escape task is then embedded into a variety of sometimes quite
complicated training and testing protocols to investigate specic theoretical issues.
APPARATUS

The apparatus consists of a large circular pool, generally 1.5 to 2.0 m in diameter, containing
water at around 25C made opaque by adding milk or other substance (e.g., latex) that helps
hide the submerged platform. The water in the pool is lled and drained daily via an automated
lling and draining system. The choice of water temperature, around 13C below body temper-
ature, is sufficiently stressful to motivate the animals to escape but insufficiently stressful to
inhibit learning. There is a stress reaction (corticosterone release) on day 1 of training, but this
habituates over days. If the pool temperature drops to 19C, performance improves, but when it
drops to 12C, it gets worse, reecting the inverse U-shaped function relating stress to cognitive
function. The pool is located in a laboratory room with distinctive two- and three-dimensional
distal cues that aid orientation or surrounded with hanging curtains that occlude these room
cues. Cues may then be hung inside the curtains so they can be rotated relative to the room
when this degree of experimental control is required. A video camera is placed above the center
of the pool to capture images of the swimming animal and is connected to a video or DVD
recorder and an online computer system running specialized tracking software. The top surface
of the hidden platform, usually about 10 to 15 cm in diameter and thus between 1/50th and
1/100th of the surface area of the pool, is 1.5 cm below the water surface (Fig.1323A).
PROTOCOLS

Reference memory protocols have been widely used in which the platform is in a xed loca-
tion relative to the room cues across trials and days. The animals are placed in the water at and
facing the side walls of the pool at different start positions across trials, and they quickly learn
to swim to the correct location with decreasing escape latencies and more direct swim paths
(Fig. 1323B). The tracking system measures the gradually declining escape latency across trials
and parameters such as path length, swim speed, directionality in relation to platform location,
and so on. Observation of the animals reveals that, having climbed onto the escape platform,
they often rear up and look around, as if trying to identify their location in space. Rearing
habituates over trials but then dishabituates if the hidden platform is moved to a new location.

Figure 1323. Watermaze. A. Axonometric drawing of a typical watermaze setup with overhead
video-camera and rat swimming to nd the hidden platform. B. Representative escape latency
graph and swim paths across various stages of training: initial swimming at the side walls, then
circuitous paths across the area of the pool, and nally directed path navigation. C. The hidden
platform is removed for posttraining probe tests. Whereas normal or sham-lesioned controls
swim to the target quadrant (NE, within dotted gray lines), rats with hippocampus, subiculum,
or combined lesions do not. (Source: After Morris et al., 1990.) D. Overtraining of hippocam-
pus-lesioned rats can result in quite focused search patterns during a probe test. (Source: After
Morris et al., 1990.) E. Atlantis platform. The hidden platform is at the bottom of the pool
where the swimming rat cannot bump into it by chance. Online automated data capture of
swim paths is used to determine whether the rat swims within a virtual zone around the plat-
forms location, raising the platform to within 1.5 cm of the water surface when a criterion is
reached. This protocol trains highly focused search patterns. (Source: After Spooner et al., 1994.)
F. Reversible hippocampal inactivation with a glutamate antagonist during training (encoding)
or at retention (retrieval) results in poor probe test performance compared to the controls,
which have the hippocampus working continuously. Analysis of the search patterns show that
rats trained with the hippocampus on during encoding, but off at retrieval display searching
at the wrong location in the pool, as quantied using zone analysis. (Source: After Riedel et al.,
1999.)

630
 Theories of Hippocampal Function 631

A The watermaze B Paths and latency during place navigation

video
camera start mid end

Escape latency (sec)

hidden
platform
pool

base board

C Post-training probe tests (no platform) D Overtraining

control hippocampus subiculum hippocampus & hippocampus
lesioned lesioned subiculum lesioned lesioned

E Atlantis platform F Dissociating spatial memory and search strategies

by inactivation at encoding or retrieval.
on/on on/off off/on

Water surface
% time in trained quadrant

50

40

30 chance
20
zone
10 analysis
Platform Platform
down up 0
on/on on/off off/on off/off

G Delayed matching to place (DMP)

2h TI
I
IT
I
s
15

Day T1 T2 T3 T4 80 sham
ITI lesion
latency (sec)

N 60
short
N+x 40
medium 20
N+y 2h
long 0
T1 T2 T3 T4 T1 T2 T3 T4
Trials within a day, averaged across days

Figure 1323. (Continued) G. Delayed match-to-place training involves four trials per day with
the location of the hidden platform moved between days. Training can continue indenitely
with this protocol, enabling within-subject drug manipulations throughout the life-span and
averaging across days. Acquisition typically takes 8 to 10 days. For trial 1 of each day, the ani-
mals search for the platform, typically taking 60 seconds to nd it, and encode its location; they
then show fast escape latencies during trials 2 to 4. Hippocampus-lesioned rats cannot learn
this task irrespective of the intertrial interval (ITI) between trials 1 and 2. The shaded zone
shows the ITI between T1 and T2 extended to 2 hours. (Source: After Steele and Morris, 1999.)
(Continued)
 Box 135
Watermaze (Continued)

During or after training is complete, the experimenter conducts a probe trial in which the
escape platform is removed from the pool, and the animal is allowed to swim for 60 seconds.
Typically, a well trained rat swims to the target quadrant of the pool and repeatedly across the
former location of the platform until it starts to search elsewhere (Fig. 1323C). This spatial
bias, measured in various ways, constitutes evidence for spatial memory. Rats with lesions of
the hippocampus and dentate gyrus, subiculum, or combined lesions, do poorly in posttraining
probe tests.
However, if rats with hippocampal lesions are given overtraining (typically consisting of a
large number of trials over many days), performance can improve in probe tests. Even rats with
hippocampal lesions can show quite localized searching (Fig. 1323D).
Numerous other protocols have been developed to test specic hypotheses. Many involve
cryptically moving the hidden platform. This might be a reversal procedure in which, after
one location has been thoroughly trained the platform is moved to a different quadrant of the
pool. Because it is hidden, it is not apparent that anything has changed until the animal fails to
nd the platform in its usual place. The focus is on how the animal reacts to this change and
how quickly it learns the new location. The relearning that occurs with the reversal protocol has
been used in a major factor analysis of the determinants of watermaze behavior across strains
of mice.
As the animals sometimes bump into the submerged platform by chance, one useful inno-
vation is an on-demand or Atlantis platform that is initially at the bottom of the pool and
becomes available only when the animal swims in its vicinity for some predetermined time. An
automatic release system allows the platform to rise gently to near the surface of the water (it
remains hidden). This procedure results in acquisition of a highly focused searching strategy
that focused on the target location during training. Reversible inactivation of the hippocampus
with a drug that blocks excitatory neurotransmission after the training is completed results in
animals displaying localized searching at inappropriate places in the pool. Pharmacological
inactivation during training results in failure to develop this search strategy because the ani-
mals cannot learn the place at which to execute the strategy in the pool (Fig. 1323E). The
accuracy of the searching can also be measured using a zone analysis that measures time spent
in a virtual zone around the place where the platform is located.
With other protocols, the platform can be moved to a new location each day, creating what is
called the delayed matching to place (DMP) procedure (Fig. 1323F). The animal cannot
know, with this procedure, where the platform is hidden on trial 1 of each day; it can only
search for it. However, if it can encode this new location in one trial, the animal nds the plat-
form much faster during trial 2 and subsequent trials of that day. The memory delay interval
between trials 1 and 2 (ITI) can then be systematically varied to explore how well one-trial spa-
tial memory is remembered, a procedure with some similarities to delayed matching and non-
matching tasks used to examine recognition memory. Rats with complete hippocampal lesions
cannot show the rapid, one-trial learning required for the DMP task and are just as poor at a
short ITI between trials 1 and 2 as a long one (Fig. 1323G). Treatment with an NMDA antago-
nist such as D-AP5 results in a selective decit in memory at a long ITI, but the animals can
learn with short memory intervals between trials (see Chapter 10, Section 10.9).
Other variants include alterations to the apparatus, such as constraining the path of the
swimming animal to minimize navigational demands (e.g., an annular watermaze), decreasing
the number of available extramaze cues between training and testing to look at pattern comple-
tion, the use of oating platforms, and yet other manipulations. A radial watermaze has also
been introduced, combining the virtues of the radial maze with the ease of training to escape
from water. This has proved invaluable for testing transgenic mice expressing familial
Alzheimer mutations.
TREATMENTS AND CONTROLS

A wide variety of treatments have been explored including lesions, drugs, and molecular-
genetic alterations. They alter watermaze performancez in various ways, but experimenters
must be cautious, as such alterations need not be specic decits (or improvements) in spatial
learning or memory processes per se. Lesions or drugs may have a direct effect on learning
mechanisms, and many seem to do so, but they may also affect an animals ability to see the
extramaze cues or their motivation to escape from the water rather than learning per se. Factor
analytical studies reveal that many molecular-genetic alterations inuence the probability of
mice to stay at the side walls (thigmotaxis) instead of swimming into the center of the pool and
that this effect is statistically independent of spatial memory.

632
 Theories of Hippocampal Function 633

Accordingly, treatments must be accompanied by relevant control conditions. A common

protocol is to include trials in which the escape platform is made visible, the idea being that
treatments that merely affect motivation to escape should impair performance in this task as
well as the basic task. It is unclear how sensitive this assay is, but it does provide a rst pass at
detecting gross sensorimotor abnormalities. As blind rats have been claimed to do surprisingly
well in the watermaze (except in probe trials), other psychophysical techniques have been intro-
duced offering precise control of the spatial frequency of cues that are more sensitive to subtle
visual decits. The use of sham-lesioned animals, vehicle-infusion conditions, oxed mice,
and other pertinent manipulations has also become widespread to ensure that any alterations
in spatial learning in the experimental group are not an unintended by-product of achieving
the treatment.

WATERMAZE AS AN ASSAY

As our understanding of the impact of various treatments has developed, the watermaze has
gradually been subsumed into test batteries and assays for investigating aspects of brain func-
tion other than just hippocampal function. This is a strength of the assay but also an analytical
weakness. The task is so sensitive to manipulations of normal brain function in many brain
areas, not just the hippocampus, that it can be used almost like a litmus test of the normal-
ity of brain function. This is valuable, as it brings behavioral observations of function into
elds of neuroscience that have historically relied exclusively on stress (corticosterone release),
neuropathology (stroke research), biochemical analyses (Alzheimers disease), or electrophysiol-
ogy (development of cognitive enhancing drugs). The analytical weakness is that a task affected
by such a wide variety of treatments is gradually being revealed as having less specicity than
was once believed.
Innovation in the development of new training protocols is enabling the simplicity and
speed of learning to escape from water to be retained in the arsenal of behavioral tools at the
disposal of neuroscientists. Arguably, the watermaze illustrates both themes of this book: (1) a
tool to study hippocampal function and (2) an illustration of how studies of hippocampal
function have had a wide impact on neuroscience in general.

SELECTED REFERENCES

Morris, 1981, 1984; Sutherland and Dyck, 1984; Brandeis et al., 1989; Good and Morris, 1994;
Nunn et al., 1994; Whishaw and Jarrard, 1996; Gallagher and Rapp, 1997; Lindner et al., 1997;
Sandi et al., 1997; Lipp and Wolfer, 1998; Morgan et al., 2000; Prusky et al., 2000a; DHooge
and De Deyn, 2001.

Morris et al. (1982) trained rats to nd a hidden escape plat- decit in escape latency disappeared when the platform was
form in a circular pool of opaque water. Some had been sub- made visible, indicating that the impaired place navigation
jected to hippocampal aspiration lesioning (damaging the was unlikely to be secondary to any change in motivation to
hippocampus, dentate gyrus, and subiculum), others cortical escape from the water and reappeared when it was subse-
control lesioning (damaging the same small amount of corti- quently hidden again. Similar results were obtained in the rst
cal tissue in the region of the parietal cortex as was necessar- study to use neurotoxic lesions in the watermaze by
ily damaged when making the hippocampal lesion), or no Sutherland et al. (1983), investigators who were primarily
surgery (unoperated control). The cortical control lesion responsible for popularizing the task (Box 135).
was included in this and many other early rat studies (in con- Over the years since the watermaze was developed, there
trast to the monkey work discussed in Section 13.3). In the have been numerous replications of the decit in place-
watermaze, both the normal and the cortical tissue-lesioned navigation learning caused by hippocampal lesions. Other
groups quickly learned to swim toward the hidden platform pertinent ndings concern the impact of lesions to structures
from any starting location; only the hippocampus-lesioned afferent or efferent to the hippocampus and of ber pathways
group was impaired in executing directed swim-paths. interconnecting these structures, as comprehensively reviewed
Performance during a posttraining probe test showed that
There have been differences in opinion about the name watermaze for
both control groups swam persistently across the exact spot
an apparatus that is not obviously a maze. Some describe it as a water
where the platform had been located during training, a pat-
task, others as a milk bath, and many use an acronym MWM. The use
tern of behavior that could not occur during training and of the term maze is appropriate, as it is an apparatus in which the ani-
was arguably reminiscent of free recall. In contrast, the mal has to nd its way aroundand that is what one does in a maze. By
hippocampus-lesioned group swam all over the pool. The analogy, golfers play on a golf course not golf elds.
634 The Hippocampus Book

by DHooge and De Deyn (2001). For example, decits in cial navigation system (Trullier et al., 1997). These machines
place navigation have been reported after selective lesions of clean the oor as they work their way around a living space
the entorhinal cortex (Schenk and Morris, 1985), perforant and typically incorporate two key features of spatial memory.
path (Skelton and McNamara, 1992), dentate gyrus They remember where in the room to go to get their battery
(Sutherland et al., 1983), area CA1 (Nunn et al., 1994), medial recharged and to avoid recleaning places that they have
septum (Hagan et al., 1988; Kelsey and Landry, 1988), fornix already moved across.
(Morris, 1983), and subiculum (Taube et al., 1992). Decits
are generally seen only after bilateral lesions. The overwhelm- Varied Training Protocols Reveal the Importance
ing consensus is that the integrity of the hippocampal forma- of the Dorsal Hippocampus for Spatial Navigation
tion and its afferent and efferent connections is essential for
the acquisition and/or consolidation of normal, appropriately A key distinction regarding the watermaze is between thinking
directed spatial navigation. of it as an apparatus and as a task. Sometimes described as the
However, as with the radial maze, further study has revealed latter, it is really an apparatus in which a wide variety of train-
lesion dissociations that are not predicted by the cognitive map ing protocols can be carried out to address specic issues.
theory and point to a functional heterogeneity that underlies Some modications have come into widespread use, others
normal performance. Dysfunction in a variety of structures not. One was a modication of the original spatial reference
has an impact on performance, notably damage to neocortical memory protocol (Morris et al., 1986a) that for various rea-
structures such as the retrosplenial and parietal cortices, sons was not taken up by other laboratories. It involves train-
frontal cortex, basal forebrain, striatum, and cerebelluman ing rats to choose between two visible platforms, one of which
issue addressed in the critique below. A problem with the is rigid and offers escape from the water, the other oating and
watermaze is that its procedural simplicity belies an underly- so offers insufficient buoyancy for escape. Local-cue and spa-
ing complexity with respect to the numerous sensorimotor tial versions of this task were compared. Hippocampal lesions
and cognitive processes that, in practice, it engages. That it has impair the spatial task but have no effect on the procedurally
been classied as a spatial memory task should not lull users similar cue task. In many respects this is a more rigorous test
into believing that other sensorimotor and cognitive processes of the dissociation between locale and taxon strategies pro-
are not involved. It would be tidy if the impact of lesions or posed in the cognitive map theory than the comparison
other dysfunctions of distinct brain areas or neurotransmitter between the hidden and visible platforms of the standard
systems is isomorphic to the spatial mapping or navigation watermaze task because it compares two tasks of comparable
processes identied within the theory, but the data point to a difficulty. It also removes the recall element of the standard
more complex picture in which the hippocampus engages with procedure because the navigational demands are minimized
other brain structures to mediate effective performance. and the animal need only approach either platform and then
Before considering navigation in the watermaze in detail, it use spatial recognition to decide whether it or the other plat-
should be noted that many dry-land place-nding tasks have form is associated with escape. Interestingly, the hippocam-
also been developed, including the earlier Barnes arena task pus-lesioned group displayed rst-choice performance that,
rst developed in the context of studies linking hippocampal starting at 50%, gradually rose above chance. The effect was
LTP to spatial learning (Barnes, 1979), a place preference task not large, but it was both detectable and statistically signi-
in which rats are systematically rewarded by intracranial brain canta rst suggestion that allocentric place navigation is
stimulation for visiting a particular location (Fukuda et al., possible after hippocampal dysfunction. One possibility is
1992), similar appetitive place preference tasks (Rossier et al., that it is easier in this task to bridge temporally discontiguous
2000), and a place avoidance task. These tasks are conceptually gaps in the processing of relevant sensory information
similar but, being on dry land, the animal can stop moving (Rawlins, 1985).
and is also more likely to use olfactory and somatosensory Although a consensus emerged that the standard hidden
(vibrissa) cues. This is likely to make them less dependent on platform reference memory version of watermaze spatial nav-
visual cues than the watermaze. With place avoidance, the igation is a hippocampal taskbeing easy to train and
mirror image of place navigation, animals are trained stay producing reliable, replicable results in different laborato-
away from a location to avoid the punishment of mild foot- riesthe impact of such lesions is more variable than many
shock (Bures et al., 1998). Place avoidance has been used may appreciate. First, the sensitivity of the cue task (which is
extensively to examine the differential role of allocentric ver- unimpaired by hippocampal lesions) has been called into
sus ideothetic cues in location learning by allowing the circu- question following the observation that rats with experimen-
lar arena to rotate slowly in relation to extramaze cues. Rats tally reduced visual acuity display impaired place navigation
can be successfully trained to avoid a sector dened in relation but are unimpaired in the cue task (Prusky et al., 2000b). This
to room cues or in relation to its location on the rotating dissociation falsely masquerades as a decit in place learning.
arena. An interesting commercial twist on place navigation Although a nice proof-of-principle, it is unlikely that hip-
and place avoidance is incorporated into modern robot vac- pocampal lesions cause any change in visual acuity. However,
uum cleaners, a good example of a biologically inspired arti- regionally nonselective transgenic manipulations and many
 Theories of Hippocampal Function 635

peripherally administered drugs may affect visual acuity. islabs of dorsal hippocampus intact allow normal learning. In
Psychophysical procedures have therefore been introduced to contrast, lesions of the dorsal hippocampus that left the ven-
provide a more rigorous test of sensory acuity than the cue tral hippocampus apparently intact did not. They suggested
task (Prusky et al., 2000a; Robinson et al., 2001), and it would that a relatively small number of lamellae in the dorsal hip-
be valuable to see these used with selected transgenic lines. pocampus, if adequately connected to entorhinal cortex, the
Those with cell- and region-specic mutations are also septum, and other brain regions, were sufficient for learning.
unlikely to be impaired, but the important lesson that needs to A follow-up study using ibotenic acid lesions conrmed these
be taken on board from these sensory studies is that caution ndings (Moser et al., 1995). The second study was conducted
must be exercised about treating the watermaze as an ostensi- because the aspiration lesions of the original study may have
bly selective assay of memory function. Interestingly, the cue inadvertently left the ventral hippocampus deafferented, and
task is impaired by dorsal striatum lesions (McDonald and thus the apparent dissociation between dorsal and ventral hip-
White, 1994), a sensitivity that is more likely due to the role of pocampus could have been an artifact. In fact, similar results
this structure in habit learning than any sensory impairment. were obtained with as little as 27% of dorsal hippocampus
Second, as in the two-platform task, some place navigation being required for normal spatial learning that was indistin-
does take place in rats with hippocampal dysfunction. Using guishable from that shown by sham-lesioned controls (Fig.
overtraining in the standard, reference-memory single plat- 1324). In parallel, Jung et al. (1994) observed that the ring
form procedure, Morris et al. (1990) found that rats with elds of place cells were less spatially selective in the ventral
lesions in the hippocampal formation could learn to escape to hippocampus. Taken together, these observations t well with
the platform with short escape latencies after about 80 trials of anatomical studies of the connectivity between the entorhinal
training. Overtraining never obviated slightly elevated escape cortex and hippocampus, with particular zones of the dorso-
latencies relative to those of the controls; but after a mixed medial entorhinal cortex preferentially projecting to the dor-
series of hidden and visible platform trials, probe test per- sal (septal) hippocampus of the rat. Detailed anatomical
formance was spatially focused in the correct training quad- analyses have suggested the concept of parallel closed loops
rant and quantitatively indistinguishable from that of the as the substrate for both reverberation and the processing of
controls (Fig. 1323D). This slow, incremental spatial learning functionally distinct sets of cortical information (Amaral and
was apparent in animals with hippocampal or subiculum Witter, 1989; Witter et al., 2000) and of functional differenti-
lesions, whereas those with combined hippocampus and ation along the long axis of the hippocampus (Moser and
subiculum lesions failed to learn despite extensive overtrain- Moser, 1998b). (For further discussion, see Chapter 3.)
ing. Eichenbaum et al. (1990) also showed incremental learn- Other studies using the watermaze, T maze alternation,
ing in rats given fornix lesions when the animals were and contextual fear conditioning are consistent with this
consistently trained from a single location. Probe test per- dorsal-ventral (i.e., septo-temporal) gradation of function
formance was, however, sensitive to the use of novel starting (Bannerman et al., 1999; Richmond et al., 1999). The search
locations around the pool. Animals subjected to retrohip- for a specic function for the ventral hippocampus has
pocampal lesioning that encompassed the region of the included a return to J.A. Grays ideas about a possible role in
entorhinal cortex containing the entorhinal grid cells also fear and anxiety (Bannerman et al., 2004). For example,
showed gradually improving probe test performance, but Kjelstrup et al. (2002) found that rats with selective lesions
analysis of the paths taken revealed accuracy in swimming located in the ventral pole of the hippocampus were less fear-
across the correct platform location but little indication of ful on an elevated cross maze and showed decreased neuroen-
recognizing that this was the correct place to stop and search docrine stress responses. These animals did, however, display
(Schenk and Morris, 1985). However, not all studies have normal contextual fear conditioning. Bannerman et al. (2002)
observed effects of entorhinal lesions on spatial reference have also implicated the ventral hippocampus in anxiety as
memory in the watermaze (Bannerman et al., 2001), raising ventrally lesioned animals freeze less to unsignaled foot-shock
the possibility of functional heterogeneity across the medial, in a test chamber. They also show less neophagia (the fear of
intermediate, and lateral regions of the entorhinal cortex. A novel foods). However, the notion that the ventral hippocam-
key issue may be whether such lesions encroach on the dorsal pus is not involved in spatial learning at all can be overstated.
region containing the grid cells (Hafting et al., 2005). De Hoz et al. (2003) observed that changing the reference
A third important issue, rst raised by Andersen in relation memory training protocol to one involving fewer trials but
to the lamella theory of hippocampal information process- over a larger number of days enabled rats with large dorsal
ing (Andersen et al., 1971, 2000), is how much hippocampal hippocampus lesions (i.e., only ventral tissue intact) to learn
tissue is actually required for learning or retention. This was a just as effectively as rats with only spared dorsal tissue, as
fresh way of looking at the impact of lesions on function measured by posttraining probe tests. Similarly, Broadbent et
focusing less on whether a lesion causes a decit than if nor- al. (2004) found that ventral hippocampal lesions encompass-
mal performance is still possible with a partial lesion of a ing approximately 50% of total hippocampal volume can
specic size. Moser et al. (1993) discovered that aspiration impair spatial memory in a watermaze reference memory
lesions of the ventral hippocampus that leave only small min- task. Thus, the dorsal/ventral distinction is not clear-cut.
636 The Hippocampus Book

A Search patterns during probe tests of representative animals with partial or complete lesions

intact HPC dorsal 20-40% intact ventral 40-60% intact absent HPC

B Relation between size of residual tissue and performance during the probe test

dorsal HPC intact

70
intact HPC
60
absent HPC
time in quadrant (%)

50
40

30

20

10
ventral HPC intact
0
0-20 20-40 40-60 60-80 100
residual HPC tissue (%)
Figure 1324. Partial lesions reveal a dissociation of function along performance in animals with dorsal tissue. B. Quantitative assess-
the longitudinal axis of the hippocampus. A. After training by ani- ment of the probe tests show near-normal performance of rats with
mals with spared tissue volumes of different sizes that began at dif- as little as 20% to 40% of the dorsal hippocampus intact. Inset.
ferent ends of the longitudinal axis of the hippocampus, posttraining Parasagittal diagrams of the hippocampus show the locations of
probe tests (hidden platform absent) revealed a graded change in intact (white) and lesioned (black) tissue along the longitudinal axis.
animals in which only ventrally located tissue was intact but good (Source: After Moser et al., 1995.)

A fourth important issue is asymmetry in the impact of To test this idea more rigorously, Steffenach et al. (2002) inves-
lesion size on initial acquisition versus retention. Retention is tigated whether the integrity of the longitudinal-association
more sensitive. If normal rats are trained in the watermaze pathway of CA3 pyramidal cells is critical for integrating such
and then given small lesions of the mid- or ventral hippocam- distributed information. In animals previously trained as nor-
pus immediately after training, memory retention is impaired. mal rats in the standard reference memory task, a single trans-
However, such lesions have no effect on learning or later versely oriented cut of no more than 3% of the hippocampal
retention when given before training (Moser and Moser, volume was made bilaterally through the dorsal CA3 region
1998a). This dissociation is paradoxical from the perspective (the lesion shown in Figure 131, panel 5B). Memory reten-
of thinking about the hippocampus as a distributed associa- tion was impaired. These knife cuts inevitably damage cells as
tive memory system (Rolls and Treves, 1998) because such well and possibly some Schaffer collaterals coursing to CA1,
systems are generally robust in the face of partial damage. but no impairment was observed in controls deliberately
However, the nding is not necessarily incompatible with given small, equivalent-sized neurotoxic lesions of CA3 cells
such a perspective if encoding and trace storage takes place but leaving the longitudinal bers intact. This study is unusual
across an extended region of the longitudinal axis in normal amongst behavioral studies in using FluoroJade tract-tracing
rats. Animals given lesions before training would use the to observe degenerating bres on either side of the cut.
remaining tissue, and normal learning should occur provided Taken together, these studies led by the Trondheim group
there was enough of it. Small lesions created after training have denitively established that the efficient acquisition and
may indeed damage only a small proportion of the synaptic retention of spatial memory requires only a small minislab
storage sites but may nonetheless interrupt inputs to and out- of hippocampal tissue, preferentially although not exclusively
puts from such sites, even when neurotoxic lesions are used. in the dorsal hippocampus. The integrity of longitudinally
 Theories of Hippocampal Function 637

oriented, translamellar connections of CA3 pyramidal cells in Prompted by tantalizing observations on the relation
this minislab is essential for normal functioning. It is not yet between place cells and navigation by OKeefe and Speakman
known if this is a decit in memory retrieval or consolidation, (1987) and their dependence on the spatial reference frames
but varying the interval after initial training before the longi- used by the animals for specic tasks (Gothard et al., 1996;
tudinal cut is made would be one way of investigating this Zinyuk et al., 2000), Poucet and colleagues embarked on a sys-
question (see discussion of memory consolidation below). tematic research program to identify whether the ring of
place cells is critical for navigational decisions and goal iden-
Spatial Recognition Does Not Require CA3 Cells tication (Lenck-Santini et al., 2001, 2002). As behavioral and
cell ring had to be obtained simultaneously, they began by
If cognitive mapping involves both spatial representation and returning to the use of traditional mazes. The principle guid-
navigation, the question arises of whether different parts of ing this work was that altering place cell ring elds by rotat-
the hippocampal formation are involved in recognizing a ing or otherwise changing environmental cues should also be
place versus navigating to it. Using a modication of the associated with systematic changes in navigational behavior if
watermaze that involved changing it to no more than a circu- place cell ring and navigation are causally related.
lar swimming track, Brun et al. (2002) have shown that CA3 Lenck-Santini et al. (2001) trained rats in a continuous
lesions can leave CA1 place cell ring elds and spatial recog- spatial alternation task on a Y maze. One arm of the maze (G)
nition intact. Rats were intensively trained to swim in an served as the goal where food could be obtained. The animal
annular watermaze for a hidden platform that could rise from had to alternate its visits to the other arms (A, B) to secure
the bottom of the pool at a specic location once the animals food in G. Thus, the correct sequence would be A G* B
had completed two or three laps of the pool. They displayed G* A G* B and so on that secured reward (*) upon
knowledge of this potentially safe place in the pool by hesi- every visit to G, whereas an incorrect sequence might be A
tating or by swimming more slowly there during the succes- G* A G, which did not (Fig. 1325). This incorrect
sive laps. Complete CA3 lesions leave this spatial recognition sequence was classied as an alternation error because the
intact, while complete lesions of the hippocampus impair this animal returned to A when it should have gone to B. A differ-
ability (just as they impair navigation in the open pool). This ent, and initially much rarer, sequence might be A G* B
nding complements the further nding that CA1 place cells A in which, having correctly alternated between A and B
can be normal in rats with CA3 lesions, which points to the after a visit to the goal, the animal then returned to A instead
importance of the direct pathway from layer III of the entorhi- of going back to the goal. This was called an orientation
nal cortex into CA1 (Brun et al., 2002). Taken together with error. A prominent cue card was located on one side of the
the study of Steffenach et al. (2002), there is therefore a dou- maze. Once all the animals were trained and performing well
ble dissociation, reecting the importance of the longitudinal (around 80% correct choices), a total of 47 single units were
connections in CA3 for navigation and the direct perforant recorded while the animals continued to run on the maze.
path input to CA1 for spatial recognition. It would, nonethe- Various manipulations of the cue card (rotation, removal)
less, also be valuable to damage the cells of origin of the layer were systematically made to observe the impact on both place
III input to investigate the impact on CA1 place cell ring and, cell ring and behavior. As expected, the probability of correct
if the lesion could made be large enough while remaining alternation sequences declined during these manipulations;
selective (a tall order), on behavior as well. but more importantly, they did so in a systematic way. Pairs of
recording/behavior sessions were compared. The second ses-
Place Cells, Head-direction Cells, and Navigation sion of each pair was categorized as a consistent session
when the location of a place eld on the Y maze relative to the
Although these lesion studies indicate that the integrity of at goal arm was the same as during the rst session; it was cate-
least part of the hippocampus is necessary for spatial naviga- gorized as inconsistent when the cue manipulation between
tion, it does not require that place and head-direction units the pairs of sessions caused some shift in the relative location
are the critical neural mediators of navigation. Indirect evi- of the place eld to the goal. Behavioral choice performance
dence that these cells are important has come from a number was signicantly poorer on inconsistent sessions. Moreover, an
of studies in which pharmacological or genetic manipulations analysis of the kinds of errors made by the animals indicated
cause alterations in both place cell ring and spatial learning that alternation errors occurred occasionally on consistent
(Mizumori et al., 1994; Wilson and Tonegawa, 1997; Kentros sessions, whereas inconsistent sessions were characterized by a
et al., 1998; Rotenberg et al., 2000; Cooper and Mizumori, striking increase in orientation errors. These ndings indicate
2001). However, an alternative possibility is that these cells that the locations of place elds in a learned spatial navigation
only provide information about current location and direc- task are relevant to performance. Inconsistent place elds were
tion, and other still unidentied cells mediate the representa- statistically associated with disorganized spatial behavior.
tion of goals and carry out the information processing Similar observations were also made by Dudchenko and
responsible for navigational choicesthe problem identied Taube (1997) with respect to head-direction cells.
in Figure 1316B. Such cells may or may not be located in the In a follow-up study, Lenck-Santini et al. (2002) attempted
hippocampal formation. to compare the value of correct place cell ring in a locale
638 The Hippocampus Book

The relationship between the spatial firing of CA1 cells and spatial navigation

cue card

A B A B A B
correct alternation error orientation error

120

cue card cue card

goal goal goal

A G B G A G A A G B A

Figure 1325. Relation between place cell ring and spatial naviga- tended to be alternation errors (middle panel). However, when the
tion. Rats from whom CA1 cell recordings were made simultane- cue card was rotated (as shown) or removed or the goal location
ously were trained to alternate on a Y maze such that the correct was rotated, orientation errors increased in frequency. They
order of arms would be A to Goal, then B, then back to Goal, and so occurred when the cells were induced to display place elds that
on (A G B G, and so on) (left panel). With the cue card in became out of register to their standard positions. (Source: After
its training position, errors occurred rarely but when they did occur Lenck-Santini et al., 2001.)

task that required spatial memory with that in a taxon task mals alternate between locale and taxon tasks (Trullier et al.,
that did not. They reasoned, in keeping with the cognitive 1999).
map theory, that inappropriate place elds induced by However, as noted by Jeffery (2003), the ndings of these
cue manipulations should have a greater effect on behavior in studies are only correlational. Experiments in which neural
the locale task. They used a place preference task, one of the representations of location (e.g., place elds) are manipulated
appetitive analogues of the watermaze (Rossier et al., 2000), in directly, rather than indirectly via environmental manipula-
which rats search around a large circular arena for tions, are required to provide a yet more rigorous test of the
food they receive if they enter a virtual circular zone located at role of place cells in mediating behavior. It is not obvious,
one particular position in the arena (the food drops into the however, how this might be done. Dudchenko (2003) took the
zone from a hidden feeder abovea scientic version of same view with respect to head-direction cells and suggested
manna from heaven). In animals from which place cell that selective disconnection lesions might be a way of teasing
recordings were being taken simultaneously, this task was apart cause and effect.
conducted in one of three ways: with the virtual zone dened
as at a distance from a cue card at the periphery of the Knowing Where, Getting There, and the Puzzle
arena, with it placed adjacent to the cue card, and with of Hippocampal Involvement in Path Integration
the zone dened by a local cue explicitly outlining its area. The
key nding was that certain manipulations, such as visibly The term spatial navigation has been unpacked in various
rotating the cue card while the animal was in the apparatus, ways. The original theory made only a binary distinction
created shifts between the expected location of the virtual between locale and taxon processing during navigation, as
zone as dened by the animals place elds and by the if there are only two ways to get from A to B. However, evolu-
now rotated cue card. In brief situations of ambiguity (a few tion has discovered numerous other ways to navigate, ranging
minutes), most rats searched in a region dened by the from path integration to local views (Benhamou, 1997;
various place elds that were being recorded rather than as Biegler, 2000; Rodrigo, 2002). Inspired in part by ideas relating
determined by the now rotated cue card. This did not occur in the hippocampal theta rhythm to aspects of movement mon-
the taxon task when the zone was adjacent to the cue card itoring and control (Vanderwolf et al., 1973), a series of stud-
or visibly identiable by local cues. Poucet et al. (2003) argued ies by Whishaw explored the possibility of functional
that spatial searching behavior correlates well with place ring dissociation in navigation between getting there and know-
elds only when the rat is likely to be using a place navigation ing where (Whishaw et al., 1995). They recognized that spa-
strategy. Data showing a correlation between the ability of tial learning in the watermaze involves both learning the
an animal to nd a goal on a linear track and the alignment layout of distal cues and the development of navigational
of place elds with salient room cues is consistent with strategies to use this learned information. Following the sev-
Poucets interpretation (Rosenzweig et al., 2003). Place ring eral observations that hippocampus-lesioned rats may be able
elds, however, may be maintained in a taxon task when ani- to acquire spatial information slowly, Whishaw et al (1995)
 Theories of Hippocampal Function 639

wondered if mbria-fornix lesions might disrupt only the entire area (Worden, 1992). Path integration operates using
getting there component and not the knowing where. This ideothetic cues (i.e., cues generated during the course of
line of reasoning was to lead to unraveling the different senses movement) that may be vestibular, kinesthetic, or visual ow
of getting there. in origin. McNaughton et al. (1991) and Bures et al. (1998)
Although rats with mbria-fornix lesions normally have a have joined Whishaw in emphasizing the importance of path
profound decit in watermaze place navigation when trained integration to hippocampal function. Redish (1999) shares
to a hidden platform from several starting points, they display this view of its importance but places the path integrator out-
several indications of having learned the platform location side the hippocampus.
when trained using a careful shaping procedure. They learn to A special case of path integration, referred to as homing,
head in the correct direction, slow their speed of swimming has been extensively studied in hamsters (Etienne and Jeffery,
when approaching the vicinity of the platform, and search in 2004). It occurs when an animal, having left its nest site to for-
the correct quadrant when the platform is removed. Although age for food, somehow keeps a continuous record of the dis-
not as effective as control animals on all measures, there is tance and direction it has moved and uses this information to
clear evidence that mbria/fornix-lesioned animals can learn take a relatively direct path back to the nest. This is not a mat-
something about location. However, a substantial decit reap- ter of retracing its steps as the paths out may be circuitous,
pears when the animals are switched to the delayed match- whereas those back are direct. Field studies with desert ants
ing-to-place (DMP) protocol in which the hidden platform is have also revealed the accuracy of this form of home-base
moved from one location to another between days. Whishaw navigation and imply that it is an evolutionarily old form of
et al. (1995) suggested that this prole is consistent with dis- navigation. It solves the problem of getting food and then get-
ruption of a process associated with getting to a hidden plat- ting home again without recourse to any map. Laboratory
form, not one of learning its location. This interpretation was studies of mammals using rotating arenas, movable nest
backed up by a later study that, following visible platform boxes, and foraging trips in the dark have established that
training from three starting locations around the pool, homing can function in total darkness, that it accumulates
included probe tests from a novel start location. Once again, errors over both time and the circuitous movements of the
mbria/fornix-lesioned rats swam in the correct direction animal, and that it typically returns the animal to a point near
toward the now hidden platform, implying some knowledge but just short of the home-base by a vector likely to intersect
of where it was located. Neither they nor the control rats could the initial part of the outward path. The animal can then
do this if vision was temporarily occluded (Whishaw, 1998). switch to searching using local cues (e.g., olfaction) to nd the
In both studies, section of the mbria-fornix was complete, exact spot of a nest or burrow entrance. As path integration
and the ndings were in no way due to partial damage. accumulates errors over time, it is not surprising that the vec-
Whishaw (1998) also drew attention to similar data demon- tor processor can be reset by visual or olfactory cues, such as
strating some indication of effective place navigation after familiar landmarks or territory-dening odors. In addition, as
prior visible cue training in the studies of Morris et al. (1982). many rodents are nocturnal and live in burrow systems, a self-
Morris et al. (1990), in contrast, had argued that acquiring a motion monitoring system of this kind is likely to be helpful
spatial representation that encoded a single escape location in many situations. Etiennes view is that the neural algo-
might occur slowly in rats with hippocampal lesions, with the rithms used to yield directional and positional information in
map stored in the neocortex, provided interference could be rodents remain elusive. Path integration capacity develops
minimized. Visible platform training may help in this respect, postnatally, presumably as specic local circuits that process
particularly if many trials are undertaken. Both the getting the relevant sensory information become functional, and it is
there and the slow learning ideas can account for the failure rst observed as young pups abandon their sole use of olfac-
of rapid learning in the DMP protocol in both Whishaws and tory cues to nd the dam in the nest box in favor of forays in
Morriss studies. Moreover, the data concerning a novel start the world beyond. Right from the start, the very rapid and
location are ambiguous because the experimenter cannot con- very direct return paths they display appear to be guided by
strain the animals swimming patterns in the open pool dur- path integration.
ing earlier training. The view from the ostensibly novel start The idea that the hippocampal formation might be
location is likely to be familiar. involved in path navigation has been offered by a number of
The focus by Whishaw (1998) on the getting there com- investigators (Wiener, 1993; McNaughton et al., 1996; Bures et
ponent of spatial navigation is important. One the many ways al., 1998; Whishaw, 1998). Whishaw and colleagues (Whishaw
of getting from A to B is path integration (often called dead- and Tomie, 1997; Whishaw et al., 2001) trained rats on an
reckoning), a process that has been extensively studied in a arena similar to that used by Etienne in which they sponta-
variety of animal species (Mittelstaedt and Mittelstaedt, 1982; neously carried large food pellets back to their next box that
Wehner et al., 1996; Etienne et al., 1998). Path integration was was either visible or hidden at the edge of the arena. Search
identied as the evolutionary building block of the frag- paths from the home-base were circuitous and marked by sev-
ments within the fragment-tting extension of the cognitive eral pauses as the rats tried to nd the cryptic food pellet, but
mapping theory in which small parts of space (the fragments) the return path was much more direct. In normal animals, the
are assembled by an annealing process into a larger map of an return paths continued to be direct in the dark, suggesting
640 The Hippocampus Book

guidance by a homing vector. However, return paths were non-food-storing great tit, even though the animal and the
clearly disrupted in mbria/fornix-lesioned animals, which rest of its forebrain are much smaller. The importance of this
often took circuitous routes around the circumference of the correlation between the relative size of the hippocampus and
arena and quantitative changes with respect to both direction cache recovery derives from the argument of Andersson and
and velocity. However, data from Alyan and McNaughton Krebs (1978) that food-storing is an evolutionarily stable
(1999) have cast doubt on this nding because rats with hip- strategy provided the animal that stores seeds utilizes the pri-
pocampal lesions seem to be able to home accurately in the vacy of memory to relocate seed caches. Not surprisingly, the
dark, a nding that led Redish to place the path integration alternative strategy of tagging sites with some visible marker
machinery outside the hippocampal formation (Redish, to avoid having to use memory is vulnerable to intruders,
1999). The precise role of the hippocampus in path integra- which then steal the caches. Ingenious eld studies of tits in a
tion remains unclear, but the topic has been of considerable wood near Oxford using radiolabeled seeds to enable the relo-
interest for neural network modeling (see Chapter 14). cation of nonrecovered caches provided evidence that some
kind of location memory is involved (Shettleworth and Krebs,
13.4.4 Comparative Studies of Spatial 1982; Shettleworth, 1998). Similar observations were made in
Memory and the Distinction Between corvids that cache in the autumn but do not retrieve until the
Spatial and Associative Learning winter or early spring (Balda and Kamil, 1989). Effective food
caching by scatter hoarders therefore relies on memory. A nice
The third key feature to highlight about the cognitive map feature of this eldwork is the implied reminder that memory
theory is the dual claim (1) that allocentric spatial mapping can be allied to privacy. Like us, there are things animals do
evolved as a distinct, independent memory system of the and remember and keep to themselves.
vertebrate brain mediated by a specic brain area and (2) that Food caching is not an isolated case. Sherry et al. (1992)
spatial learning is qualitatively distinct from associative learn- pointed out that articial selection among pigeon breeds
ing in that it is rapid, exible, and uniquely encodes informa- appears to have had an effect on hippocampal volume similar
tion into map-like representations. These two strands of to that of natural selection among passerine birdsthe
an adaptive specialization hypothesis of hippocampal func- hippocampus of homing pigeons being larger than that
tion (Sherry and Schacter, 1987; Papini, 2002) have been inves- of nonhoming domestic breeds. Studies of brood parasitism
tigated in two contrasting research constituencies: by students are also interesting: Female cowbirds that lay their eggs in
of comparative cognition (who are broadly sympathetic) and the nests of other unsuspecting birds have a larger hippocam-
by general process learning theorists investigating possible dif- pus than male cowbirds (Lee et al., 1998). Cowbirds doing
ferences between spatial and associative learning (who this have to seek out and remember the locations of possible
are not). nests and then time their egg laying to coincide with that of
the host bird. The occurrence of a common neuroanatomical
Comparative Studies Using Avian feature in unrelated species ostensibly exposed to similar
Species and in Naturalistic Situations selection pressures raises the possibility that convergent evo-
lution is responsiblethe development of an adaptive spe-
The original cognitive map theory put value on adopting a cialization.
neuroethological as well as a neuropsychological approach A potential problem with naturalistic studies is that they
to structure/function relations, an aspect reaffirmed by can be overly suggestive. As Shettleworth somewhat guardedly
(Nadel, 1991. p. 224). It is important, he wrote, to adopt a put it, animals show many behaviors for which explanations
perspective on brain and behavioral organisation that takes in terms of human-like understanding readily come to
into account ecological, evolutionary and ethological aspects mind: the animal has a cognitive map, a concept, a theory of
of the animal under study. The argument is that if space is mind . . . a leap of imagination may be necessary to grasp that
important because we live in it, move through it, explore it, behaviors so readily explicable by such intuitively appealing
defend it, there will have been selection pressures of various mechanisms are accomplished in completely different ways
kinds favoring the evolution of effective cognitive mecha- (Shettleworth, 2001, p. 279). She and Krebs were among the
nisms for representing and navigating through space. Since rst to appreciate that eld studies had to be followed up,
1978, study of the hippocampus in relation to naturalistic where possible, by rigorous laboratory analogues. They went
spatial behavior, in both eld settings and the laboratory, has on to create an avian laboratory setting in which food storing
thrown up some striking brain/behavior correlations in sev- and retrieval could indeed be studied under controlled condi-
eral avian and mammalian species. tions, conrming the observations of the role of spatial mem-
One nding is that the relative volume of the hippocampus ory in cache recovery rst discovered in the wild. A typical
in certain species of passerine birds (e.g., tit species in Britain, avian food-caching laboratory is an indoor room with trees
chickadees in North America) is larger in the subspecies that where seeds can be hidden (Fig. 1326B). Such rooms are
store seeds than in those that do not (Krebs et al., 1989; Sherry often described as large, but they are of course much smaller
et al., 1989) (Fig. 1326A). For example, the hippocampal for- than the areas over which the birds normally forage in their
mation of the food-storing marsh tit is 31% larger than the natural habitat, an issue always to be borne in mind when a
 Theories of Hippocampal Function 641

A Relative HPC volume in storers/non-stores B Laboratory aviary to study food storing

0.4 non-storers
2
storers
hippocampus residuals
0.3
0.2 9
1
0.1 3 9
0.0 1
-0.1 4
6
-0.2 8 7 5

-0.3
-0.35 -0.25 -0.15 -0.05 0.05 0.15 0.25
telencephalon residuals

C Spatial memory error over time D Development of HPC size with experience
pre-experimental baseline
125 0.10

residual hippocanmpal volume

non-storers inexperienced storers
storers experienced storers
distance error (mm)

100 0.05

75 0.00

50 -0.05

25 -0.01

0 -0.15
2.5 5 10 20 35-59 60-83 115-138
stage of experiment
retention interval (seconds)
(days posthatch)

Figure 1326. Relation between relative hippocampal size and spa- more persistent over time in coal tits (a storing species) than great
tial memory in birds. A. Study of 35 species or subspecies of passer- tits (a nonstoring species). (Source: After Biegler et al., 2001). D.
ine birds revealed that hippocampal volume, relative to the rest of Relative hippocampal volume at different stages of development.
the telencephalon (residuals), is larger in storing species. (Source: The structure becomes larger if the young marsh tits are allowed to
After Krebs et al (1989). B. Laboratory aviary used to study storing, store sunower seeds (black bars), but it regresses and shows more
window-shopping (see text), and other food-caching protocols in apoptotic cells if the birds are given powdered seeds that cannot be
birds. (Source: Courtesy of Nicola Clayton.) C. Spatial memory is cached. (Source: After Clayton and Krebs, 1995.)

laboratory study yields a null result. Researchers using this form better than great tits (a nonstoring species) on a task
kind of facility were faced with the challenge of how to com- assessing the persistence of memory over time (from 2.5 to 25
pare cache memory by a species that does store seeds with the seconds) but not on tasks that assess the ne-grained resolu-
spatial memory displayed by a species that does not. tion or, somewhat surprisingly, the capacity of memory (Fig.
Shettleworth and Krebs did this by creating a paradigm called 1326C). This is an important reminder that not all parame-
window shopping in which seeds that could be retrieved ters of spatial memory are better in storers than nonstorers.
later were placed on display behind Perspex screens in holes Another question that seminaturalistic studies have made
drilled into the trees. Thus, during the initial sampling trials, possible concerns the ontogenetic development of the
the birds were allowed to y into the aviary without storing or caching-associated difference in hippocampal size. How does
retrievingmerely to see what was on offer and where. Later, experience interact with any genetic predisposition? In a clas-
the Perspex screens were removed, and the birds (presumably sic study, Clayton (1995) reared juvenile marsh tits (a food-
armed with their credit cards) could y back and collect what storing species) under conditions in which they were either
they had seen earlier. In this and other one-trial associative allowed to store seeds or prevented from doing so (by giving
paradigms (Clayton and Krebs, 1995), the act of caching was them only powdered food). Tits allowed to store had the usual
nessed while still allowing comparisons of spatial memory large hippocampal volume (Fig. 1326D). However, birds that
between storers and nonstorers. Other approaches include were prevented from storing showed a gradual decline in hip-
operant paradigms with peck screens (an avian touch- pocampal size and the highest proportion of apoptotic neu-
screen) to cast light on what qualitative aspects of memory rons. These ndings suggest that hippocampal volume, and
might be better with a large hippocampus. In this way, Biegler perhaps neurogenesis also (Patel et al., 1997), may be partially
et al. (2001) found that coal tits (a food-storing species) per- regulated by activity-dependent factors such as the behavior
642 The Hippocampus Book

of the animal, a concept captured by the slogan use it or lose series of visual-sighting and radio-tracking experiments by a
it. If so, it might be expected that relative volume in a species group based in Pisa in Italy over the past two decades. Early
might also be seasonal. Specically, the onset of hoarding studies suggested that the integrity of the hippocampal for-
behavior might the trigger for an increase in volume, with mation was critical for aspects of the navigation required, par-
regression later. This has been conrmed in black-capped ticularly near the home loft, but more recent work has pointed
chickadees (Smulders et al., 1995) and appears to be due to a to a more complex picture.
net increase in cell number (Smulders et al., 2000). The possi- Pigeon homing seems to be guided by a multiplicity of
bility that it is merely due to a general seasonal mechanism strategies and mechanisms, in part because of their need to
operating on both food-storing and nonstoring species is less cope with differing environments and potential variation in
likely given that seasonal changes in volume have been shown weather conditions. When released from an unfamiliar loca-
not to occur in at least one nonstoring species (Lee et al., tion, experienced homing pigeons y off with a vanishing
2001). Males of this same species do, however, show changes bearing in the correct direction of home (this being the com-
in the size of song-related nuclei, such as the higher vocal cen- pass bearing that the bird takes measured from the release site
ter (HVC) and archistriatum (RA) across the seasons that cor- to last sighting after release). To do this, it was suggested by
related with song stereotopy. It therefore seems reasonable to Kramer (cited in Rodrigo, 2002) that they use a long-range
infer that changes in the relative volume of the hippocampus navigational map that may, at least in Italy, be partly reliant
of food-storing avian species results from a species-specic on odors (Papi, 1990). A problem with this concept is that it is
and regionally specic form of structural plasticity that occurs far from clear how the pigeons learn such a map. In any event,
in response to seasonal pressures to engage in food-caching they also keep ying in the correct direction and have long
and later retrieval. One slightly puzzling nding is that the been thought to have some kind of a compass to do it
seasonal variation in size is, in black-capped chickadees, asso- (Schmidt-Koenig, 1961, cited in Gagliardo et al., 1999). Once
ciated with the time of year when caching occurs, but hip- they approach the vicinity of the loft, they recognize the land-
pocampal volume regresses in the winterthe very time marks and the layout of the area they are used to ying around
when caches are recovered. Why retrieval places less of a at home and so nd their way back to their perch in the loft.
demand on the need for a large hippocampal volume than It is not just the loft area that may be spatially familiar; the
memory encoding is unclear. birds can also learn about the landmarks at familiar release
Allometric brain/behavior relations in the avian brain are sites and even the roads over which they may y. Indeed,
intriguing, but they are a chicken and egg problem! What Italian pigeons stylishly display a classic superiority in matters
causes what? Lesion studies were a rst step in investigating navigational by following old Roman roads such as the SS
causal relations, it being essential to check that lesions of Aurelia (Lipp et al., 2004).
the avian hippocampus impaired location memory, just The early lesion studies established that pigeons with telen-
as such lesions do in mammals. These studies conrmed cephalic lesions, including the hippocampal formation, have
this prediction (Sherry and Vaccarino, 1989). Hampton and vanishing bearings from an unfamiliar release site that are as
Shettleworth (1996) went on to show specicity to location accurate as controls (Bingman et al., 1984). Their long-range
rather than color memory after hippocampal lesions, and navigational map is therefore intact. They also reach the vicin-
work using reversible inactivation in chickadees has both ity of the loft but sometimes have difficulty nding itunless
borne out this claim and established that it operates at the they have been given sufficient postoperative experience of the
time of memory encoding (Shiett et al., 2003). These and loft vicinity (Bingman et al., 1987). This implies that the sun
other studies (Columbo et al., 1997) have established that the compass can work as well, this having been learned through
functional integrity of the avian hippocampus is required for sightings of the sun at different times of day when outdoors at
spatial memory (Macphail, 2002), but they do not really nail the loft site. More recent studies have conrmed that the hip-
down whether the variation in the size of the hippocampus is pocampus plays only a small role in the navigational map.
causally related to the very changes in behavior with which Acquisition and retention is independent of the hippocampus
they are correlated. Nor is it clear to what aspect of spatial or if the birds acquire their navigational map during free ights
other information processing the avian hippocampus con- (Ioale et al., 2000). However, young birds can also learn the
tributes. Work on homing pigeons has shed light on the sec- navigational map when held in an outdoor aviary. Under these
ond of these issues. conditions, learning seems to be hippocampus-dependent; but
perhaps surprisingly, its retention is not (Gagliardo et al.,
Homing Pigeons Reveal Hippocampus- 2004). The reason for this difference is that hippocampal
dependent and Hippocampus-independent lesions disrupt the process by which sun compass information
Components of Navigation is used only in the context of new learning (Bingman and
Jones, 1994). Use of this compass requires that the animals
Homing pigeons have a remarkable ability to y back to their keep track of time and orient in an appropriate direction to
lofts over long distances from both familiar and unfamiliar the angle of the sun given the time of day. When clock-
release sites. The possible contribution of the hippocampus to shifted by being kept on articial lightdark schedules, the
homing behavior has been the subject of a comprehensive orientational bearing is shifted from the correct homeward
 Theories of Hippocampal Function 643

direction. Experienced pigeons later subjected to hippocam- learned about the sun compass long before attaining hip-
pal lesioning show the usual clock-shift effects, indicating pocampal lesions. Given the multiplicity of ways that pigeons
their successful use of the sun compass. However, the hip- can navigate, these data represent an unambiguous demon-
pocampus is necessary for learning to use this sun-compass stration of the critical role of the hippocampal formation in
information. landmark learning in a naturalistic setting.
Landmarks they have learned about at a familiar release site Although ostensibly supporting a close link between spa-
can be used for homing in two distinct ways: pilotage in tial memory and hippocampal function, one must recognize
which they use the perceived layout of the landmarks to head that the cytoarchitectonic appearance and intrinsic circuitry
toward the loft or site-specic compass orientation in which of the avian hippocampus is very different from that of the
they use the landmarks to recall the sun compass direction in typical mammalian hippocampus described in Chapter 3.
which to y. Whereas pilotage requires quite a sophisticated There are both similarities and differences between the avian
spatial representation, site-specic compass orientation does and mammalian brain (Jarvis et al., 2005), but both Lee et al.
not. Although pigeons with lesions of the hippocampal for- (1998) and Macphail (2002) make the caselargely on
mation can learn to use familiar landmarks to navigate, it is embryological groundsfor the avian hippocampal forma-
possible that they do it in only one of these two ways. tion being considered homologous to the hippocampus and
Gagliardo et al. (1999) studied experienced pigeons rendered dentate gyrus of mammals. Certain arguments seem uncon-
anosmic by means of nasal zinc sulfate (a procedure that pre- vincing or circumstantial, such as the presence of synaptic
vented the use of their navigational map) to compare the man- plasticity (this also occurs outside the hippocampus) and the
ner in which control and lesioned birds used familiar claimed similarity in neurotransmitters used (but similar neu-
landmarks. As shown in Figure 1327A, the birds were rotransmitters are used all over the brain in numerous orders
released on a number of occasions from La Costanza, north of of animals). Electrophysiological studies are also equivocal.
Pisa, and from Livorno, situated southeast of Pisa. Prior to the For example, Bingman et al. (2003) reported some evidence of
critical test days, the birds were clock-shifted in their lofts by 6 space-specific firing of hippocampal cells in homing
hours. This should have no effect on the vanishing bearings pigeons trained in a small plus maze, but the correlate was
taken from a release site if the birds use a representation of the transitory and less clearly tied to space than in recordings
spatial layout of the landmarks to nd their way home. from rats. The possibility remains that greater spatial speci-
However, if they use site of the landmarks around each now city would be observed in a task in which the pigeons were
familiar release site to work out a site-specic compass orien- allowed to engage in homing, but this is not yet technically
tation, a 6-hour clock shift should result in an approximately feasible. One prominent avian behavioral biologist, writing
120 error in the initial ight direction. The results of test about the pigeon brain, noted that: the pigeon and rat hip-
releases from both sites indicated a striking difference between pocampal formation reside in different forebrain environ-
the controls and the birds with hippocampal lesions (Fig. ments characterized by a wulst and neocortex, respectively.
1327B). The controls headed in the correct direction, with Differences in the forebrain organization of pigeons and rats,
only a small deviation of 34 in the direction expected given and of birds and mammals in general, must be considered in
the clock shift; the lesioned birds showed a mean bearing error making sense of the possible species differences in how the
of 154 in the counterclockwise direction. Thus, the controls hippocampal formation participates in the representation of
were using pilotage, whereas the lesioned birds were not. There space (Bingman et al., 2003, p. 117). Clearly, there are differ-
is no conict here with the earlier ndings demonstrating that ences of opinion on the similarity of the avian and mam-
learning to use the sun compass requires the hippocampus malian brain, and it is tempting to wonder if some of those
because these were experienced homing pigeons that had who work on birds are making a virtue of necessity in sup-

Figure 1327. Avian hippocampus and landmark learning. A Home and two release sites B Vanishing bearings of anosmic
A. The home loft of the pigeons was near Pisa. The animals birds released from Livorno
were extensively trained to y home (to Pisa) from either
of two release sites, La Costanza and Livorno. B. When ren- La Costanza

dered anosmic and clock-shifted by 6 hours, the controls

headed in the correct direction whereas the hippocampus-
lesioned birds showed greater variability around a mean, 16.7 km anosmic controls
reecting the expected shift of 120. The reason a clock birds

shift of 6 hours is expected to produce a deviation of 120

rather than 90 is because at around midday in summer
Home (Pisa)
the azimuth of the sun moves about 20 per hour. (Source:
After Gagliardo et al., 1999.) HPC anosmic
13.6 km lesioned birds

Livorno
644 The Hippocampus Book

porting a claim for homology, aware of wider interest in the are used to establish a sense of direction) and a sketch map
brains of mammals. Although understandable, the issue has to (which is an allocentric representation of local landmarks).
be resolved on the basis of all the evidence. The two mapping systems are held to be mediated by and
to interact at different subregions of the hippocampal forma-
Comparative Studies of Spatial tion; the primary evidence for this is the consistent patterns
Memory in Rodents of distinctive search strategies seen in spatial tasks after
selective lesions or other pharmacological forms of interven-
Hippocampal-oriented neuroethology studies have been con- tion. It is too early to assess the status of this proposed modi-
ducted in mammals as well, but the research is less well devel- cation of the original cognitive map theory, but Jacobs
oped. One idea about mammalian hippocampal evolution is (2003) made the case that it is pertinent to issues of evolution
that sexual selection may have led to a relative increase in hip- and ontogeny across a range of species, including other
pocampal volume of certain polygymous male rodents, an vertebrates. Jacobs and Schenk also used it to offer novel
important case because sex differences evolve slowly and thus explanations of certain hitherto puzzling lesion and pharma-
any correlated differences in cognitive function and neu- cologically induced dissociations in task performance by lab-
roanatomy are especially strong evidence of an adaptive mod- oratory rats.
ication of the brain. In support of this idea, Jacobs et al. The relative volume of a brain area is a limited measure. It
(1990) found that polygymous male meadow voles, which is important to know whether this reects variation in cell
have an area over which they range about ve times larger number (and, if so, of what cell type), cell proliferation or cell
than that of females, show a corresponding sex difference in death, dendritic arborization, synapse number or density, or
hippocampal volume (Fig. 1328). In contrast, in monoga- other neuronal processes. Finer grain neuroanatomical or bio-
mous prairie and pine voles, male and female hippocampal chemical measures could offer insights into precisely what co-
size is equivalent. Earlier laboratory studies revealed that varies with the type of information storedcapacity or
males of the polygymous meadow vole are superior to females temporal persistence. Given that neurogenesis may be under
in place learning in Tolmans sunburst and other mazes hormonal control (Galea and McEwen, 1999), there has been
(Gaulin and Fitzgerald, 1989; Gaulin et al., 1990). The inter- considerable interest in the possibility that seasonal changes in
action between natural and sexual selection has also been hippocampal volume reflect neurogenesis (Barnea and
revealed in studies of subspecies of kangaroo rats, some of Nottebohm, 1994). Neurogenesis is inuenced by environ-
which store food and some of which do not, some polygy- mental enrichment in some strains of mice (Kempermannet
mous and some monogamous. In comparisons across species, al., 1998) and by voluntary activity in running wheels (van
hippocampal size was found to be largest in males and scatter Praag et al., 1999). There is evidence for seasonal changes in
hoarding in polygymous Merriams kangaroo rats (Sherry et spatial learning ability and ne-grained aspects of hippocam-
al., 1992; Jacobs and Spencer, 1994). pal anatomy (Jacobs, 1995). However, Jacobs own studies of
One proposed modication of the cognitive map theory, possible seasonal variation in wild caught squirrels revealed
inspired by the allometric studies, pigeon homing work and no difference in cell proliferation or total neuron number in
by the discovery of head-direction cells, is the parallel-map the hippocampus that co-varied with seasonal variation in
theory of Jacobs and Schenk (2003). This explicitly identies demands on spatial memory (Lavenex et al., 2000). Chapter 9
a bearing map (in which olfactory gradients or distal cues discusses neurogenesis in the dentate gyrus in more detail.

Figure 1328. Sexual selection inuences hippocampal size. A. pal volume in polygamous meadow voles and monogamous pine
Space use by mammalian species with different mating systems. voles. The argument is that the hippocampus needs to be larger in a
Under polygamy, male home ranges encompass several smaller male meadow vole, which has to patrol a much larger territory.
female ranges; under monogamy, males and females share a joint (Source: After Jacobs et al., 1990.)
home territory. (Source: After Jacobs, 1995.) B. Relative hippocam-

A Male and female range size B Relative hippocampal volume

polygamy: meadow vole monogamy: pine vole 0.06 meadow vole

pine vole
hippocampal volume
brain volume

0.05

0.04
 Theories of Hippocampal Function 645

Neuroanatomical studies have nonetheless revealed consis- To try to unravel the reported correlations and take a step
tent genetically stable variations in the intrinsic circuitry toward causation, postnatal injections of thyroxine (Fig.
of the hippocampal formation across strains of mice (Lipp et 1329C) have been successfully used to induce variations in
al., 1999). Schwegler and Lipp (1981) discovered that the infrapyramidal mossy bers within a single strain and corre-
length of the infrapyramidal mossy bers (IIP-MFs) is sponding variations in the learning rate (Lipp et al., 1988;
negatively correlated with two-way avoidance learning (a task Schwegler et al., 1991). Still, the anatomical disquiet remains:
impaired by hippocampal lesions) (Fig.1329A,B), a nding These injections may be having other unmeasured effects.
that holds across a number of mouse (and rat) strains Some measure of behavioral specicity has been realized by
(Schwegler and Lipp, 1983). Later work showed that it is the nding that the correlations with infrapyramidal mossy
positively correlated with performance in spatial tasks, includ- ber length holds for spatial working memory on a radial
ing the radial maze and reversal learning in the water- maze but not cue working memory (Crusio et al., 1987;
maze (Crusio et al., 1987; Bernasconi-Guastalla et al., 1994). Schwegler et al., 1990). However, later studies revealed strain-
Mouse strains, such as C57/BL6, known to be good at spatial specic covariations between the extent of the infrapyramidal
learning have an extensive infrapyramidal mossy ber system. mossy ber distribution and both intermale aggression
Other strains, such as DBAs and BALB/C, have a limited (Guillot et al., 1994) and paw preference (Lipp et al., 1996),
infrapyramidal mossy ber system. However, such correla- which are clearly unrelated to spatial learning and memory;
tions, although striking, should be treated cautiously not and there is a report of failure to observe the correlation with
least because there may be other covariates. For example, radial maze performance (Roullet and Lassalle, 1992).
BALB/C mice have retinal problems and limited vision; Lipp et al. (1999) recognized that these noncognitive
and DBA mice are reported to have deciencies in pro- ndings may not t well with some prevailing theories of hip-
tein kinase C (PKC), an enzyme implicated in synaptic plas- pocampal function but are relevant to a multifunctional
ticity. view in which the dorsal hippocampus mediates spatial

Figure 1329. Avoidance learning and infrapyramidal mossy bers. A. Relative location of the
supra- and infrapyramidal mossy bers (IIP-MF) projecting from the dentate gyrus (DG) to
area CA3. B. Correlation between two-way avoidance learning and IIP-MF extent across mouse
strains. C. Thyroxine induces alterations in both two-way avoidance learning and extent of IIP-
MF. (Source: After Lipp et al., 1999.)

A Mossy fibre pathway (black)

showing supra- and infra- CA1
pyramidal parts of the
pathway. PYR LM
FIM
RAD
SP GC
OR
CA4
CA3

SGL
MOL
Infrapyramidal mossy fibres (IIP-MF)

B Strain Comparisons C DBA/2 Thyroxine Postnatal

80 100
DBA/2 90
70
80
% avoidance

60 70
BALB/c
60
50
50
40 40
NMRI C57BL/6
30 30
SM/J
20 FEMALE
20
MEAN
SEM ICR C3H 10 MALE
10 0
0 1 2 3 4 5 .5 1 1.5 2 2.5 3 3.5
% IIP-MF in CA3/CA4 % IIP-MF in CA3/CA4
646 The Hippocampus Book

memory functions whereas the more ventral region is more learning in being rapid, exible, and uniquely encoding infor-
relevant to emotional and aggressive behavior. They also sug- mation into maps. As noted above, the animal learning con-
gested, partly on the basis of studies of the survival of various stituency is suspicious of this idea.
mouse strains through successive harsh winters at a Russian The claim that spatial learning is qualitatively different
eld station that behavioral reactivity rather than spatial would be stronger if the psychological properties of either
memory may be the parameter of greatest adaptive signi- type of learning could be shown to be specic to that type of
cance in relation to the infrapyramidal mossy bers. learning. A particular sticking point has been whether certain
To conclude this subsection, allometric brain/behavior phenomena known to occur in classical conditioning, such as
correlations in birds and mammals have revealed several nd- blocking and overshadowing, do or do not occur in the
ings consistent with spatial interpretation of hippocampal spatial domain. Blocking refers to the ability of one stimulus
function, but they have not denitively established that it is to block learning about another, and overshadowing is a
only spatial, rather than other kinds of memory, with which differential sharing of associative strength when two stimuli
changes in hippocampal volume or structure are correlated. are conditioned together. Modern animal learning theories,
More analytical studies are required to unravel precisely what discussed in more detail in Section 13.5, rely on error-cor-
features of memory correlate with changes in hippocampal recting learning rules in which changes in the associative
volume or circuitry. Identifying evolutionary adaptations, strength of a stimulus occur only when the expected outcome
attributing them denitively to particular selective pressures, of that stimulus differs from the actual outcome. Although
and then relating them to specic mosaic variations in brain such rules enable learning to be rapid and exible, stimuli
size or circuitry is a slow, painstaking task (Lipp et al., 1999). compete with each other for associative strength rather than
The criticism has been made that studies of evolutionary become linked together in anything like a map. The condi-
function not only do not, but cannot, provide information tioning phenomenon that most directly established the idea of
about neurobiological mechanisms (Bolhuis and Macphail, error-correcting learning and captured everyones attention
2001; Macphail and Bolhuis, 2001). However, even if this log- when it was rst discovered is blocking (Kamin, 1968). A stim-
ical point were to be accepted, the implications are hotly ulus, call it A, is arranged to predict reinforcement (R*). After
debated (Hampton et al., 2002). Shettleworth (2003) offered a several pairings of A and R* have occurred and the associative
thoughtful defense of studies of the neuroecology of learn- strength of A has reached an asymptotic value, a second stim-
ing. She highlighted, by way of example, striking data from ulus is added (B). Stimulus B now also predicts R*; and one
two populations of chickadees by Pravosudov and Clayton might expect, given the contiguous pairing of B and R*, that
(2002). Instead of comparing different strains or species, they this association would be readily learned also. Often, however,
tested the adaptive specialization hypothesis in a single little appears to be encoded about the relation between B and
species, black-capped chickadees. One population was living R*. The prior learning of the A R* association is said to
in Alaska, where winters are erce, and the other in Colorado. block learning about B. An important control is one in which
The different latitudes at which these distinct populations of a stimuli A and B are introduced together for the rst time in
common species live might have resulted in a quantitative dif- the second phase of a blocking experiment and arranged to
ference in the degree of selection pressure. Birds were caught predict R*. Under these conditions, both the A R* and the
in the autumn, around the time that seasonal volume changes B R* associations are learned (although overshadowing
should be maximal, and brought to a laboratory in California. sometimes occurs and is determined by such factors as stimu-
Once there, the Alaska population was observed to be better at lus salience or proximity). Many other control procedures
food storing and retrieval than the Colorado birds, better on have been conducted among a vast array of experiments
spatial but not color memory (an allometric correlation anal- investigating blocking and the processes responsible for it
ogous to earlier lesion data (Hampton and Shettleworth, (Mackintosh, 1983).
1996), and had relatively larger hippocampal volumes. It is not One reason hippocampus-dependent learning might be
unreasonable to conclude that mosaic evolution of brain thought to be qualitatively different from conventional asso-
structure driven by natural selection on particular behavioral ciative conditioning is because hippocampal lesions do not
capacities (Barton and Harvey, 2000) will continue to con- reliably affect blocking. An early conditioning study indicated
tribute to understanding qualitative and quantitative differ- that there may be an effect of partial hippocampal lesions on
ences in the hippocampus within and across species. blocking (Ross et al., 1984), but at later more denitive exper-
iments indicated that blocking can occur normally in animals
Maps, Spatial Representations, with complete hippocampal lesions (Garrud et al., 1984;
and Geometry: Is Spatial Learning Qualitatively Holland and Fox, 2003). Of greater concern for the cognitive
Different from Associative Learning? map theory is whether blocking occurs between locale and
taxon learning or even within the spatial domain. A compre-
The other strand of the adaptive specialization idea is the issue hensive program of work led by Chamizo has shown that
of whether hippocampus-dependent spatial learning is quali- blocking can occur in the radial maze between intra- and
tatively distinct from associative learning. The cognitive map extramaze cues (Chamizo et al., 1985) and entirely within the
theory asserts that spatial learning is different from taxon spatial domain in the watermaze (Rodrigo et al., 1997). In the
 Theories of Hippocampal Function 647

later experiment, analogous to the studies using classical con- directional polarization provided by a specic arrangement of
ditioning, the pool is surrounded by heavy black curtains and distal extramaze curtains. The task involved learning the loca-
two-dimensional cues hung at specic locations within it (A, tion of food in relation to two identical landmarks, with a
B, C, and so on). This spatial array of cues can then be rotated third and later fourth cue added to disambiguate two possible
collectively from trial to trial; in this way, the experimenter places where food might be found. The paths taken by the ani-
can be condent that a rats directed navigation in the pool is mals were tracked automatically. Blocking was again seen in
determined by this array rather than by other, experimenter- the spatial domain, and this occurred despite the animals
uncontrolled cues. In addition, although swimming is permit- noticing and exploring the added landmark during the block-
ted at certain phases of the experiments (to enable learning ing phase (as shown by the tracking data) but then choosing
and later testing of what has been learned), the animals can not to learn about its association with reward. The use of an
be either allowed to swim or placed on the platform during instrumental rather than a classic conditioning paradigm
the critical blocking phase when another cue (X) is added. introduced complications, as the conditions across groups
Blocking occurs normally. Chamizo (2003) summarized a could not be exactly matched on the rst trial of the blocking
wide range of such experiments and concluded that there is phase, though Biegler and Morris (1999) argued that this did
as yet no basis to suppose that spatial learning differs from not invalidate their conclusion. Unpublished lesion studies
conventional conditioning with respect to the use of an error- were also conducted, but performance was too poor in ani-
correcting learning rule. Mackintosh (2002) recognized cer- mals given lesions before training for any useful information
tain tensions in the associative account between the to be obtained about the role of the hippocampus in blocking
occurrence of blocking, as required by animal learning theory, using this paradigm.
and other evidence that rats often learn about the locations of One point noted by Biegler and Morris (1999) concerns a
all available landmarks in simple exploratory tasks, as we saw potentially interesting difference between blocking in the
in the work of Save et al. (1992a,b) earlier. However, he shared spatial and conditional domains. With conventional condi-
Chamizos skepticism and asserted that the behavioral evi- tioning, the relations learned, such as A R*, are condi-
dence does not seem to have supported OKeefe and Nadels tional (ifthen) associations (and not necessarily temporal
original hypothesis that true spatial learning or locale learning although often described as such). Relations must be more
is quite distinct from associative learning (Mackintosh, 2002, than that in spatial learning because the Landmark R*
p. 165). association must provide some information about the dis-
There remain some difficulties in accepting these conclu- tance and direction from a landmark (L) that food is to be
sions without qualication. First, no lesion studies of blocking found. Under circumstances in which a memory representa-
in the spatial domain have been conducted, and we must take tion must be formed of the nature of the relation between two
on trust that the spatial memory seen in these studies is cues, blocking may obey different rules. For example, when
hippocampus-dependent. This seems likely, but we have the relation between two spatially dispersed landmarks L1 and
already noted instances in which multitrial spatial reference L2 and food has been learned, the addition of a third land-
memory tasks in the watermaze can be learned by rats with mark, L3, may not always result in blocking simply because
hippocampal lesions. The denitive hippocampus-dependent the location of R* is already predicted. The relative spatial
task in the watermaze is now recognized as being the DMP location of the new cue L3 to the old cues L1 and L2 may be
paradigm, and one-trial blocking experiments have not been critical to its incorporation into an associative or map-like
conducted. Second, placement on the hidden platform may representation. Blocking may be more or less likely as a func-
be insufficient for much learning with the kind of hanging tion of whether the vector (argument and direction) from L3
cues used in Chamizos laboratory. This is because two- R* is similar to or different from the vectors from L1 and L2
dimensional cues at the periphery are known to be much less to R*. The conclusion of Chamizo (2002, 2003) that spatial
effective than three-dimensional cues, the animals cannot information interacts during learning in apparently the same
explore them proximally (e.g., using vibrissa movements), and way as it does during conventional conditioning is probably
the constellation of cues in the blocking phase of training secure, but there are issues to do with vectors and geometry
(A, B, X) may be sufficiently similar to those in the earlier that deserve further investigation.
training phase (A,B) that the animals do not notice the differ- Swimming against the tide of this developing experimental
ence. The outcome is then apparent rather than true blocking. skepticism in the animal learning constituency, Cheng and
Although various controls argue against this, the data meas- Gallistel have gone farther than OKeefe and Nadel (1978)
ures reported are sometimes the proportion or ranking of ani- in suggesting that the brain possesses a geometric module
mals showing an effect rather than absolute tracking data for space that operates according to distinctive learning rules
(such as time spent in a quadrant or zone). (Gallistel, 1980; Cheng, 1986; Margules and Gallistel, 1988).
Partly for these reasons, Biegler and Morris (1999) devel- This followed observations of the behavior of rats searching
oped a food reward task in a large arena in which to explore for food hidden at one corner of a rectangular enclosure.
blocking under conditions in which (1) the local cues were Despite the presence of disambiguating cues located at
three-dimensional objects that could be explored by the ani- the four corners (visual or olfactory), the rats ignored them
mals in a multisensory manner, and (2) there was explicit and searched systematically at both the correct corner and
648 The Hippocampus Book

the diagonally opposite corner. These two corners are geomet- 1330B). Training continued to the one corner with a long
rically equivalent in a symmetrical rectangle. These ndings wall on the right and short wall on the left. In posttraining
seemed to imply that rats were more sensitive to environ- probe tests, preferential searching was now observed at the
mental shape than to local features, such as the distinc- correct location but also at the apex of the maze (Fig.
tive cueshence the proposal of a geometric module. 1330C). This nding makes no sense if the animals are using
Observations and models of place cellsspecifically a geometric module representing the shape of the kite but is
their determination by boundary vectorsare consistent readily explicable if the animals have learned to search prefer-
with both this and OKeefes preferred local feature view entially at the left-hand end of any long wall. Interestingly, rats
(OKeefe and Burgess, 1996; Hartley et al., 2000; Lever et al., with hippocampal lesions were poorer at learning to search at
2002). the correct corner unless prominent disambiguating local
In an ingenious test of the geometric module idea, Pearce cues were added to walls on either side of the apex during
et al. (2004) wondered if the symmetrical search patterns seen training (Fig. 1330D,E). Once these local cues were added,
in a rectangular array reected the operation of a specialized the training corner and the apex were easily distinguished
geometric module or something simpler. The animal learning visually and hippocampus-lesioned animals then learned as
constituency enjoys wielding Occams razor, and this issue well as sham-lesioned controls. The implication of this ingen-
represented another opportunity. Might the rats have instead ious study is that there need be no recourse to an abstract geo-
learned only the rule of searching at a corner dened by one metric module to learn these kinds of tasks and no need to
long wall on the right and one short wall on the left? To test suppose that the hippocampus mediates such a module. The
this, they rst trained rats to search for a hidden escape plat- data are, however, consistent with the boundary vector model
form in a rectangular watermaze and observed, in a conven- of place cells in which their location is xed by intersecting
tional probe test with the platform absent, focused searching vectors from local cues.
at both the correct corner and the geometrically equivalent Yet more abstract geometry is implied by the observation
corner (Fig. 1330A). This result is identical to that found by that rats trained to forage for food in the center of a square
Cheng and Gallistel in the food-rewarded task. The animals arena can transfer to searching at the geometric centre of
were then transferred to training in a kite-shaped maze (Fig. rectangular and triangular arenas (Tommasi and Thinus-

Figure 1330. Geometry and the shape of spatial learning to come. sham-lesioned controls when learning to search correctly in the kite
A. Rats trained to nd an escape platform in one corner of a rectan- maze. E. When disambiguating local cues were added to the long
gular watermaze also search in the geometrically equivalent corner. walls of the rectangle and on either side of the apex, hippocampus-
B. The rats were then transferred to swimming in a kite-shaped lesioned animals could learn as rapidly as controls. The bar graph
arena, with the platform again located at the corner with a long wall data have been simplied to the long walls of the rectangle and
on the left and a short wall on the right. C. During posttraining either side aid the description of this study. (Source: After Pearce et
probe tests, the rats now searched at the correct location and at the al., 2004.)
apex. D. Hippocampus-lesioned rats (gray shading) are poorer than

A Hidden platform in basic rectangle B Transfer to kite-shaped arena

G
A B
incorrect
geometrically incorrect
correct
F apex obtuse H

locally
incorrect correct correct

D C E

C Searching at left-hand D HPC lesion deficit E Successful local cue learning

end of a long wall.
Performance

Performance

Performance

Chance

Sham HPC Sham HPC

ct

e
ct
ex

us
re

re
Ap
or

bt
r
co

O
C

In
 Theories of Hippocampal Function 649

Blanc, 2004). This result is not, however, always observed. One rather than inside. Astonishingly, a signicant proportion of
classic series experiments contrasted a map-like representa- the search visits during a posttraining probe test with the
tion of space to one based on vectors (Collett et al., 1986). rotated array were outside the landmark array. This striking
Vectors encode distance and direction. In a directionally pattern is not to be expected had the animals learned a map
polarized environment, a vector representation would be one of space with food at the center of the landmark array. It is
in which object A may be represented as a vector with a spe- consistent with the encoding of three independent vectors
cic argument (i.e., length) and angle from landmark P. specifying the distance and direction of food from each of the
Many objects (A, B, C, and so on) both visible and hidden, three landmarks. During training, these vectors meet at the
may be represented in such a way, but it does not require there single, central point; after rotation they do not. Collett et al.
to be any map of space. Specically, there may be no encod- (1986) did not go on to conduct lesion studies, so we do not
ing of the relative locations of A, B, and C to each other. To test know the impact of the hippocampal dysfunction. However,
this idea, gerbils were trained to nd sunower seeds at par- in a conceptually similar lesion study using the watermaze,
ticular distances and directions from landmarks in a room Pearce et al. (1998) found that hippocampus-lesioned rats
that provided cues enabling directional polarization but not could be trained to locate a hidden platform at a specic dis-
localization. The landmarks were then manipulated in various tance and direction from a visible landmark that moved places
ways. In one manipulation, animals were rst trained to nd within the pool between days. It seems that the hippocampus
food hidden halfway between two identical landmarks. The is not required to encode vectors across multiple trials but
two landmarks were then moved farther apart. If the animals may be necessary to assemble multiple vectors into a map-like
had learned to nd reward at some abstract centre of the representation.
array, they would be expected to search preferentially at a With respect to multiple types of memory, there are good
common point, now a bit farther from each of the landmarks. grounds to believe that there has been selection pressure
Instead, searching occurred at two locations, each at the to develop map-like spatial representation and navigation sys-
appropriate distance from each landmark along the line tems in vertebrates. However, the studies comparing spatial
between them (Fig. 1331A). In a yet more striking example learning with conventional associative conditioning have
of vector-based representation, the gerbils were trained to nd revealed fewer differences than might have been expected
sunower seeds at the center of a triangular array of three dis- on the basis of the original cognitive map theory. The possi-
tinctive landmarks (Fig. 1331B). The landmark array was bility that spatial information is fundamentally associative,
then rotated by 60 such that the vectors representing reward in contrast to the position adopted within cognitive mapping
location from each landmark were now outside the array theory, has to be considered. Section 13.5 takes up this

Figure 1331. Maps or vectors? A. Gerbils were trained to nd center of an array of three distinctive landmarks in an environment
food halfway between two landmarks (large circles). When the land- in which directionally polarization is determined by other distal
marks were moved farther apart in a posttraining probe test, the cues. When the landmark array was rotated by 60 in a probe test, a
animals searched at two distinct locations (small squares denote substantial proportion of the searching was now outside the arena.
time spent searching). B. Gerbils were trained to nd food in the (Source: After Collett et al., 1986.)

A Searching for food between B Searching for food at the centre of 3 distinctive
2 landmarks landmarks

training non-reward training non-reward

probe tests probe tests

Food

Food

Directionally polarised
environment

Landmark

Locations of searching in
non-reward probe tests
650 The Hippocampus Book

issue further; but before turning to that, we discuss the is usually engaged is in the spatial domain, where it may be
issue of where allocentric spatial information is stored in the helpful in updating during navigation.
brain. Reecting on these putative functions of systems-level con-
solidation, Moscovitch wondered if memory retrieval with
13.4.5 Storage and Consolidation recovered consciousness might be disentangled from other
of Spatial Memory features of the standard model. This led on to a series of arti-
cles in which he and Nadel proposed an alternative multiple
A fourth major proposition of cognitive map theory is that the trace model of memory persistence (Nadel and Moscovitch,
memory traces required for allocentric representations 1997, 1998; Moscovitch and Nadel, 1998). In this model, a
through a familiar place are stored in the hippocampal forma- subset of memories is permanently mediated by corticohip-
tion itself. This storage is held to be distributed, possibly along pocampal circuitry. Specically, the theory holds that each
both the longitudinal and transverse axes of at least the dorsal reactivation of memory leads toor at least can lead tothe
hippocampus for even a single memory, such that damage to formation of additional traces that are located at different
the structure should cause a loss of both processing capacity points along the longitudinal axis of the hippocampus
(encoding, retrieval, and navigation) and storage sites (local (anterior-posterior in humans). Temporal information is
alterations in synaptic weights). Hippocampal damage should thought to be stored in the prefrontal lobe, at least in humans.
therefore cause a at gradient of retrograde amnesia for spa- Consolidation is thus a reactivationrestorage cycle that
tial information, as rst shown by Bolhuis et al. (1994). In occurs in a punctuated manner over time, rather than as a
contrast to the retrieval of factual information by humans or gradual, inexorable biological processan analogy might be
contextual fear by rats, where gradients are observed (see drawn with the concept of punctuated equilibrium in evolu-
Section 13.3), the integrity of structures in the hippocampal tionary theory (Gould and Eldredge, 1977). There would then
formation is also claimed to be essential for successful recall of be multiple traces of a common past event, but these traces
episodic memory in humans (Viskontas et al., 2000; Cipolotti could entail different associations with other information that
et al., 2001; Maguire, 2001). Moreover, very remote episodic reect their distinct provenance. Temporally graded retro-
memories may still elicit neural activation of the hippocam- grade amnesia for spatial information can then be explained
pus long after they were rst acquired (Nadel et al., 2000), by the proliferation of memory traces in the hippocampus
although the use of reminders prior to brain scanning has without recourse to a neocortical consolidation process, a pre-
raised a query about the true of age of ostensibly old memo- diction that has been successfully modeled with a connection-
ries in imaging work. Given the very different perspective on ist network (Nadel et al., 2000). Remote spatial memories are
these matters in declarative memory theory, it is no surprise represented by a larger number of traces and so be more
that memory consolidation has become one of the battleelds accessible than recently acquired memories. According to a
of contemporary cognitive neuroscience. later version of this model (Nadel and Bohbot, 2001;
We have seen that the standard model of memory per- Rosenbaum et al., 2001), this multiple-trace theory applies
sistence (see Section 13.3) is that some systems-level process particularly to contextually rich memories but not those that
of hippocampus-dependent memory consolidation takes are context-free or semantic in character. Whether these pro-
place after learning that, by enabling an interaction of hip- posals should be seen as extensions of the cognitive map the-
pocampal and neocortical ensembles, secures the stabilization ory or as independent ideas inspired by it is unclear. They are
of memory traces outside the hippocampus (Squire and probably more the latter.
Alvarez, 1995a; Kapur and Brooks, 1999; Manns et al., 2003). The declarative memory and the multiple trace theories
The neurological function of such systems-level consolidation make contrasting predictions about retrograde amnesia in the
presumably has nothing to do with the protection of memory particular case of partial lesions. Unfortunately for theories of
against brain damage. Exploiting the unfortunate occurrence memory, the capricious nature of human brain injury results
of relevant brain damage in humans, or its experimental cre- in most MTL amnesic subjects having only partial damage to
ation in animals, is just a technique for revealing this neuro- the hippocampus (Rempel-Clower et al., 1996; Bayley et al.,
logical process. Rather, consolidation is thought to have two 2005b). Such patients do not discriminate the two theories.
functions. First, it engages mechanisms that gate whether One neurological case does exist where the damage is appar-
memory traces are to be retained. Second, once consolidation ently extensive yet there remains clear evidence of sparing of
is complete, a lasting neocortical memory site exists that spatial memory from a much earlier time period in the
would enable much faster retrieval of information than would patients lifepatient E.P. (Teng and Squire, 1999). However,
be possible were it always necessary for there to be a neural we have seen that the testing of remote memory in humans
loop: through the hippocampus. A consolidation process and the identication of an explicit temporal gradient is
might also allow retrieval of information from neocortical fraught with difficulty (see Chapter 12). Can animal studies of
traces when it is required without recourse to consciousness, spatial memory shed light on this important theoretical issue?
but the activation of this neural loop may be a signature of Studies using the watermaze have consistently shown a at
conscious awareness in the specic domain of memory gradient of retrograde amnesia for spatial memory when it is
(Moscovitch, 1995). One situation where conscious awareness rst tested postoperatively. In such studies, rats have typically
 Theories of Hippocampal Function 651

rst been trained on a reference memory version of the task dorsal or ventral hippocampus (Fig. 1332A). This rst nd-
(i.e., to nd a single hidden platform); then, at periods rang- ing of Martin et al.s (2005) study is inconsistent with both the
ing from a few days to up to 3 months later, they have been declarative memory and the multiple-trace theories but con-
subjected to large excitotoxic or radiofrequency lesioning of sistent with the original statement of the cognitive map the-
the hippocampal formation (Bolhuis et al., 1994; Mumby et ory, subject to the proviso of there being some threshold of
al., 1999; Sutherland et al., 2001; Clark et al., 2004, 2005b). intact residual tissue ( 35%) necessary to mediate effective
Subsequent to surgery, the animals have been retested in the navigation.
watermaze either with the platform in its original training However, in these and other studies, the performance of
location, with it moved to a new location, or using probe tests sham-lesioned rats in the watermaze tends to be poor after
(platform absent). Martin et al. (2005) used such a design surgery, possibly reecting a oor effect that masks any con-
with animals given partial hippocampal lesions (circa 35% solidation or multiple trace process. To address this problem,
sparing of either the dorsal or the ventral hippocampus) Bolhuis et al. (1994) followed up the initial probe test with
and others complete lesions ( 5% sparing). They used an retraining to unmask cryptic spatial memory traces. If any-
Atlantis platform during training to encourage highly focused thing, their savings data favored better relearning by recently
searching and better memory. An initial posttraining probe lesioned groups (made 3 days after original training) than the
test provided no evidence of better retention of remote spatial remote groups. Rats given subiculum lesions that left the hip-
memory, even in the group with a partial lesion of either the pocampus intact were also capable of relearning to a high

Figure 1332. Retrograde amnesia for spatial memory. A. An initial lesioned group. The reminding procedure was shown not to induce
pre-reminder probe test conducted 2 weeks after sham (white), relearning. (Source: After Martin et al., 2005.) C. Extensive training
and either partial (gray) or complete (black) hippocampal lesions from an early age failed to reveal spared spatial memory in animals
revealed no evidence of spared spatial memory recall in either with complete hippocampal lesions. (Source: After Clark et al.,
lesioned group. This study conrms the earlier ndings of Bolhuis 2005a.) D. Spatial recognition memory tested using an annular
et al. (1994) but under conditions in which a oor effect in the watermaze that obviates the need for hippocampal updating during
sham control group did not occur. (Source: After Martin et al., navigation also failed to reveal spared spatial memory after com-
2005.) B. Reminding given in the form of successive probe tests plete hippocampal lesions. (Source: After Clark et al., 2005b.)
revealed some sparing of spatial memory recall in the partially

A Spatial recall before reminding B Spatial recall with reminding

first probe test acts as a reminder post-reminder probe test

100 100
relative memory

relative memory
sparing (%)

sparing (%)

60 60
20 20
-20 -20
recent remote recent remote

C Spatial recall with extensive training

probe test 60
correct quadrant
% time in the

Age (days) 40
21 90
20 chance Lesion groups
NC and sham
0 H (partial)
v.remote H (complete)
D Spatial recognition in annular watermaze
probe test 30 Training-surgery intervals
% time in the
correct zone

recent: 0-3 days

20 remote: 6-9 weeks
v.remote: 14-16 weeks
10

0
remote v.remote
652 The Hippocampus Book

degree of prociency. This is consistent with the integrity of cular corridor until they nd an escape platform. The use of
hippocampal circuitry being important for some aspects of the annular watermaze is interesting as the need for naviga-
spatial memory storage. Mumby et al. (1999) also used tional updating is minimized; the swimming rats have only to
retraining and a within-subjects protocol with two watermaze recognize a potentially safe place along the ring and slow
tasks that were trained at different time points prior to sur- down when they get there. However, using a single probe test,
gery. They found little evidence of spatial memory in their both tasks again showed a at gradient (Fig. 1332D).
complete hippocampal lesion group for either of the two The animals joy in eventually nding water in the Oasis
tasks. Sutherland et al. (2001) used both retraining and probe maze cannot have been matched by any exhilaration on the
tests and observed a signicant trend for longer training part of proponents of the declarative memory theory by the
surgery intervals to be associated with poorer retentiononce outcome of this series of experiments. Clark et al. (2004) sug-
again the opposite of what is predicted by the declarative gested that a feature that may distinguish tasks where remote
memory theory. These studies reect a consistent pattern, but spatial memory is spared (cross and radial maze tasks) com-
the use of retraining after surgery confounded two factors: the pared to those where it is impaired (watermaze tasks) is that
impact of the lesion on consolidation (the focus of interest) only in the latter does the expression of memory require the
and its impact on new learning (a separate matter). now-lesioned animal to recall by navigating through open
In their comparison of partial and complete lesions, Martin space. That the hippocampus might be essential for expressing
et al. (2005) used a different approach: a series of rewarded spatial memory, rather than storing or retrieving it, is remi-
probe trials at 1-hour intervals. This gradually revealed above- niscent of Whishaws distinction between getting there
chance performance by the partial hippocampus-lesioned (impaired by lesions) and knowing where (which may not
group that had undergone surgery immediately after training always be). A processing decit of this kind conveniently pre-
(i.e., their recent group). No recovery of remote memory was dicts that any underlying hippocampal-neocortical consolida-
observed in the group for which there was a much longer 6- tion process would remain cryptic.
week interval between the end of training and surgery (Fig. There are, however, problems with this account. First, it
1332B). This apparent unmasking of latent spatial memory incorrectly predicts that remote memory should be intact in
was not due to any relearning during the series of probe trials, the annular watermaze because its navigational demands are
as de Hoz et al. (2004) had separately shown that the conduct minimized. The animal does not have to learn or access any
of a probe trial with an Atlantis platform serves, in animals navigational strategies for getting thereit has merely to
with partial lesions, only as a reminder cue. The result vindi- swim in a constrained circle. Second, remember that reminder
cates the interpretation drawn by Bolhuis et al. (1994) much cues revealed latent spatial memory in partially lesioned ani-
earlier. mals (Martin et al., 2005). Presumably. sufficient hippocampal
Noting the same problem of weak postsurgery memory, tissue was spared for the expression of memory. However, this
Clark et al. (2004) extended the presurgery training to as latent spatial memory was in the recent-memory group, not
many as 80 trials, nding that it improved the performance of the remote. If cryptic consolidation does occur, the effects on
the sham/control group. The point of doing this is to increase trace strength must therefore be substantially smaller than
the opportunity of seeing an upward consolidation function those of forgetting. Third, remember that different results are
in the data (along the lines of Fig. 1310B3). However, using obtained after small partial lesions as a function of whether
large radiofrequency lesions restricted to the hippocampus, they are made before or after training (Moser et al., 1995;
they again observed both poor memory and a completely at Moser and Moser, 1998b). When spatial memory is acquired
gradient of retrograde amnesia in postsurgery probe tests. by normal animals and then partial lesions are made in the
Another attack on the problem was suggested by the fact that hippocampus, performance is seriously compromised and
patient E.P. had learned his home neighborhood as a young needs to be relearned. This makes sense if trace storage of spa-
man. Clark et al. (2005a) therefore trained normal rats from tial memory is within the hippocampus, it being fully distrib-
very early in life (21 days) for a total of 392 trials over 69 days. uted along the longitudinal axis when created in an intact
Hippocampal and sham lesions were made at 90 days of age, hippocampus. Relearning would be possible with a partially
and the animals were tested 2 weeks later. Once again, the intact hippocampus, and this would also be enough to express
complete lesion group failed to display above-chance per- the memory once retrieved.
formance and that despite the use of a reminding procedure a The last twist of the consolidation story comes from the
few days before testing (Fig. 1332C). Undeterred, Clark et al. use of novel techniques. Using a behavioral training proto-
(2004) wondered if there was something unusual about the col similar to that of Martin et al. (2005), indirect evidence for
standard watermaze task, as other nonspatial tasks had found the presence of a systems level consolidation process after
evidence of a temporal gradient, albeit only using a savings or watermaze training was obtained using a pharmacological
choice measure (Cho et al., 1993, 1995; Ramos, 1998; Maviel approach. Riedel et al. (1999) observed that chronic reversible
et al., 2004). Accordingly, they also trained rats in an Oasis inactivation of the hippocampus for 1 week after training
maze, in which they have to search for water in a small well through bilateral infusion of a GluR15 antagonist impaired
located on a desert-like table top, and in an annular water- spatial memory when the animals were tested 16 days later
maze (Hollup et al., 2001) in which the animals swim in a cir- (i.e., long after the effects of the drug had worn off). The drug
 Theories of Hippocampal Function 653

demonstrably blocked fast synaptic transmission in the hip- essary role in this process. Memories are widely thought to be
pocampus, implicating the necessity for hippocampal neural stored in a distributed manner (i.e., their component traces
activity for some period after training if spatial memory was are stored across different brain areas). Retrieval of the right
to persist. Riedel et al. (1999) explored the impact of delaying combination of traces might be possible under circumstances
the shut-down by a few days and got the same results, but in which the retrieval cues are either particularly apposite or
they did not examine a longer time course after training. This sufficiently rich to disambiguate cortically based traces with
reversible but chronic inactivation approach is promising ana- overlapping components. However, under circumstances in
lytically; but even if hippocampal inactivation long after which the retrieval cues do not permit easy disambiguation or
training was found to have minimal impact on remote mem- where they have to be generated indirectly through a process
ory, it would not mean that the neural activity required for of recovered consciousness (Moscovitch, 1995), the process-
memory persistence is necessarily a hippocampal-neocortical ing capacity of the hippocampus and its stored indices would
dialogue of the kind presumed with the standard model. It remain essential. Disambiguation would generally be critical
could also be the creation of multiple traces in the hippocam- for context-dependent episodic memory in which a unique
pus, or some other intrahippocampal consolidation process, combination of traces corresponds to the particular event
that requires neural activity. Reversible inactivation may from the past that needs to be reactivated. Consistent with this
not be quite the help for distinguishing the rival theories it idea, the expression of human episodic memory is sometimes
rst promised to be. A similar ambiguity pertains to the strik- found to be permanently affected by hippocampal damage, as
ing evidence that sectioning the direct input from layer III noted earlier. Notwithstanding our focus on memory retrieval
of the entorhinal cortex to area CA1 of the hippocampus processes, this account is consistent with Rosenbaum et al.s
(the temporo-ammonic tract) is essential for memory consol- (2001) revision of the multiple trace theory. It distinguishes
idation (Remondes and Schuman, 2004). When lesions were between context-dependent and context-free memories and
made before 7 days of watermaze training, the animals showed suggests that only the former remain dependent on the hip-
good probe test performance 1 day after training but poor pocampus for our lifetime. Winocur et al. (2005) have
memory 28 days later; sham-lesioned animals showed no reported data consistent with this view in a study of rats that
decline. The same result was obtained when the lesions were lived in a rodent village for 3 months prior to explicit train-
made 1 day after the end of training. However, if the lesions ing on a spatial problem in the village. Posttraining hip-
were delayed until 21 days after the end of training, both the pocampal lesions had no effect on effective navigation to the
sham and temporo-ammonic tract-lesioned animals showed a correct location in the village, which they interpreted as
continued, albeit weaker, tendency to swim in the correct implying the existence of a detailed semantic map stored in
quadrant 7 days later. Collectively, these three experiments are the neocortex. Rats with less experience of the village were
consistent with the idea that ongoing cortical input conveyed impaired by posttraining hippocampal lesions because they
by the temporo-ammonic path is required to consolidate presumably lacked a context-free memory of the village.
long-term spatial memory (Remondes and Schuman, 2004, Unfortunately, the lesions used in this study were partial
p. 702), the implication being that this pathway is necessary for (around 50%), allowing some use of spared hippocampal tis-
the hippocampal-cortical dialogue implicated in the declara- sue in memory retrieval. Winocur et al. (2005) argued against
tive memory theory. However, this is not absolutely required this being of signicance on the grounds of nding no corre-
by the data. It could still be that temporo-ammonic input is lation between lesion size and retrieval, but this is a weak
essential for ongoing consolidation in the hippocampus. argument based on only a small number of subjects with dif-
A nal expression theory, as discussed by Martin et al. ferent sized lesions.
(2005), relates to the role of the hippocampus in the retrieval Studies of interactions and correlations between single-unit
and expression of consolidated memory traces that reside in and eld-potential recordings in hippocampus and neocortex
the neocortex but for which the hippocampus plays a part in during and after sleep are highly suggestive of consolidation-
their retrieval. There are a variety of possibilities. One is that like processes (Siapas and Wilson, 1998; Sirota et al., 2003).
the hippocampus does store information but stores only the Targeted molecular studies can help resolve some of the ambi-
indices (Teyler and DiScenna, 1986) or the cartoons guities at the mechanistic level, although not necessarily all of
needed to retrieve a consolidated cortical memory trace rather them, as a daunting combination of novel behavioral proto-
than detailed sensory-perceptual information (this remains in cols, lesions, and molecular interventions will probably prove
the cortex). With this view, the hippocampus might retrieve a necessary to unravel the complexities. Frankland et al. (2001)
cortical memory in a manner analogous to conducting a key- made the intriguing observation that heterozygous CAMKII
word search on an electronic document. Partial lesions would mutant mice show normal LTP in the hippocampus, decaying
compromise the indices rather than the detailed sensory- LTP in the cortex, and failure to consolidate both spatial and
perceptual memory traces and so limit the ability to retrieve contextual information. They suggested that the instability of
cortical memory traces when a recall process is necessary. The synaptic potentiation in the cortex could be the basis of the
circumstances surrounding memory retrieval and the very consolidation failure. This is an interesting idea, but it is
character of the information retrieved (Nadel and Bohbot, equally consistent with the idea that effective retrieval relies on
2001) may determine whether the hippocampus plays a nec- the synergy of different kinds of information permanently
654 The Hippocampus Book

stored in both hippocampus and neocortex. Frankland and have come back to a central prediction of the theory in rela-
Bontempi (2005) summarized these developing new lines of tion to brain activity associated with exploration and spatial
research using transgenic animals and IEG markers to plot the novelty, most recently in showing differential activation of the
time course and regional contributions of neocortex and hip- human hippocampus proper when people take shortcuts in a
pocampus to memory consolidation. virtual reality environment (Maguire et al., 1998; Burgess et
al., 2002). Quantitative models are precise and testable, such as
13.4.6 Critique the boundary vector model of place cells (Hartley et al., 2000).
However, the concept of a geometric module is undergoing
The cognitive map theory has been highly inuential. The role something of a forensic examination with experiments in
of the hippocampus in spatial learning and memory is rodents (Golob and Taube, 2002; McGregor et al., 2004b;
accepted, particularly in rodents; and such learning is widely Pearce et al., 2004) and humans (Wang and Spelke, 2002),
used as an assay for investigating the cellular and molecular raising questions about whether there is any need for a
mechanisms of learning and its dysfunction in models of global representation of environmental shape, as in the
neurodegenerative disease. Spatial memory has become a study of the kite-shaped watermaze discussed earlier. Pearces
keyword at international meetings, such as those of the study, part of a larger series, breaks new ground because it
Society for Neuroscience and the Federation of European suggests that earlier data pointing to the existence of a global
Neuroscience Societies, and there are now laboratories all over geometric module centered on the hippocampus is also con-
the world investigating the neurobiological basis of allocentric sistent with a different account that sees no need for such a
spatial representation and spatial navigation. Although much representation. Learning may require only a perceptual system
of this reects a general interest in spatial learning and mem- that can distinguish long barriers from short ones (i.e., per-
ory rather than commitment to one particular theory, it is ceptual features of local cues in general) and bridge the tem-
perhaps not surprising that it has been criticized as the theory poral delay between turning appropriately as the animal
that wouldnt die. approaches the long wall and the subsequent receipt of
However, the theory clearly has major difficulties, both delayed reward. It remains to be seen where this research on
conceptual and empirical. This chapters presentation of the local features versus boundaries and geometry will lead.
theory was built around four themesspatial representation,
spatial navigation, comparative work, and studies of memory Successful Spatial Navigation
storage and consolidationand has already included a meas- Without the Hippocampus
ure of critical discussion. A few additional points should be
noted in relation to these same four themes. A second set of difficulties for the theory is that rodents with
nearly complete cell loss in the hippocampal formation can,
Spatial Representation, Maps, with certain training procedures, learn allocentric spatial
and the Geometric Module tasks. This is the longstanding claim of the working memory
theory of Olton et al. (1979), who argued that spatial reference
With respect to representation, one set of problems center memory tasks can be learned by lesioned rats. We have seen
around whether the concept of a cognitive map is an expla- that this claim is correct, but the rate of learning is much
nation of anything or merely a beguiling metaphor (Healy, slower. Regardless of whether it is slower, the inescapable
1998). A map is an easily understood concept, but maps are implication is that extrahippocampal structures can support
things that people look at to extract information. Adopting spatial learning. This nding that the integrity of the hip-
this term for the neural activity of a region of the brain seems pocampus, or of hippocampal synaptic plasticity, is not essen-
to carry with it the mental baggage that there must be some tial for spatial reference memory is not necessarily fatal for the
cryptic homunculus that is looking at the map to do like- cognitive map theory because recent experimental and theo-
wise. In the absence of a mechanistic account of how informa- retical studies have established that effective spatial navigation
tion is extracted from the map, the explanation it offers is can be realized in many ways, including strategies other than
incomplete. With respect to the concept of cognitive mapping map-guided navigation. However, certain observations, such
itself, Bennett (1996) argued that the now widespread use of as normal probe test performance in the watermaze by rats
the term in many different ways by investigators in ethology, with complete hippocampal lesions (Morris et al., 1990) must
psychology, and brain science is confusing and that the con- be accepted as problematic unless the theory is to descend into
cept has outlived its usefulness. This feels unpersuasive. That irrefutability. That rapid, one-trial spatial learning in the
some people use a scientic term loosely is not grounds for DMP task is always impaired (Steele and Morris, 1999) sug-
castigating those who use it more precisely. As Bennett (1996) gests that the integrity of the hippocampus is essential for
himself conceded, Tolman (1948) introduced it and OKeefe rapid encoding of spatial information (spatial events?), but
and Nadel (1978) used it explicitly in relation to the ability to the gradual accumulation of information about spatial regu-
represent the environment in a viewer-independent allocen- larities of an environment might be information that can,
tric manner and to solve such problems as novel shortcuts. albeit slowly, be encoded, stored, and retrieved in the neocor-
This denition is concise. Time and again, OKeefes group tex without recourse to the hippocampus. That retrograde
 Theories of Hippocampal Function 655

amnesia is apparently always seen after both partial and com- aspects of human memory, its adequacy as a general theory of
plete hippocampal lesions in rats appears to be contradictory, multiple types of memory is limited. It has little to say about
but the initial learning in such tasks always takes place in nor- the relations between propositional (declarative) and non-
mal animals. propositional forms of long-term memory (e.g., skill learn-
In passing, it is relevant to note that knocking out the ing), between working and long-term memory, or, more
capacity to express a particular form of NMDA receptor- curiously, about the nature of perceptual representations. Its
dependent LTP through mutation of the GluR-A receptor proponents would argue that the theory was never intended to
selectively impairs one-trial spatial working memory but not be a general theory of memory and cognition and that Figure
incremental spatial reference memory (Zamanillo et al., 1999; 1317 reects their primary focus. It is a theory built around
Reisel et al., 2002). The knockout is not hippocampus- the discovery of place cells and the sensitivity of spatial navi-
specic, but a genetic rescue of the receptor mutation that gation by rodents to various kinds of brain damage. It has
appears to be largely expressed in the dorsal hippocampus imaginatively extended from these beginnings, but its place as
(Schmitt et al., 2005) is sufficient for partially rescue of effi- a building block of a more general neurobiological theory
cient spatial working memory performance. This body of remains a task for the future.
work is discussed in Chapter 7. For the present, note only that However, even within the domain of the cognitive analysis
it illustrates how new techniques are shedding light on what it of space, there are problems with respect to the claim that spa-
means to classify a task as hippocampus-dependent. A mouse tial learning is fundamentally different from associative learn-
in which GluR-A is absent shows apparently normal fast ing. In 1978, it was valuable to make the point that associative
synaptic transmission in the hippocampus but fails to show a conditioning is generally slow and inexible, whereas place
rapidly induced form of LTP. Such an animal is like a lesioned learning can occur during one trial; but strictly behavioral
animal in that it cannot display types of memory encoding studies of allocentric place learning have now revealed that
that occur in the hippocampus; but it is more subtle than a spatial learning displays several qualitative characteristics of
lesioned animal, as it allows information throughput in the associative learning, such as blocking (Chamizo, 2003).
hippocampus, which may be essential for neocortical encod- Spatial learning may be associative after all.
ing of incremental learning.
If this is correct, one would expect that damage to certain Integrity of the Hippocampus Is Required
neocortical brain areas (and not just the entorhinal cortex) for Many Nonspatial Learning Tasks
would have effects on spatial navigation that are clearly not
predicted by the theory. These have indeed been observed. If rats with hippocampal lesions can learn certain spatial
Decits in spatial navigation in a variety of tasks follows tasks, a separate concern is that such lesions are now also
lesions in the posterior parietal cortex (DiMattia and Kesner, known to affect a range of nonspatial tasks, including classic
1988a; Kesner, 1998) and the midline retrosplenial/cingulate operant tasks with a temporal component, such as differential
cortex (Sutherland et al., 1988). Harker and Whishaw (2004) reinforcement of low rates of response (DRL). Although it has
summarized a body of conicting data, including strain differ- always been possible to construct a spatial account of DRL
ences, regarding involvement of the retrosplenial cortex in var- performance, a temporal one seems more reasonable
ious spatial tasks including navigation. Aggleton and Vann (Rawlins, 1985). However, there a number of other tasks, such
(2004) concurred that retrosplenial lesions do impair allocen- as nonlinear discrimination learning and socially transmitted
tric spatial navigation in the watermaze, though lesion size can food preferences, for which a spatial account would strain
be a critical factor (Vann and Aggleton, 2004). Focusing on a credibility. Section 13.5 is a detailed examination of such tasks
different part of the neocortex, Kesner (2000) argued that there in the context of the predictable ambiguity theories for
is parallel processing of spatial information in the hippocam- which they are relevant.
pus and parietal cortex, with the former more important for
spatial events and the latter part of a neocortical knowl- Coda
edge system. This distinction echoes the episodic/semantic
distinction to which we return later. It is also is relevant to the As a last word, it is worth reecting on the fact that the cogni-
issue of whether memory storage is in the hippocampus or tive map theorythe theory that refuses to dieenjoys the
there is some hippocampally guided consolidation process. security of a number of key physiological ndings that are not
Kesner (1998) is not alone in regarding a strictly spatial view of enjoyed by the other main theory discussed so far, the declar-
hippocampal function as too narrowly drawn. ative memory theory. The theoretical battleeld between these
two theories would be very different if the discovery of other
Multiple Types of Memory? spatially responsive neurons (e.g., head-direction units, view
cells, grid cells) had not occurred. All of these discoveries were
Another conceptual problem is that the simple division of made since 1978. Hippocampal neurons are not differentially
learning processes into spatial versus nonspatial now appears active as a function of stimulus familiarity, but they clearly are
dated. Although there have been many developments of the responsive to an animals location or view of the world. This
cognitive map theory, including its application to various is a major if somewhat neglected obstacle to the credibility of
656 The Hippocampus Book

declarative memory theory. We have also seen a number of offers insights into the way in which animals learn about the
studies in which interventions in the induction and expres- causal texture of their environment (Dickinson and
sion of LTP have an effect on spatial learning and retention Mackintosh, 1978). This implies that conditioning could have
(see Chapter 10), although the extent to which these effects are to do with the acquisition of knowledge that could be used
cognitively selective is not well understood. Some comfort for inferentially (Mackintosh, 1983; Rescorla, 1988; Pearce and
the declarative memory theory in the physiological domain Bouton, 2001; Pickens and Holland, 2004) as rst discussed
can be derived from glucose uptake and IEG studies indicat- explicitly by Dickinson (1980). These intellectual develop-
ing greater neuronal activity in the hippocampus soon after ments are relevant here because psychological processes have
learning than later, perhaps reecting the dynamic process of been identied in modern learning theory for which the hip-
memory consolidation. These observations are pertinent to pocampus is a potential anatomical substrate.
the task of trying to build a neurobiological account of hip- Paradoxically, it has long been known that animals with
pocampal function to which we return in Section 13.6. hippocampal lesions are generally normal regarding simple
associative conditioning tasks. These tasks involve no more
than the pairing of an initially neutral stimulus (CS) with,
after a delay, the presentation of reinforcement (US). The
13.5 Predictable Ambiguity: Congural, result is the gradual development of associative strength and
Relational, and Contextual Theories of its expression as a conditioned response (CR). In general, sim-
Hippocampal Function ple delay conditioning depends on other brain circuits. For
example, hippocampal lesions are without effect in nictitating
The two major theories discussed so far command strong sup- membrane (NMR) delay conditioning in which a CS starts
port and continue to be rened in the light of new ndings. several hundred milliseconds before and then overlaps with
Neither will be discarded until a cogent alternative is widely the presentation of a strong puff of air to the eye (US). NMR
seen to account for a larger body of data. However, numerous learning was shown independently by several groups (using
other theories of hippocampus function have been proposed. lesion, inactivation, and unit-recording techniques) to be
They include hypotheses that build on modern associative mediated by circuits in the spinal cord and cerebellum (Krupa
learning theories whose origins, unlike the concept of multi- et al., 1993; Hardiman et al., 1996; Beggs et al., 1999). As noted
ple memory systems, lie in the phenomenon of classical and in Chapter 11, hippocampal pyramidal cells nonetheless show
instrumental conditioning rather than human amnesia. In ring patterns during NMR conditioning that are correlated
one way or another, their applications to thinking about hip- in time with the learned response (Berger et al., 1983), but this
pocampus function all have to do with dealing with the prob- neural activity is not usually necessary for normal perform-
lem of predictable ambiguity. This phrase is intended to ance. The exception to this is when a trace interval is
capture the sense that a stimulus can consistently mean one inserted between the end of the CS and the onset of US, as we
thing in one situation but something else in a different one. saw in the discussion of the declarative memory theory; alter-
Learning theory research is characterized by exacting train- ations in excitability are then seen in both young animals
ing protocols that explore conditioning phenomena far (Moyer et al., 1996) and aged animals (Moyer et al., 2000).
removed from the classic account of learned salivation to the Simple fear conditioning to a punctate CS such as a tone (in
sound of a bell. As we touched on briey in the preceding sec- which it is sounded for a period prior to and overlapping the
tion, examples include protocols for examining whether learn- delivery of an aversive stimulus such as a mild electric shock)
ing occurs when an animals expectations are violated or that is thought to involve storage of traces in the amygdala (Davis
it alters the associability of stimuli or the contexts in which et al., 1993; LeDoux, 2000), although others have argued that
they occur (Rescorla and Wagner, 1972; Pearce and Hall, the role of the amygdala is primarily neuromodulatory
1980). Such learning could include both excitatory and (Vazdarjanova and McGaugh, 1999; McGaugh, 2000). This
inhibitory associations (i.e., that a predicted event is more or alternative is important because, as we saw in the discussion of
less likely to happen). Beyond simple associations, there are late LTP (see Chapter 10), activity in the amygdala can modu-
also experience-dependent attentional phenomena and late persistence of LTP in the hippocampus. These and many
occasion-setting paradigms in which a stimulus (or more other studies have broadened our understanding of the role of
broadly an entire context) inuences the retrieval of other various brain areas in learning and memory.
memories quite apart from any associations in which it may The hippocampus may, however, become denitively
also become engaged. Such phenomena have led to the sugges- engaged in conditioning situations in which the associations
tion that associative conditioning, far from deserving its low- into which the initially neutral CS enters are predictably incon-
level status as a nondeclarative form of stimulus-response sistent. In Section 13.2 we saw how a specic behavioral task is
habit learning (within declarative memory theory) or of said to be ambiguous if there are two or more ways to solve it.
mere taxon learning (in the cognitive map theory), is more A different sense of ambiguity is when the outcome of a spe-
complex and might offer a conceptually economical account cic CS is sometimes one US but on other occasions either
of a variety of ostensibly cognitive phenomena. Proponents of another US or the absence of the rst US (often abbreviated as
this approach have introduced the idea that conditioning NoUS). For example, a light and a tone may each predict the
 Theories of Hippocampal Function 657

US when presented on their own but not when presented also emphasized the possibility that such stimulus relations
together. How can an organism cope with this and other forms can be exibly retrieved in an appropriate way in novel situ-
of predictable ambiguity and so behave appropriately, as ations (e.g., allowing inferences such as that A is better than
many behavioral studies have shown that they can? This class C). Ideas about the involvement of the hippocampus in con-
of problem is called the XOR problem in the articial intelli- textual retrieval (Hirsh, 1974; Good and Honey, 1991) are
gence and cognitive science communities (Rumelhart and closely related. What we remember about a stimulus and its
McClelland, 1986) and a nonlinear task by students of the ani- associations depends critically on the context in which it
mal learning theory. The nonlinearity arises because it had occurs, including the absence of an expected stimulus as
been thought that CSs acquire associative strength as a func- occurs in the extinction of conditioning when a reinforcing
tion of the reinforcement with which they are linked. However, US no longer follows a CS. Wide-ranging ideas about the con-
if this were all there is to learning, the net associative strength tribution that context cues make to encoding and retrieval,
of the combined light and tone should always be higher than including the notion that contexts contain and predict rather
that acquired by either CS on its own despite the combination than simply compete with explicit CSs (Nadel and Willner,
of CSs always being followed by nonreward. A nonlinear solu- 1980, p. 218), helped spur the development of now widely
tion to the problem seems to be required. used learning paradigms such as context fear conditioning.
Suppose, however, that the brain has a learning device that Although intellectually distinct and occasionally at logger-
can resolve such ambiguity. It would then be possible to learn heads with each other, the congural-association, relational-
the differential outcome of single and combined stimulus pre- processing, and contextual-retrieval theories have two
sentations and so react appropriately on every occasion. The common threads: They all set out to solve problems involving
congural-association theory of Sutherland and Rudy (1989b; predictable ambiguity, and they all argue that the hip-
Rudy and Sutherland, 1995b) was the rst of several hypothe- pocampal formation is central to achieving it.
ses about hippocampal (and cortical) function that speci- One of the early theorists to recognize the problem was
cally addressed this kind of problem. The relational-processing Hirsh (1974). He looked upon conditioning as a process in
theory (Cohen and Eichenbaum, 1993; Eichenbaum and which stimulus input came to trigger motor output along a
Cohen, 2001), a development of Squires original declarative performance line as a function of simple CS-US associations
memory theory, also tackled the issue of ambiguity. It took (Fig. 1333A). He argued that it is independent of the hip-
things further in focusing on relationships being encoded pocampus. On the other hand, he also supposed that a sepa-
between stimuli (e.g., that A is better than B and that B is bet- rate context memory could somehow interact with the
ter than C), rather than stimulus congurations (AB, BC). It performance line, particularly during memory retrieval, and

Figure 1333. Coping with predictable ambiguity. A. Hirshs model different in another. (Source: After Hirsh, 1974.) B. Modern learning
recognized a distinction between a performance line, which stores theory has developed an elaborate framework that is relevant to
simple CSUS relations and expresses them as conditioned thinking about the ways in which the hippocampus, functioning as
responses, and a context memory, which modulates the perform- a context memory, could modulate internal representations of asso-
ance line when a CS predicts a US in one context but something ciations. See text for discussion.

A Associative encoding and retrieval B Contemporary representational framework for

associative memory

Stimulus Motor
CSrep CSrep
Input Output
Performance Line Stimulus Action Motor
Input Representations Output

USrep USrep NoUS

rep

Stimulus
Input Context Memory Context Memory
Context - CS associations
Stimulus
Input Context - US associations
Context modulation

CS - US associations
658 The Hippocampus Book

so ensure appropriate motor output when a specic stimulus

predicted reinforcement in one context but the absence of Box 136
Congural Association Theory
reinforcement in another. Both context memory and the
process of contextual retrieval were identied as mediated by 1. A class of nonlinear associative problems can be solved
the hippocampus. Contemporary research has elaborated this by the process of stimulus conguring.
framework in numerous ways (Fig. 1333B). First, stimulus 2. The hippocampus builds and stores congural associa-
input is said to give rise to internal representations of CSs and tions between elemental stimuli (1989), serves an
USs that can enter into a variety of associative relations rang- enabling role in their encoding and eventual storage in
ing from rst-order associations (CSrepUsrep where rep neocortex (1995), or helps pattern-separate stimuli rap-
refers to an internal representation of these stimulus events), idly to aid slow cortical learning (2001).
to second-order relations (CS1CS2), and even predictions of
the absence of a previously expected stimulus (CSNoUS).
These associative relations constitute knowledge and (some- Ambiguity, Conditioning, and the Hippocampus
how) give rise to goal-oriented action representations that
constitute appropriate behavior (e.g., approach food) and Consider the case of two auditory CSs such as tones (call them
then to motor output (the particular movements produced to A1 and A2) that, through repeated pairings, become selec-
fulll an action). The concept of contextual modulation tively associated with two types of food (F1 and F2). The
remains, but the manner in which contexts are represented experimenter arranges matters such that when a light (L) is
and enter into associations is now recognized as quite compli- illuminated in the test chamber A1 signals F1 and A2 predicts
cated. The representations (lower part of Fig. 1333B) may be F2. However, when the animals are in the dark (D), A1 is
elemental or congural, and these in turn may become arranged to signal F2 (rather than F1) and A2 predicts F1. Rats
associated with specic CSs or USs that occur in their pres- and other animals can resolve this predictive ambiguity and
ence (context-event associations). In addition to direct associ- learn what leads to what in each situation.
ations between contexts and other stimuli, contexts may also Rescorlas proposal was that animals (mentally) create
modulate the expression of other CS-US associations, such as stimuli that are congural associations between L and A1, L
those expressed within the knowledge system. and A2, D and A1, and D and A2. These would be the cong-
Several theories addressing the problems of predictable ural cues (LA1), (LA2), and so on. If these congural cues
ambiguity suppose that the hippocampus is a form of cong- are discriminable, they could enter into unambiguous associ-
ural, relational, or context memory; but they differ with ations with reinforcement without interference. Learning
respect to the precise set of processes by which such stimuli would otherwise obey established principles of associative
enter into or modulate other associations. A consequence of learning, namely the gradual accumulation of associative
this is that the eld has become somewhat baroque and strength to each congural stimulus as a function of how well
impenetrable. However, it is a serious endeavor that attempts the sum total of stimuli available at any one time predicts the
to go beyond descriptive taxonomy to a more detailed and outcome (Rescorla and Wagner, 1972; Mackintosh, 1983). The
predictive understanding of process and mechanism. The sec- animal would have solved the problem and could express its
tion that follows is a selective representation of this eld of knowledge of associative relations in performance.
research. In its original form, Sutherland and Rudy (1989) proposed
that the hippocampus was the physical substrate for encoding
13.5.1 Congural Association Theory and storing congural associations. They also proposed that
the congurations could be quite arbitrary combinations of
Learning is often complicated by ambiguity. A stimulus may individual stimulus events (e.g., of a light and a tone) occur-
signify one event in one situation and another event elsewhere. ring together in time, one after another in sequence, or in spa-
It is then a fact that a tone predicts food in one situation and tial proximity. The latter aspect of this proposal neatly
a fact that it does not in anothera complication that the extended the domain of the theory to tasks such as place
declarative memory theory may have recognized but offered learning where multiple stimuli in a given context might be
no mechanistic way of addressing. Animals can interpret said to dene a place. This represented something of a volte
ambiguous associations such as these provided they are given face for Sutherland, who had earlier argued, in keeping with
sufficient additional information and a learning device sophis- cognitive map theory, that the learning of a mapping strat-
ticated enough to do it. In a series of papers, Rudy and egy in the watermaze was distinct from associative learning
Sutherland (Sutherland and Rudy, 1989; Rudy and Sutherland, (Sutherland and Dyck, 1984). His new idea, with Rudy, was
1995; OReilly and Rudy, 2001) proposed a theory of hip- that place learning might actually be a special case of cong-
pocampal function, building upon earlier work with animal ural learning in which the individual extramaze cues around a
learning theory in which the conguring of stimuli together as watermaze or radial maze enter into stimulus congurations
fused compound CSs is the way that animals solve such prob- and so (somehow) guide the animal to the correct target. Data
lems (Rescorla, 1973). The essence of Sutherland and Rudys that had hitherto offered support for the spatial mapping the-
and later OReillys proposals are as follows (Box 136). ory could then be radically reconsidered as potentially sup-
 Theories of Hippocampal Function 659

porting the new hypothesis instead. Similarly, the supposition lever-pressing tasks and a swimming task with cues hanging
that the hippocampus encodes congurations takes the the- above potential escape locations (Alvarado and Rudy, 1995a).
ory into territory not considered in the original version of the This nding did not, however, prove replicable by others.
declarative memory theory. With that theory, the consolidated Davidson et al. (1993) observed no decit in negative pattern-
trace strengths of the associations between A1 and F1 and ing in animals with two types of hippocampal lesion. In
between A1 and F2 (in the example above) would be equal, Davidson et al.s version of the task, the rats were required to
and the animal would have no way of resolving the predictive press a lever several times during a tone (T) or light (L)
ambiguity of its factual knowledge. for occasional reward (). This requirement provided a more
An important feature of the original 1989 congural asso- effective baseline against which to observe a decrease in the
ciation theory was that it made a number of clear-cut predic- rate of responding during the nonreinforced TL compound.
tions about when deleterious effects of hippocampal lesions They also used a posttraining transfer test with a novel clicker
would and would not be observed in associative learning. This stimulus to check that learning of the original task did depend
is because Sutherland and Rudy recognized that not all kinds on the formation of a true TL congural cue. Although we
of associative learning require congurations to be formed cannot be certain why different outcomes were obtained in
(simple delay conditioning being such a case), whereas others the two studies, Davidson et al. (1993) made the further
do (nonlinear problems). In short, the theory could be observation that animals with the same kainic acid 
testedan attractive, widely respected but always precarious colchicine lesions used by Alvarado and Rudy, but not those
feature of any good theory. with ibotenic acid lesions, displayed disinhibited rates of oper-
ant responding during the intertrial interval between tone and
Tests of the Original Congural light presentations. Both types of lesion are excitotoxic and
Association Theory should therefore spare bers of passage, unlike older lesion
techniques. However, kainic acid also produces lesions distant
The behavioral tasks Sutherland and Rudy considered to be to the hippocampal formation, such as in the piriform and
valid tests are our rst taste of the somewhat baroque nature periamygdaloid cortex. This additional damage could have led
of this eld of research. Arguably they are problems that an to disinhibited responding, and its potential presence raises
animal is unlikely to face in its natural habitat, a worrying fea- the possibility that altered motor performance rather than
ture because it is not clear what might have been the evolu- impaired learning contributed to the positive result in some
tionary selection pressure to develop an ambiguity learning of the earlier Rudy and Sutherland studies. An operant
module, However, the apparent articiality of these tasks lever-pressing task used by McDonald et al. (1997) indicated
reects a creditable desire on the part of the theorys architects that hippocampal lesions slowed down but did not ultimately
for their hypothesis to be tested rigorously. Four types of task prevent the rate of learning of a negative patterning task,
were studied intensively. They are called negative patterning, and fornix lesions had no detectable effect on learning at
transverse patterning, biconditional discrimination, and all. Lesions of the avian hippocampus were also ineffective
feature-neutral discrimination learning (Fig. 1334). An in impairing negative patterning in pigeons (Broadbent et
explanation of each follows, with reference to exemplar exper- al., 1999; Papadimitriou and Wynne, 1999). Bussey et al.
iments. (2000) found no decit in a congural conditioning task
With negative patterning, two stimuli are each paired indi- (conditioned approach to a place where food was available)
vidually with reinforcement, but the combination of the two after radiofrequency fornix lesions or neurotoxic perirhi-
stimuli is nonrewarded. This is annotated in Figure 1334 as nal/postrhinal (PRPH) lesions. All groups learned to
A, B, AB
, where A and B are individual stimulus ele- approach in response to either an auditory tone or a brief
ments,  refers to reward, and
is the absence of reward. As period of darkness and to withhold approach when these two
summarized by Rudy and Sutherland (1995), several studies CSs were presented together. They also used posttraining
from their laboratory from 1989 onward indicated that neu- transfer tests to establish that a congural solution had been
rotoxic hippocampal lesions made in either of two ways used for the task, and behavioral histology that took the
(kainic acid plus colchicine or ibotenic acid alone) cause a form of showing that the fornix-lesioned rats were impaired
decit in negative patterning. These tasks included operant in T maze alternation and the PRPH-lesioned rats in object

Figure 1334. Four tasks used to test critical pre-

dictions of the congural-association theory. The A Negative patterning B Transverse patterning
letters A, B, and so on refer to arbitrary stimuli; and
the symbols  and
refer to the availability of
A+ , B+ , AB- A+B- , B+C- , C+A-
reinforcement.

C Bidirectional discrimination D Feature-neutral discrimination

AB+ , CD+ , AC- , DB- AC+ , C- , AB- , B+

660 The Hippocampus Book

recognition memory. Overall, the available data from these rations and may be a form of learning that depends on other
and other studies on negative patterning represents, at best, brain structures (e.g., the caudate). Conditional rules are not
only mixed support for the theory. easy for rats to learn; but with sufficient training and provided
Transverse patterning fares better. It involves a series of the information with which they are provided is consistent,
simultaneous discrimination tasks of the form A versus B
, they can do it.
B/C
, and C/A
in which the stimuli A, B, and C are Later work on transverse patterning failed to replicate the
equally often rewarded () and nonrewarded (
). This set of deleterious effects of hippocampal dysfunction. For example,
problems, as noted by Moses and Ryan (2006) when compar- using an operant task that made it less likely that the animals
ing predictions of the congural and relational theories, is would process the cues as congural scenes, Bussey et al.
akin to the childrens game of rock, paper, scissors. Alvarado (1998) found that fornix lesions signicantly facilitated trans-
and Rudy 1995b) reported that rats with hippocampal lesions verse patterning. An identied weakness of a rst experiment
could readily learn to swim toward A and avoid B
and to was the relatively poor performance of sham-lesioned rats on
learn B/C
when either task was presented over successive the critical third phase when all three problems were trained
trials on its own (Fig. 1335). However, they failed to learn the together. This was addressed in a second study with all three
full transverse pattern when all three problems were presented phases trained concurrently. The sham-lesioned group was
together. This result is beautifully consistent with the theory now above chance, but the trend toward a fornix lesion-
provided it is the AB congural stimulus that triggers induced facilitation remained. We come back to transverse
approaches to stimulus A, the BC conguration to B, and the patterning in the discussion (below) of relational processing
CA conguration to C. Unfortunately, it is also a task that may theory.
be amenable to solution without recourse to explicit congur- Biconditional discrimination tasks can also be solved using
ing. This would be a conditional solution in which the ani- congural cues, but unfortunately they may also be learned
mals learn that one stimulus (e.g., A) is correct given the through the application of a conditional rule. Whishaw and
presence of B, that B is correct given C, and that C is correct Tomie (1991) trained rats to pull on strings of different widths
given A. The distinction between congural learning and rule (A, B) and different odors (C, D) in combinations that led to
learning may seem subtle, but the acquisition of an ifthen reward (e.g., AB, CD) or no reward (e.g., AC
, BD
).
conditional rule obviates the need to form stimulus congu- These pairings meant that any one elemental cue predicted

Figure 1335. Transverse patterning. A. Rats were trained to swim toward hanging cue cards that
were either placed in front of a possible escape location (e.g., A) or merely hanging over the pool
(e.g., B
). The spatial locations of A and B
would be randomized across trials. B. The stimu-
lus patterns used for the full transverse patterning problem. C. Decit in hippocampus-lesioned
rats (H) compared to sham lesion controls. (Source: After Alvarado and Rudy, 1995b.)

A Apparatus B Stimuli C Performance

one pair at a time all pairs together

100 100
90 90
% correct

% correct

80 80
A B 70 70
60 60
50 50
+ _ 40 40

100 100
90 90
% correct

% correct

80 80
70 70
60 60
50 50
+ _ 40 40
NC H 100
90
% correct

80
70
60
50
_ 40
+
 Theories of Hippocampal Function 661

reward and nonreward equally often, but the creation of four ulus congurations are ever formed but that their construc-
congural cues would enable a solution. Here the congural- tion is more likely to have taken place in the neocortex.
association theory fared less well: Rats with hippocampal
damage and sham-lesioned animals were both able to learn to Revised Theory
pull the correct string to get the reward. Similar negative
results had also been found in a similar earlier study using pri- In response to the mixed picture of initial results, Rudy and
mates (Saunders and Weiskrantz, 1989). Sutherland presented a revised theory that wisely put the con-
The fourth category of ambiguous problem that Suther- guring part of the system outside the hippocampus in neo-
land and Rudy (1989) considered is the feature-neutral dis- cortex. However, they still asserted that the hippocampus
crimination task. An explanation of the name for this protocol plays a role in learning congural associations by selectively
may help clarify what is being examined. The feature of the enhancing the salience of congural representations (Rudy
task is a particular stimulus; let us call it stimulus A (as in Fig. and Sutherland, 1995, p. 382). A yet later change in the theory
1334). In conjunction with stimulus C, the feature A signies emphasizes, in a similar vein, the role that the hippocampus
that a reward is available. Conditioning should lead to stimu- could play in rapidly separating similar patterns for effective
lus A acquiring some associative strength. However, in con- processing in the cortex (OReilly and Rudy, 2001). These
junction with stimulus B, feature A now signies that a reward changes were in response to observations that ambiguous
is unavailable. Stimulus A should therefore become tasks that could eventually be learned by animals with hip-
inhibitory. By scheduling all four types of trial, stimulus A pocampal damage were often learned more slowly than by
should remain neutral with respect to its scheduled associa- controls. It was a set of changes that would rescue aspects of
tion with reward (it is as often paired with reward as not) the theory but at the cost of making it more difficult to test
while still being the critical feature that enables the rat to learn denitively. In some respects, this is unfortunate because,
the discrimination. Hence, it is called a feature-neutral dis- arguably more than other theorists, Rudy and Sutherland have
crimination task. As in negative patterning, no combination been rigorously honorable in demanding explicit predictions
of elementary associations between stimuli and a reward can and denitive tests.
solve the problem, and the congural association theory Going down with the ship is all very well, but a key issue is
uniquely predicts that rats with hippocampal lesions should whether the logical structure of tasks, such as those laid out in
fail. Unfortunately, two studies using this design have both Figure 1334, actually constrains the way in which an animal
found that rats with hippocampal lesions can learn it. In fact, solves any one task. The original version of the congural
Gallagher and Holland (1992) found a facilitation of some association theory implied that this was the case. However,
aspects of feature-neutral discrimination learning by rats with when comparing negative patterning with feature-neutral dis-
neurotoxic hippocampal lesions, a result that echoes Bussey et criminations and biconditional tasks, Rudy and Sutherland
al.s (1998) results on transverse patterning after fornix (1995) noted a parametric gradation across the tasks. This
lesions. gradation was from negative patterning that absolutely
Taken together, this pattern of results from the four sets of required animals to learn to withhold responding to a cong-
tasks provided only qualied support for the congural asso- ural stimulus composed of two elements (i.e., AB
) that are
ciation theory. The results of the negative and transverse pat- also presented alone and reinforced (A, B) to bicondi-
terning tasks sometimes upheld predictions of the theory, tional tasks in which none of the elements are ever presented
whereas the outcome of the biconditional and feature-neutral alone and reinforced (i.e., always X with Y and either  or
).
discrimination tasks did not. It is perhaps worth reecting They argued that this parametric variation is important
further on the curious paradox that whereas a normal rat can because it relates to the extent to which individual stimuli
solve these problems we ostensibly smarter humans often acquire associative strength during conditioning that is likely
struggle to understand them. The struggle is nonetheless to conict or oppose the strength that should be shown by
worthwhile, as understanding ambiguity and its role in learn- congural stimuli. If this gradation is signicant, it might be
ing is important. All these tasks are ones in which an implicit that the task with the greatest conict would be denitively
learning device is assumed to encode a stimulus conguration impaired by hippocampal damage, whereas that with less con-
that acquires associative strength through its consistent asso- ict would be least affected. To some extent, the data available
ciation with reward. Ostensibly, no new principles of associa- in 1995 did t this pattern.
tive learning are involvedonly a conguring device. The However, one aspect not considered was the possibility that
question therefore arises of whether this device is likely to be distinct CSs may not interact only by forming congural cues
in the hippocampal formation. One reason to be skeptical is but in more subtle ways as well. For example, under some cir-
that the anatomical and electrophysiological data discussed in cumstances, one CS may modulate or set the occasion for
Chapters 3, 4, and 8 strongly suggest that the sensory infor- the other CS to enter into a conditioned association. A good
mation to which the hippocampal formation has access is illustration of this is the feature negative discrimination task in
already polymodal and already highly processed. Therefore, a which a stimulus A is reinforced on its own A but not rein-
neurobiological concern about the theory, quite apart from forced when B is presented before A (i.e., B A
). Under
the inconsistencies in the behavioral data, is not whether stim- these circumstances, instead of an AB congural stimulus
662 The Hippocampus Book

being created (which may happen when all the stimuli are pre- Cohen and Eichenbaum (1993) and later revised by them
sented concurrently), B acts in a different manner, modulating (Eichenbaum and Cohen, 2001). It has been developed further
whether the presentation of A retrieves a memory of the US or in relation to episodic memory processing (Eichenbaum,
the absence of the US. Occasion-setting is an associative 2004), which we address later. The theory also places empha-
process in which one CS serves as a signal for an association sis on the encoding of relations between stimuli. This is
between other stimuli in a hierarchical manner. Work by quite different from forming congurations (as just dis-
Holland and his colleagues indicates that hippocampal lesions cussed), and more akin to the idea of associations between
can disrupt one type of negative occasion-setting (Holland et facts, such as that one fact can remind us of another. The
al., 1999), raising the possibility that the memory traces information processing necessary for exible relations is
responsible for conditioned associations between stimuli are held to occur at the time of encoding but only becomes appar-
encoded and stored in the cortex (including congural stim- ent when retrieval is required in circumstances different from
uli) but can be subject to hierarchical control by stimuli those of the original learning. Like the original version of the
processed in the hippocampus. However, an alternative inter- declarative memory theory, this theory holds that relational
pretation remains feasible. This is that although the hip- processing is within the capacity of (at least) higher mam-
pocampus is not involved in forming excitatory CS-US mals, that it depends on activity in the MTL, and that the
connections it is involved in inhibitory learning (i.e., that a engagement of the hippocampal formation is time-limited.
particular CS does not lead to an otherwise expected US). Relational processing is held to be implemented by the hip-
Chan et al. (2001) had somewhat unfashionably defended an pocampal formation, whereas anatomically distinct neocorti-
inhibitory view of hippocampus function. They looked on the cal regions of the MTL mediate the processing and storage of
feature negative occasion-setting task as one in which, in nor- individual stimuli. This theory led to a number of distinctive
mal rats, a CS forms both excitatory and inhibitory associa- lines of experimentation that require discussion separate from
tions, and the occasion-setter helps to choose between them. those already outlined in the discussion of the declarative
The disruption caused by hippocampal lesions may then be a memory theory but to which the congural association theory
decit in inhibitory, rather than hierarchical, occasion-setting. is, as we shall see, also relevant. The distinctive propositions of
The problem with this view is that there are many occasions in the theory are in Box 137.
which hippocampal lesions have no effect on the extinction of Relational representations were dened as memory rep-
CS-US associationsthe classic test of an inhibition theory resentations that are created by and can be used for compar-
(Wilson et al., 1995). ing and contrasting individual items in memory, and weaving
In the further renement of the congural framework, new items into the existing organization of memories. This
OReilly and Rudy (2001) have semantically recast the theory form of representation maintains the compositionality of the
to refer to conjunctive representations rather than congu- items, that is, the encoding of items both as perceptually dis-
rations and suggested that there are two kinds of conjunctive tinct objects and as parts of larger scale scenes and events that
learning. One type is thought to occur incidentally or auto- capture the relevant relations between them (Cohen and
matically, irrespective of task demands; this is the type that is Eichenbaum, 1993).
rapidly learned and involves the hippocampus. The other type The gist of the rst proposition is that the relations that
is slower and more deliberate, and the kind of learning that humans and animals are able to store and recall go beyond the
emerges as a consequence of a deliberate type of problem- mere fact that two stimuli were experienced together in tem-
solving. This new incidental versus deliberate framework is poral contiguitythe CS predicts US type of association so
also explicit about the contribution that local circuits in the extensively studied in conditioning paradigms. Comparing
hippocampal formation make to algorithms such as pattern and contrasting implies that relations between stimuli can be
separation and pattern completion (in the dentate gyrus more sophisticated. They may be causal, but they may also
and area CA3 respectivelysee Chapter 14), and to the simi-
larities between tasks involving congurations of individual
CSs with context-dependent associations. We return to issues Box 137
to which these revisions of the theory are relevant in Section Relational Processing Theory
13.6, together with reference to the important issue of auto-
maticity in hippocampus-dependent learning. 1. Declarative memory generally involves processing the
relations between different items. Relational processing at
13.5.2 Relational Processing Theory: Renement encoding enables exible access to information in situa-
tions quite different from those of the original learning.
of the Declarative Memory Theory
2. Relational processing is carried out by the hippocampal
formation, but storage of individual items in intermediate
A signicant feature of memory is the ability to recall facts
memory takes place in the perirhinal and parahippocam-
and events in circumstances different from those in which the pal cortex.
information was acquired in the rst place. The inherent ex- 3. The role of the hippocampus in memory is temporary, as
ibility of this form of memory was particularly emphasized in the declarative memory theory.
in a revision of the declarative memory theory advanced by
 Theories of Hippocampal Function 663

relate to physical relations (e.g., that A is near to, or bigger lesions spare the acquisition and expression of motor, proce-
than, B; as in Versailles is near Paris) or to abstract relations dural, and cognitive skills. The renement they offer is what
(as in the ugly sisters were unfriendly to Cinderella). amounts to fractionation of declarative memory into two sub-
Moreover, relations can extend beyond one-to-one pairings to components: (1) the processing and storage in intermediate
elaborate networks of interconnections between facts and memory (ITM) of stimulus items in isolation carried out in
eventsthe very networks that constitute our personal the perirhinal and parahippocampal gyrus; and (2) the pro-
understanding of the world. The theorys proponents asserted cessing of stimulus relations in such a way as to enable exible
that their view echoes William James, who described memory access, carried out by the hippocampal formation (Fig.
as an elaborated network of associations that can be applied 1336). As long-term storage is held to be in other areas of
across a broad range of situations (James, 1890). Although the association cortex, the stimulus representations in the MTL
circumstances or conditions of learning are not well specied are asserted to be compressed. This is a metaphor taken from
in the theory, in the manner of contemporary learning theory, modern computer science; but a claim with no direct evidence
the emphasis on relations is clearly a step beyond reference to to support it requires a much deeper understanding of neural
the facts and events in Squires (1992) original version of the representations in the hippocampal formation than we
declarative-memory theory. Cohen and Eichenbaum (1993) presently possess.
have put their nger on an important issue; and, indeed, There are three other features to highlight. First, because
Squire has now incorporated some of their ideas into his spatial relations are only one of the many types of exible rela-
modern writings (e.g., Squire et al., 2004). tions that humans and other mammals can encode, the theory
Whereas some types of factual information tend to be embraces some of the positive evidence hitherto held to sup-
recalled in a rote-like fashion, as in a childs recitation of port the cognitive map theory in an ostensibly more general
arithmetic (two-plus-two is four, four-plus-four is eight . . .), theoretical position. In fact, Eichenbaum (1996) argued that
remembering facts and events is often deductive. They may be spatial learning fractionates into exible and inexible forms,
recalled in different contexts from those in which they were with only the former sensitive to hippocampal disruption
originally experienced and in a manner that permits access to (reminiscent of Oltons working memory theory in which
other perceptual features of a current situation. Indeed, cer- spatial reference memory is insensitive to such lesions). This
tain facts may not have been stored at all (such as knowing may account for successful spatial learning by hippocampus-
the number of windows in your house, which most people
work out inferentially by mentally walking through the build-
ing and counting them). The overall network of relations does Figure 1336. Three functional components of the relational pro-
not need to be present or to be recalled to access a smaller sub- cessing memory system: cortical-hippocampal connections showing
set of facts. Flexibility also allows deductive inference. An the three main components. The cortical areas store short-term
illustrative example is inferring geographical relations. memory (STM) and long-term memory (LTM) traces of specic
items; the parahippocampal region serves as intermediate memory
Suppose an American schoolboy living in South Carolina has
(ITM) for specic items and does the job of cue compression. The
learned in school that Paris is the capital of France, that
hippocampal formation computes relational representations in a
Madrid is the capital of Spain, and that France is farther manner that enables representational exibility.
north than Spain. Based on these facts, the child should be
able to infer the additional fact that Paris is north of Relational processing theory - processes
Madrid. If then told that Madrid is on the same latitude as and anatomical mediation
Washington, DC, he might, assuming he knows his American
geography, also draw the further inference that Paris must be
north of where I live because Madrid and Washington are on Cortical areas
the same latitude. This illustrates how new spatial facts can be
inferred even though never explicitly learned. Of course, ex- STM & LTM storage
ibility is a hallmark of the cognitive map theory as well, but
there it is held to be a specic, unique property of spatial
memory. The supposition now is that exibility includes the
spatial domain but extends well beyond it to size, social rela- Parahippocampal region
tionships, and yet others. The information processing pro-
ITM
vided by the hippocampus is not domain-specic.
cue-compression
The second and third propositions of the theory concern
the relation between the hippocampal formation and neocor-
tex. Like Squires theory, Cohen and Eichenbaum argued that
damage to the hippocampal formation causes loss of the Hippocampus
capacity to store new explicit memories without having an relational representations &
effect on implicit memory (i.e., for stimulus relations of which representational flexibility
we are not consciously aware. They agreed that hippocampal
664 The Hippocampus Book

lesioned rats during overtraining (Morris et al., 1990) because circuitry for social cognition (Adolphs, 2003). In a further set-
after extended training the animals can learn discrete back for the relational processing account of human hip-
approach responses from each of several starting locations pocampal function, a sharp distinction between item and
in the pool. However, this nding is also consistent with the associative memory is also not fully supported by other con-
hippocampus-independent taxon approach strategies of the temporary fMRI studies, as reviewed in Chapter 12.
cognitive map theory. Second, the theory states that the role of Hippocampal activation can sometimes be seen with individ-
the hippocampus in the formation of exible representations ual memory items, particularly when they are novel, though
is time-limited. Such representations are then consolidated in the novelty-driven nature of such activations is often the
the neocortex, where they can be accessed exibly without unusual spatial context in which a stimulus is presented and
further hippocampal input. Thus, the machinery for creating thus cryptically relational.
exibility is in the hippocampus, but (somewhat clumsily) the Animal studies have played a bigger part in driving the the-
machinery for exible recall is in the neocortex. Third, if the ory. Bunsey and Eichenbaum (1996) developed an odor-
hippocampal formation is damaged, stimulus items that are guided paired-associate learning task for rats to examine
presented together may become fused into congurations. whether learned information could be used inferentially in
Associations can still be created, but the theory supposes that novel testing situations. The animals could dig through food
they are not relational but always congural. This idea capi- cups containing a mixture of ground rat chow and sand to
talizes on the same kind of data that led Rudy and Sutherland secure a fruit-loop cereal reward. One of several odors was
(1995a) to propose their revised theory. added to the sand/chow mixture so the rats experience of dig-
ging was associated with the odor. The animals took to the
Evidence for Hippocampus-dependent task quickly, perhaps because digging for food is as much a
Relational Representations and Flexible Access part of their natural foraging strategy as remembering where
they are (Fig. 1337).
The rst study directly investigating relational processing in Paired-associate learning tasks are most extensively used in
humans involved brain imaging using positron emission human cognitive psychology and typically involve presenta-
tomography (PET) (Henke et al., 1997). This study required tion of the rst member of each pair (the so-called cue item),
subjects to look at slides containing a person or a house and followed by either free recall of the second item or a two-alter-
asked them to judge the gender of the person and whether an native forced-choice test between two potential associates of
inside or outside view of the house was being displayed or to which only one is correct. The pairs may be word pairs, words
guess whether the person was likely to live in the house or be and faces, or other combinations of stimuli. Bunsey and
a regular visitor to it. Only the latter instructions encouraged Eichenbaums paired-associate training protocol for rats used
comparing and contrasting of the two pictures and so trig- the two-alternative forced-choice test. The initial training
gered relational processing. Comparison of the relative brain phase consisted of digging through the sample item, in which
activation in the two conditions revealed greater activation in a single digging cup with a single odor was presented (e.g.,
the right MTL during the relational condition. This nding cocoa  cue A), followed immediately by the two-alternative
supports the relational processing account, all the more so choice test in which two cups were presented and the animals
because the associations formed were nonspatial. Other task was to dig in the cup whose odor was to be associated
human studies, using fMRI, have partially supported the idea with the initial cue item (e.g., coffee  cue B). That is, follow-
that relational processing can drive hippocampal activation ing cue A the animal was rewarded during the choice phase if
(Cohen et al., 1999). However, in an explicit comparison of it dug through cue B but was unrewarded if it dug through the
two matched spatial and social relationship tasks, Kumaran other cup (e.g., salt  cue Y). Conversely, if the animal was
and Maguire (2005) saw activation of the human hippocam- presented with turmeric as a rst-item cue (cue X), digging in
pal formation only during the spatial relation task. A network the salt cup (cue Y) in the choice phase secured reward
of real people, some of whom were friends (or friends of whereas digging through cue B (coffee) did not. Having
friends) and all of whom lived at various places in the same learned in this way that A goes with B and that X goes with Y
city, was identied in an initial brieng session. The main (over many training trials), a second set of paired associates
tasks to be conducted in the scanner included getting a crate was then taught in which the rats learned that odor B goes
of wine from one person to another (together with various with C (not Z) and that odor Y goes with Z (not C). Sham and
control tasks). In the social version of the main task, the crate ibotenate hippocampus-lesioned rats learned both sets of
could be passed from friend to friend. In the spatial version, it premise paired associates in this way quite rapidly and at the
was passed to the nearest neighbor. Hippocampus activation same rate (Fig. 1337B). An ingenious feature of the experi-
was observed when subjects were required to focus on and ment was inherent in the design. The two sets of paired asso-
navigate spatially (e.g., their friends houses). A quite different ciates deliberately contained common elements: odors B
network of brain areas was activated when subjects wandered (coffee) and Y (salt). This commonality afforded the opportu-
mentally within their social networkretrieving knowledge nity of exploring whether rats that had learned that A
about their friends and their relationships to each other goes with B (and B goes with C) could access these stimu-
including the medial prefrontal cortex, insula, superior tem- lusstimulus representations exibly and so reveal transitive
poral sulcus, and other areas implicated in a proposed knowledge about the relation between A and C. In a probe test
 Theories of Hippocampal Function 665

A Odor paired associates B Normal learning of paired associates C Hippocampal deficit in transitivity

sample
90 0.45

80 0.35
70 0.25 C
C

errors to criterion
H

preference index
60 H 0.15

50 0.05
choice 40 -0.05

30 -0.15

20 -0.25

10 -0.35

0 -0.45
AB BC
XY YZ training set
training set 1: AB & XY
sample A X
D Primate study of inference in paired-associate learning
100
relational discontinuous
choice B vs Y B vs Y
training set 2: BC & YZ 90 C
Entorhinal
% correct

sample B Y
80

choice C vs Z C vs Z
70
training set 2: AC & XZ
sample A X 60
? ?
choice C vs Z C vs Z 50
premise probe premise probe

Figure 1337. Flexible retrieval of learned paired-associate infor- ciations. The question at issue is the nature of the relations that
mation. A. The rat approaches and digs through the scented were encoded during this training. C. The test for transitivity
sand/chow mixture during a sample trial until it gets its fruit-loop showed that the sham rats were above chance in choosing C given
reward. A short while later, it is presented by two other scented food A, and Z given X, whereas the hippocampus-lesioned animals could
cups in a two-alternative forced choice test. The scent associated not draw these inferences. (Source: After Bunsey and Eichenbaum,
with the previous sample indicates in which of the choice cups the 1996.) D. Findings from an analogous primate study revealed selec-
animal should dig to get another reward. B. Sham-lesioned and tive impairment on the probe but not premise choice trials after
lesioned rats were equally capable of learning the odor-paired asso- entorhinal cortex lesions. (Source: After Buckmaster et al., 2004.)

for this, the rats were rst given either odor A or X as the ini- After they had mastered these four problems, the animals were
tial single-cup cue item and then presented with a choice presented with two other novel problems: choosing between
between cups containing odors C and Z. The striking result odors B and D and between A and E. Both involved a choice
was that the sham-lesioned animals chose appropriately (i.e., between odors that had never previously been presented
C after A and Z after X), but hippocampus-lesioned rats did together, but only the former pair (B  D) required knowl-
not (Fig. 1337C). edge of a serial order. This is because the probe odors B and
In further experiments also using the same sand-digging D had each served equally often as rewarded or as nonre-
method, Dusek and Eichenbaum (1997) considered the issue warded odors during training, whereas the end-anchor
of hippocampal involvement in building ordered representa- odors, A and E, were always positive or negative odors, respec-
tions using a classic Piagetian seriation paradigm. It was tively. Sham-lesioned rats chose both B over D and A over E,
adapted for use in work with primates by McGonigle and whereas rats with fornix or perirhinal/entorhinal lesions
Chalmers (1977). Later work established that pigeons and rats could only do the A versus E comparison (Fig. 1338).
could also behave logically, but Eichenbaums group was the Work in species other than the rat has revealed mixed sup-
rst to explore the anatomical substrate. Sham-, fornix-, and port for the relational processing idea. A study in pigeons
perirhinal/entorhinal-lesioned rats were trained in a transi- revealed that they could learn the paired-associate task and
tive inference task to choose odor A over odor B and then successfully solve the inferential probe tests, but hippocampal
B  C, C  D, and D  E using a training protocol that began lesions were without effect (Strasser et al., 2004). Using objects
with successive trials of each discrete problem and ended with in a WGTA, Saunders and Weiskrantz (1989) trained monkeys
the individual problems in random sequence (see Box 138). on a biconditional task (like those described in the previous
666 The Hippocampus Book

Transitive inference task trols succeeded in doing this. This nding suggests that the
normal controls had acquired an abstract representation of
the object pairings that was more exible than that acquired
by the lesioned monkeys. The successful learning of the bicon-
100 C ditional task with poor inferential probe test performance is
90 FX
arguably more consistent with the relational processing
Percent correct

PRER
80 account than the congural-association theory.
Buckmaster et al. (2004) trained monkeys on a series of
70
procedures intended to emphasize relational representations:
60 a paired associate task analogous to that of the Bunsey and
chance Eichenbaum (1996) study, a transitive inference task based on
50
the original primate work of McGonigle and Chalmers
40
(1977), and a novel spatial delayed recognition span proce-
BD AE dure. They were also received training in object discrimina-
relational end-anchored
probe pair probe pair tion learning and DNMS (simple associations and recognition
memory). Monkeys with large ( 90%) entorhinal lesions
Figure 1338. Hippocampal participation in transitive inference. acquired and performed these latter tasks normally but were
Choice performance by controls and fornix-lesioned (FX) and
impaired on each of the three tasks that required relational
perirhinal/entorhinal-lesioned (PRER) groups during the two
processing. The entorhinal cortex-lesioned monkeys gradually
choice tests after seriation training. Both lesioned groups are at
chance on the B versus D comparison, but both perform as well as
acquired the difficult discriminations required in the paired-
controls on the A versus E end-anchored choice test. (Source: After associate task, but acquisition was slower than normal,
Dusek and Eichenbaum, 1997.) suggesting that a different, less exible congural learning
strategy was being used. Performance by normal monkeys
in both the premise and probe trials was more than 80% cor-
rect (Buckmaster et al., 2004), but there was selective impair-
section) followed by an interrogation of what the monkeys
ment on the probe trials after entorhinal cortex lesions (Fig.
knew about which object was related to which other object.
1338D). The results on the transitive inference tasks were
The animals were trained, through a series of stages, with pair-
consistent with those obtained in rats, but aberrant poor
ings in which each object was equally often rewarded or non-
performance by one normal control monkey in this task
rewarded, but specic pairings were consistently rewarded
make it difficult to draw rm conclusions. The perennial
(e.g., AB, CD) or consistently nonrewarded (AC
, BD
).
problem of the numbers of subjects used in primate studies is
Hippocampal disruption (fornix lesions or combined fornix,
often quite low. The spatial span task could not be learned
hippocampal, and mammillary body damage) did not affect
by the lesioned monkeys, a nding that would be consistent
acquisition of the task but did impair performance in the
with the cognitive map theory were it not for the decits in
posttraining probe test of what goes with what. In this
the other relational tasks that were explicitly nonspatial.
probe, the monkeys were confronted with three objects (A, B,
These are important results, as studies of entorhinal cortex
C) and, having displaced A to receive a reward, got a second
lesions had previously failed to reveal any lasting decit in
reward on the same trial only if they then displaced B (AB
DNMTS (Leonard et al., 1995), but the commonality of
having been a rewarded pair during training). Only the con-
the decit here with hippocampal lesions is consistent with
the proposal made in Chapter 3 that the entorhinal cortex
should be viewed as a major component of the hippocampal
Box 138 formation.
Transitive Inference Premise Training A potential difficulty is that the transitive inference task is
a classic example of a procedure in which the devil is in the
A  B Individual premise problems detail, both conceptually and procedurally (McGonigle and
B  C trained separately Chalmers, 2003). Like other ambiguous tasks, it is a task that
C  D
can be solved in several ways. Van Elzakker et al. (2003) noted
D  E
at least four theoretical accounts of successful task perform-
ance and argued that a simple excitatory stimulus value
Ordered mental representation of relations
account cannot be ruled out. This may seem surprising as, on
the face of it, the novel B  D probe compares two stimuli that
A  B  C  D  E
should be of equal excitatory stimulus value as they are
equally often paired with reward and nonreward during train-
Probe choice tests
ing. However, there are some cryptic asymmetries to the train-
B vs. D Test of transitivity
ing protocol that may render this assumption false. When
A vs. E Nontransitive novel pairing learning D  E in the ve-problem series, the pairing is with
a stimulus E that is never reinforced. D may therefore not have
 Theories of Hippocampal Function 667

to acquire a very high net excitatory stimulus value before the that requires further investigation in analytically exacting
difference between D and E crosses the threshold necessary for tasks. However, a follow-up study (Dudchenko et al., 2000)
criterion performance. Similarly, A is always reinforced, and revealed this continuous recognition task to be insensitive to
thus the A  B difference might also be realized despite B hippocampal lesions, again raising doubts about the extent to
actually having a fairly high associative strength. Given this, which a relational description of the cellular correlates is
the relative values of B and D may be different despite the justied. The parallel to what is seen in NMR conditioning is
ostensibly equal pairing with reward and nonreward. In addi- beguilingas if the hippocampus mischeviously likes to lis-
tion, one might then expect a B versus D probe to be easier in ten in on things it does not have anything to do with. We later
a four-problem series (A  B, B  C, C  D, D  E) than in argue that this is exactly what an incidental learning system
a ve-problem series in which an E  F problem is added. may have to do.
Exactly these results were found by Van Elzakker et al. (2003), These experiments are taken by Eichenbaum and Cohen
who therefore queried whether relational processing and (2001) to indicate that: (1) rats develop a knowledge of stim-
inferential memory retrieval were being used to perform the ulus relations that can be retrieved exibly to guide inferences
task. Computational modeling by Frank et al. (2003) has taken from memory; (2) hippocampal cells are responsive to
this further, with a detailed treatment of the impact of the such task-related attributes; and (3) the integrity of the fornix,
end-anchors on the associative strength of stimuli with hippocampus, and perirhinal/entorhinal cortices are essential
which they are routinely paired, the role of congural repre- for representing these attributes exibly. This interpretation
sentations in the performance of normal (unlesioned) ani- clearly goes beyond the domain of both the original
mals, and the balance between effects of pattern separation declarative memory and cognitive map theories but leaves
and pattern completion in the task. They also queried the open whether the hippocampus is engaged in relational pro-
adequacy of a relational processing account of transitive infer- cessing at encoding, retrieval, or both. As just noted in the
ence and made the interesting prediction that hippocampus contrast with a congural association account of transitive
or dentate gyrus lesions created after training should have no inference, making lesions before training does not distinguish
effect on performance of the BD probe, in contrast to the dele- these two alternatives. Reversible inactivation, using drugs
terious effect on choice when made before training. This is an that inhibit synaptic transmission cell ring, would also pro-
important prediction because, in contrast, the relational pro- vide an opportunity to explore whether the hippocampus
cessing theory would require the hippocampus to be intact at needs be active at encoding and retrieval, as it does for spatial
the time of a probe test. This valuable pretraining versus post- reference memory (Riedel et al., 1999), and whether it plays a
training comparison of the congural and relational process- time-limited role in systems-level memory consolidation as
ing accounts has not yet been reported. the relational processing theory predicts. If time-limited, it
The lesion studies that have at least partially supported the would imply that exibility can be displayed by neocortical
relational processing theory of hippocampus function have circuits also, provided information is subject to both hip-
been complemented by unit-recording data using similar par- pocampus-dependent encoding and a period of hippocam-
adigms (Wood et al., 1999). Multiple single-unit recordings pal-neocortical consolidation.
were taken of CA1 pyramidal cells ring during a 1-second The notion of representational exibility and its depend-
period just prior to a rat deciding whether to dig in a contin- ence on the hippocampus is also illustrated by experiments on
uous recognition variant of the paradigms just described. the social transmission of food preferences using a paradigm
Food cups containing sand were repeatedly presented to ani- rst developed by Galef and Wigmore (1983). Two rats are put
mals in different locations of an open arena. For trials in together for a short period. One of these rats, the demonstra-
which the cups were scented with a novel odor (half the tri- tor, is arranged to have recently eaten a novel foodstuff. The
als), the sand had a fruit-loop reward buried at the bottom; for animals are then separated and, after a memory delay, the
trials in which the odor was repeated (i.e., no longer novel), other animal, the observer, is tested with two novel foods. One
no fruit loop was present. The rats quickly learned the logic of novel food is the same as the one the demonstrator has eaten,
the situation: to dig when it was worth it and refrain from and the other novel food is not. In this two-alternative forced-
doing so when it was not. The unit-recording data taken dur- choice test, the observer rat shows a preference for the novel
ing the Shakespearean to dig or not to dig decision period food eaten by the demonstrator over the other (Fig. 1339A).
indicated that approximately one-third of the hippocampal Some social feature of the situation conquers the animals
cells had task-related nonspatial correlates. Place-specic r- usual neophobia. An early series of analytical studies (Galef,
ing was certainly seen in some cells (about another third). Of 1990) had established that this social transmission of food
the nonspatial cells, some were associated with approach preferences is mediated by carbon disulde (CS2) in the breath
behavior, some with approaching a particular odor irrespec- of the demonstrator rat that acts a carrier of the smell of the
tive of its location in space, and some with whether the trial food that the demonstrator has eaten. Although this social-
was with an odor that was familiar or novel (match/non- transmission task is not obviously relational in quite the same
match)all relations of a nonspatial character. These data way as the paired-associate task, it is exible with respect to
indicate that the functional correlates of CA1 pyramidal cell how the learned information is expressed. Information about
ring may depend in part on the task a rat is being trained to the novel food is acquired by the observer on the breath of the
perform, not just on its location in space. This is a deep idea demonstrator rather than through actual consumption, and
668 The Hippocampus Book

A Odor-guided, paired associate learning B Retrograde amnesia after hippocampal lesions

Phase I 100 C
demonstrator

% correct (diet sample eaten)

H
90

Phase II
80
CS2+food odour
demonstrator observer

70

60
Phase III observer
?
50
0 2 5 10
consolidation period (days from training to test)
Figure 1339. Social transmission of food preferences. A. Odor- form of a preference in a two-alternative choice test over another
guided paired associate learning in which information about a food food (white). (Source: After Galef and Wigmore, 1983.) B.
odor that is safe to eat (gray food) is acquired by an observer rat Retrograde amnesia for social transmission after hippocampal
smelling the breath of a demonstrator rat that has previously lesions. (Source: After Winocur, 1990.)
eaten that food. This acquired knowledge is later expressed in the

this knowledge is later expressed by the observer rat. mation projected to the perirhinal and parahippocampal cor-
Procedural learning systems in which knowledge is embedded tex is recognized (Suzuki and Amaral, 1994a,b). In contrast,
in performance would not support such learning. the relational processing theory requires the hippocampal for-
Winocur (1990) discovered that this task is sensitive to mation to perform a role in declarative memory distinct from
hippocampal lesions created either before or shortly after the that performed by the surrounding neocortical structures.
demonstratorobserver interaction. Faster forgetting, over 6 The latter structures are held to be an intermediate memory
days, was observed in the animals lesioned before training. In store (ITM) for individual items, whereas the hippocampal
animals subjected to lesioning after the social transmission formation is the relational processor. This modication
phase but before the preference test, a consolidation gradient accommodates the nding that DNMTS for trial-unique cues
was observed: Those with hippocampal lesions displayed ret- is impaired by neocortical damage to the perirhinal and
rograde amnesia 1 day after lesion creation, whereas those parahippocampal gyrus but is unaffected (or only very mildly
whose lesions were created 5 to 10 days later performed sig- affected) by hippocampal lesions. The new ndings, discussed
nicantly better (Fig. 1339B). Bunsey and Eichenbaum above, support the relational processing modications of the
(1995) conrmed Winocurs nding of faster forgetting using declarative memory theory.
rats with ibotenate lesions that encompassed the hippocam- One point to note in passing is that, whereas most studies
pus, dentate gyrus, and subiculum. Rats with more selective of DNMTS have been conducted using visual cues and
hippocampus  dentate lesions showed little forgetting over objects, olfactory cues have been used extensively in
24 hours, but the longer 6-day interval used in Winocurs Eichenbaums rodent studies. Faster forgetting of olfactory
study was not examined and these more selective lesions were cues after entorhinal lesions was rst observed by Staubli et al.
incomplete at the temporal pole. In the light of data concern- (1984). Otto and Eichenbaum (1992) conrmed this nding
ing differential functions of the dorsal and ventral hippocam- using a continuous delayed nonmatching task (cDNM) with
pus, it is possible that social transmission of food preference odors. Fornix lesions had no effect, whereas perirhinal and
could be one that is specically sensitive to ventral hippocam- parahippocampal cortex lesions caused faster forgetting.
pal lesions. Thus, for the type of information used most extensively in the
studies on which the theory has been built (odors) an inter-
Evidence for Functional Dissociations Between mediate memory does appear to exist in the neocortical
Anatomical Components of Declarative Memory: regions of the medial temporal area.
Role of the Hippocampus
13.5.3 Contextual Encoding and Retrieval
One point of departure between this and Squires (1992) ver-
sion of declarative memory theory has to do with functional The original version of spatial mapping theory focused on
dissociations between components of the MTL memory sys- how environmental landmarks of a specic context are
tem. Squire argued against functional distinctions, except at explored and integrated into spatial maps for the purposes of
the input stage where some differentiation between the infor- navigation (OKeefe and Nadel, 1978). However, context can
 Theories of Hippocampal Function 669

be important in ways other than as a space to be moved Fear conditioning to a discrete stimulus is mediated by the
around. Memories of discrete events could include associa- amygdala (Davis et al., 1993; LeDoux, 2000; Maren, 2005).
tions between an event and its context, such as the place or Whether shown as an altered startle response or somatic
time at which the event happened (Nadel and Willner, 1980). immobility (freezing), circuits interconnecting the lateral,
Generally, a context has also to be learned about before the basolateral, and central nuclei in the amygdala have been
representation of it can enter into other associations implicated in fear conditioning using lesion, pharmacological,
(Fanselow, 1990), but although this representation is often single-unit, and LTP studies. However, the amygdala interacts
spatial it need not be (Jeffery et al., 2004). It could include with the hippocampus in certain forms of fear conditioning.
temporal cues, particularly during extinction of condi- The most widely used context-freezing task, as developed
tioned associations (Bouton, 2004); it may include feelings by Fanselow, involves three conceptually separate but often
associated with administered drugs (Overton, 1964) or natu- overlapping phases (Fig. 1340A). First, the rat or mouse is
rally occurring motivational states such as hunger or thirst placed into a distinctive box with a grid oor (the context)
(Davidson, 1993). A context may also be cognitive as in the and allowed to explore it with a view to forming an integrated
mental frameworks or schema associated with particular representation of its shape, appearance, odor, and somatosen-
classes of problem (Morris, 2006). The importance of context- sory characteristics. This learning takes a minute or two, a
dependent memory is well demonstrated by occasions on fact that points to a somewhat neglected link between this
which it fails, such as when we embarrassingly fail to recog- form of associative conditioning and exploratory behavior.
nize a person we have recently met in a different context. Second, a weak or intermediate-intensity electric shock is
Contextual encoding is clearly an effective, often automatic delivered to the animal through the grid oor. This may be
associative mechanism for binding events into memory. during the same session as the initial exploration or a later one
As shown in Figure 1333B, the modulation of memory by (protocols differ slightly across laboratories). In either case,
contexts can be mediated by distinct associations and associa- the animals reaction to the shock may be barely detectable, it
tive processes (Balsam and Tomie, 1985; Bouton and Moody, may run around for a bit, or it may immediately remain
2004). As with the formation of stimulus congurations or immobile. Whatever its immediate or unconditioned reac-
relational associations, contexts can inuence the attention tion to the shock, this behavior gradually gives way to sus-
paid to a stimulus, set the occasion for a stimulus to predict tained periods of immobility in which practically the only
one outcome or another, disambiguate stimuli and their pre- observable movement is breathing. This conditioned reaction
dictive signicance in other ways, or provide an incidental or (CR) is a species-specic defense reaction (Bolles, 1970)
deliberate way of organizing attended information. In general, shown by various rodent species and is generally called freez-
the ability to form context-event associations is parasitic upon ing. The probability of freezing varies as a function of the
animals having previously developed a memory representa- number and intensity of the shocks delivered. If the initial
tion of the context if that is necessary (which may take time) exploratory period is not included, contextual fear condition-
and upon the occurrence of rapid one-trial learning (because ing does not occur (Kiernan and Westbrook, 1993), a phe-
unique events do not happen twice). nomenon that Fanselow (1990) referred to as the immediate
It turns out that some of the distinct aspects of contextual shock effect (Fig. 1340B). However, when sufficient
encoding, modulation, and recall depend on the hippocam- exploratory time in a context is allowed, subsequent fear con-
pus, but others do not. Accordingly, a simple formula such as ditioning occurs successfully. This learning phase ends with
context-learning  hippocampus-dependent learning is the animals return to its home cage. The third phase of test-
inadequate. Experiments on the role of the hippocampal for- ing involves returning the animal to the context at a pre-
mation in context-associated information processing have scribed memory interval after training (e.g., the next day).
now developed a number of subtleties, and their discussion The experimenter monitors the level of freezing (Fig.
leads us on toward a more episodic account of hippocampal 1340C). Little freezing is seen during a postoperative mem-
function. ory test in rats subjected to hippocampal lesioning soon after
context fear conditioning. However, the extent of freezing
Context Fear Conditioning increases systematically as a function of the interval between
context fear conditioning and the time when a hippocampal
A learning paradigm called context fear conditioning is lesion is created (Kim and Fanselow, 1992). Animals that have
now widely used to study context-event learning. It is rap- conditioned well typically freeze for as much as 60% of a short
idly learned (Fanselow, 1990), generally though not always testing session. This freezing response is only slowly forgotten
hippocampus-dependent (Winocur et al., 1987; Selden et al., over time (weeks or longer) and only extinguished by repeated
1991; Gisquet-Verrier et al., 1999), and has proved ideal as a context exposures in the absence of shock. Studies of the
rapid assay of lesion, drug, and forebrain-specic genetic extinction of conditioned fear, including contextual fear, are
manipulations that might differentially affect short- and long- currently a focus of considerable interest for their transla-
term memory for fear and the signal-transduction mecha- tional implications for the treatment of posttraumatic stress
nisms engaged in its consolidation (Aiba et al., 1994; disorder (Cahill and McGaugh, 1996; Ressler et al., 2002;
Frankland et al., 2001; Ohno et al., 2001). Nader, 2003).
670 The Hippocampus Book

A Contextual fear conditioning

context post-training
exploration
conditioning memory test

!!!!!! !!??

electric shock

B Immediate shock effect C Context freezing D Preference for a safe context

freezing level place-preference
30 5 100 250
C C
freezing % H

seconds on safe side

4 80 200 H
defecation (#)
defecation (#)

freezing %

20
freezing %

3 60 150

2 40 100
10
1 20 50

0 0 0 0
1 3 9 27 81 1 7 14 28 days
placement-shock interval (s)

Figure 1340. Context fear conditioning. A. The three phases of Fanselow, 2000.) C. The extent of freezing as a function of the inter-
training: context exploration, fear conditioning, and posttraining val after context conditioning before a hippocampal lesion is made.
memory testing (Source: After Fanselow, 1990.) B. The immediate (Source: Kim and Fanselow, 1992.) D. The impact of hippocampal
shock effect that reects the absence of context fear conditioning lesions on place-preference as a measure context fear. (Source:
unless sufficient time is allowed to explore the context (Source: After Selden et al., 1991.)

Although by far the most popular, freezing is not the only freezing on the integrity of the hippocampus and of tone-
measure used to assess fear conditioning. Other measures elicited freezing on the amygdala (Kim and Fanselow, 1992;
used include the relative preference for a safe unshocked Phillips and LeDoux, 1992). The hippocampal dependence of
chamber compared to the conditioning chamber to which it is contextual freezing seems to be unrelated to the level of activ-
connected or the extent to which fear, conditioned to a dis- ity the animal displays in the conditioning chamber prior to
crete CS, inhibits appetitive or consummatory behavior. shock delivery (Gewirtz et al., 2000). This lack of correlation
Selden et al. (1991) examined both the extent to which a is important, as it is known that hippocampal lesions can
clicker stimulus (the discrete CS) would evoke fear after pair- sometimes cause hyperactivity (e.g., during exploration of a
ings with shock and the extent to which an initially nonpre- novel environment), and the decit in context freezing might
ferred and brightly lit white chamber would be favored over a therefore have been no more than a secondary, nonmnemonic
black one in which the clicker and shock had been presented consequence of a lesion-induced alteration in motor behavior.
(the safe and conditioning contexts, respectively). They Two arguments marshaled by Fanselow (2000) against this
observed a double dissociation between the effects of excito- possibility include the selectivity of the hippocampal lesion
toxic amygdala and hippocampal lesions on fear conditioning decit to context but not discrete CS-associated fear and the
to explicit and contextual aversive cues. Amygdala lesions presence of a within-subject temporal gradient of retrograde
selectively impaired conditioning to the clicker measured by amnesia for discriminable contexts (see below). Thus,
its capacity to reduce drinking (i.e., inhibition of consumma- although other measures have been developed that would not
tory behavior). Hippocampal lesions selectively impaired con- be subject to the hyperactivity objection, freezing has retained
ditioning to the context measured by the proportion of time its popularity, and the research eld has stuck to it.
spent in the black and white chambers in a preference test The difference between fear conditioned to a discrete
(Fig. 1340D). Independent work by other groups conrmed CS (amygdala-dependent) and to a context (hippocampus-
the double dissociation between the dependence of contextual dependent) reects the prior learning of the context. Fanselow
 Theories of Hippocampal Function 671

suggested that the rat must first form an integrated son, a one-to-one association between hippocampal process-
mnemonic representation of the many features of the context ing and context freezing is an oversimplication.
and the rat must have this representation in active memory at
the time of the shock (Fanselow, 2000, p. 75). He likened it to Differential Effects of Hippocampal Lesions
a Gestalt memory of the environment, analogous to the spa- Before and After Fear Conditioning
tial maps that OKeefe and Nadel (1978) had suggested are
formed during active exploration. It is therefore noteworthy The importance of exploratory learning about the context is
that this necessary time period of exposure to a novel condi- relevant to a puzzling feature of studies examining the impact
tioning chamber is similar to that which Bostock et al. (1991) of hippocampal dysfunction. Hippocampal lesions made
rst reported to be necessary for rats to form a stable hip- shortly after conditioning are effective in blocking contextual
pocampal ring eld in a novel recording chamber. This uni- freezing (Kim and Fanselow, 1992; Phillips and LeDoux, 1992,
ed representation of the environment then functions as an 1994; Maren et al., 1997; Frankland et al., 1998), whereas those
congural stimulus that, through classical conditioning, is made prior to conditioning have varied effects (Winocur et al.,
associated with shockin a manner essential identical to that 1987; Maren et al., 1997; Gisquet-Verrier et al., 1999; Maren,
implied by the congural association theory. Rudy and 2001). This dissociation with respect to the relative timing of
OReilly (1999) provide direct support for this view in exper- training and the lesion is not because the lesions are incom-
iments with normal rats in which they showed that preexpo- plete (although often they are). Clearly, were they incomplete,
sure to the conditioning context, but not to the individual there may be sufficient spared tissue for hippocampus-
stimulus elements that made up the context presented one dependent learning to occur when the lesion is made prior to
after the other, facilitated later contextual fear conditioning. conditioning. That this is not the explanation for the differen-
They also established that the extent to which fear condition- tial effects of lesions made pretraining versus posttraining is
ing generalizes to other contexts that had not been used for because if an identical incomplete lesion is made after train-
conditioning is inuenced by preexposure in a manner sug- ing a substantial decit in context fear conditioning is still
gestive of pattern completion during conditioning (see observed.
Chapter 14). Specically, if contexts X and Y contain common The likely explanation for the differential effect, as outlined
elements and context Z is quite different, preexposure to con- by Maren et al. (1997), Rudy and OReilly (1999), and
text X but not to context Z facilitates generalization to context Anagnostaras et al. (2001), is that an animal with a hip-
X after fear conditioning in context Y. These observations, pocampal lesion prior to conditioning is predisposed to asso-
reminiscent of Guzowskis data obtained using the immediate ciate the fearful shock with some elemental feature of the
early gene Arc (see Section 13.4) suggest that contextual freez- conditioning chamber (e.g., its smell) rather than the unied
ing reects fear elicited by a stimulus that is a mental entity: representation that because of the lesion it is unable to form.
It is a conjunctive memory representation of the environment As this type of single-stimulus classical conditioning is unaf-
based on polymodal sensory information. This led Rudy to an fected by hippocampal lesions, less disruption of contextual
ingenious test of his context preexposure facilitation (CPF) freezing is to be expected in the anterograde domain. This
concept. Rats were led to expect, over a series of exposure possibility alerts us to the importance of the exact design of
days, that being transported in a distinctive container around the chambers in which these sorts of experiments are con-
the laboratory would always end in them being taken to a par- ducted. Different contexts are sometimes readily distinguish-
ticular test chamber (context A). On the critical conditioning able in terms of an elemental feature; in other cases, they may
day, they were transported in the container but unexpectedly only (or more readily) be distinguished by the set of cues that
put into a distinctively different test chamber (context B) and forms a single unied representation. The devil is again in the
given an immediate shock. Remember that immediate shock detail with respect to understanding the pattern of results,
does not ordinarily result in context fear conditioning; and, particularly in relation to the impact of genetic manipulations
consistent with this, no contextual fear conditioning occurred that, unless inducible, occur long before conditioning. As
to context B. However, the rats displayed freezing when placed inducible, region-specic mutations are rare, we do not
in context A, the context they had expected to arrive at after include a discussion here of this developing and important
transport in the distinctive container but not the one in which eld but do recognize its likely importance in the years ahead.
they actually received the shock. This CPF effect was not Creating lesions after training offers the opportunity to use
observed in rats with hippocampal lesions (Rudy et al., 2002) contextual freezing as a way of investigating the temporal gra-
A complication to the story that the hippocampus is criti- dient of retrograde amnesia. Rapid learning (during a single
cal only for contextual fear conditioning is that several studies session) and the fact that retention is extremely robust over
have shown that disruption of the ventral hippocampus (with time are features that render the task highly suitable. Kim and
lesions or drugs) can disrupt fear conditioning to a discrete CS Fanselow (1992) produced electrolytic hippocampal lesions in
such as a tone (Bast et al., 2001; Zhang et al., 2001). This may, rats 1, 7, 14, or 28 days after cued and context fear condition-
in part, be due to the presence of direct afferents from the ven- ing. As noted above (Fig. 1340C), a graded decit in contex-
tral hippocampus to the amygdala (Pitkanen et al., 2000) that tual freezing is seen as a function of the traininglesion
ordinarily mediates CS fear conditioning. Whatever the rea- interval, with animals given lesions 28 days after training
672 The Hippocampus Book

being indistinguishable from those that underwent sham cues (tones and clickers) and were equally reinforced in rela-
surgery. This nding was interpreted in terms of a time- tion to both sets of cues. Rats were trained in an appetitive
dependent consolidation process for contextual fear, consis- paradigm (collecting food pellets) in which one discrete stim-
tent with declarative memory theory. Maren and Fanselow ulus (A) was reinforced in one context (X), whereas a second
(1997) obtained similar results, but only over much a longer stimulus (B) was reinforced in Y. The critical test of the ability
time interval, in animals with excitoxic hippocampal lesions. of either context to retrieve the appropriate signicance of
An elegant within-subjects study by Anagnostaras et al. (1999) these cues was to present, for the rst time, A in context Y and
examined the effects of electrolytic hippocampal lesions on B in context X. The phenomenon of context specicity was
animals conditioned in two highly distinctive contexts with displayed as a reduction in the appetitive responses elicited by
training scheduled 50 days apart. Conducted at UCLA, the A and B in their now inappropriate contexts. Rats subjected
two contexts were named after two famous streets in to electrolytic (Good and Honey, 1991) or neurotoxic (Honey
Hollywood. The Sunset Room was a darkened laboratory and Good, 1993) hippocampal lesioning failed to show this
containing conditioning chambers that were triangular, had reduction of conditioned responding. This failure was not
white plastic walls, and were scented with acetic acid. The because the subjects with lesions could not discriminate the
Wilshire Room was a distinctive, well lit laboratory with two contexts; a separate experiment established that they had
conditioning chambers that were rectangular, made of alu- no difficulty in doing this when only one of the two contexts
minum and plexiglass, and scented with ammonium hydrox- predicted the availability of reinforcement.
ide. Previous work had indicated that there is little The implication appeared to be that the integrity of the
generalization between these two environments; the animals hippocampus is necessary for a process associated with the
would have readily distinguished them during training and contextual control of associative conditioningassociative
testing. The key nding was that lesions made shortly after retrievalrather than learning a unied representation, as
context fear conditioning in the Sunset Room but 7 weeks described by Fanselow. However, this implication is insecure
after conditioning in the Wilshire Room (or vice versa) for two reasons. First, remember that even the ostensibly
impaired later expression of the recent memory but not robust decit in contextual freezing induced by hippocampal
remote memory. Location, location, location is, not surpris- lesions does not always occur if the lesions are produced
inglyas important in Hollywood as anywhere else. before training as they were in the Good and Honey (1991)
experiment. Second, Good and Honeys decit in contextual
Contextual Control of Stimulus Signicance? retrieval has not proven easy to replicate in other laboratories.
A series of experiments by Hall et al. (1996) failed to do so, and
There are ways in which context can affect behavior other later work showed that hippocampal lesions have no effect on
than by entering into associations with the US as in context the rate of learning a conditional context discrimination test
freezing. One way is to modulate the effectiveness of other over a series of test sessions (McDonald et al., 1997).
associations. As noted long ago by Nadel and Willner (1980), Good et al. (1998) put forward the suggestion that there
contexts can be said to contain events that occur within may be a subtle but important difference between the inci-
them. The relation between cue and context is then hierar- dental processing of contextual cues (when their processing
chical rather than associative in the traditional sense. Work is not formally necessary to learn a task) and the intentional
by Moita et al. (2003) nicely illustrates the impact of contex- or contingent processing of contextual cues (when their pro-
tual cues on CA1 pyramidal cell ring during auditory fear cessing is essential for learning that task). Their results indi-
conditioning (see Chapter 11). Place elds were rst observed cated that when rats with hippocampal lesions were required
in the conditioning chamber prior to fear conditioning. Rapid to use context cues over several sessions to differentiate the
fear conditioning then took place over a few trials in which stimulus signicance of discrete cues, there was no decit.
shock was presented at the end of a white-noise CS, with a However, when the animals were presented with a single
control group receiving temporally random presentations of unexpected context specicity test, with context novelty con-
the white noise and shock. Freezing was observed in response trolled, the hippocampus-lesioned group was impaired (Fig.
to the CS. In parallel, there was an increase in CA1 cell respon- 1341). The possibility that there may be something critical
siveness that was specic to the place elds in the chamber about the hippocampus in relation to the distinction between
where a specic cell normally red. This suggests that the spa- automatic and deliberate processing is directly relevant to the
tial representation encoded and stored during the earlier conjunctive representations theory (OReilly and Rudy, 2001),
exploration can incorporate elements of the task-related fear an issue we revisit in the section of episodic memory, below.
conditioning events into its representation.
Contexts can also set the occasion for a particular cue Hunger and Thirst as Contexts
having a particular signicance or meaning. Good and Honey
(1991) developed the rst satisfactory experimental design to Patient H.M. is unable to report whether he is hungry. He does
investigate the role of the hippocampus in such a process. not ask for meals and, quite soon after eating, attempts to eat
Their protocol ensured that subjects became familiar with two again if a plate of food is placed before him (Hebben et al.,
contexts (operant conditioning chambers) and two distinct 1985). That he may have forgotten that it is a long time since
 Theories of Hippocampal Function 673

A Incidental context specificity B Contingent context-specificity

15 15
Control
H
mean response rates

mean response rates

12 12

9 9 H S+
H S-
6 6 control S+
control S-
3 3

0 0
ffe e

ffe e
nt

nt 1 2 3 4 5 6
m

m
re

re
sa

sa

2 days blocks
di

di

Figure 1341. Incidental versus contingent processing of context presented in both contexts and differentially rewarded in each of
cues. A. When the discrete CSs predicting the availability of food them, both control and hippocampus-lesioned rats readily learned
in the conditioning chamber were unexpectedly presented in the to discriminate their signicance appropriately. (Source: After Good
other context, controls but not hippocampus-lesioned rats showed et al., 1998.)
a decrease in responding. B. When the discrete CSs were repeatedly

he last ate or that he has just had a recent meal is unsurpris- which placement in an operant chamber was associated with
ingyet another indication of his memory decitbut that footshock when the rats were hungry but not when nonde-
he has apparently no awareness of the internal state of hunger prived (or vice versa). An A/(AB) type of discrimination
or its absence is an additional decit. Motivational states ratio was used to measure the relative degree of conditioned
could contribute to the contextual control of behavior. freezing in the two situations, and the data in Figure 1342A
A number of animal experiments have addressed this ques- indicate that normal animals could learn this discrimination,
tion, with early studies using spatial tasks in which a rat was whereas rats with neurotoxic lesions of the hippocampus and
required to perform one spatial response when hungry and a dentate gyrus could not. A later study by Kennedy and Shapiro
different one when thirsty. Rats could learn this interesting (2004) sought to distinguish involvement of the hippocampus
contextual-dependent task, but any lesion-associated decit in learning such a discrimination and a role in contextual
could arise because of a failure to discriminate the motiva- retrieval. Three visual distinct goal-boxes (shown graphically
tional states, failure to use this information to retrieve the as black, gray, and white in Fig. 1342B) were, across succes-
appropriate response, or failure of spatial memory itself. sive trials, randomly placed at the ends of the arms of a three-
Davidson and Jarrard (1993) examined a nonspatial task in choice pitchfork-type maze and the animals deprivation

Figure 1342. Motivational states as hippocampus-dependent con- fornix or hippocampal lesions were no longer able to use their dep-
textual cues. A. Control but not hippocampus-lesioned rats learned rivation state to retrieve a memory of the appropriate response.
to discriminate whether to be afraid in a place as a function of their They did, however, continue to avoid the never-rewarded goal box.
state of hunger. (Source: After Davidson and Jarrard, 1993.) B. After Goal boxes were moved randomly across trials. H, food available
preoperative training to choose a visually distinctive goal box to to a hungry rat; NoR, never-rewarded box; Th, water available to a
secure an appropriate reward when hungry or thirsty, rats with thirsty rat. (Source: After Kennedy and Shapiro, 2004.)

A Hunger as a context for fear B Hunger and thirst as contexts for discrimination

0.7 100 NoR

Th- H+
mean percent correct (1st trial)
discrimination ratio (A/A + B)

80
C
C F
0.6 H Trial N
H 60
H-
NoR Th+
40
chance
0.5
20
Trial N+1

Blocks of hungry and sated states pre-lesion post-lesion

674 The Hippocampus Book

state switched between hunger and thirst on alternate days. conditional exception to the ruleone that depends on the
One goal-box provided food if the animals were hungry (but current context (Bouton and Moody, 2004, p. 665).
not otherwise), another supplied water if they were thirsty, Studies by Bouton have implicated the hippocampus in
and the third never provided reward. Before any lesions were certain extinction phenomena but not others. Wilson et al.
made, training showed that the animals could learn to make (1995) using fornix lesions and Frohardt et al. (2000) using
the appropriate choice as they ran the maze. After fornix neurotoxic hippocampal lesions have provided evidence that
lesioning or neurotoxic hippocampal lesioning, the ability to the integrity of hippocampal function is needed for a context-
choose between the hunger and thirst boxes fell to chance, processing phenomenon called reinstatement but is not
although the animals continued to avoid the goal-box that required for another phenomenon called renewal. The dis-
never provided reward. The sham-lesioned rats continued to tinction between them is important. Both are instances in
perform effectively. Occasional probe trials were scheduled in which an initially conditioned and then extinguished CR
which hunger was repeated across two successive days; choice recovers. Reinstatement is said to occur after independent pre-
behavior continued to be controlled by motivational state sentations of the US in the same context as original training,
rather than any alternation pattern in the sham-lesioned con- with memory test consisting of CS presentations in this same
trols. Different results in a similar task were reported by context (reinstatement of the CR does not occur if the US pre-
Deacon et al. (2001), but all training was postoperative and a sentations occur in a different context). In renewal of the CR,
reward was consistently available independent of motivational the previously trained CS is extinguished in a different context
state. The rats with hippocampal lesions may therefore have and then tested in the original training context (renewal does
solved the task by updating their predicted value of the not occur if CS extinction occurs in the same context
reward (i.e., food when hungry and food when sated) using an throughout). These experiments used the classic paradigm of
amygdala-dependent learning process. Accordingly, it seems conditioned suppression in which rats were rst trained to
secure to conclude that the hippocampus is required for using press a lever for food reward in each of two contexts. They
internal motivational states in a contextual manner for were then presented with a CS (lights turning off) on a num-
memory retrieval. ber of occasions, this stimulus being paired with weak electric
shock (US). Rats reduce their rate of appetitive responding
Contextual Control of Extinction during the CS (conditioned suppression), and this decline
in operant responding served as a quantiable measure of fear
The phenomenon of extinction refers to changes in learned in much the same way as freezing. Once trained in this way,
behavior that occur when a stimulus or a response is no longer the CS was then extinguished by being presented several
followed by the reinforcement with which it was previously times without shock over a series of lever-pressing sessions.
associated. In the extinction of classical fear conditioning, a Rates of operant responding recovered until they were as high
tone (CS) that has previously predicted the imminent arrival during the CS as in its absence (this is expressed as a sup-
of shock (US) is now followed by nothing. What then hap- pression ratio calculated as CSrate/(CSrate  preCSrate), a
pens, in terms of performance, is that the overt expression of measure that tends toward 0.5 when rates of responding have
the prior conditioning becomes weaker over successive extinc- equilibrated). Reinstatement was observed as the selective
tion trials. An old-fashioned view of this change is that extinc- recovery of the fear response to the CS (i.e., somewhat con-
tion is some kind of unlearning process, a breaking of the fusingly, this is a return to a low suppression ratio) after rats
associative bonds between cue and consequence. However, had been exposed several times to the US alone in the training
various phenomena discovered about the determinants of context. Critically, this reinstatement effect occurred in con-
extinction have led many to doubt that unlearning is what trol animals but not in subjected to fornix or neurotoxic hip-
normally occurs. Extinction is more often a learning process pocampal lesioning (Fig. 1343A). Conversely, renewal of the
in which contextual factors play a particularly important CR occurred just as much in control as in lesioned rats using
modulatory rolethe extinction context altering what US an ABA design in which the CS was extinguished in a differ-
memory is retrieved when the previously trained CS is pre- ent context (B) from that in which it was initially conditioned
sented (Bouton and Moody, 2004). During training the CS and later tested (i.e., context A) (Fig. 1343B).
comes to evoke a memory representation of the US; during Both phenomena are context-dependent, but Bouton
extinction it may acquire a memory representation of no- argued that the reason reinstatement is hippocampus-
US (as in Fig. 1333), but importantly there is no overwrit- dependent, but renewal is not, is because the hippocampus is
ing of the previous CSUS association. Which memory is important for contextUS associations but not for all forms of
retrieved in a given memory test depends on the context of contextual modulation of retrieval such as positive occasion-
testing, with the nature of the environment and how it differs setting or conditional rules. A strong contextUS association
from other environments, the passage of time, or other factors rapidly created by the unsignaled US presentations retrieves a
inuencing the operative context as it is perceived by the memory representation of the shock and, consequently,
animal. As Bouton and Moody put it, the fact that extinction reminds the animals of the original CSUS association. That
is more context-dependent than conditioning is consistent is, the re-presentation of the ostensibly extinguished CS then
with the idea that the animal codes extinction as a kind of retrieves a memory of the US (with which it was associated
 Theories of Hippocampal Function 675

A Reinstatement (Hippocampus-dependent) B Renewal (Hippocampus-independent)

CS-US training and then Extinction; CS-US training in first context

US-only in same or different contexts; Extinction in same or different contexts;
Test CS in original (same) context Test CS in original (same) context
neurotoxic
radiofrequency lesions
lesions
0.5 0.5 0.5
suppression ratio

0.4 0.4 0.4

0.3 0.3 0.3

0.2 control different 0.2 0.2

fornix different
0.1 control same 0.1 0.1
fornix same
0.0 0.0 0.0
S D S D
after 1 2 3 4 after 1 2 3 4
extinction test trial Sham HPC extinction test trial

Figure 1343. Reinstatement and renewal after the extinction of those that had fornix (left) or neurotoxic hippocampal (right)
classic fear conditioning. A. Using conditioned suppression of oper- lesions. B. Renewal of fear to a conditioned CS was observed in
ant responding as a measure of conditioning, presentation of the both controls and fornix-lesioned groups after the CS was extin-
US only in the same context as that in which CS-US conditioning guished in a different context. (Source: After Wilson et al., 1995 and
had taken place reinstated conditioned fear in control rats but not Frohardt et al., 2000.)

during training) rather than of the absence of the US (as had some isolated or elemental feature of the testing chamber. One
developed during extinction). This inferential feature of way to reduce the likelihood of this happening in animals that
context-dependent modulation of conditioning critically already have lesions is to ensure that two contexts are used in
involves context-US associations and thus the hippocampus. the behavioral protocol and that the critical manipulations are
In renewal, the subject also learns a CSUS association during scheduled for only a short period (e.g., one session). The
training and a CSNoUS association during extinction. Bouton studies used two contexts that differed in quite subtle
However, in contrast to reinstatement, there is no strong ways (e.g., the orientation of the grid bars), whereas the appet-
hippocampus-dependent contextUS association formed. itive study by Fox and Holland did not (they used the control
Instead, the animals representations of the two contexts procedure of no US reexposure). Thus, the animals in the
(training and extinction) modulate whether the CSUS asso- appetitive study of reinstatement after extinction would have
ciation or the CSnoUS association is evoked, and this modu- been able to form an elemental contextUS association
lation does not require the hippocampus. It is most likely using neocortical circuitry, and this may have been sufficient
mediated by neocortical circuits on their own. Frohardt et al. to remind the animals of which CSUS association to recall as
(2000) offer several reasons from the animal learning litera- effectively in lesioned as in normal rats.
ture for recognizing renewal and reinstatement as separate Another problem is that the renewal phenomenon is sen-
context effects. sitive to muscimol-induced reversible inactivation of the hip-
This tidy picture is complicated in two ways. First, a study pocampus (Corcoran and Maren, 2001, 2004; Corcoran et al.,
by Fox and Holland, (1998) that found no effects of neuro- 2005). In contrast to the argument just given, Maren argued
toxic hippocampus lesions on reinstatement in an appetitively that Boutons use of pretraining lesions would have allowed
motivated task. It is possible that contextUS associations in elemental representation of the two contexts and thus a
appetitive conditioning are weaker than in aversive condition- hippocampus-independent form of contextual-dependent
ing because food also occurs in the home cage, whereas shock renewal of extinction in the lesioned rats. His use of post-
is specic to the testing chamber. However, a clear reinstate- training muscimol infusions would have made it more likely
ment effect was observed in the appetitive task, the difference that a congural representation, acquired normally during
being that it was just as great in the rats with hippocampal fear conditioning and extinction, would have been disrupted.
lesions. An alternative and more parsimonious explanation However, this explanation does not seem to account for the
concerns the complication, noted earlier, of elemental versus dissociation between renewal and reinstatement observed in
congural/conjunctive representations of contexts. In these the permanent lesion studies. Another concern is that the spa-
extinction studies, the hippocampal lesions are made prior to tial extent of the inhibition of hippocampal activity by the
conditioning. We have noted that contextual fear conditioning muscimol infusions was revealed in autoradiographic studies
sometimes occurs normally in such animals, as they have the to be quite small and extended as much into cortex as it did
opportunity of representing the context with reference to along the longitudinal axis of the dorsal hippocampus
676 The Hippocampus Book

(Corcoran et al., 2005). If renewal is mediated by a form of that somehow triggers freezing. The speed of learning a
contextual modulation controlled by extrahippocampal cir- toneshock relationship and its ready expression in novel test-
cuits (Bouton and Moody, 2004), these infusions may have ing situations points to it being, by Cohen and Eichenbaums
disrupted a cortex-mediated component of that process also. denition, a exible representation. Conversely, its slow rate
As noted before, the use of reversible inactivation is analyti- of extinction suggests it is inexible. In this and other exam-
cally more powerful than permanent lesions, but there are ples, independent rules for identifying into which category an
technical problems to its use that should be recognized association should be classied other than observed effects of
(Anagnostaras et al., 2002). hippocampus lesions do not seem to exist. We are back to the
issue that has long haunted declarative memory theories:
13.5.4 Critique irrefutability arising from logical circularity.
Third, both the congural and relational theories also lack
The valuable common feature of the three theories just dis- detail at the level of neurobiological implementation. We are
cussed is their serious attempt to tackle the fundamental still at an early stage of thinking about how the circuitry of the
problem of dealing with ambiguity in memory. However, hippocampal formation is used for forming or storing cong-
stepping back from the details of individual experiments, ural/relational associations, how it could implement the selec-
aspects of each of the ideas are problematic. tive enhancement of the salience of congural units elsewhere
First, although it is easy to understand the selection pres- in the brain, or how it could realize memory representations
sure for memory systems for remembering facts and events or that would somehow afford exibility at the time of retrieval.
nding ones way around, it is not clear why mammals would What are the relative roles of the dentate gyrus, CA3, and CA1
have evolved a specialized neural system for solving nonlinear in achieving any of these tasks? How might the organization
discrimination tasks. Ambiguity of cue signicance may be a of the mossy bers and the longitudinal associational pathway
task a scientist can invent, but is it rife in nature? To be sure, of area CA3 help implement stimulus conguring (if it does)?
there are observations indicating that monkeys execute differ- What role might associative LTP play in relational processing?
ent behavioral actions in response to distinct fear-evoking To be fair, the relational theory has taken an important step in
stimuli (e.g., snakes, eagles), but the stimuli from the animals this direction in asserting an anatomically distinct ITM and
and the calls then given by troop members are in no way LTM for isolated stimuli compared to relations.
ambiguous (Seyfarth et al., 1980). Where ambiguities do Finally, fourth, ideas about contextual processing are not
occur, they generally arise as a function of the spatiotemporal immune from some of the same criticisms. When is a stimu-
context of events and can therefore be solved without recourse lus a context and when a discrete stimulus? In the animal lit-
to conguring. Whether an animal should freeze or ee upon erature, a context tends to be a box. In the human literature, it
hearing the sound or smell of a predator more likely depends is as often the color of the ink in which a word is presented.
on the relative safety of the location it presently occupies. This is deeply confusing. Furthermore, how precisely can a
Conguring stimuli might help solve tasks of a particular log- contextual representation function in the several ways we have
ical structure; however, for example, the occasional failure of discussed? Why is the hippocampal formation involved in
rats with hippocampus lesions to learn transitive discrimina- some of these aspects of contextual processing but not others?
tion tasks could actually be secondary to decits of a different In short, how can a contextual account avoid the same charge
kindthe inability to form relational representations or to of occasional circularity that we have leveled against the
allow a context to modulate retrieval. declarative memory theory? There are two strands to any
Second, the relational processing theory is built around the answer to these questions. One is that there is a formal body
idea that associative relations may be more abstract and more of theoretical work on associative conditioning that has delin-
exible than implied by traditional theories of conditioning, eated distinct ways in which contexts can enter into associa-
with their emphasis on associative strength and unlabeled tions with CSs and USs or modulate other associations, as
associations. There are, however, aspects of the theory that represented in Figure 13.33B. Reinstatement and renewal are
seem vague. Critics such as Mackintosh (2002), prefer the pre- good exemplars, as their associative basis is operationally quite
cision of quantitatively precise theories of associative learning, different, and a specic prediction could then be made about
which assert that a variety of qualitatively different associa- the outcome of a hippocampal lesion. Neuroscientists trying
tions can occur such as excitatory links, inhibitory links, and to understand mechanisms at the neural level can capitalize
stimulus representations that set the occasion for the retrieval on this prior psychological researchas they have done with
of others. This complexity of conditioning aside, consideration such phenomena as the immediate shock effect, the context
of even the simplest of associationsthe pairing of a tone with preexposure facilitation effect, and the distinction between
shockillustrates the problem of identifying the type of asso- renewal and reinstatement.
ciation. Is a toneshock pairing relational or nonrelational? The second strand of defense for the context ideas is the
Can it be either depending on the training situation? In terms sheer plausibility and simplicity of the idea that context plays
of the animals phenomenal experience, the tone may come to a critical role in coping with ambiguity. Events happenat
evoke the memory representation of something painful about specic moments in time and space and in specic social
to happen, or it may simply assume high associative strength contexts.Their meaning and signicance is often context-
 Theories of Hippocampal Function 677

dependent, including semantic knowledge, just as so many

aspects of our behavior. In addition, events often happen Box 139
Hippocampus: Episodic and Episodic-like Memory
unpredictably. This simple fact requires that any system whose
role is to keep track of the important events of our lives must 1. Conceptual issues: Episodic memory is the recall, by
be able to respond instantly, diverting attentional resources as humans, of discrete events that happened at a particular
required and encoding information rapidly as it happens and place and a particular time. Such recall entails mental
with reference to the context in which it occurs. This is why time travel. Episodic-like memory in animals is the mem-
the distinction between incidental and contingent processing ory of what, where, and when with respect to events.
of contextual information may be important. Context pro- 2. Hippocampal involvement: The hippocampus is one of a
cessing, as OKeefe and Nadel (1978) recognized long ago, has network of brain structures that mediate the automatic
the potential to be one building block of a true episodic mem- encoding and retrieval of attended events and the contexts
ory system. It is to that issue which we now turn. in which they occur (episodic-like memory). Distinct
brain structures, including the prefrontal lobe, mediate
different and more volitional aspects of this form of
memory.
3. Differential role of subcomponents: Components of the
13.6 Episodic Memory, Hippocampus, hippocampal formation (e.g., CA1, CA3, dentate gyrus)
and Neurobiology of Rapid Context- are differentially involved in dissociable components of
specic Memory episodic-like memory, such as pattern separation and pat-
tern completion.
The last hippocampal theory to be considered, and one in 4. Awareness: Episodic memory and episodic-like memory
which there is growing interest, is that the hippocampal for- are distinguishable, as only the former requires auto-
mation participates in certain aspects of episodic memory (in noetic consciousness. This cannot, at present, be studied
humans) and episodic-like memory (in animals). This idea has neurobiologically. Awareness may also be an attribute of
episodic-like memory in animals, but this need not be
emerged in an integrative manner from neuropsychological
autonoetic.
studies of developmental amnesia (Vargha-Khadem et al.,
1997) and work concerning scene memory, objectscene asso-
ciations, and top-down control of memory processing in
monkeys (Gaffan, 1994a; Miyashita, 2004). In other animals, tic memoryfactual knowledgewhich is not bound to any
behavioral studies of food caching by avian species (Clayton temporospatial context.
and Dickinson, 1998), studies of sequence learning in rodents Episodic memory has several important properties. First,
(Fortin et al., 2002), single-unit correlates of prospective and the encoding of attended events often occurs automatically;
retrospective memory (Wood et al., 2000; Ferbinteanu and and by virtue of this the temporal-spatial relations of
Shapiro, 2003), pharmacological interventions affecting the Tulvings denition cannot be deliberately selected or ignored.
encoding and retrieval of rapid paired-associate learning Automaticity implies that humans cannot normally turn
(Day et al., 2003), and genetic interventions targeted condi- the system off and voluntarily decide which attended events
tionally at specic components of hippocampal circuitry or what features of the context in which they occur are
(Nakazawa et al., 2003) are all contributing to a developing encoded (Martin et al., 1997; Morris and Frey, 1997). This par-
perspective. That a range of approaches are converging on the adoxically reexive feature can be illustrated with reference
same general idea indicates that the theory may be on the to an example. Imagine that during a routine activity such as a
right lines (Box 139). shopping trip you witness a car crash. You were not expecting
it to happen nor planning ahead of time to remember it. Still,
13.6.1 Concept of Episodic Memory what happens in front of you automatically commands your
attention and it then becomes impossible not to encode and
First introduced by Tulving (1972) and elaborated in a num- later remember that it happened, where it happened, and
ber of ways since (Tulving, 1983, 2004; Schacter and Tulving, when. This inevitability of the associations formed at the time
1994), episodic memory refers to the memory of a unique seems to be a characteristic of the brain system that processes
event and/or a temporal sequence of events that collectively this information. You get home and tell your family that you
comprise an episode. As originally dened, the content of saw a car crash. They ask about what happened. You describe
episodic memory is a mental representation of what hap- the scene of the crash, the crash itself, what happened next,
pened during an event, where it happened, and when. and anything else that was distinctive or unusual. This is
Writing in 1983, Tulving asserted that episodic memory straightforward. But how strange it would be to tell your fam-
receives and stores information about temporally-dated ily of having seen a car crash, but then to say that you cannot
episodes or events, and temporal-spatial relations among remember anything about where it happened, stating only that
these events (Tulving, 1983, p. 385). Thus, remembering what it was half-an-hour ago and involved two cars and a bicycle.
you did on holiday, where you went, and with whom would be During ordinary discourse, the What happened? enquiry
an instance of episodic memory. It is contrasted with seman- carries with it the implicit trilogy of what, where, and when.
678 The Hippocampus Book

This automatic binding of event, place, and time infor- ory by adding that the observer needed to have a particular
mation is a characteristic of episodic memory that has neuro- sense of self-identity. He called it autonoetic consciousness.
biological implications. It points to the need for a memory This form of self-consciousness is necessary for memories of
system that has anatomical access to highly processed infor- the form I remember scoring a goal because such a memory
mation from all sensory modalities, is downstream of the entails a sense of self-identity. Opinion is divided, however, on
attentional lters of the neocortex and the perceptual mecha- the need for so sophisticated a form of awareness for all types
nisms for object-centered representations, and is downstream of event memory. It is far from clear that one needs a sense of
of but reciprocally connected with semantic memory. It must self-identity to recall having seen a dangerous animal, yet one
also be a system that stays online, can encode information rap- can readily imagine evolutionary selection pressures that led
idly, and is generally independent of volitional control. This to a brain system, present in nonhuman animals, for the effec-
last qualication of the automaticity property arises because tive memory of having seen a predator at a particular place
some aspects of episodic memory can be volitional. For exam- and a specic time. Animals that stay that much safer by hav-
ple, subjects in a neuropsychology experiment asked to learn a ing such a memory system may or may not have a sense of
set of novel paired-associates or to remember a passage of self-identity. Tulving nonetheless insisted on his autonoetic
prose are deliberately cooperating in a learning exercise. Their consciousness criterion for true episodic memory, but this is
later memory of what happened is clearly an instance of impossible to apply to animals because we are unlikely ever to
episodic memory, but it occurs in the context of an intentional know whether animals have a sense of self-identity (Clayton et
course of action, unlike the event of witnessing a car crash. al., 2003b). Episodic memory may or may not be uniquely
The what-where-when denition also captures a second human and so may or may not reect distinctive features of
important feature of episodic memory that marks it off from the human brain.
semantic memory, the other major category of declarative Given this somewhat metaphysical state of affairs, Griffiths
memory. This is the need for mental time travel as a part of et al. (1999) introduced the term episodic-like memory. This
our phenomenological experience of remembering. This con- is something that can be dened behaviorally in terms of
cept may feel a bit fuzzy at rst, but it is crucial. A person Tulvings older 1972 denition as memory for what, where,
remembering an event, as distinct from merely declaring they and when with the autonoetic aspect nessed for the time
know something, travels in their mind through time to that being. Meeting even this ostensibly more modest requirement
moment in the past when the event happened. Their experi- for demonstrating episodic memory remains far from easy.
ence is then one of being simultaneously in the present while Considerable ingenuity is required of behavioral scientists to
also reexperiencing the past, a mental coexistence that tran- invent novel memory tasks with the appropriate attributes to
scends time. This is one of those psychological juxtapositions do it successfully (Morris, 2001). Much of what is available so
that may never have seemed paradoxical until it is pointed far, including the many hippocampus-dependent tasks dis-
out: Mental time travel is a capacity of mind that we all take cussed in this chapter are not up to the task (Hampton and
for granted, is apparently effortless, and yet is extraordinarily Schwartz, 2004).
complex. Efforts to get a handle on this experimentally is However, suppose that an animal model of episodic-like
sometimes done by asking subjects to make a subjective memory were possible. Its value would be that we would then
assessment of whether they remember or merely know be able to get a handle on the neurobiology of rapid, context-
something. We do not yet have the ability to pose such a ques- specic memory formation with access to the usual range of
tion to animals in a neurobiological study. anatomical, physiological, and biochemical data necessary for
Tulving (2004) made the strong claim that the capacity for making causal inferences. It could then be classied as the class
mental time travel is unique to humans. Suddendorf and of vertebrate memory that refers to unique events and for
Corballis (1997) shared this view, arguing that animals are which the representational content varies in a species-specic
forever mentally trapped in the present or, as Roberts (2000) manner. There is nothing special about such a comparative
so nicely put it, they are stuck in time. This is a controversial perspective. Procedural learning is just the samethe motor
claim, as biological scientists are rarely at ease with unsup- skills that can be learned by a bird, bee, or human vary con-
ported claims of human uniqueness. Like Pinkers arguments siderablyyet Tulving would not deny that all three species
for the uniqueness of human language (Pinker, 1994), Tulving have a nonpropositional form of learning that can be classied
asserted the human uniqueness of mental time travel in an as sensorimotor procedural skills. By analogy, perhaps
arresting way. There are, he wrote, things that bees, birds and episodic-like memory may vary as a function of phylogeny
humans all do. But there are also things that bees do but birds with respect to both content and degree of awareness. It may
and humans do not. There are things that birds do that bees have evolved in vertebrates as a one-shot memory because it
and humans do not. And there are things that humans do that was evolutionary adaptive to be able to encode and then later
bees and birds do not. recall unique events, such as seeing an unexpected predator.
A third feature of episodic memory is that the event being This eclectic approach opens the prospect of a neurobio-
remembered may be one in which subjects actually partici- logical analysis provided the distinctive features of episodic-
pated themselves (e.g., playing in a game of football) or one like memory can be tied down operationally. Various research
that was merely observed (e.g., seeing a dangerous animal). groups are feeling their way toward a range of tasks that may
Tulving (1983) revised his earlier denition of episodic mem- be suitable assays. An unfortunate problem is that several such
 Theories of Hippocampal Function 679

tasks are asserted to be episodic in character when the clas- the idea that fornix lesions leave associative memory unaf-
sication is barely deserved. In the context of the theories that fected following a series of ingenious experiments reported
have been considered in this chapter, episodic-like memory is in somewhat impenetrable papers published during the early
over and beyond being merely declarative and something 1980s. The insight buried within this work was a new set of
more than a spatial memory because, in both cases, the ideas about what it means to say that an animal has learned
memory must include information about what and when an association. The traditional view was that associations
something happened as well as information about a specic were, to all intents and purposes, conditional reexes or
location. Episodic-like memory is also more than one-trial autonomous habits, but Gaffans studies and modern animal
memory. Certain emotional dispositions, such as learned learning theory have considerably enriched our view of asso-
fear associated with a specic stimulus, can be acquired dur- ciative learning.
ing one trial. There may be congural or relational compo- Inspired in part by the work of the animal learning theorist
nents of episodic memory as well, yet neither of these two (Capaldi, 1971), Gaffan explored tasks in which an object
categories quite captures the distinctive features of the full might lead to food the rst time it was seen but not the next
episodic concept or the possible role of the hippocampus in time it was presented (Gaffan et al., 1984). That is, reward was
mediating it. available when the object was novel but not when it was famil-
iar. Normal monkeys could learn this type of task, correctly
13.6.2 Scene Memory as a Basis for reaching for the other object of a pair in two-alternative
Episodic Memory and Top-down choice tests as quickly as in a simple association task. This
Control by the Prefrontal Cortex novelty association task is sensitive to fornix lesions. The
implication was that certain types of rapidly acquired associ-
One idea about the neurobiological basis of episodic memory ations between an object and reward are not autonomous
was Gaffans proposal that it was a form of scene-specic habits at all; they are bits of stored knowledge. Gaffan went
encoding (Gaffan, 1991, 1994). His hypothesis included state- on to suggest that the type of actions a monkey is required to
ments about the role of hippocampal formation and related perform during an association task may be critical to whether
structures, but the framework is anatomically more wide- a fornix lesion that partially disconnects the hippocampus
ranging (Box 1310). from anterior parts of the brain has a deleterious effect
These and related ideas were developed gradually through (Gaffan, 1983). However, this hypothesis, although supported
an extensive series of experiments on Old World primates, by studies suggesting that spatially directed actions are sensi-
beginning with Gaffans seminal study of recognition memory tive to fornix transection (Rupniak and Gaffan, 1987), failed
published in 1974 (see Section 13.3). to attract general support.
A turning point that led to a key insight about episodic-like
Recognition Memory, Associative Memory, memory was the outcome of a study on conditional learning
and Scene Memory (Gaffan and Harrison, 1989). Monkeys were trained to reach
for one of two objects (A or B), such that A was correct when
The distinction between recognition and association (Gaffan, the animals were facing one way in the laboratory room, but
1974) was, in certain respects, a forerunner of Mishkins dis- B was correct when they were facing in the other direction.
tinction between memory and habit (Mishkin, 1982). This task was relatively easy for control animals to learn, but a
However, later work by Gaffan and his colleagues overturned severe decit occurred after fornix lesioning. Interestingly, the
decit was not a general problem of conditional learning
because there was no decit if the objects were presented on
trays of different colors with object A correct on an orange
Box 1310
tray but B correct on a white tray. From a strictly logical per-
Scene Memory Basis of Episodic Memory
spective, the two tasks are identical, but they were differen-
1. Part of a circuit: The hippocampus is part of a circuit that tially sensitive to fornix lesions. Gaffan suspected that it might
includes the fornix, mammillary bodies, and anteror thal- have something to do with the spatial organization of the
amus (the Delay-Brion circuit), which collectively encodes scene before the animal. When the monkey had to face in
events with respect to the spatially organized scenes in different directions, the scene before him changed even
which they occur. though many elements of it were still there (e.g., the window).
2. Spatial basis: There is a spatial component to episodic When the monkey looked at objects on trays of different col-
memory (scene-encoding) that is neuropsychologically ors, however, the global scene was unchanged despite the
dissociable from place memory per se and from spatial minor change of a specic element in it. A third experiment
navigation. contrasted scenes with places. To compare them, a new set
3. Anatomical dissociation of function: The functions of the
of monkeys was trained to reach for one of ve objects as a
Delay-Brion circuit are also dissociable from the roles in
memory played by the frontal lobe and the medial tempo-
function of their place in the room. The places were chosen to
ral lobe (which it connects), including the amygdala and be as visually different from one another as possible so loca-
perirhinal cortex. tion rather than the spatial arrangement of the scene would
differentiate the places (e.g., in front of bookshelves, by the
680 The Hippocampus Book

door, by the window). Surprisingly, no fornix lesion decit too easily ignore features of the surrounding scene because
was seen. Accordingly, Gaffan argued for a distinction between they do not seem relevant to the solution of the task.
places and scenes, arguing that a fornix lesion disrupts some Recognition memory is, once again, a case in point. Although
aspect of scene encoding or the capacity of scene information the surrounding features of the testing apparatus (the WGTA
to set the occasion for other stimuli to be rewarded or nonre- or computer touchscreen) may not help the monkey remem-
warded. This analysis was a forerunner of Aggleton and ber whether object X or image Y was seen most recently,
Browns later description of hippocampal memory as involv- an incidental and noncontingent memory of the context
ing the spatial organization of familiar objects (Aggleton and may still be important. Normal animals automatically form
Brown, 1999). The elegance of Gaffan and Harrisons (1989) objectcontext associations, and the context may then act as a
classic paper is marred only by a minor confounding between retrieval cue or as an element of the remembered visual scene.
the selectivity of the decit and the change from two scenes to This was, it should be remembered, the basis of Nadels con-
ve places, with a fornix lesion impairment in the two scenes cern about removing monkeys from a WGTA during long
experiment but no such effect in the ve places experiment. memory delays in some of the primate studies of recognition
However, subsequent work went on to establish the sensitivity memory (Nadel, 1995).
of scene memory to fornix transection in monkeys that were The second point to note is methodological. The rst
not moved from place to place. scene-encoding study (Gaffan and Harrison, 1989) and other
The emphasis on scenes rather than places is a subtle but studies conducted around the same time involved moving the
important way in which Gaffans hypothesis differs from the transport cage containing the monkey to change the scene
cognitive map theory. What is important is not where the around the animal. This procedural burden, coupled with the
monkey is itself located but what he is looking at. Later unit- laborious nature of training monkeys by hand in WGTAs,
recording work by Rolls revealing the existence of view cells encouraged the Oxford laboratory to be pioneers in the devel-
in the hippocampus (Robertson et al., 1998) raised the possi- opment of automated stimulus presentation systems.
bility that fornix transection disrupts a view-specic physio- Automated testing was introduced by the National Institutes
logical process that is implemented in or, at least projected to, of Health (NIH) group around the same time (Murray et al.,
the hippocampus. Gaffan, being apparently uninterested in 1993), following innovative developments of T. Aigner, and
localization of function as a scientic goal, was noncommittal later by a San Diego group to test monkeys at very short mem-
about what aspect of cognitive function is mediated within ory delays in DNMTS (Alvarez et al., 1994). Horels group also
the hippocampus itselfperhaps uncertain that it would map developed systems that used slide projectors under computer
onto any psychologically dened category. His interest has control. The development of touchscreen technology was
generally been restricted to whether damage to area X disrupts partly a matter of convenience, but it was also to have an
some psychological process or, in disconnection studies, that impact on Gaffans developing theoretical ideas. He began by
conjoint unilateral damage to area X and contralateral damage using laser disks to show a large number of scenes to monkeys
to area Y does likewise (e.g., Easton et al., 2001). Thus, despite (e.g., scenes from a lm) and later digital graphics software to
the extensive use of lesions as a tool to dissociate putative generate a nearly innite variety of abstract designs as scenes
memory systems and a detailed specication of relevant pri- (typically of colored circles, ellipses, and typographical char-
mate neuroanatomy, the scene-processing theory does not acters of different sizes).
identify the episodic system as located in one or more struc-
tures. The closest Gaffans theory comes to anatomical local-
ization is with reference to the Delay-Brion circuit, a group of Scene Memory, Place Memory, and Object-in-Place
structures that includes the hippocampal formation as well as Memory Are Sensitive to Fornix Lesions
the fornix, mammillary bodies, and anterior thalamus. This
In one study, monkeys were shown scene after scene from
emphasis on the anterior connections of the hippocampal for-
George Lucass lm Raiders of the Lost Ark (Gaffan, 1992).
mation, not least the connectivity with the prefrontal lobe
Although we may wonder what they made of it, their task was
(Browning et al., 2005), is strikingly different from the per-
to remember which scenes were associated with reward and
spective offered in the declarative memory theory, where the
which were not.* A highly signicant decit in the rate of
emphasis is much more on posterior neocortical connections
learning was observed in fornix-lesioned monkeys. This result
(as in Chapter 3). Gaffans ideas can be said to be forward-
is important because it suggests that scene learning and object
looking.
discrimination learning may be quite different. From a strictly
Subsequent experiments have elaborated and developed
the idea of scene-specic memory as the essential basis of
episodic-like memory. Before describing them, two points *An unpublished study by Nicholas Humphrey should be mentioned in
should be noted. First, the hypothesis captures an important passing. He showed cartoon movies to monkeys on a television screen. To
test whether there was any understanding of what was happening, he
phenomenological feature of episodic memoryour ability
allowed the animals to choose between seeing the lm run normally or in
to remember the context in which an event occurred. This is the reverse direction. Although the scenes the animals saw were visually
relevant to many memory tasks given to humans and animals similar, only the normal movie made sense. The monkeys preferred the
where the description provided by the experimenter can all regular lm.
 Theories of Hippocampal Function 681

logical point of view, learning the reward signicance of a set impairment of comparable magnitude; and (2) adding a
of scenes is identical to concurrent object discrimination fornix lesion to a preexisting mammillary body lesion does
learning in a WGTA. We saw earlier that concurrent discrimi- not exacerbate the decit. It therefore seems likely that these
nation tasks are sensitive to MTL lesions when multiple trials structures make their contribution to processing in a serial
of each pair of objects are given each day but not when only manner. Later studies have gone on to implicate the entorhi-
one trial of each object is given per day (Malamut et al., 1984). nal and perirhinal cortex in object-in-place memory as well
However, somewhat paradoxically, when several hundred (Gaffan and Parker, 1996; Murray et al., 1998; Easton and
scenes are presented each day, each for no more than one or a Gaffan, 2000; Charles et al., 2004b).
few trials, a clear decit is observed with a fornix lesion that A possible weakness is that although described as an ani-
typically produces only mild amnesia in humans and little or mal model of episodic memory there is nothing very event-
no decit in many classic primate declarative memory tasks. like in remembering which of two alpha-numeric characters
Gaffans nding of a positive effect of fornix lesions therefore is rewarded with food within particular scenes. The task could
supports his case that there may be something special about be learned semantically and stored as a fact rather than as an
scene processing in the Delay-Brion circuit compared to episode. There is also no recall element to the task. The cues
merely discriminating objects. are always provided on the computer screen, and the animal
Computer-generated abstract scenes provided another way can recognize them rather than bring them to mind. To be
to compare three conditions: place memory, object memory sure, learning is remarkably rapid such that, in experienced
and object-in-place memory (Gaffan, 1994a). In the object- monkeys, it takes only two to ve trials to learn each object-in-
in-place task, for example, the correct response for the mon- place task. At that speed of learningclearly much faster than
key is to reach out and touch the one object of a pair that traditional visual discrimination tasksthe animal is remem-
always occupies a particular position in a unique background bering correctly what to do on the basis of only one or at most
of abstract colors and shapes (Fig. 1344A). The use of the a few trials. However, the episodic character of the task, like
term object is slightly misleading because, in practice, the paired-associate learning in humans, is being inferred more
monkeys are often required to reach toward nothing more from the sheer speed of learning than its formal logic. It does
than alpha-numeric characters on the screen. However, work not absolutely require memory of what, where, and when.
by Buckley and Gaffan (1998b) showed that monkeys are very
good at equating objects seen on a computer screen with real Memory Functions of the Temporal Lobe,
three-dimensional objects they can handle and pick up. In a Delay-Brion Circuit, and Prefrontal Cortex
series of experiments using such abstract scenes, a central
nding has been that hippocampal, fornix, mammillary body, The proposal that the hippocampus is part of the Delay-Brion
or anterior thalamic lesions each cause a decit in recalling circuit and that this circuit has a highly specialized role in
scenes learned preoperatively and in learning new scenes (Fig. memory is only one aspect of Gaffans wider ideas about
1344B). The reasons for arguing that the Delay-Brion struc- multiple types of memory mediated by circuits interconnect-
tures form a circuit responsible for this type of memory are ing the temporal and frontal lobes. His scene-specic theory
(1) lesions to each structure individually cause a learning of episodic memory differs from the declarative memory

Figure 1344. Scene-encoding as a component of episodic mem- (e.g., the letter u in the right-hand side) when the presented against
ory. A. The monkey faces a computer screen that typically consists a different background scene (not shown). B. Monkeys soon learn
of a number of randomly placed shapes and alpha-numeric charac- these types of problem rapidly, whereas fornix-lesioned animals
ters. The task is to reach for one object (e.g., the number 7 in the show a clear decit. (Source: After Gaffan, 1994.)
bottom-left corner) against one background and for another object

A. Greyscaled display of object-in-place task B. Mean performance during

learning
3
Errors per scene (early trials)

2 

7
X u
1

0





NC Fx
682 The Hippocampus Book

theory in two major respects: First, it regards hippocampus- Miyashitas analysis extended beyond the domain of episodic
dependent scene memory as one of several qualitatively dis- memory to paired-associate learning following his identica-
tinct propositional memory systems. Like the position taken tion of neocortical pair-coding neurons that re differen-
by Horel a generation ago (Horel, 1978b, 1994), Gaffan there- tially as a function of the association that binds two stimuli
fore takes exception to the concept of a monolithic MTL together (Sakai and Miyashita, 1991). His suggestion is that
memory system (Gaffan, 2002). Second, unlike both the automatic retrieval of episodic or semantic information
declarative memory and relational processing theories, there requires no more than activation in the MTL or area 36,
is apparently no explicit consolidation mechanism. Gaffan respectively, but effortful retrieval also involves a top-down
believed that the Delay-Brion circuit is important for memory signal from the prefrontal lobe. Two main lines of evidence
encoding and memory retrieval, even if retrieval is not from animal studies support this idea. First, the latency at
required until months after original training, but it is non- which unit-recording signals are observed during automatic
committal on sites of storage. The perspective on the role of retrieval is shortest in the most medial structures and spreads
the MTL in memory is difficult to pin down. This is because backward from the hippocampus (for episodic information)
the dissociations that have emerged from his experiments and from area 36 (for semantic information) to area TE where
(using a variety of tasks and both bilateral and crossed unilat- neurons are gradually recruited that provide a fuller represen-
eral lesion techniques) are generally described only in terms of tation of remembered information, be it an event or a fact.
the necessity of a given area/pathway for a given task. No Second, during active retrieval, neurons in the inferotemporal
statement of what cognitive operations are carried out by each cortex received categorical rather than stimulus-specic
structure in the temporal lobe has been presented, nor state- information from the frontal cortex to help guide the effort-
ments about the nature of different memory systems, their ful process of memory searching (Tomita et al., 1999).
rules of operation, or the type(s) of information on which Miyashita saw the challenge ahead as being to understand
they operate. Even for the best studied part of the system, the the hierarchical interactions between multiple cortical areas
Delay-Brion circuit, there is no statement of the presumably (Miyashita, 2004, p. 435) in the expression of memory.
distinct contributions to scene-specic episodic encoding and Notwithstanding differences of approach and detail, Gaffans
recall made by the four distinct structures of the circuit. This perspective was similar.
is puzzling because the local circuitry, numbers of cells, and
extrinsic connections of the mammillary body differ radically 13.6.3 What, Where, and When: Studies
from those of the hippocampus. Even if Gaffan is correct of Food-caching and Sequence Learning
that they work together serially to enable a common func-
tion, it would be helpful to unpack this statement by outlin- Food Caching by Corvids as a Model
ing their distinct contributions to that common function. of What, Where, and When
Electrophysiological studies may be helpful here.
These points may seem unduly harsh criticisms. However, A different approach to thinking about the recollection of past
the picture that is emerging from Gaffans seminal work is that experience was taken by Clayton and her colleagues, who con-
he envisages a widespread but loose hierarchy of cortical areas ducted a series of studies examining food caching and recov-
involved in memory (Murray et al., 2005). The middle and ery by a corvid species, the Californian scrub-jay. Clayton and
inferior temporal gyri have the job of analyzing visual object Dickinson (1998) allowed adult hand-raised scrub jays to
features, such as shape and color. The most anterior neocorti- cache food items in sand-lled ice-cube trays surrounded by
cal region, the perirhinal cortex, associates these object fea- trial-unique arrays of distinctive visual objects (Lego bricks).
tures with nonvisual features of objects (e.g., reward The two food items were (1) preferred but perishable wax
signicance) to form perceptual representations of unique moth larvae and (2) less-preferred but nonperishable peanuts,
individual objects (Buckley and Gaffan, 1998a). Information each of which were in plentiful supply in bowls at the time of
about unique objects processed in the perirhinal cortex is put caching. During two successive caching periods separated by
together, in circuits involving the hippocampus, with spatial an interval of 120 hours, the birds transported and then
information to form representations of unique complex cached wax moths on one side of the caching tray and peanuts
scenes (Murray et al., 2005). Finally, the entire system has on the other, with each side being covered by a Perspex strip
important functional connections with the prefrontal lobe. during the caching of the other food item (Fig. 1345A,B).
By way of a coda, it should be noted that Gaffans empha- These caching periods constituted the opportunity for the
sis on networks for memory processing has points of similar- animals to encode into memory what food item was being
ity to Miyashitas distinction between automatic and active cached where and when. Recovery was permitted either 4 or
retrieval of information (Miyashita, 2004). Although 124 hours later with both sides of the ice-cube tray uncovered
Miyashita accepted the idea of an MTL memory system, he and available. Over a series of pretraining trials extending over
also identied a semantic associational system in area 36 of several weeks, with the visual arrangement of Lego bricks and
the limbic cortex and the important role of top-down con- the assignment of tray sides for the two foods randomly
nections from the prefrontal cortex in guiding the intentional changed on each trial, birds in the degrade group had the
or active retrieval of information from these systems. opportunity to learn that wax moths recovered after 4 hours
 Theories of Hippocampal Function 683

A Experimental Design for Training Trials

Caching experience 1 Caching experience 2 Recovery

Worms Degraded
Worms
120 hours later 4 hours
Peanuts
Peanuts

or on other trials....

Worms
120 hours later 4 hours Worms

Peanuts Peanuts

B Scrub-jay retrieving food caches C First Search in non-rewarded Probe

100

Percentage of Birds

Peanuts
Worms
50

0
Retrieval after... 4 hr 124 hr

Figure 1345. Memory for what, where, and when in California the degrade group showing proportions of birds making their rst
scrub-jays. A. Experimental design for the training trials in which pecks to the wax-moth or peanut side of the caching tray. (Source:
the birds in the degrade condition (shown) had the opportunity After Clayton and Dickinson, 1998. Photograph courtesy of N.
to learn that wax-moths become unpalatable after a period of 4 Clayton.)
days. B. California scrub-jay performing the task. C. Results from

are tasty and worth eating, whereas those retrieved after 124 free recall at cache recovery obviates the use of a familiarity
hours are not. Birds assigned to a separate replenish control strategy that has been so problematic in studies of recognition
group always found fresh wax worms at recovery whether it memory. Clayton and Dickinsons claim that their initial study
occurred at the short or the long recovery interval. The main provides the rst conclusive behavioral evidence of episodic-
nding of this ingenious study was that in nonrewarded probe like memory in animals other than humans (Clayton and
tests in which neither food was actually available (thereby pre- Dickinson, 1998, p. 274) was vindicated by later studies in
venting any olfactory or visual cues to guide search location) which a number of detailed issues were taken further (Clayton
the scrub-jays in the degrade group searched preferentially for et al., 2001). In one of the studies, scrub-jays were allowed
wax moths when recovery occurred 4 hours after caching but to cache preferred crickets (rather than wax moths) and less-
switched their search to peanuts when recovery did not occur preferred peanuts (Clayton et al., 2003a). After caching but
until 124 hours had elapsed (Fig. 1345C). This switchover before recovery, the birds were taught that the crickets, which
was not shown by the birds in the replenish condition but used to remain edible for at least 3 days, now degraded rapidly
partially shown by birds in a third, pilfer, group in which the and became inedible. During cache recovery, the birds now
preferred wax worms were missing after a 124-hour delay. avoided the crickets and sought out the less preferred peanuts.
This pattern of results implies that, at recovery and using free Thus, the birds seem to be able to apply information retro-
recall, the birds can recollect what food item was stored where spectively at the time of retrieval to guide their choice of food
and, at least in some sense, how long ago. item. Other studies have investigated prospective aspects of
In addition to careful controls, such as nonrewarded probe food caching with the scrub-jays (Emery and Clayton, 2001)
tests, there are several innovations in this experimental and rened the criteria for identifying mental time travel in
approach. The use of food caching is helpful as it ensures the animals (Clayton et al., 2003b).
animal is attentive and engaged in the task at the time of
memory encoding; the act of caching is not, however, a Sequence-dependent and Context-
requirement for the formation of episodic-like memory. The dependent ObjectPlace Memory in Rats
use of very long memory intervals over hours and days
ensures that memory retrieval is outside the range over which The scrub-jays are not suitable subjects for neurobiological
interval timing mechanisms, mediated by short-term mem- studies, particularly because these studies use hand-raised
ory, can operate (Hampton and Schwartz, 2004). The use of birds to ensure control of the caching experience. Unfortu-
684 The Hippocampus Book

nately, efforts to demonstrate what, where, and when in that rats with hippocampal lesions could eventually acquire
other species, particularly laboratory animals such as rats, some sequential information about repeated sequences with
have not been successful so far. This has led some to question extended training but never well enough to disambiguate
whether animals can undertake mental time travel (Roberts, overlapping sequences. That the hippocampus could be
2000). Others have raised the possibility that the when com- involved in rapidly learning sequences has also been the sub-
ponent should be thought about in other ways, such as with ject of modeling studies focusing on the recurrent associative
reference to the context in which they occur or their sequence network organization of area CA3 (Jensen and Lisman, 1996;
of occurrence. Both sequence and context sometimes act as Levy, 1996). Nakazawas observation that mice with a knock-
mental surrogates for the experience of time. out of the NMDA receptor specic to area CA3 are poor at
With respect to context, which we have already discussed one-shot learning but successfully master a spatial reference
extensively (see Section 13.5), Eacott made the interesting memory task (Nakazawa et al., 2003) represents partial sup-
argument that when asked about when an event happened port for this view, but it would be interesting if such mice
people often work out where they were at the time and then could also be shown to be poor at learning explicit sequences
use the retrieved contextual information to remember events. of information. Given these ndings, Eichenbaum (2004)
In this mental exercise, there is not necessarily any mental extended his relational processing theory into the episodic
time travel as such (Eacott et al., 2005). Based on this thought, domain by arguing that the hippocampus is also the neural
she has developed a modication of Ennaceurs novel object substrate for mediating the sequential organization of mem-
recognition (NOR) task for rats in which context, place, and ory. In his view, the mediation of memory for time can often
object can be a triad of associations. That is, the rat has to be achieved by reference to memory of the order in which
remember whether a specic object was in a specic location items occurred rather than some absolute memory of the time
in a particular context. Using this task, she reported that at which they were initially presented.
fornix lesions disrupt recollection of the whatwherewhere In a complementary approach to thinking about sequential
triad while, somewhat surprisingly, appearing not to affect a aspects of events, single-unit recording studies in rats have
simpler whatwhere task (Eacott and Gaffan, 2005). This is a indicated that hippocampal pyramidal cells can sometimes
challenging result being, on the face of it, very different from mediate more than place information. Chapter 11 docu-
the extensive results reported by Gaffans group using object- mented the extensive evidence for the spatial sensitivity of
in-place memory. However, given the importance of testing single-units in rodents, including place cells, theta units, head-
numerous memory delays in NOR tasks before claiming selec- direction cells, and grid cells. However, we have already seen
tivity (Clark and Martin, 2005), it would be valuable for vari- that a growing body of evidence indicates that CA1 and CA3
ous memory delays to be investigated in her task. On a pyramidal cells may have cognitive and behavioral correlates
conceptual level also, that people sometimes cannot remem- that extend beyond the domain of space (Eichenbaum et al.,
ber time and use the surrogate of context does not imply that 1999; Wood et al., 1999; Jeffery, 2003).
they always do so, nor does it undermine the central impor- Wood et al. (2000) recorded hippocampal complex-spike
tance of mental time travel to true episodic memory. neurons in rats running a gure-of-eight maze in which, after
With respect to memory for the sequence of events, Honey moving through the common stem of the maze, the animals
et al. (1998) showed that rats with hippocampal lesions were had to alternate between turning left or right to secure reward
less able than sham-lesioned animals to notice any change in (Fig. 1346). A high proportion of the cells (up to 67%) with
the order of two sequentially presented stimuli during an ori- place elds on the stem red differentially as the rat traversed
enting task. In another study, Fortin et al. (2002) showed that the common stem on left-turn and right-turn trials, even
rats could remember some information about the sequence in when potentially confounding variations in running speed
which a series of distinctive odors was presented. Cups of sand and exact heading direction were taken into account. Other
through which rats could dig to nd food (like the paired- cells red similarly on both trial types (i.e., were place cells),
associate and transitive inference experiments of Section 13.5) and other factors contributed to the ring determinants of the
were presented at 2.5-minute intervals. Common household remaining cells. They argued that hippocampal representa-
odors were mixed into the sand to determine its smell, with tions, as represented by cell ring, encoded information about
different novel sequences of odors given during each series of other nonspatial features of specic memory episodes. An ear-
trials. For a given sequence, call it A-B-C-D-E, the animals lier study had also found evidence of patterns of nonspatial
were later tested for whether they could remember that sand ring correlates in an olfactory nonmatching task (Wood et
odor A came before odor C, B before D, and so on; they were al., 1999). CA1 cell ring was related consistently to percep-
also tested for whether odors AE were familiar compared to tual, behavioral, or cognitive events, irrespective of the loca-
other novel odors. The results showed that whereas the sham- tion where these events occurred.
lesioned animals could perform above chance on both the Ferbinteanu and Shapiro (2003) went on to disentangle
sequence and recognition components of these tests, the rats factors that could be responsible for journey sensitivity, show-
with radiofrequency-induced hippocampal lesions could ing that CA1 complex-spike cells certainly display retrospec-
remember the individual odors but not the sequential infor- tive coding and possibly also prospective coding. They used
mation. A follow-up study by Agster et al. (2002) indicated a hippocampus-dependent plus maze task (Packard and
 Theories of Hippocampal Function 685

A Figure-of-Eight Maze B Sensitivity to Memory Delay

Choice Point
100
R R
NC

Percent correct choices

90 HPC

80

70

60

50 Chance
Base
0 2 10 (sec)

C Differential CA1 unit firing as a function of journey

Figure 1346. Task-dependent modulation of CA1 pyramidal cell as a function of an imposed delay before entering the stem of the
ring. A. Figure-of-eight maze as used by Wood et al. (2000) to maze. C. Differential ring as a function of whether the rat has
establish additional properties of complex-spike cells than just place come from the left side of the maze and is about to turn right (lled
and direction sensitivity. The dotted and continuous lines refer to symbols) or has come from the right side and is about to turn left
alternating traverses through the central stem as the animal runs the (open symbols). (Source: After Wood et al., 2000 and courtesy of
maze, receiving a reward (R) soon after each alternating choice. B. James Ainge.)
Choice performance of normal and hippocampus-lesioned animals

McGaugh, 1996) in which the start arms of the maze were rewards (e.g., different avors of food) at the east and west
switched between north and south within a block of trials, and arms of the plus maze, respectively. Then, through cueing of
the goal arm switched from east to west between blocks of tri- the animals with a pre-taste of one or other food reward in
als. In this way, the animals could make one of four kinds of the start box, it might be possible to secure recordings within
journey: from north to either east or west and from south to a block of trials in which the animals turned toward the east
either east or west. Thus, in traversing the east arm up to the when cued with avor 1 (from either north or south) and
goal, the rats could have come either from north or from turned to the west when cued with avor 2. Evidence for
south. During recording studies, many pure place cells were prospective coding uncontaminated by the use of a reversal
observed, as described by Wood et al. (2000), but differential schedule might then be obtained.
cell ring in the east and west arms was also observed, What do these unit-recording ndings imply for hip-
depending on whether entry was from the north or south pocampal involvement in episodic-like memory? With respect
arms, reflecting journey-sensitive retrospective coding. to neurobiological mechanisms mediating the what-where-
Prospective coding is also claimed in the paper but accept- when trilogy, it is clearly necessary for place representations
ing that it is slightly problematic because the recordings were to be supplemented by temporal or sequence information
then compared across blocks of trials in which the reward (time-tagging or relative order) and by representations of
location had been switched. Although unlikely, this goal rever- objects that are to be found at each place. These new studies of
sal might itself alter the ring properties of the cells, a com- the correlates of hippocampal cells raise the possibility that at
plication that does not arise with retrospective coding in least when and where might both be represented in pop-
which reward location is xed and recordings are taken from ulation vectors of such cell ring. To date, the data are limited
within a single block of trials in which some journeys began in to short time periods (seconds) and to the simplest descrip-
the north and others in the south. One way in which this con- tors of sequence (e.g., south-before-east is encoded by cells
founding might be addressed is through the use of two different from those representing north-before-east).
686 The Hippocampus Book

13.6.4 Problem of Awareness Hits are recorded when a subject correctly identies a previ-
ously presented stimulus (i.e., a signal); false alarms are
Experiments with animals that lay claim to revealing a recol- recorded when the subject incorrectly identies a stimulus as
lective component of memory should somehow discriminate seen before when it had not in fact been presented earlier (i.e.,
explicit and implicit processing. We should try to face up to noise). The range of conditions involves systematic variation
the formidable problem of understanding the neurobiology of of relative value, that is, the costs and benets of correctly
awareness. Following Gaffans work, asking a monkey if it reporting hits and mistakenly making false alarms. If the cost
saw a specic image on a computer screen 40 minutes ago is of missing a correct stimulus is high and that of making a false
now routine. Another example is the seminal work on blind- alarm is low, subjects adjust their reporting criterion in the
sight in monkeys (Cowey and Stoerig, 1995) that was dis- direction of reporting hits (i.e., to the right of the graphs in
cussed earlier (Fig. 1313). The discovery of mirror neurons Figure 1347A.) Conversely, if the risks associated with incor-
in primates (Rizzolatti and Craighero, 2004) further extends rect memory are high (as in eye-witness testimony), then sub-
our appreciation of a possible neurobiology of animal aware- jects shift their response criterion downward (i.e., to the left in
ness, as do Irikis studies of tool use (Maravita and Iriki, 2004). the graphs).
These are telling examples because they are precisely the kinds How can this seemingly overcomplicated way of thinking
of phenomenon that an honorable and scholarly skeptic who about memory performance be helpful? Yonelinass important
asserts the importance of language to awareness, such as contribution was to recognize that the divergence from the
Macphail (1998), might have predicted to be impossible to diagonal can take two forms. It may be symmetrical or asym-
demonstrate in animals. metrical. In the former case, p (hit) initially rises faster than p
To be sure, cogent arguments have been put forward to (false alarm) until both reach 0.5, and then the increase in p
suggest that we may never know whether any animal can ever (false alarm) catches up. This creates a symmetrical ROC
be said to possess autonoetic consciousness or engage in curve and reects the familiarity component of recognition
mental time travel and that seeking true episodic memory in (curved line in Fig. 1347B). In the latter case, p (hit) rises
animals is therefore a forlorn exercise (Suddendorf and Busby, rapidly at low values of p (false alarm), an asymmetry thought
2003). However, there is value in taking a positive view of the to be due to the second component of recognition memory
possibility of identifying aspects of memory awareness in ani- recollection (dotted line in Fig. 1347B). This component
mals using behavioral criteria (Clayton et al., 2003b). operates as a threshold process in which items and their spa-
Language may not be central to its expression, a sense of the tiotemporal context are either recollected (above threshold)
self may not be required, and new behavioral protocols will or not recollected (below threshold). There is no gradation of
hopefully continue to reveal more about animal awareness subjective familiarity as occurs in other components of recog-
and the role, if any, of the hippocampus in it. In this way, the nition memory. Like pregnancy, there are no half-measures.
problem of awareness is being broken down into bite-sized Fortin et al. (2004) have now successfully established that
chunks at which experimentalists can nibble away. We have ROC curves for recognition memory can also be plotted for
already seen (Fig. 1314) that monkeys can reveal an aware- rats. This was done taking advantage of their spectacular
ness of whether they can remember (Hampton, 2001). Can memory for odors, using a continuous recognition task in
animals also display quantitative characteristics of memory which the rats had to dig in sand wells for reward when the
that parallel those shown by humans when they are using rec- odor of the adulterated sand was novel and refrain from doing
ollection or familiarity to solve a recognition task? Can they so when it was familiar (and then search elsewhere for
do one-trial cued recall? For each of these, we can go on to ask reward). Wood et al. (1999) had previously shown that this is
whether performance is affected by hippocampal dysfunction. a task that displays a number of hippocampal unit correlates.
Costs and benets were manipulated by altering the height of
Recollection, Familiarity, the sand wells (making digging easier or more difficult) and
and Region of Interest Curves the size of the reward the animals got when they dug success-
fully (from one-fourth of a pellet to three pellets of food
Studies examining the hippocampal contribution to the phe- reward). Their ndings indicate that when normal rats were
nomenological experience of memory have also been inspired tested 30 minutes after exposure to a list of odors (analogous
by developments in cognitive psychology. Fortin et al. (2004) to a list of words in a human recognition experiment), the
exploited a quantitative approach for distinguishing the two ROC curve was asymmetrical (Fig. 1347C). When half the
components of dual-process theories of recognition mem- rats were then subjected to hippocampal lesioning, the ROC
oryrecollection and familiaritythat was developed during curve became symmetrical (Figure 13.47D). Together with
the 1990s by Yonelinas (2001). His approach, discussed in other data in this study, it seems that recognition memory in
Chapter 12, involved plotting graphs called receiver operating rats complies with dual-process models of recognition mem-
characteristic (ROC) curves, which are an important part of ory in humans and that hippocampal lesions may remove
the signal-detection theory (Tanner and Swets, 1954). Two the recollection component of this memory. The task of
sets of probability functions, usually Gaussian, are plotted that realizing a neurobiological understanding of episodic and
vary as a function of cost and benet: hits and false alarms. episodic-like memory and the nature of hippocampal involve-
 Theories of Hippocampal Function 687

A Gaussian noise and signal B Two components of ROC curves

1.0
ty
iari
mil

Probability of hits
0.8 Fa
noise signal
0.6
tion
collec
0.4 Re

0.2

0.0

C Preoperative odor recognition D Postoperative odor recognition
1.0 1.0 Controls
Hippocampus
Probability of hits

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability of false alarms Probability of false alarms

Figure 1347. Receiver operating characteristic (ROC) analysis of the right) or more risky (to the left) detection policy, as described
recognition memory in the rat. A. Hypothetical Gaussian functions in the text. B. Quantitative analysis of ROC functions reveals a lin-
plotting the probability of signal and noise during memory of ear recollection component and a symmetrical familiarity compo-
a prior event. The receiver operates at various points along the x- nent of memory. The sum of the two components is a curved but
axis and can so make estimates of what constitutes a hit (i.e., a asymmetrical function. C. Normal rats show asymmetrical function
true signal) and what constitutes a false-alarm (i.e., from the noise in odor recognition memory. D. After hippocampal lesions, rats
distribution but claimed to be a memory). The threshold  can show a symmetrical ROC function. (Source: After Fortin et al.,
be adjusted up or down the x-axis to enable a more conservative (to 2004.)

ment in it seems no longer quite as forlorn an enterprise as Automatic Recording and Retrieval
once it did. It is to one developing theory of this type that we of Attended Experience
now turn.
One neurobiological theory of hippocampal function does
13.6.5 Elements of a Neurobiological attempt to specify a role for synaptic plasticity in memory for-
Theory of the Role of the Hippocampus mation (Morris et al., 2003). This theory supposes that hip-
in Episodic-like Memory pocampal memory can, like other forms of memory, be
divided into four processes: encoding, storage, consolidation,
It is premature but not inappropriate to end this chapter with and retrieval. A key aspect of the neurobiological approach is
an attempt to tie several strands together. The focus in this the assertion that activity-dependent synaptic plasticity (e.g.,
chapter has been on the major historical contribution that LTP) is critical for encoding and the intermediate storage of
animal lesioning studies over the past 30 years have made to memory traces in the hippocampus that correspond to the
our understanding of hippocampal function. Modern work indices or pointers to cortical regions where detailed per-
has greatly expanded the levels of analysis at which neurosci- ceptual information is temporarily stored (as discussed in
entists think about function and has been converging on the Section 13.3). These hippocampal indices are cartoons that
idea of a role in the automatic aspects of episodic-like mem- bind event information to the context in which they occur. A
ory. There has been little discussion so far of the specic role novel feature of the theory is that these index memory traces
hippocampal synaptic plasticity might play in this process or decay rapidly but are sometimes rendered more persistent by
how the cell biological determinants of that plasticity t with the process of cellular consolidation. That is, LTP is actually
neuropsychological ideas about memory consolidation. Given something of a misnomer as many, perhaps most, memories
the shift toward more interdisciplinary approaches in neuro- of events captured online are soon lost. Protein synthesis-
science, it is tting to end on this note. independent E-LTP is more of a system for creating the
688 The Hippocampus Book

potential for memory with synaptic tagging at the time of This framework builds unashamedly on the functional/
E-LTP induction (see Chapter 10) playing a part in determin- neuropsychological theories discussed earlier in the chapter,
ing what traces are rendered more persistent. The rapid snap- but it also makes a number of novel predictions, including
shot creation of a temporary event-context index involves some counterintuitive predictions, at other levels of analysis.
synaptic potentiation at a large number of synapses across Some examples of the shoulders on which it stands and these
parts of the hippocampal network (probably in CA1 and predictions are as follows. From a neuropsychological per-
CA3). Tags set automatically at the time of an event, irrespec- spective, it recognizes that events happen in particular places
tive of how trivial that event may be, can capture the products at particular times, and their later recall generally includes the
of the somatic or dendritic synthesis of plasticity proteins set memory of where and when an event occurred (Gaffan,
in train by events before, during, or even after an event to be 1994b; Clayton et al., 2001). Thus, event encoding is necessar-
remembered. This capturing process renders the temporary ily associative and contextual in character, a point empha-
indices in the hippocampus more permanent and thus allows sized in several functional theories (e.g., cognitive mapping,
sufficient time for systems-level consolidation to occur. This congural associative, relational). Many events cannot be
two-stage process of consolidation serves as a protective l- anticipated, occur only once, and may contain distinct fea-
ter in the sense that rapidly decaying index traces do not last tures that, in sequence, form short episodesa point about
long enough to guide neocortical consolidation; and thus episodic memory emphasized by Eichenbaum (2004). It is
events that are not subject to cellular consolidation in the hip- vital that traces representing information about such episodes
pocampus do not become represented in a persistent manner are encoded and stored in real time (as they happen), a
in the cortex. Although hippocampal synaptic plasticity is process that Frey and Morris characterized as the automatic
critical for encoding and its expression subject to intrahip- recording of attended experience (Morris and Frey, 1997).
pocampal cellular consolidation, the theory also asserts that it Not all events are remembered for any length of time, and to
is not involved in memory retrieval. It is also unlikely to be do so is not only unnecessary but might also saturate the stor-
involved in the widely discussed systems-level consolidation age capacity of relevant areas of the brain. Moser et al. (1998)
process that depends on hippocampusneocortex interac- exploited this feature of hippocampal memory in their contri-
tions, although it does not preclude the possibility that neo- bution to studies examining the functional signicance of
cortical synaptic plasticity is critical. More formally, the hippocampal LTP (see Chapter 10). With respect to the level
propositions of this developing theory are presented in Box of representational detail, it is also unclear whether the hip-
1311. pocampus need receive, via its extrinsic afferents, detailed sen-
sory/perceptual information pertaining to individual objects
or events. The networks and local circuits for detailed percep-
Box 1311 tual processing are in the cortex. Rather, all it needs to remem-
Neurobiological Theory of Hippocampal Function ber events and the sequence in which they occur is binding
1. Some animals have episodic-like memory, and the hip- information about the locations in the neocortex where this
pocampal formation is one group of brain structures that detail is processed and temporarily encodeda point about
can encode and retrieve such memories. hippocampusneocortex interactions discussed in Chapter 3.
2. Activity-dependent, associative, NMDA receptor-depend- Later processing, guided by these indices, determines whether
ent synaptic plasticity in the hippocampus is the primary the temporary neocortical encoding becomes permanent.
neural mechanism responsible for inducing temporary This view of hippocampal memory processing capitalizes on
encoding and storage of hippocampal indices of rapidly the advances made in other theories; and, like them, it implies
acquired event-context memories. The detailed perceptual that the hippocampus is far from being an island. Realizing its
information associated with events is represented in the functions absolutely requires the synergistic activity of other
cortex.
brain areas both upstream and downstream. This is therefore
3. An essential feature of cellular consolidation processes to
which these index traces are subject is the interaction
a framework that, unlike some others (McClelland et al.,
between local synaptic tags(set by glutamatergic activa- 1995), implies that the cortex must be able to encode infor-
tion), and diffusely targeted plasticity proteins (which mation rapidly online. The view here is not only that the cor-
can be triggered by heterosynaptic activation of neuro- tex can do this but that it loses information rapidly unless new
modulatory and glutamatergic inputs). connections are stabilized by systems-level consolidation. In
4. Systems-level consolidation requires both hippocampus this sense, consolidation is permissive rather than instructive.
and neocortical neural activity and is therefore not solely The next step of this framework takes us farther into the
time-dependent although certainly a process that takes neurobiological domain. If events are to be encoded, there
time. It does not require hippocampal plasticity but may must exist physiological mechanisms for capturing events as
engage mechanisms of neocortical plasticity. they happen and encoding traces that could later enable
5. The hippocampal formation is one of a number of brain
retrieval of event or event-associated information. A promi-
regions that detects novelty, and its neural activity alters
in ways that enable a representation of novel events to be
nent candidate for the neural substrate of event memory is
encoded rapidly online. hippocampal NMDA receptor-dependent synaptic plasticity,
so-called Hebbian plasticity. Such plasticity, as revealed by the
 Theories of Hippocampal Function 689

physiological phenomenon of LTP, exhibits many properties and storage capacity of such networks. One of them, connec-
that are suitable for a role in memory formation (Bliss and tivity density (i.e., the number of connections per cell), pro-
Collingridge, 1993; Martin et al., 2000), and a growing body of vides an anatomical basis for understanding an important
evidence offers support for this view (see Chapter 10). Index feature of the relation between hippocampus and neocortex
memory traces are achieved by an alteration in the connectiv- (McNaughton et al., 2003). Specically, the average connectiv-
ity between neurons such that retrieval cues, expressed as pat- ity in the cortex is, in general, too low to support the encoding
terns of neural activity, are able to induce new patterns of of arbitrary associations (Rolls and Treves, 1998). The cortical
neural activity that are one part of the substrate of memo- mantle contains on the order of 1010 neurons, but each corti-
ries of prior events. cal principal neuron receives only about 104 connections.
Another strand relates to the persistence of such traces. A Thus, the average connection probability in cortex is only
key supposition is that, notwithstanding the fantastic storage 1:106. To overcome this apparent biological limitation, mam-
capacity of the brain, there is no need to store everything per- mals seem to have evolved an arrangement whereby distrib-
manently, and the system is absolutely not evolved to do so. uted associative memory between items represented in
Indeed, most memories fade and are lost. Of those that persist, different sensory modalities can be accomplished through
retrieval processes contribute to their accessibility such that indirect associations mediated by a hierarchical organization
forgetting is a dual function of trace decay and retrieval fail- (McNaughton et al., 2003). In such a scheme, neocortical
ure. If most automatically encoded event memories are tem- modules at the base of the hierarchy are reciprocally con-
porary, there must be psychological processes and neural nected via modiable synapses with one or more hippocam-
mechanisms for selecting the subset of traces that are to be pal modules at the apex. The hippocampal modules include
rendered longer lasting or even permanent, such as the emo- CA3 as well as the dentate hilus, both characterized by high
tional signicance of the event to be remembered or, more internal connectivity as well as modiable synapses. In CA3,
interstingly, the signicance of other events happening close each pyramidal cell is contacted by ~ 4% of the pyramidal
together in time or space. Experiments on behavioral tag- cells of the same subeld (Amaral, 1990), implying that most
ging are needed to explore the predictions of the synaptic CA3 pyramidal cells are connected via two or three synaptic
tagging framework in the context of real memory (see steps (Rolls and Treves, 1998). This high degree of internal
Chapter 10, Section 10.10). There is also evidence that the per- recurrent connectivity is probably sufficient to allow autoas-
sistence of memory can also be determined by the relevance of sociation, or association among individual elements of a pat-
ongoing events to the existing knowledge structures and terned input (Marr, 1971). Activity patterns reecting sensory
interests of the person witnessing them (Bartlett, 1932; detail in neocortical modules may generate a unique identify-
Bransford and Johnson, 1972; Bransford, 1979), an idea long ing index in such a network (Teyler and DiScenna, 1987). This
known in educational circles but that has had curiously little high-level hippocampal trace index is no longer sensory in
impact on neuroscience until recently (Maguire et al., 1999). any strict sense but is stored associatively with other indices
Mediating these psychological processes of persistence are two and the output fed back to neocortical modules via modiable
neuronal mechanisms of memory consolidation: (1) cellular synapses. Activation of a cortical pattern (e.g., a specic avor
consolidation mechanisms, which include the synthesis and of food) could then result in activation of its index in the hip-
synaptic capture of plasticity proteins that stabilize memory pocampus. In turn, this enables retrieval of associated indices
traces at the level of the individual synapse in neurons (Goelet and thence the complementary pattern in the other cortical
et al., 1986; Frey and Morris, 1998; Sweatt, 2003); and (2) sys- modules (e.g., where the food is found). Indirect associations
tems-level consolidation mechanisms, as discussed in previ- enable memory retrieval between cortical modules that are
ous sections of this chapter, which reflect a dynamic too sparsely connected to do this directly.
interaction between populations of neurons in the hippocam- This principle of indirect association in memory places
pus and neocortex (Dudai and Morris, 2001). These forms of high demands on the synaptic storage capacity of the hip-
consolidation are distinct but also interdependent. This inter- pocampus that, as noted briey above, is otherwise in danger
dependence arises because cellular consolidation enables of becoming saturated. Once saturated, learning can no longer
memory indices in hippocampus to last long enough for the proceed effectively (McNaughton and Barnes, 1986; Moser et
slower systems-level consolidation process to work and thus al., 1998). There are several ways in which this burden may be
for their time courses to dovetail. limited. One, as already noted, is via rapid decay of a high pro-
The dening functional characteristics of associative net- portion of the traces that are automatically encoded online (a
works such as the hippocampus are believed by several theo- property of E-LTP). Heterosynaptic depression may also serve
rists to include distributed representations, interleaved storage a normalizing function and so increase effective storage
across multiple synapses, and associative retrieval. These capacity (Willshaw and Dayan, 1990). A third way to limit the
processes enable patterns of activity to be stored as a matrix of rate of use of storage space, also an element of the present the-
synaptic changes and retrieval to be completed from partial oretical ideas, would be to ensure that what is stored in the
fragments of the original input (Marr, 1971; McNaughton and hippocampus is only a cartoon, with the full sensory/percep-
Morris, 1987; Paulsen and Moser, 1998; Nakazawa et al., tual details encoded in the cortex. The fourth waythe
2002). Several factors determine the operating characteristics process of systems-level memory consolidation itselfwould
690 The Hippocampus Book

be to enable hippocampal associations that are strongly arena (Fig. 1348). In each of two sample trials on each day a
and/or repeatedly recalled, such as those representing envi- few minutes apart, rats left the start box to nd a single open
ronmental regularities, to trigger the development and stabi- sand well where they could dig for a avored food. No choices
lization of low-level intermodular connections in the were required on this sample trial, and there was nothing in
neocortex. It is these connections that could later enable cor- the procedure that would have required the animals to engage
tical retrieval in the absence of neural activity in the hip- a contingent learning mechanism. The animals merely ran out
pocampus. To work, it is vital that the intermodular into the arena, found a sand well, dug in it, and found food.
connections that develop are appropriate to the associations This constituted one event. The two sample trials did, how-
represented. This requires the gradual interleaving of appro- ever, use two locations and two foods. This gave the animals
priate connections (McClelland et al., 1995), perhaps during the opportunity of automatically encodingin one trial for
sleep (McNaughton et al., 2003). Interestingly, the rate of con- each sandellthat one avor of chow was at one location and
solidation may not be strictly time-dependent. Rather, the another avor at another (Fig. 1348A). They therefore had
process may be cladistic, with the rate of consolidation two events prior to the daily choice trial, which was started
affected by the frequency with which hippocampal indices are 5 to 20 minutes later. It began by giving the rats one of the two
reactivated. It may also depend on an activated neocortical avors to eat in the start box (their recall cue). A short while
schema to receive the new information. later, they were allowed to enter the arena where they now
The last step to be considered relates to the detection of found two sand wells (Fig. 1348B). On choice trials during
novelty and acting on it. If the hippocampus is critical for training, going to the location associated with the avor cue
building context representations and encoding events in rela- was rewarded with more of that same avor. Nonrewarded
tion to them, it is likely that its neural activity reects changes probe trials were also run. Up to 30 locations and avors were
to the environment. This is not a new idea, as Gray (1982), paired in novel combinations at the rate of two pairs per day.
Vinogradova (1995), and more recently Gray (2000) have The results obtained with this whatwhere paradigm pro-
extensively discussed possible comparator functions of the vided evidence that rats are capable of a limited form of
hippocampus. Novelty detection is by no means unique to the episodic-like memoryand in a species more amenable to
hippocampus, it being a process required in several cognitive neurobiological study than the scrub-jays used by Clayton.
systems; but evidence in favor of a hippocampal contribution The protocol lacks the dimension of time (the when ele-
has come from many experiments on different forms of nov- ment), which is a concern; but it shares with the avian food-
elty detection, including the studies of exploratory behavior in caching paradigms that recall is required rather than
which rats examine objects and landmarks in simple arenas recognition, and this represents a step forward from the
(see Section 13.4). The misplace cells of OKeefe (1976) and object-in-place tasks used by Gaffan.
the more recent observations of neural ring in response to Day et al. (2003) also established that the one-trial encod-
the movement of a hidden escape platform in a watermaze ing of such paired associates lasted no longer than 60 minutes
(Fyhn et al., 2002) also suggest that the hippocampus, perhaps before recall performance dropped to chance (Fig. 1348C).
acting in concert with neuromodulatory systems, detects and Nonrewarded probe trials established that there was a prefer-
represent changes in the conjunction of objects and their loca- ence for digging at the cued location despite the limited nature
tions. Several cells red vigorously at the new platform loca- of the event that had uniquely taken place there that day.
tion, despite previously having been silent. Others that red in Encoding was sensitive to the acute intrahippocampal infu-
different locations around the maze continued to do so after sion of an NMDA receptor antagonist (D-AP5) without any
platform relocation, arguing against spatial remapping. The effect on retrieval (Fig. 1348D,E), whereas retrieval was
new activity was paralleled by reduced discharge in a subset of insensitive to the AMPA receptor antagonist CNQX (Fig.
simultaneously recorded interneurons. The pattern of activity 1348E). These pharmacological observations were made by
largely returned toward its original conguration as the rat timing the infusions to be either just before or just after the
learned the new location. However, a few of the newly encoding trial. That a difference between encoding and
recruited neurons remained active. This persistent ring may retrieval was obtained when the dorsal hippocampus was
reect facilitated synaptic plasticity during the temporary infused with D-AP5 indicates that the mere presence of the
reduction in inhibition (Wigstrom and Gustafsson, 1983; drug has no effect on memory retrieval, but it denitively
Paulsen and Moser, 1998). NMDA receptor-dependent LTP blocks encoding. It seems more likely, and is in keeping with
may be necessary for these permanent modications in ring physiological studies revealing its importance for associative
patterns when novel events occur in a familiar environment. LTP, that an NMDA receptor-dependent mechanism in the
hippocampus is involved in encoding the paired associate of
Hippocampal Synaptic Plasticity, Neural an event and its location into memory in such a manner that
Activity, and Rapid Paired-associate Learning the reactivation of one member of the pair can retrieve the
memory of the other.
Day et al. (2003) developed a one-trial paired-associate task in The event arena has the potential to serve as a exible test
which rats were trained to recall in which of two locations a bed for examining the neurobiology of rapid associative
particular avor of rat chow was to be found in a large event memory of the kind that may underlie human episodic mem-
 Theories of Hippocampal Function 691

A Sample trial 1 (encoding) Sample trial 2 (encoding) B Choice trial (retrieval)

Start-boxes

F1 F1+

F1
F2 No F2
sand wells

door

C Within-day forgetting D Drug infusion prior to encoding E Drug infusion prior to retrieval
70 p < 0.025 NS
100
Dig times at correct
Sand-well digging (%)

p < 0.01
80
sand-well (%)

NS
60
60
Chance
40
50
Chance 20

40 0
5 min 90 min 540 min aCSF AP5 CNQX aCSF AP5 CNQX

Figure 1348. Cued recall of the location of a single event. A. The selectively for digging at the sand well containing this same food.
1.6 m2 event arena made of plexiglass consisted of 49 sand wells, 2 C. Rapid decay of event-place memory over the course of the day.
intramaze landmarks, and 4 start boxes. On sample trial 1, the door D, E. Nonrewarded probe trials show that the animals preferentially
to a start box is drawn back, and the rat runs out into the arena digs in the sand well appropriate to the cue avor given in the start
(dotted line) where it displays occasional lateral head movements box. Memory encoding in this task is sensitive to blockade of
to nd food 1 at the single open well. Sample trial 2 is to a different NMDA receptors (D), whereas memory expression requires only
food (F2) at a different location. B. The cued-recall choice trial fast synaptic transmission via AMPA receptors (E). (Source: After
begins with presentation of either of the two sample trial foods in Day et al., 2003.)
the start box (food F1 is shown) followed by the rat being rewarded

ory (what, where, and when). However, there are several fea- several weeks; and once fully encoded and stored, new infor-
tures of behavioral performance in this task that need to be mation can be added to the schema rapidly. If the schema can
claried and improved. First, the temporal component has yet be shown to be neocortical, which preliminary evidence indi-
been addressed, though it may be possible to use the surrogate cates they are, such a study may offer some handle on whether
of contextual specicity. Second, no hippocampal lesion study neocortical storage of information in a hippocampus-
has yet been reported, and there is the difficulty that the drugs dependent learning task can be achieved during a single trial.
used may be directly affecting only one component of the This would provide vindication for the unusual prediction of
what-where-when triad, such as the spatial component. That this neurobiological theory that cortical learning is sometimes
is, the drugs apparent impact on associative encoding or just as fast as hippocampal learning. Fourth, it would also be
retrieval could be misleading. Studies of one-trial spatial valuable to complement lesion and inactivation studies with
memory in the event arena have helped clarify this point (Bast physiological studies investigating whether there is differential
et al., 2005). Third, alterations in the protocol to look at the c-fos or Arc activation in the hippocampus during this task,
concurrent training of multiple places and avors, analogous and if it is possible to identify neural representations of cued
to learning a list of objectplace pairs, are also underway. recall using ensemble single-unit recording. The latter would
Conducted in the spatial domain, such a study would cast be a formidable undertaking but a valuable one, as it might
light on how multiple spatial associates could become inte- reveal anticipatory components of place-specic unit ring
grated into spatial schema that might be constructed and driven by the task demands in which a avor cue evokes a
stored in the neocortex. Preliminary data from such a study by memory of a place. This would be a logical extension of the
Tse, Kakeyama and Langston (personal communication) indi- concepts of prospective and retrospective coding introduced
cates that such spatial schema can be gradually acquired over by Ferbinteanu and Shapiro (2003).
692 The Hippocampus Book

Storage of Spatial/Contextual Information that most principal cells in entorhinal cortex exhibit view-
Outside the Hippocampus? independent spatial modulation suggests that, by this stage, a
fundamental spatial computation may already have been
An important milestone in the understanding of spatial cog- made. It is also possible, however, that spatial ring in super-
nition was the discovery that most pyramidal cells in the hip- cial entorhinal neurons depends on spatial input from cells
pocampus exhibit location-specic activity and that the in deep layers, which in turn may rely on associative compu-
activity of such place cells is inuenced by the training history tations in afferent hippocampal structures. The manner in
of the animal (see Chapter 11). The predominantly spatial which entorhinal cells code space remains unclear, but that
nature of hippocampal neuronal activity led to the proposal they do helps to understand why pyramidal cells in CA1
that place cells form a distributed map-like representation of exhibit spatial ring both after selective lesions of the dentate
the spatial environment that an animal uses for efficient nav- gyrus (McNaughton et al., 1989) and after disconnection of
igation (OKeefe and Nadel, 1978). When a rat is exposed to a CA1 from CA3 (Brun et al., 2002). In CA3-lesioned animals,
new environment, pyramidal cells develop distinct ring elds CA1 pyramidal neurons receive cortical input only via direct
within a few minutes (Hill, 1978; Wilson and McNaughton, connections from the entorhinal cortex. The presence of place
1993b). The place elds then remain stable for weeks or more elds in these preparations suggests that direct entorhinal
if the environment is constant (Thompson and Best, 1990; hippocampal circuitry has signicant capacity for transform-
Lever et al., 2002), as predicted if these cells contribute to a ing weak location-modulated signals from supercial layers of
particular spatial memory. the entorhinal cortex into accurate spatial ring in CA1.
It is commonly believed that the development of hip- Several simple lter mechanisms could accomplish such a
pocampal ring patterns, like hippocampal memory, might transformation. For example, ring rates of perforant path
depend on LTP at hippocampal excitatory synapses. Several bers to CA1 could be thresholded by feedforward inhibition,
studies have investigated the contribution of LTP to spatial r- such that only the highest afferent ring rates (i.e., those in the
ing in hippocampal pyramidal cells. Surprisingly, interven- center of the entorhinal place eld, are able to drive postsy-
tions that abolish hippocampal LTP have only subtle effects on naptic neurons in the hippocampus. Alternatively, single exci-
place elds. For example, place-specic ring elds develop tatory postsynaptic potentials (EPSPs) in distal pyramidal cell
normally in rats treated systemically with an NMDA receptor dendrites of CA1 may often not be sufficient to trigger
antagonist at a dose that prevents new LTP in the hippocam- somatic action potentials in these cells; reliable discharge may
pus (Kentros et al., 1998). Despite blockade of the NMDA require summation of EPSPs (i.e., high afferent ring rates
receptor, place elds can be maintained between consecutive (Golding and Spruston, 1998; Golding et al., 2002).
test sessions in the same environment for at least 1.5 hours. It is important to note that the computation and storage of
They do, however, show instability across days. Place elds positional information outside the hippocampus does not
recorded in CA1 in mice with mutations of NMDA receptors preclude long-term storage of some spatial information (or
in either CA3 or CA1 are somewhat less distinct than in the indices of such information) in the hippocampus. Indeed,
control mice under certain conditions; but in contrast to the internal recurrent connectivity of hippocampal area CA3
the pharmacological ndings, ring elds remain stable across makes the region highly suitable for storage of just these types
repeated tests on the same day once they have been established of patterned cartoons, at least for an intermediate period of
(McHugh et al., 1996; Nakazawa et al., 2002). Similar results time and perhaps longer in some cases. Other results suggest
were obtained when LTP was blocked by interference with that plasticity in associational synapses of CA3 is necessary for
Ca2/calmodulin-dependent protein kinase II (Rotenberg et successful snapshot encoding of spatial information in a
al., 1996) or protein kinase A (Rotenberg et al., 1996). manner that later allows recall with partial cues (Nakazawa et
However, these interventions did decrease the long-term al., 2002). Longitudinal axon collaterals in CA3 may be
stability of the place elds, as measured 24 hours after the ini- important for successful retrieval of such information, as
tial exposure to the environment. Together, these studies sug- memory retention may be impaired by a single transversely
gest that NMDA receptor activation may not be necessary for oriented cut through the dorsal CA3 region of each hip-
the development or initial maintenance of place-related activ- pocampus (Steffenach et al., 2002).
ity in hippocampal neurons, although it may contribute to
the ne-tuning of place elds and aspects of the long-term sta- A Last Word
bilization of spatial representations (see Chapter 11). The lat-
ter function could involve the neocortex as well as the It remains to be seen how this and other neurobiological the-
hippocampus. ories of episodic-like memory (Hasselmo et al., 2002;
The development of place elds during NMDA receptor McNaughton et al., 2003; Nakazawa et al., 2004; Buzsaki, 2005;
blockade or other forms of disruption of LTP is consistent Hasselmo and Howard, 2005) will handle the growing body of
with several lines of work suggesting that spatial information data that point to a specic role for the hippocampus in at
can be generated and stored upstream of the hippocampus. least aspects of episodic-like memory. The approach just
First, location-specic ring is already expressed in the super- described builds on the neuropsychological framework devel-
cial layers of the entorhinal cortex (Quirk et al., 1992; Frank oped in this chapter. It accepts that understanding the neu-
et al., 2000; Fyhn et al., 2004; Hafting et al., 2005). The fact roanatomical basis of recognizing a previously seen stimulus
 Theories of Hippocampal Function 693

is an important foundation stone of declarative memory (see thinking about the problems of trace stabilization and mem-
Section 13.3) but asserts that there is more to memory than ory consolidation. A major obstacle ahead is understanding
this (as, indeed, declarative memory theory has long recog- better how information is represented as patterns of neural
nized). It also accepts that hippocampal cells agree that space activity, and progress in unraveling the physiology of hip-
is special but argues that they can do more than encode spa- pocampal neurons will aid in that formidable task (see
tial information and calls into question the supposition that Chapters 5 and 6). In this research program, a wide range of
allocentric spatial memory is fundamentally different from behavioral tasks with animals will continue to play a key part
other forms of associative memory (see Section 13.4). It tries in unraveling the functions and neural mechanisms of the
to take on board the somewhat impenetrable developments in hippocampal formation. This chapter has drawn on data
contemporary animal learning theory and the progress that derived from primate experiments investigating recognition
has been made in understanding predictable ambiguity (see memory and object-in-place memory and from rats running
Section 13.5). Perhaps most difficult of all, it is part of an in radial mazes, digging in odorized sand wells for food, or
international effort seeking to integrate functional and mech- swimming in watermazes; and it has touched on new behav-
anistic ideas about rapid, one-trial, context-specic learning. ioral techniques to investigate episodic-like memory. It is t-
This task requires an exquisite understanding of hippocampal ting to acknowledge and celebrate the contribution that
neuroanatomy, including that of the myriad inhibitory laboratory animals have made to this effort (Fig. 1349, see
interneurons and the local circuits of which they are a part color insert). Progress is being made in tackling the mysteries
that surely make a major contribution to the information pro- of the hippocampal memory machine. The challenges ahead
cessing algorithms computed in the hippocampus. Cell bio- are as formidable as have been the achievements in character-
logical approaches will also prove increasingly valuable when izing its role in memory over the past three decades.

Figure 1349. Examples of common behavioral tasks used to study hippocampal function in
animals. A. Classic Wisconsin General Testing Apparatus for macaque monkeys. B. Screen shot
of modern touchscreen display. C. Radial maze foraging task for rats, shown with doors. D.
Rat making a choice in a two-pot olfactory discrimination task. Only one odor is familiar or
may be cued by another odor in a paired-associate manner.
694 The Hippocampus Book

Figure 1349. (Continued) E. Foraging monkey exploring objects in what-where-when memory task in which it caches various foods
a laboratory. F. Watermaze. Rat swimming in a large 2-m pool try- in an ice-cube tray and then retrieves them selectively later. H.
ing to nd a hidden escape platform under the water surface (just Paired-associate event arena task for rats. Different foods are found
visible on the right). Inset (below). The rat is on the platform, rear- at specic locations (e.g., one shown) in the apparatus.
ing and working out his location. G. Scrub-jay in an episodic-like

ACKNOWLEDGMENTS Guzowski, Lucia Jacobs, Len Jarrard, Kazu Nakazawa, and Simon
Rempel). Patrick Spoooner helped with many aspects of the comput-
I am grateful to many people for their help in preparing this chapter. ing. I apologize to yet others who helped but whom I have not men-
In addition to my co-editors of this book, Mark Baxter (Oxford tioned. Finally, one institution should not be forgotten. I am eternally
University) read a draft of the entire manuscript and offered numer- grateful to the UK Medical Research Council without whose support
ous comments and criticisms. Tobias Bast, Stephen Martin throughout my career this work would not have been possible.
(University of Edinburgh and other members of the Laboratory for
Cognitive Neuroscience), Edvard Moser and May-Britt Moser
(Center for the Biology of Memory, Trondheim), and Craig Stark REFERENCES
(Johns Hopkins) also read large parts and directed me to additional
literature. I am grateful to Helen Knight, Marie Pezze, and Carolyn Abel T, Lattal KM (2001) Molecular mechanisms of memory acqui-
Smith for collecting material for and assisting with many of the g- sition, consolidation and retrieval. Curr Opin Neurobiol
ures. I am also grateful to many neuroscience colleagues around the 11:180187.
world who provided material for the gures or advice of other kinds Adams CD, Dickinson A (1981) Instrumental responding following
(including Jamie Ainge, Verner Bingman, Norbert Fortin, John reinforcer devaluation. Q J Exp Psychol 33B:109122.
 Theories of Hippocampal Function 695

Adolphs R (2003) Cognitive neuroscience of human social behav- Anagnostaras SG, Gale GD, Fanselow MS (2002) The hippocam-
iour. Nat Rev Neurosci 4:165178. pus and pavlovian fear conditioning: reply to Bast et al.
Aggleton JP (1985) One-trial object recognition by rats. Q J Exp Hippocampus 12:561565.
Psychol 37B:279294. Andersen P, Bliss TVP, Skrede KK (1971) Lamellar organization of
Aggleton JP (1999) Mapping recognition memory in the primate hippocampal excitatory pathways. Exp Brain Res 13:222238.
brain: why its sometimes right to be wrong. Brain Res Bull Andersen P, Soleng AF, Raastad M (2000) The hippocampal lamella
50:447448. hypothesis revisited. Brain Res 886:165171.
Aggleton JP, Brown MW (1999) Episodic memory, amnesia, and the Andersson M, Krebs JR (1978) On the evolution of hoarding behav-
hippocampal-anterior thalamic axis. Behav Brain Sci 22: iour. Anim Behav 26:707711.
425489. Angeli SJ, Murray EA, Mishkin M (1993) Hippocampetized monkeys
Aggleton JP, Pearce JM (2001) Neural systems underlying episodic can remember one place but not two. Neuropsychologia
memory: insights from animal research. Philos Trans R Soc Lond 31:10211030.
B Biol Sci 356:14671482. Archer J, Birke L (1983) Exploration in animals and humans. London:
Aggleton JP, Shaw C (1996) Amnesia and recognition memory: a re- Van Nostrand Reinhold.
analysis of psychometric data. Neuropsychologia 34:5162. Attwell PJ, Rahman S, Yeo CH (2001) Acquisition of eyeblink condi-
Aggleton JP, Vann SD (2004) Testing the importance of the retrosple- tioning is critically dependent on normal function in cerebellar
nial navigation system: lesion size but not strain matters: a reply cortical lobule HVI. J Neurosci 21:57155722.
to Harker and Whishaw. Neurosci Biobehav Rev 28:525-531. Bachevalier J, Brickson M, Hagger C (1993) Limbic-dependent
Aggleton JP, Hunt PR, Rawlins JN (1986) The effects of hippocampal recognition memory in monkeys develops early in infancy.
lesions upon spatial and non-spatial tests of working memory. Neuroreport 4:7780.
Behav Brain Res 19:133146. Bachevalier J, Nemanic S, Alvarado MC (2002) The medial temporal
Agster KL, Fortin NJ, Eichenbaum H (2002) The hippocampus and lobe structures and object recognition memory in nonhuman
disambiguation of overlapping sequences. J Neurosci 22: primates. In: Neuropsychology of memory, 3rd ed. (Squire LR,
57605768. Schacter DL, eds), pp 326339. New York: Guildford Press.
Aiba A, Chen C, Herrup K, Rosenmund C, Stevens CF, Tonegawa S Bachevalier J, Vargha-Khadem F (2005) The primate hippocampus:
(1994) Reduced hippocampal long-term potentiation and con- ontogeny, early insult and memory. Curr Opin Neurobiol
text-specic decit in associative learning in mGluR1 mutant 15:168174.
mice. Cell 79:365375. Baddeley A (2001) The concept of episodic memory. Philos Trans R
Alvarado MC, Rudy JW (1995a) A comparison of kainic acid plus Soc Lond B Biol Sci 356:13451350.
colchicine and ibotenic acid-induced hippocampal formation Baddeley A, Conway M, Aggleton JP (2002) Episodic memory: new
damage on four congural tasks in rats. Behav Neurosci directions in research. Oxford, UK: Oxford University Press.
109:10521062. Balda RP, Kamil AC (1989) A comparative study of cache recovery by
Alvarado MC, Rudy JW (1995b) Rats with damage to the hippocam- three corvid species. Anim Behav 38:486495.
pal-formation are impaired on the tranverse-patterning prob- Balsam PD, Tomie A (1985) Context and learning. Hillsdale, NJ:
lem but not on elemental discriminations. Behav Neurosci Erlbaum.
109:204211. Bannerman DM, Yee BK, Good MA, Heupel MJ, Iversen SD, Rawlins
Alvarez-Royo P, Zola-Morgan S, Squire LR (1992) Impairment of JN (1999) Double dissociation of function within the hip-
long-term memory and sparing of short-term memory in mon- pocampus: a comparison of dorsal, ventral, and complete hip-
keys with medial temporal lobe lesions: a response to Ringo. pocampal cytotoxic lesions. Behav Neurosci 113:11701188.
Behav Brain Res 52:15. Bannerman DM, Yee BK, Lemaire M, Wilbrecht L, Jarrard L, Iversen
Alvarez P, Zola-Morgan S, Squire LR (1994) The animal model of SD, Rawlins JN, Good MA (2001) The role of the entorhinal
human amnesia: long-term memory impaired and short-term cortex in two forms of spatial learning and memory. Exp Brain
memory intact. Proc Natl Acad Sci USA 91:56375641. Res 141:281303.
Alvarez P, Zola-Morgan S, Squire LR (1995) Damage limited to the Bannerman DM, Deacon RM, Offen S, Friswell J, Grubb M, Rawlins
hippocampal region produces long-lasting memory impair- JN (2002) Double dissociation of function within the hip-
ment in monkeys. J Neurosci 15:37963807. pocampus: spatial memory and hyponeophagia. Behav Neurosci
Alyan S, McNaughton BL (1999) Hippocampectomized rats are capa- 116:884901.
ble of homing by path integration. Behav Neurosci 113:19-31. Bannerman DM, Rawlins JN, McHugh SB, Deacon RM, Yee BK, Bast
Amaral DG (1990) Neurons, numbers and the hippocampal net- T, Zhang WN, Pothuizen HH, Feldon J (2004) Regional dissoci-
work. Prog Brain Res 83:111. ations within the hippocampusmemory and anxiety. Neurosci
Amaral DG, Witter MP (1989) The three-dimensional organization Biobehav Rev 28:273283.
of the hippocampal formation: a review of anatomical data. Barnea A, Nottebohm F (1994) Seasonal recruitment of hippocampal
Neuroscience 31:571591. neurons in adult free-ranging black-capped chickadees. Proc
Amaral DG, Insausti R, Cowan WM (1987) The entorhinal cortex of Natl Acad Sci USA 91:1121711221.
the monkey. I. Cytoarchitectonic organization. J Comp Neurol Barnes CA (1979) Memory decits associated with senescence: a neu-
264:326355. rophysiological and behavioral study in the rat. J Comp Physiol
Anagnostaras SG, Maren S, Fanselow MS (1999) Temporally graded Psychol 93:74104.
retrograde amnesia of contextual fear after hippocampal damage Barnes CA (1988) Spatial learning and memory processes: the search
in rats: within-subjects examination. J Neurosci 19:11061114. for their neurobiological mechanisms in the rat. Trends Neurosci
Anagnostaras SG, Gale GD, Fanselow MS (2001) Hippocampus and 11:163169.
contextual fear conditioning: recent controversies and advances. Bartlett FC (1932) Remembering. Cambridge, UK: Cambridge
Hippocampus 11:817. University Press.
696 The Hippocampus Book

Barton RA, Harvey PH (2000) Mosaic evolution of brain structure in Bingman VP, Ioale P, Casini G, Bagnoli P (1987) Impaired retention
mammals. Nature 405:10551058. of preoperatively acquired spatial reference memory in homing
Bast T, Zhang WN, Feldon J (2001) Hippocampus and classical fear pigeons following hippocampal ablation. Behav Brain Res
conditioning. Hippocampus 11:828831. 24:147156.
Bast T, da Silva BM, Morris RGM (2005) Distinct contributions of Bingman VP, Hough GE 2nd, Kahn MC, Siegel JJ (2003) The homing
hippocampal NMDA and AMPA receptors to encoding and pigeon hippocampus and space: in search of adaptive special-
retrieval of one-trial place memory. J Neurosci 25:58455856. ization. Brain Behav Evol 62:117127.
Baxter MG, Murray EA (2001a) Effects of hippocampal lesions on Bliss TVP, Collingridge GL (1993) A synaptic model of memory:
delayed nonmatching-to-sample in monkeys: a reply to Zola long-term potentiation in the hippocampus. Nature 361:3139.
and Squire (2001). Hippocampus 11:201203. Bliss TVP, Collingridge GL, Morris RGM (2004) Long term potentia-
Baxter MG, Murray EA (2001b) Opposite relationship of hip- tion: enhancing neuroscience for 30 years. Oxford, UK: Oxford
pocampal and rhinal cortex damage to delayed nonmatching- University Press.
to-sample decits in monkeys. Hippocampus 11:6171. Bolhuis JJ, Macphail EM (2001) A critique of the neuroecology of
Bayley PJ, Frascino JC, Squire LR (2005a) Robust habit learning in learning and memory. Trends Cogn Sci 5:426433.
the absence of awareness and independent of the medial tem- Bolhuis JJ, Stewart CA, Forrest EM (1994) Retrograde amnesia and
poral lobe. Nature 436:550553. memory reactivation in rats with ibotenate lesions to the hip-
Bayley PJ, Gold JJ, Hopkins RO, Squire LR (2005b) The neu- pocampus or subiculum. Q J Exp Psychol B 47:129150.
roanatomy of remote memory. Neuron 46:799810. Bolles RC (1970) Species-specic defense reactions and avoidance
Bear MF, Connors BW, Paradiso MA (2001) Neuroscience: exploring learning. Psychol Rev 77:3248.
the brain. Philadelphia: Lippincott. Bontempi B, Laurent-Demir C, Destrade C, Jaffard R (1999) Time-
Beason-Held LL, Rosene DL, Killiany RJ, Moss MB (1999) dependent reorganization of brain circuitry underlying long-
Hippocampal formation lesions produce memory impairment term memory storage. Nature 400:671675.
in the rhesus monkey. Hippocampus 9:562574. Bostock E, Muller RU, Kubie JL (1991) Experience-dependent
Beggs J, Brown T, Byrne J, Crow T, LeDoux J, LeBar K, Thompson R modications of hippocampal place cell ring. Hippocampus
(1999) Learning and memory: basic mechanisms. In: Funda- 1:193205.
mental neuroscience, pp 14111454. London: Academic Press. Bouton ME (2004) Context and behavioral processes in extinction.
Behrmann M, Geng JJ, Shomstein S (2004) Parietal cortex and atten- Learn Mem 11:485494.
tion. Curr Opin Neurobiol 14:212217. Bouton ME, Moody EW (2004) Memory processes in classical condi-
Benhamou S (1997) On systems of reference involved in spatial tioning. Neurosci Biobehav Rev 28:663674.
memory. Behav Processes 40:149163. Brandeis R, Brandys Y, Yehuda S (1989) The use of the Morris water-
Bennett AT (1996) Do animals have cognitive maps? J Exp Biol 199 maze in the study of memory and learning. Int J Neurosci
(Pt 1):219224. 48:2969.
Berger TW, Rinaldi PC, Weisz DJ, Thompson RF (1983) Single-unit Bransford JD (1979) Human cognition: learning, understanding and
analysis of different hippocampal cell types during classical remembering. Belmont, CA: Wadsworth.
conditioning of rabbit nictitating membrane response. J Bransford JD, Johnson MK (1972) Contextual prerequisites for
Neurophysiol 50:11971219. understanding: some investigations of comprehension and
Berlyne DE (1950) Novelty and curiousity as determinants of recall. J Verb Learn Verb Behav 11:717726.
exploratory behavior. Br J Psychol 41:6880. Brasted PJ, Bussey TJ, Murray EA, Wise SP (2002) Fornix transec-
Berlyne DE (1966) Curiosity and exploration. Science 153:2533. tion impairs conditional visuomotor learning in tasks involv-
Bernasconi-Guastalla S, Wolfer DP, Lipp HP (1994) Hippocampal ing nonspatially differentiated responses. J Neurophysiol 87:
mossy bers and swimming navigation in mice: correlations 631633.
with size and left-right asymmetries. Hippocampus 4:5363. Broadbent NJ, Gallagher S, Colombo M (1999) Hippocampal lesions
Biegler R (2000) Possible uses of path integration in animal naviga- and negative patterning in pigeons. Psychobiology 27:5156.
tion. Anim Learn Behav 28:257277. Broadbent NJ, Squire LR, Clark RE (2004) Spatial memory, recogni-
Biegler R (2003) Reading cognitive and other maps: how to avoid get- tion memory, and the hippocampus. Proc Natl Acad Sci USA
ting buried in thought. In: The neurobiology of spatial behaviour 101:1451514520.
(Jeffery KJ, ed), pp 259274. New York: Oxford University Press. Brown MF (1992) Does a cognitive map guide choices in the radial-
Biegler R, Morris RGM (1993) Landmark stability is a prerequisite arm maze? J Exp Psychol Anim Behav Process 21:147156.
for spatial but not discrimination learning. Nature 361:631633. Brown MF, Rish PA, Von Culin JE, Edberg JA (1993) Spatial guidance
Biegler R, Morris RGM (1999) Blocking in the spatial domain with of choice behavior in the radial arm maze. J Exp Psychol Anim
arrays of discrete landmarks. J Exp Psychol Anim Behav Process Behav Process 19:195214.
25:334351. Brown MW (1996) Neuronal responses and recognition memory.
Biegler R, McGregor A, Krebs JR, Healy SD (2001) A larger hip- Semin Neurosci 8:2332.
pocampus is associated with longer-lasting spatial memory. Proc Brown MW, Aggleton JP (2001) Recognition memory: what are the
Natl Acad Sci USA 98:69416944. roles of the perirhinal cortex and hippocampus? Nat Rev
Bingman VP, Jones TJ (1994) Sun compass-based spatial learning Neurosci 2:5161.
impaired in homing pigeons with hippocampal lesions. J Brown MW, Xiang J-Z (1998) Recognition memory: neuronal sub-
Neurosci 14:66876694. strates of the judgement of prior occurrence. Prog Neurobiol
Bingman VP, Bagnoli P, Ioale P, Casini G (1984) Homing behav- 55:149189.
ior of pigeons after telencephalic ablations. Brain Behav Evol Brown MW, Wilson FAW, Riches IP (1987) Neuronal evidence
24:94108. that inferomedial temporal cortex is more important than
 Theories of Hippocampal Function 697

hippocampus in certain processes underlying recognition mem- Cassaday HJ, Rawlins JN (1995) Fornix-mbria section and working
ory. Brain Res 409:158162. memory decits in rats: stimulus complexity and stimulus size.
Browning PG, Easton A, Buckley MJ, Gaffan D (2005) The role of Behav Neurosci 109:594606.
prefrontal cortex in object-in-place learning in monkeys. Eur J Cassaday HJ, Rawlins JN (1997) The hippocampus, objects, and their
Neurosci 22:32813291. contexts. Behav Neurosci 111:12281244.
Brun VH, Otnass MK, Molden S, Steffenach HA, Witter MP, Moser Cavaco S, Anderson SW, Allen JS, Castro-Caldas A, Damasio H
MB, Moser EI (2002) Place cells and place recognition main- (2004) The scope of preserved procedural memory in amnesia.
tained by direct entorhinal-hippocampal circuitry. Science Brain 127:18531867.
296:22432246. Chamizo VD (2002) Spatial learning: conditions and basic effects.
Buckley MJ, Gaffan D (1998a) Perirhinal cortex ablation impairs Psicologica 23:3357.
visual object identication. J Neurosci 18:22682275. Chamizo VD (2003) Acquisition of knowledge about spatial location:
Buckley MJ, Gaffan D (1998b) Learning and transfer of object- assessing the generality of the mechanism of learning. Q J Exp
reward associations and the role of the perirhinal cortex. Behav Psychol B 56:102113.
Neurosci 112:1523. Chamizo VD, Sterio D, Mackintosh NJ (1985) Blocking and over-
Buckmaster CA, Eichenbaum H, Amaral DG, Suzuki WA, Rapp PR shadowing between intra-maze and extra-maze cues: a test of
(2004) Entorhinal cortex lesions disrupt the relational organiza- the independence of locale and guidance learning. Q J Exp
tion of memory in monkeys. J Neurosci 24:98119825. Psychol 37B:235263.
Buffalo EA, Reber PJ, Squire LR (1998a) The human perirhinal cor- Chan KH, Morell JR, Jarrard LE, Davidson TL (2001) Reconsideration
tex and recognition memory. Hippocampus 8:330339. of the role of the hippocampus in learned inhibition. Behav
Buffalo EA, Stefanacci L, Squire LR, Zola SM (1998b) A reexamina- Brain Res 119:111130.
tion of the concurrent discrimination learning task: the impor- Charles DP, Gaffan D, Buckley MJ (2004a) Impaired recency judg-
tance of anterior inferotemporal cortex, area TE. Behav Neurosci ments and intact novelty judgments after fornix transection in
112:314. monkeys. J Neurosci 24:20372044.
Bunsey M, Eichenbaum H (1995) Selective damage to the hip- Charles DP, Browning PG, Gaffan D (2004b) Entorhinal cortex
pocampal region blocks long-term retention of a natural contributes to object-in-place scene memory. Eur J Neurosci
and nonspatial stimulus-stimulus association. Hippocampus 20:31573164.
5:546556. Cheng K (1986) A purely geometric module in the rats spatial repre-
Bunsey M, Eichenbaum H (1996) Conservation of hippocampal sentation. Cognition 23:149178.
memory function in rats and humans. Nature 379:255257. Cho YH, Beracochea D, Jaffard R (1993) Extended temporal
Bures J, Fenton AA, Kaminsky Y, Wesierska M, Zahalka A (1998) gradi-ent for the retrograde and anterograde amnesia produced
Rodent navigation after dissociation of the allocentric and idio- by ibotenate entorhinal cortex lesions in mice. J Neurosci
thetic representations of space. Neuropharmacology 37:689699. 13:17591766.
Burgess N, Maguire EA, OKeefe J (2002) The human hippocampus Cho YH, Kesner RP, Brodale S (1995) Retrograde and anterograde
and spatial and episodic memory. Neuron 35:625641. amnesia for spatial discrimination in rats: role of hippocam-
Bussey TJ, Clea Warburton E, Aggleton JP, Muir JL (1998) Fornix pus, entorhinal cortex, and parietal cortex. Psychobiology 23:
lesions can facilitate acquisition of the transverse patterning 185194.
task: a challenge for congural theories of hippocampal func- Cipolotti L, Shallice T, Chan D, Fox N, Scahill R, Harrison G, Stevens
tion. J Neurosci 18:16221631. J, Rudge P (2001) Long-term retrograde amnesia...the crucial
Bussey TJ, Muir JL, Aggleton JP (1999) Functionally dissociating role of the hippocampus. Neuropsychologia 39:151172.
aspects of event memory: the effects of combined perirhinal Cirillo RA, Horel JA, George PJ (1989) Lesions of the anterior tem-
and postrhinal cortex lesions on object and place memory in the poral stem and the performance of delayed match-to-sample
rat. J Neurosci 19:495502. and visual discriminations in monkeys. Behav Brain Res
Bussey TJ, Dias R, Redhead ES, Pearse JM, Muir JL, Aggleton JP 34:5569.
(2000) Intact negative patterning in rats with fornix or com- Claparede E (1911) Recognition et moite. Arch Psychol Geneva
bined perirhinal and postrhinal cortex lesions. Exp Brain Res 11:7990.
134:506519. Clark GA, McCormick DA, Lavond DG, Thompson RF (1984)
Bussey TJ, Saksida LM, Murray EA (2002) Perirhinal cortex resolves Effects of lesions of cerebellar nuclei on conditioned behav-
feature ambiguity in complex visual discriminations. Eur J ioral and hippocampal neuronal responses. Brain Res
Neurosci 15:365374. 291:125136.
Bussey TJ, Saksida LM, Murray EA (2003) Impairments in visual dis- Clark RE, Martin SJ (2005) Interrogating rodents regarding their
crimination after perirhinal cortex lesions: testing declarative object and spatial memory. Curr Opin Neurobiol 15:593598.
vs. perceptual-mnemonic views of perirhinal cortex function. Clark RE, Squire LR (1998) Classical conditioning and brain systems:
Eur J Neurosci 17:649660. the role of awareness. Science 280:7781.
Buzsaki G (2005) Theta rhythm of navigation: link between path Clark RE, Zola SM, Squire LR (2000) Impaired recognition memory
integration and landmark navigation, episodic and semantic in rats after damage to the hippocampus. J Neurosci 20:
memory. Hippocampus 15:827840. 88538860.
Cahill L, McGaugh JL (1996) Modulation of memory storage. Curr Clark RE, West AN, Zola SM, Squire LR (2001) Rats with lesions of
Opin Neurobiol 6:237242. the hippocampus are impaired on the delayed nonmatching-to-
Capaldi EJ (1971) Memory and learning: a sequential view. In: sample task. Hippocampus 11:176186.
Animal memory (Honig WK, James PHR, eds), pp 111154. San Clark RE, Broadbent NJ, Squire LR (2004) Hippocampus and remote
Diego: Academic Press. spatial memory in rats. Hippocampus 15:260272.
698 The Hippocampus Book

Clark RE, Broadbent NJ, Squire LR (2005a) Impaired remote spatial Corkin S (1984) Lasting consequences of bilateral medial temporal
memory after hippocampal lesions despite extensive training lobectomy: clinical course and experimental ndings in H.M.
beginning early in life. Hippocampus 15:340346. Semin Neurol 4:249259.
Clark RE, Broadbent NJ, Squire LR (2005b) Hippocampus and Cowey A, Stoerig P (1995) Blindsight in monkeys. Nature
remote spatial memory in rats. Hippocampus 15:260272. 373:247249.
Clayton NS (1995) Development of memory and the hippocampus: Crabbe JC, Morris RGM (2004) Festina lente: late-night thoughts on
comparison of food-storing and nonstoring birds on a one-trial high-throughput screening of mouse behavior. Nat Neurosci
associative memory task. J Neurosci 15:27962807. 7:11751179.
Clayton NS, Dickinson A (1998) Episodic-like memory during cache Crusio WE, Schwegler H, Lipp HP (1987) Radial-maze performance
recovery by scrub jays. Nature 395:272274. and structural variation of the hippocampus in mice: a correla-
Clayton NS, Krebs JR (1995) Memory in food-storing birds: from tion with mossy bre distribution. Brain Res 425:182185.
behaviour to brain. Curr Opin Neurobiol 5:149154. Dale N, Roberts A (1985) Dual-component amino-acid-mediated
Clayton NS, Griffiths DP, Emery N, Dickinson A (2001) Elements of synaptic potentials: excitatory drive for swimming in Xenopus
episodic-like memory in animals. In: Episodic memory: new embryos. J Physiol 363:3559.
directions in research (Baddeley A, Conway M, Aggleton JP, eds), Davidson TL (1993) The nature and function of interoceptive signals
pp 232248. Oxford, UK: Oxford University Press. to feed: toward integration of physiological and learning per-
Clayton NS, Yu KS, Dickinson A (2003a) Interacting cache memories: spectives. Psychol Rev 100:640657.
evidence for flexible memory use by Western scrub-jays Davidson TL, Jarrard LE (1993) A role for hippocampus in the uti-
(Aphelocoma californica). J Exp Psychol Anim Behav Process lization of hunger signals. Behav Neural Biol 59:167171.
29:1422. Davidson TL, Jarrard LE (2004) The hippocampus and inhibitory
Clayton NS, Bussey TJ, Emery NJ, Dickinson A (2003b) Prometheus learning: a gray area? Neurosci Biobehav Rev 28:261271.
to Proust: the case for behavioural criteria for mental time Davidson TL, McKernan MG, Jarrard LE (1993) Hippocampal
travel. Trends Cogn Sci 7:436437. lesions do not impair negative patterning: a challenge to cong-
Cohen NJ, Eichenbaum HE (1993) Memory, amnesia and the hip- ural association theory. Behav Neurosci 107:227-234.
pocampal system, 1st ed. Cambridge, MA: MIT Press. Davis M, Falls WA, Campeau S, Kim M (1993) Fear-potentiated star-
Cohen NJ, Squire LR (1980) Preserved learning and retention of tle: a neural and pharmacological analysis. Behav Brain Res
pattern-analyzing skill in amnesia: dissociation of knowing how 58:175-198.
and knowing that. Science 210:207210. Day M, Langston RF, Morris RGM (2003) Glutamate receptor
Cohen NJ, Ryan J, Hunt C, Romine L, Wszalek T, Nash C (1999) dependent encoding and retrieval of paired associate learning.
Hippocampal system and declarative (relational) memory: Nature 424:205-209.
summarizing the data from functional neuroimaging studies. Deacon RM, Bannerman DM, Rawlins NP (2001) Conditional dis-
Hippocampus 9:8398. criminations based on external and internal cues in rats with
Collett TS (1992) Landmark learning and guidance in insects. Philos cytotoxic hippocampal lesions. Behav Neurosci 115:4357.
Trans R Lond B 337:295303. Dean P, Weiskrantz L (1974) Loss of preoperative habits in rhesus
Collett TS, Cartwright BA, Smith BA (1986) Landmark learning monkeys with inferotemporal lesions: recognition failure or
and visuo-spatial memories in gerbils. J Comp Psychol A relearning decit? Neuropsychologia 12:299311.
158:835851. De Hoz L, Knox J, Morris RGM (2003) Longitudinal axis of the hip-
Collins AM, Quillian MR (1969) Retrieval time from semantic mem- pocampus: both septal and temporal poles of the hippocampus
ory. J Verbal Learn Verbal Behav 8:240247. support watermaze spatial learning depending on the training
Columbo M, Cawley S, Broadbent NJ (1997) The effects of hip- protocol. Hippocampus 13:587603.
pocampal and area parahippocampalis lesions in pigeons. II. De Hoz L, Martin SJ, Morris RGM (2004) Forgetting, reminding, and
Concurrent discrimination and spatial memory. Q J Exp Psychol remembering: the retrieval of lost spatial memory. PLoS Biol
50B:172189. 2:E225.
Conference B (1997) Mutant mice and neuroscience: recommenda- De Kloet ER, Oitzl MS, Joels M (1999) Stress and cognition: are cor-
tions concerning genetic background; Banbury conference on ticosteroids good or bad guys? Trends Neurosci 22:422426.
genetic background in mice. Neuron 19:755759. DHooge R, De Deyn PP (2001) Applications of the Morris water-
Consortium RGSP (2004) Genome sequence of the brown maze in the study of learning and memory. Brain Res Brain Res
Norway rat yields insights into mammalian evolution. Nature Rev 36:6090.
428:493521. Dickinson A (1980) Contemporary animal learning theory.
Cooper BG, Mizumori SJ (2001) Temporary inactivation of the ret- Cambridge: Cambridge University Press.
rosplenial cortex causes a transient reorganization of spatial Dickinson A, Mackintosh NJ (1978) Classical conditioning in ani-
coding in the hippocampus. J Neurosci 21:39864001. mals. Annu Rev Psychol 29:587612.
Corcoran KA, Maren S (2001) Hippocampal inactivation disrupts DiMattia BV, Kesner RP (1988a) Role of the posterior parietal associ-
contextual retrieval of fear memory after extinction. J Neurosci ation cortex in the processing of spatial event information.
21:17201726. Behav Neurosci 102:397403.
Corcoran KA, Maren S (2004) Factors regulating the effects of hip- Dix SL, Aggleton JP (1999) Extending the spontaneous preference
pocampal inactivation on renewal of conditional fear after test of recognition: evidence of object-location and object-
extinction. Learn Mem 11:598603. context recognition. Behav Brain Res 99:191200.
Corcoran KA, Desmond TJ, Frey KA, Maren S (2005) Hippocampal Dore FY, Thornton JA, White NM, Murray EA (1998) Selective hip-
inactivation disrupts the acquisition and contextual encoding of pocampal lesions yield nonspatial memory impairments in rhe-
fear extinction. J Neurosci 25:89788987. sus monkeys. Hippocampus 8:323329.
 Theories of Hippocampal Function 699

Douglas RJ, Martin KA (2004) Neuronal circuits of the neocortex. evolution of intelligence in corvids and apes. Science
Annu Rev Neurosci 27:419451. 306:19031907.
Dragunow M (1996) A role for immediate-early transcription factors Ennaceur A, Aggleton JP (1994) Spontaneous recognition of object
in learning and memory. Behav Genet 26:293299. congurations in rats: effects of fornix lesions. Exp Brain Res
Dudai Y, Morris RGM (2001) To consolidate or not to consolidate: 100:8592.
what are the questions? In: Brain, perception and memory: Ennaceur A, Delacour J (1988) A new one-trial test for neurobiolog-
advances in cognitive sciences (Bolhuis J, ed), pp 147162. ical studies of memory in rats. 1. Behavioral data. Behav Brain
Oxford, UK: Oxford University Press. Res 31:4759.
Dudchenko PA (2003) The head-direction system and navigation. In: Etienne AS, Jeffery KJ (2004) Path integration in mammals.
The neurobiology of spatial behaviour (Jeffery K, ed). Oxford, Hippocampus 14:180192.
UK: Oxford University Press. Etienne AS, Maurer R, Berlie J, Reverdin B, Rowe T, Georgakopoulos
Dudchenko PA, Taube JS (1997) Correlation between head direction J, Seguinot V (1998) Navigation through vector addition. Nature
cell activity and spatial behavior on a radial arm maze. Behav 396:161164.
Neurosci 111:319. Fanselow MS (1990) Factors governing one-trial contextual condi-
Dudchenko PA, Wood ER, Eichenbaum H (2000) Neurotoxic hip- tioning. Anim Learn Behav 18:264270.
pocampal lesions have no effect on odor span and little effect on Fanselow MS (2000) Contextual fear, gestalt memories and the hip-
odor recognition memory but produce signicant impairments pocampus. Behav Brain Res 110:7381.
on spatial span, recognition, and alternation. J Neurosci Fantz RL (1964) Visual experience in infants: decreased attention to
20:29642977. familiar patterns relative to novel ones. Science 146:668670.
Dusek JA, Eichenbaum H (1997) The hippocampus and mem- Farah MJ (2004) Visual agnosia. Cambridge, MA: MIT Press.
ory for orderly stimulus relations. Proc Natl Acad Sci USA Ferbinteanu J, Shapiro ML (2003) Prospective and retrospective
94:71097114. memory coding in the hippocampus. Neuron 40:12271239.
Dyer F (1998) Cognitive ecology of navigation. In: Cognitive ecology Finger S, Koehler PJ, Jagella C (2004) The Monakow concept of
(Dukas R, ed), pp 201260. Chicago: University of Chicago diaschisis: origins and perspectives. Arch Neurol 61:283288.
Press. Fletcher PC, Henson RN (2001) Frontal lobes and human memory:
Eacott MJ, Gaffan EA (2005) The roles of perirhinal cortex, postrhi- insights from functional neuroimaging. Brain 124:849881.
nal cortex, and the fornix in memory for objects, contexts, and Fortin NJ, Agster KL, Eichenbaum HB (2002) Critical role of the hip-
events in the rat. Q J Exp Psychol B 58:202217. pocampus in memory for sequences of events. Nat Neurosci
Eacott MJ, Norman G (2004) Integrated memory for object, place, 5:458462.
and context in rats: a possible model of episodic-like memory? Fortin NJ, Wright SP, Eichenbaum H (2004) Recollection-like mem-
J Neurosci 24:19481953. ory retrieval in rats is dependent on the hippocampus. Nature
Eacott MJ, Easton A, Zinkivskay A (2005) Recollection in an episodic- 431:188191.
like memory task in the rat. Learn Mem 12:221223. Forwood SE, Winters BD, Bussey TJ (2005) Hippocampal lesions
Easton A, Gaffan D (2000) Comparison of perirhinal cortex that abolish spatial maze performance spare object recogni-
ablation and crossed unilateral lesions of the medial forebrain tion memory at delays of up to 48 hours. Hippocampus
bundle from the inferior temporal cortex in the rhesus 15:347355.
monkey: effects on learning and retrieval. Behav Neurosci Foster TC, Castro CA, McNaughton BL (1989) Spatial selectivity of
114:10411057. rat hippocampal neurons: dependence on preparedness for
Easton A, Parker A, Gaffan D (2001) Crossed unilateral lesions of movement. Science 244:15801582.
medial forebrain bundle and either inferior temporal or frontal Fox GD, Holland PC (1998) Neurotoxic hippocampal lesions fail to
cortex impair object recognition memory in rhesus monkeys. impair reinstatement of an appetitively conditioned response.
Behav Brain Res 121:110. Behav Neurosci 112:255260.
Eichenbaum H (1996) Is the rodent hippocampus just for place? Frank LM, Brown EN, Wilson M (2000) Trajectory encoding in the
Curr Opin Neurobiol 6:187195. hippocampus and entorhinal cortex. Neuron 27:169178.
Eichenbaum H (2004) Hippocampus: cognitive processes and neural Frank MJ, Rudy JW, OReilly RC (2003) Transitivity, exibility, con-
representations that underlie declarative memory. Neuron junctive representations and the hippocampus. II. A computa-
44:109120. tional analysis. Hippocampus 13:299312.
Eichenbaum H, Cohen NJ (2001) From conditioning to conscious rec- Frankland PW, Bontempi B (2005) The organization of recent and
ollection. New York: Oxford University Press. remote memories. Nat Rev Neurosci 6:119130.
Eichenbaum H, Stewart C, Morris RGM (1990) Hippocampal repre- Frankland PW, Cestari V, Filipkowski RK, McDonald RJ, Silva AJ
sentation in place learning. J Neurosci 10:35313542. (1998) The dorsal hippocampus is essential for context discrim-
Eichenbaum H, Dudchenko P, Wood E, Shapiro M, Tanila H (1999) ination but not for contextual conditioning. Behav Neurosci
The hippocampus, memory, and place cells: is it spatial memory 112:863874.
or a memory space? Neuron 23:209226. Frankland PW, OBrien C, Ohno M, Kirkwood A, Silva AJ (2001)
Ekstrom AD, Kahana MJ, Caplan JB, Fields TA, Isham EA, Newman Alpha-CaMKII-dependent plasticity in the cortex is required
EL, Fried I (2003) Cellular networks underlying human spatial for permanent memory. Nature 411:309313.
navigation. Nature 425:184188. Frankland PW, Bontempi B, Talton LE, Kaczmarek L, Silva AJ (2004)
Emery NJ, Clayton NS (2001) Effects of experience and social context The involvement of the anterior cingulate cortex in remote con-
on prospective caching strategies by scrub jays. Nature textual fear memory. Science 304:881883.
414:443446. Frey U, Morris RGM (1997) Synaptic tagging and long-term potenti-
Emery NJ, Clayton NS (2004) The mentality of crows: convergent ation. Nature 385:533536.
700 The Hippocampus Book

Frey U, Morris RGM (1998) Synaptic tagging: implications for late Hippocampal lesions do not disrupt navigational map retention
maintenance of hippocampal long-term potentiation. Trends in homing pigeons under conditions when map acquisition is
Neurosci 21:181188. hippocampal dependent. Behav Brain Res 153:3542.
Fried I, Wilson CL, MacDonald KA, Behnke EJ (1998) Electric cur- Galea LA, McEwen BS (1999) Sex and seasonal differences in the rate
rent stimulates laughter. Nature 391:650. of cell proliferation in the dentate gyrus of adult wild meadow
Friston KJ (1995) Functional and effective connectivity in neu- voles. Neuroscience 89:955964.
roimaging: A synthesis. Hum Brain Mapp 2:5678. Galef BG (1990) Necessary and sufficient conditions for communi-
Friston KJ (2004) Models of brain function in neuroimaging. Annu cation of diet preferences by Norway rats. Anim Learn Behav
Rev Psychol 56:5787. 18:347351.
Fritz J, Mishkin M, Saunders RC (2005) In search of an auditory Galef BG, Wigmore SW (1983) Transfer of information concerning
engram. Proc Natl Acad Sci USA 102:93599364. distant foods: a laboratory investigation of the information-
Frohardt RJ, Guarraci FA, Bouton ME (2000) The effects of neuro- centre hypothesis. Anim Behav 31:748758.
toxic hippocampal lesions on two effects of context after fear Gallagher M, Holland PC (1992) Preserved congural learning and
extinction. Behav Neurosci 114:227240. spatial learning impairment in rats with hippocampal damage.
Fukuda M, Kobayashi T, Bures J, Ono T (1992) Rat exploratory Hippocampus 2:8188.
behavior controlled by intracranial self-stimulation improves Gallagher M, Rapp PR (1997) The use of animal models to study the
the study of place cell activity. J Neurosci Methods 44:121131. effects of aging on cognition. Annu Rev Psychol 48:339370.
Fyhn M, Molden S, Hollup SA, Moser M-B, Moser EI (2002) Gallistel C (1980) The organisation of action: a new synthesis.
Hippocampal neurons responding to rst-time dislocation of a Hillsdale, NJ: Erlbaum.
target object. Neuron 35:555566. Garrud P, Rawlins JN, Mackintosh NJ, Goodall G, Cotton MM,
Fyhn M, Molden S, Witter MP, Moser EI, Moser MB (2004) Feldon J (1984) Successful overshadowing and blocking in hip-
Spatial representation in the entorhinal cortex. Science pocampectomized rats. Behav Brain Res 12:3953.
305:12581264. Gattis M (2001) Spatial schemas and abstract thought. Cambridge,
Gabrieli JDE, Fleischman DA, Deane MM, Reminger SL, Morrell F MA: MIT Press.
(1995) Double dissociation between memory systems underly- Gaulin SJ, Fitzgerald RW (1989) Sexual selection for spatial-learning
ing explicit and implicit memory in the human brain. Psychol ability. Anim Behav 37:322331.
Sci 6:7682. Gaulin SJ, FitzGerald RW, Wartell MS (1990) Sex differences in
Gaffan D (1974) Recognition impaired and association intact in the spatial ability and activity in two vole species (Microtus
memory of monkeys after transection of the fornix. J Comp ochrogaster and M. pennsylvanicus). J Comp Psychol 104:8893.
Physiol Psychol 86:11001109. Gazzaniga MS (2002) Cognitive neuroscience: the biology of the mind.
Gaffan D (1983) Animal amnesia: some disconnection syndromes? New York: Norton.
In: Neurobiology of the hippocampus (Seifert W, ed), pp 513528. Gerlai R (1996) Gene-targeting studies of mammalian behavior: is it
London: Academic Press. the mutation or the background genotype? Trends Neurosci
Gaffan D (1991) Spatial organization of episodic memory. Hippocam- 19:177181.
pus 1:262264. Gewirtz JC, McNish KA, Davis M (2000) Is the hippocampus neces-
Gaffan D (1992) Amnesia for complex naturalistic scenes and for sary for contextual fear conditioning? Behav Brain Res
objects following fornix transection in the rhesus monkey. Eur J 110:8395.
Neurosci 4:381388. Gilbert P, Kesner R, DeCoteau R (1998) Memory for spatial location:
Gaffan D (1993) Additive effects of forgetting and fornix tran- role of the hippocampus in mediating spatial pattern separa-
section in the temporal gradient of retrograde amnesia. tion. J Neurosci 18:804810.
Neuropsychologia 31:10551066. Gilbert PE, Kesner RP, Lee I (2001) Dissociating hippocampal subre-
Gaffan D (1994) Scene-specic memory for objects: a model of gions: double dissociation between dentate gyrus and CA1.
episodic memory impairment in monkeys with fornix transec- Hippocampus 11:626636.
tion. J Cogn Neurosci 6:305320. Gisquet-Verrier P, Dutrieux G, Richer P, Doyere V (1999) Effects of
Gaffan D (2002) Against memory systems. Philos Trans R Soc Lond B lesions to the hippocampus on contextual fear: evidence for a
Biol Sci 357:11111121. disruption of freezing and avoidance behavior but not context-
Gaffan EA, Eacott MJ (1995) A computer-controlled maze environ- conditioning. Behav Neurosci 113:507522.
ment for testing visual memory in the rat. J Neurosci Methods Gladwell M (2002) The tipping point: how little things can make a big
60:2337. difference. Boston: Back Bay Books.
Gaffan D, Harrison S (1989) Place memory and scene memory: effects Goelet P, Castellucci VF, Schacher S, Kandel ER (1986) The long and
of fornix transection in the monkey. Exp Brain Res 74:202212. the short of long-term memorya molecular framework.
Gaffan D, Parker A (1996) Interaction of perirhinal cortex with the Nature 322:419422.
fornix-mbria: memory for objects and object-in-place Golding NL, Spruston N (1998) Dendritic sodium spikes are variable
memory. J Neurosci 16:58645869. triggers of axonal action potentials in hippocampal CA1 pyram-
Gaffan D, Gaffan EA, Harrison S (1984) Effects of fornix transection idal neurons. Neuron 21:11891200.
on spontaneous and trained non-matching by monkeys. Q J Exp Golding NL, Staff NP, Spruston N (2002) Dendritic spikes as a
Psychol 36B:285303. mechanism for cooperative long-term potentiation. Nature
Gagliardo A, Ioale P, Bingman VP (1999) Homing in pigeons: the role 418:326331.
of the hippocampal formation in the representation of land- Golob EJ, Taube JS (2002) Differences between appetitive and aver-
marks used for navigation. J Neurosci 19:311315. sive reinforcement on reorientation in a spatial working mem-
Gagliardo A, Ioale P, Odetti F, Kahn MC, Bingman VP (2004) ory task. Behav Brain Res 136:309316.
 Theories of Hippocampal Function 701

Good M, Honey RC (1991) Conditioning and contextual retrieval in Hagan JJ, Salamone JD, Simpson J, Iversen SD, Morris RG (1988)
hippocampal rats. Behav Neurosci 105:499509. Place navigation in rats is impaired by lesions of medial septum
Good M, de Hoz L, Morris RGM (1998) Contingent versus inciden- and diagonal band but not nucleus basalis magnocellularis.
tal context processing during conditioning: dissociation Behav Brain Res 27:920.
after excitotoxic hippocampal plus dentate gyrus lesions. Hall G, Purves D, Bonardi C (1996) Contextual control of condi-
Hippocampus 8:147159. tioned responding in rats with dorsal hippocampal lesions.
Good MA, Morris RGM (1994) A step linking memory to under- Behav Neurosci 110:933945.
standing? Behav Brain Sci 17:477479. Halliday MS (1968) Exploratory behaviour. In: Analysis of behav-
Gothard KM, Skaggs WE, Moore KM, McNaughton BL (1996) ioural change (Weiskrantz L, ed), pp 107126. New York: Harper
Binding of hippocampal CA1 neural activity to multiple refer- & Row.
ence frames in a landmark-based navigation task. J Neurosci Hamann SB, Squire LR (1997) Intact perceptual memory in the
16:823835. absence of conscious memory. Behav Neurosci 111:850854.
Gould SJ, Eldredge N (1977) Punctuated equilibria: the tempo and Hampton RR (2001) Rhesus monkeys know when they remember.
mode of evolution reconsidered. Paleobiology 3:115151. Proc Natl Acad Sci USA 98:53595362.
Graf P, Schacter DL (1985) Implicit and explicit memory for new Hampton RR, Schwartz BL (2004) Episodic memory in nonhumans:
associations in normal and amnesic subjects. J Exp Psychol Learn what, and where, is when? Curr Opin Neurobiol 14:192197.
Mem Cogn 11:501518. Hampton RR, Shettleworth SJ (1996) Hippocampal lesions impair
Graf P, Squire LR, Mandler G (1984) The information that amnesic memory for location but not color in passerine birds. Behav
patients do not forget. J Exp Psychol 10:164178. Neurosci 110:831835.
Graham KS, Simons JS, Pratt KH, Patterson K, Hodges JR Hampton RR, Healy SD, Shettleworth SJ, Kamil AC (2002)
(2000) Insights from semantic dementia on the relationship Neuroecologists are not made of straw. Trends Cogn Sci 6:
between episodic and semantic memory. Neuropsychologia 67.
38:313324. Hardiman MJ, Ramnani N, Yeo CH (1996) Reversible inactivities of
Grant SG, Odell TJ, Karl KA, Stein PL, Soriano P, Kandel ER the cerebellum with muscimol prevent the acquisition and
(1992) Impaired long-term potentiation, spatial learning, and extinction of conditioned nictitating membrame responses in
hippocampal development in fyn mutant mice. Science the rabbit. Exp Brain Res 110:235247.
258:19031910. Harker KT, Whishaw IQ (2004) A reaffirmation of the retrosplenial
Gray JA (1982) The neuropsychology of anxiety: an enquiry into the contribution to rodent navigation: reviewing the inuences of
functions of the septo-hippocampal system. Oxford, UK: Oxford lesion, strain, and task. Neurosci Biobehav Rev 28:485496.
University Press. Hartley T, Burgess N, Lever C, Cacucci F, OKeefe J (2000) Modeling
Gray JA (2000) The neuropsychology of anxiety: an enquiry into the place elds in terms of the cortical inputs to the hippocampus.
functions of the septo-hippocampal system, 2nd ed. Oxford: Hippocampus 10:369379.
Oxford University Press. Hasselmo ME, Howard E (2005) Hippocampal mechanisms for
Gregory RL (1961) The brain as an engineering problem. In: Current the context-dependent retrieval of episodes. Neural Netw
problems in animal behaviour (Thorpe WH, Zangwill OL, eds). 18:11721190.
Cambridge, UK: Cambridge University Press. Hasselmo ME, Bodelon C, Wyble BP (2002) A proposed function for
Griffiths D, Dickinson A, Clayton N (1999) Episodic memory: what hippocampal theta rhythm: separate phases of encoding and
can animals remember about their past? Trends Cogn Sci retrieval enhance reversal of prior learning. Neural Comput
3:7480. 14:793817.
Grillner S, Ekeberg, El Manira A, Lansner A, Parker D, Tegnr J, Hauser MD (2003) Knowing about knowing: dissociations between
Walln P (1998) Intrinsic function of a neuronal networka perception and action systems over evolution and during devel-
vertebrate central pattern generator. Brain Res Brain Res Rev opment. Ann N Y Acad Sci 1001:79103.
26:184197. Healy SD (1998) Spatial representations in animals. New York: Oxford
Guillot PV, Roubertoux PL, Crusio WE (1994) Hippocampal mossy University Press.
ber distributions and intermale aggression in seven inbred Healy SI, Braithwaite II (2000) Cognitive ecology: a eld of sub-
mouse strains. Brain Res 660:167169. stance? Trends Ecol Evol 15:2226.
Guzowski JF (2002) Insights into immediate-early gene function in Hebb DO (1949) The organization of behaviour. New York: Wiley.
hippocampal memory consolidation using antisense oligonu- Hebben N, Corkin S, Eichenbaum H, Shedlack K (1985) Dimini-
cleotide and uorescent imaging approaches. Hippocampus shed ability to interpret and report internal states after bilat-eral
12:86104. medial temporal resection: case H.M. Behav Neurosci
Guzowski JF, McNaughton BL, Barnes CA, Worley PF (1999) 99:10311039.
Environment-specic expression of the immediate-early gene Henke K, Buck A, Weber B, Wieser HG (1997) Human hippocampus
Arc in hippocampal neuronal ensembles. Nat Neurosci establishes associations in memory. Hippocampus 7:249256.
2:11201124. Herrera DG, Robertson HA (1996) Activation of c-fos in the brain.
Guzowski JF, Setlow B, Wagner EK, McGaugh JL (2001) Experience- Prog Neurobiol 50:83107.
dependent gene expression in the rat hippocampus after spatial Hess US, Lynch G, Gall CM (1995) Regional patterns of c-fos mRNA
learning: a comparison of the immediate-early genes Arc, c-fos, expression in rat hippocampus following exploration of a novel
and zif268. J Neurosci 21:50895098. environment versus performance of a well-learned discrimina-
Hafting T, Fyhn M, Molden S, Moser MB, Moser EI (2005) tion. J Neurosci 15:77967809.
Microstructure of a spatial map in the entorhinal cortex. Nature Hill AJ (1978) First occurrence of hippocampal spatial ring in a new
436:801806. environment. Exp Neurol 62:282297.
702 The Hippocampus Book

Hirsh R (1974) The hippocampus and contextual retrieval of infor- Jarrard LE (1989) On the use of ibotenic acid to lesion selectively dif-
mation from memory: a theory. Behav Biol 12:421444. ferent components of the hippocampal formation. J Neurosci
Hok V, Save E, Lenck-Santini PP, Poucet B (2005) Coding for spatial Methods 29:251259.
goals in the prelimbic/infralimbic area of the rat frontal cortex. Jarvis ED, Gunturkun O, Bruce L, Csillag A, Karten H, Kuenzel W,
Proc Natl Acad Sci USA 102:46024607. Medina L, Paxinos G, Perkel DJ, Shimizu T, Striedter G, Wild
Holland PC, Fox GD (2003) Effects of hippocampal lesions in JM, Ball GF, Dugas-Ford J, Durand SE, Hough GE, Husband S,
overshadowing and blocking procedures. Behav Neurosci Kubikova L, Lee DW, Mello CV, Powers A, Siang C, Smulders
117:650656. TV, Wada K, White SA, Yamamoto K, Yu J, Reiner A, Butler AB
Holland PC, Lamoroureux JA, Han JS, Gallagher M (1999) (2005) Avian brains and a new understanding of vertebrate
Hippocampal lesions interfere with negative occasion-setting. brain evolution. Nat Rev Neurosci 6:151159.
Hippocampus 9:143157. Jeffery KJ (2003) The neurobiology of spatial behaviour. New York:
Hollup SA, Kjelstrup KG, Hoff J, Moser MB, Moser EI (2001) Oxford University Press.
Impaired recognition of the goal location during spatial naviga- Jeffery KJ, Anderson MI, Hayman R, Chakraborty S (2004) A pro-
tion in rats with hippocampal lesions. J Neurosci 21:45054513. posed architecture for the neural representation of spatial con-
Honey RC, Good M (1993) Selective hippocampal lesions abolish the text. Neurosci Biobehav Rev 28:201218.
contextual specicity of latent inhibition and of conditioning. Jenkins TA, Amin E, Pearce JM, Brown MW, Aggleton JP (2004)
Behav Neurosci 107:2233. Novel spatial arrangements of familiar visual stimuli promote
Honey RC, Watt A, Good M (1998) Hippocampal lesions disrupt an activity in the rat hippocampal formation but not the parahip-
associative mismatch process. J Neurosci 18:22262230. pocampal cortices: a c-fos expression study. Neuroscience
Honig WK (1978) Studies of working memory in the pigeon. In: 124:4352.
Cognitive aspects of animal behavior (Hulse SH, Fowler HF, Jensen O, Lisman JE (1996) Theta/gamma networks with slow
Honig WK, eds), pp 211248. Hillside, NJ: Erlbaum. NMDA channels learn sequences and encode episodic memory:
Horel JA (1978) The neuroanatomy of amnesia: a critique of the hip- role of NMDA channels in recall. Learn Mem 3:2264278.
pocampal memory hypothesis. Brain 101:403445. Jung MW, Wiener SI, McNaughton BL (1994) Comparison of spatial
Horel JA (1994) Some comments on the special cognitive functions ring characteristics of units in dorsal and ventral hippocampus
claimed for the hippocampus. Cortex 30:269280. of the rat. J Neurosci 14:73477356.
Insausti R, Amaral DG, Cowan WM (1987a) The entorhinal cortex Kamin LJ (1968) Attention-like processes in classical conditioning. In:
of the monkey. III. Subcortical afferents. J Comp Neurol Miami symposium on the prediction of behavior: aversive stimula-
264:396408. tion (Jones MR, ed), pp 933. Miami: University of Miami Press.
Insausti R, Amaral DG, Cowan WM (1987b) The entorhinal cortex of Kaminski J, Call J, Fischer J (2004) Word learning in a domestic dog:
the monkey. II. Cortical afferents. J Comp Neurol 264:356395. evidence for fast mapping. Science 304:16821683.
Ioale P, Gagliardo A, Bingman VP (2000) Hippocampal participation Kandel ER, Schwartz JH, Jessell TM (2000) Principles of neural sci-
in navigational map learning in young homing pigeons is ence. New York: McGraw-Hill.
dependent on training experience. Eur J Neurosci 12:742750. Kao YC, Davis ES, Gabrieli JD (2005) Neural correlates of actual and
Iversen SD (1976) Do hippocampal lesions produce amnesia in ani- predicted memory formation. Nat Neurosci 8:17761783.
mals? Int Rev Neurobiol 19:149. Kapur N, Brooks DJ (1999) Temporally-specific retrograde
Jacobs LF (1995) The ecology of spatial cognition. In: Behavioural amnesia in two cases of discrete bilateral hippocampal pathol-
brain research in naturalistic and semi-naturalistic settings ogy. Hippocampus 9:247254.
(Alleva E, ed), pp 301322. Boston: Kluwer Academic. Kay RF, Ross C, Blythe A, Williams BA (1997) Anthropoid origins.
Jacobs LF (2003) The evolution of the cognitive map. Brain Behav Science 275:797804.
Evol 62:128139. Kelsey BA, Landry JE (1988) Medial septal lestions disrupt spatial
Jacobs LF, Schenk F (2003) Unpacking the cognitive map: the mapping ability in rats. Behav Neurosci 102:289293.
parallel map theory of hippocampal function. Psychol Rev Kempermann G, Brandon EP, Gage FH (1998) Environmental stim-
110:285315. ulation of 129/SvJ mice causes increased cell proliferation and
Jacobs LF, Spencer WD (1994) Natural space-use patterns and neurogenesis in the adult dentate gyrus. Curr Biol 8:939942.
hippocampal size in kangaroo rats. Brain Behav Evol 44: Kennedy PJ, Shapiro ML (2004) Retrieving memories via internal
125132. context requires the hippocampus. J Neurosci 24:69796985.
Jacobs LF, Gaulin SJ, Sherry DF, Hoffman GE (1990) Evolution Kentros C, Hargreaves E, Hawkins RD, Kandel ER, Shapiro M, Muller
of spatial cognition: sex-specic patterns of spatial behavior RV (1998) Abolition of long-term stability of new hippocam-
predict hippocampal size. Proc Natl Acad Sci USA 87: pal place cell maps by NMDA receptor blockade. Science
63496352. 280:21212126.
James W (1890) The principles of psychology, 1918 ed. New York: Kesner RP (1998) Neurobiological views of memory. In: Neurobiology
Dover. of learning and memory (Martinez JL, Kesner RP, eds), pp
Jarrard LE (1978) Selective hippocampal lesions: differential 361416. San Diego: Academic Press.
effects on performance by rats of a spatial task with preopera- Kesner RP (2000) Behavioral analysis of the contribution of the hip-
tive versus postoperative training. J Comp Physiol Psychol pocampus and parietal cortex to the processing of information:
92:11191127. interactions and dissociations. Hippocampus 10:483490.
Jarrard LE (1986) Selective hippocampal lesions and behavior: impli- Kesner RP, Bolland BL, Dakis M (1993) Memory for spatial locations,
cations for current research and theorizing. In: The hippocam- motor responses, and objects: triple dissociation among the hip-
pus (Isaacson RL, Pribram KH, eds), pp 93126. New York: pocampus, caudate nucleus, and extrastriate visual cortex. Exp
Plenum. Brain Res 93:462470.
 Theories of Hippocampal Function 703

Kiernan MJ, Westbrook RF (1993) Effects of exposure to a to-be- Transient memory impairment in monkeys with bilateral
shocked environment upon the rats freezing response: evidence lesions of the entorhinal cortex. J Neurosci 15:56375659.
for facilitation, latent inhibition and perceptual learning. Q J Leutgeb S, Leutgeb JK, Treves A, Moser MB, Moser EI (2004) Distinct
Exp Psychol 46B:271288. ensemble codes in hippocampal areas CA3 and CA1. Science
Kim JJ, Fanselow MS (1992) Modality-specic retrograde amnesia of 305:12951298.
fear. Science 256:675677. Lever C, Wills T, Cacucci F, Burgess N, OKeefe J (2002) Long-term
Kimble DP (1968) Hippocampus and internal inhibition. Psychol plasticity in hippocampal place-cell representation of environ-
Bull 70:285295. mental geometry. Nature 416:9094.
Kjelstrup KG, Tuvnes FA, Steffenach HA, Murison R, Moser Levy DA, Stark CE, Squire LR (2004) Intact conceptual priming in
EI, Moser MB (2002) Reduced fear expression after lesions the absence of declarative memory. Psychol Sci 15:680686.
of the ventral hippocampus. Proc Natl Acad Sci USA Levy WB (1996) A sequence predicting CA3 is a exible associator
99:1082510830. that learns and uses context to solve hippocampal-like tasks.
Knowlton BJ, Squire LR (1993) The learning of categories: parallel Hippocampus 6:579590.
brain systems for item memory and category knowledge. Science Lindner MD, Plone MA, Schallert T, Emerich DF (1997) Blind rats
262:17471749. are not profoundly impaired in the reference memory Morris
Knuttinen MG, Power JM, Preston AR, Disterhoft JF (2001) watermaze and cannot be clearly discriminated from rats with
Awareness in classical differential eyeblink conditioning in cognitive decits in the cued platform task. Brain Res Cogn
young and aging humans. Behav Neurosci 115:747757. Brain Res 5:329333.
Kowalska DM, Kusmierek P, Kosmal A, Mishkin M (2001) Neither Link W, Konietzko U, Kauselmann G, Krug M, Schwanke B, Frey U,
perirhinal/entorhinal nor hippocampal lesions impair Kuhl D (1995) Somatodendritic expression of an immediate
short-term auditory recognition memory in dogs. Neuroscience early gene is regulated by synaptic activity. Proc Natl Acad Sci
104:965978. USA 92:57345738.
Krebs JR, Sherry DF, Healy SD, Perry VH, Vaccarino AL (1989) Lipp HP, Wolfer DP (1998) Genetically modied mice and cognition.
Hippocampal specialization of food-storing birds. Proc Natl Curr Opin Neurobiol 8:272280.
Acad Sci USA 86:13881392. Lipp HP, Schwegler H, Heimrich B, Driscoll P (1988) Infrapyramidal
Krupa DJ, Thompson JK, Thompson RF (1993) Localization of a mossy bers and two-way avoidance learning: developmental
memory trace in the mammalian brain. Science 260:989991. modication of hippocampal circuitry and adult behavior of
Kumaran D, Maguire EA (2005) The human hippocampus: cognitive rats and mice. J Neurosci 8:19051921.
maps or relational memory? J Neurosci 25:72547259. Lipp HP, Collins RL, Hausheer-Zarmakupi Z, Leisinger-Trigona
LaBar KS, Disterhof JF (1998) Conditioning, awareness, and the hip- MC, Crusio WE, Nosten-Bertrand M, Signore P, Schwegler H,
pocampus. Hippocampus 8:620626. Wolfer DP (1996) Paw preference and intra/infrapyramidal
Lashley KS (1950) In search of the engram. In: Symposia for the mossy bers in the hippocampus of the mouse. Behav Genet
Society for Experimental Biology, pp 454482. New York: 26:379390.
Cambridge University Press. Lipp HP, Amrein I, Slominanka L, Wolfer DP (1999) Natural genetic
Lavenex P, Steele MA, Jacobs LF (2000) Sex differences, but no sea- variation of hippocampal structures and behavior. In: Cellular
sonal variations in the hippocampus of food-caching squirrels: and quantitative methods in neurogenetics: methods in life sci-
a stereological study. J Comp Neurol 425:152166. ences (Jones BC, Mormede P, eds), pp 217235. Boca Raton, FL:
LeDoux JE (2000) Emotion circuits in the brain. Annu Rev Neurosci CRC Press.
23:155184. Lipp HP, Vyssotski AL, Wolfer DP, Renaudineau S, Savini M, Troster
Lee AC, Bussey TJ, Murray EA, Saksida LM, Epstein RA, Kapur N, G, DellOmo G (2004) Pigeon homing along highways and exits.
Hodges JR, Graham KS (2005) Perceptual decits in amnesia: Curr Biol 14:12391249.
challenging the medial temporal lobe mnemonic view. Lisman JE, Zhabotinsky AM (2001) A model of synaptic memory:
Neuropsychologia 43:111. a Ca MKII/PP1 switch that potentiates transmission by
Lee DW, Miyasato LE, Clayton NS (1998) Neurobiological bases of organizing an AMPA receptor anchoring assembly. Neuron
spatial learning in the natural environment: neurogenesis and 31:191201.
growth in the avian and mammalian hippocampus. Neuroreport Lyford GL, Yamagata K, Kaufmann WE, Barnes CA, Sanders LK,
9:R1527. Copeland NG, Gilbert DJ, Jenkins NA, Lanahan AA, Worley PF
Lee DW, Smith GT, Tramontin AD, Soma KK, Brenowitz EA, (1995) Arc, a growth factor and activity-regulated gene, encodes
Clayton NS (2001) Hippocampal volume does not change sea- a novel cytoskeleton-associated protein that is enriched in neu-
sonally in a non food-storing songbird. Neuroreport 12:1925- ronal dendrites. Neuron 14:433445.
1928. Lynch G, Deadwyler S, Cotman C (1973) Postlesion axonal
Lee I, Kesner RP (2004) Differential contributions of dorsal hip- rowth produces permanent functional connections. Science
pocampal subregions to memory acquisition and retrieval in 180:13641366.
contextual fear-conditioning. Hippocampus 14:301310. Mackintosh NJ (1983) Conditioning and associative learning. Oxford,
Lenck-Santini PP, Save E, Poucet B (2001) Evidence for a relationship UK: Clarendon Press.
between place-cell spatial ring and spatial memory perform- Mackintosh NJ (2002) Do not ask whether they have a cognitive map,
ance. Hippocampus 11:377390. but how they nd their way about. Psicologica 23:165185.
Lenck-Santini PP, Muller RU, Save E, Poucet B (2002) Relationships Macphail EM (1982) Brain and intelligence in vertebrates. Oxford,
between place cell ring elds and navigational decisions by UK: Clarendon Press.
rats. J Neurosci 22:90359047. Macphail EM (1998) The evolution of consciousness. Oxford, UK:
Leonard BW, Amaral DG, Squire LR, Zola-Morgan S (1995) Oxford University Press.
704 The Hippocampus Book

Macphail EM (2002) The role of the avian hippocampus in spatial Martin SJ, Grimwood PD, Morris RGM (2000) Synaptic plasticity
memory. Psicologica 23:93108. and memory: an evaluation of the hypothesis. Annu Rev
Macphail EM, Bolhuis JJ (2001) The evolution of intelligence: adap- Neurosci 23:649711.
tive specializations versus general process. Biol Rev Camb Philos Martin SJ, de Hoz L, Morris RGM (2005) Retrograde amnesia: nei-
Soc 76:341364. ther partial nor complete hippocampal lesions in rats result in
Maguire EA (2001) Neuroimaging studies of autobiographical event preferential sparing of remote spatial memory, even after
memory. Philos Trans R Soc Lond B Biol Sci 356:14411451. reminding. Neuropsychologia 43:609624.
Maguire EA, Burgess N, Donnett JG, Frackowiak RS, Frith CD, Matsumura N, Nishijo H, Tamura R, Eifuku S, Endo S, Ono T (1999)
OKeefe J (1998) Knowing where and getting there: a human Spatial- and task-dependent neuronal responses during real and
navigation network. Science 280:921924. virtual translocation in the monkey hippocampal formation. J
Maguire EA, Frith CD, Morris RGM (1999) The functional neu- Neurosci 19:23812393.
roanatomy of comprehension and memory: the importance of Maviel T, Durkin TP, Menzaghi F, Bontempi B (2004) Sites of neo-
prior knowledge. Brain 122(Pt 10):18391850. cortical reorganization critical for remote spatial memory.
Malamut BL, Saunders RC, Mishkin M (1984) Monkeys with com- Science 305:9699.
bined amygdalo-hippocampal lesions succeed in object dis- Mayford M, Bach ME, Huang Y-Y, Wang L, Hawkins R, Kandel ER
crimination learning despite 24-hour intertrial intervals. Behav (1996) Control of memory formation through regulated
Neurosci 98:759769. expression of a CaMKII transgene. Science 274:16781683.
Malkova L, Mishkin M (2003) One-trial memory for object-place McCarthy RA, Warrington EA (1990) Cognitive neuropsychology. San
associations after separate lesions of hippocampus and Diego: Academic Press.
posterior parahippocampal region in the monkey. J Neurosci McClelland JL, Goddard NH (1996) Considerations arising from a
23:19561965. complementary learning systems perspective on hippocampus
Malkova L, Gaffan D, Murray EA (1997) Excitotoxic lesions of the and neocortex. Hippocampus 6:654665.
amygdala fail to produce impairment in visual learning for McClelland JL, McNaughton BL, OReilly RC (1995) Why there are
auditory secondary reinforcement but interfere with reinforcer complementary learning systems in the hippocampus and neo-
devaluation effects in rhesus monkeys. J Neurosci 17:60116020. cortex: insights from the successes and failures of connectionist
Malkova L, Lex CK, Mishkin M, Saunders RC (2001) MRI-based models of learning and memory. Psychol Rev 102:419457.
evaluation of locus and extent of neurotoxic lesions in monkeys. McDonald RJ, White NM (1994) Parallel information processing in
Hippocampus 11:361370. the watermaze: evidence for independent memory systems
Mandler G (1980) Recognizing: the judgement of previous occur- involving dorsal striatum and hippocampus. Behav Neural Biol
rence. Psychol Rev 87:252271. 61:260270.
Manns JR, Clark RE, Squire LR (2000a) Parallel acquisition of aware- McDonald RJ, White NM (1995) Hippocampal and nonhippocam-
ness and trace eyeblink classical conditioning. Learn Mem pal contributions to place learning in rats. Behav Neurosci
7:267272. 109:579593.
Manns JR, Stark CE, Squire LR (2000b) The visual paired- McDonald RJ, Murphy RA, Guarraci FA, Gortler JR, White NM,
comparison task as a measure of declarative memory. Proc Natl Baker AG (1997) Systematic comparison of the effects of hip-
Acad Sci USA 97:1237512379. pocampal and fornix-mbria lesions on acquisition of three
Manns JR, Hopkins RO, Reed JM, Kitchener EG, Squire LR (2003) congural discriminations. Hippocampus 7:371388.
Recognition memory and the human hippocampus. Neuron McDonald RJ, Devan BD, Hong NS (2004) Multiple memory
37:171180. systems: the power of interactions. Neurobiol Learn Mem
Maravita A, Iriki A (2004) Tools for the body (schema). Trends Cogn 82:333346.
Sci 8:7986. McEwen B, Sapolsky R (1995) Stress and cognitive function. Curr
Maren S (2001) Neurobiology of pavlovian fear conditioning. Annu Opin Neurobiol 5:205212.
Rev Neurosci 24:897931. McGaugh JL (2000) Memory-a century of consolidation. Science
Maren S (2005) Building and burying fear memories in the brain. 287:258251.
Neuroscientist 11:8999. McGlinchey-Berroth R, Carrillo MC, Gabrieli JDE, Brawn CM,
Maren S, Fanselow MS (1997) Electrolytic lesions of the m- Disterhof JF (1997) Impaired trace eyeblink conditioning
bria/fornix, dorsal hippocampus, or entorhinal cortex produce in bilateral, medial-temporal lobe amnesia. Behav Neurosci
anterograde decits in contextual fear conditioning in rats. 111:873882.
Neurobiol Learn Mem 67:142149. McGonigle BO, Chalmers M (1977) Are monkeys logical? Nature
Maren S, Aharonov G, Fanselow MS (1997) Neurotoxic lesions of the 267:694696.
dorsal hippocampus and pavlovian fear conditioning in rats. McGonigle BO, Chalmers M (2003) The growth of cognitive struc-
Behav Brain Res 88:261274. tures in monkeys and men. In: Animal cognition and sequential
Margules J, Gallistel CR (1988) Heading in the rat: determination by behavior: behavioral, biological and computational perspectives
environmental shape. Anim Learn Behav 16:404410. (Fountain SB, Bunsey MD, Danks JH, McBeath MK, eds).
Markman EM, Abelev M (2004) Word learning in dogs? Trends Cogn Boston: Kluwer Academic.
Sci 8:479481. McGregor A, Good MA, Pearce JM (2004a) Absence of an interaction
Marr D (1971) Simple memory: a theory for archicortex. Philos Trans between navigational strategies based on local and distal land-
R Soc Lond B: 262:2381. marks. J Exp Psychol Anim Behav Process 30:3444.
Martin A, Wiggs C, Weisberg J (1997) Modulation of human medial McGregor A, Hayward AJ, Pearce JM, Good MA (2004b) Hippocam-
temporal lobe activity by form, meaning and experience. pal lesions disrupt navigation based on the shape of the envi-
Hippocampus 7:587593. ronment. Behav Neurosci 118:10111021.
 Theories of Hippocampal Function 705

McHugh TJ, Blum KI, Tsien JZ, Tonegawa S, Wilson MA (1996) peptide vaccination prevents memory loss in an animal model
Impaired hippocampal representation of space in CA1-specic of Alzheimers disease. Nature 482:982986.
NMDAR1 knockout mice. Cell 87:13391349. Morgan JI, Cohen DR, Hempstead JL, Curran T (1987) Mapping pat-
McNaughton BL, Barnes CA (1986) Long-term enhancement of hip- terns of c-fos expression in the central nervous system after
pocampal synaptic transmission and the acquisition of spatial seizure. Science 237:192197.
information. J Neurosci 6:563571. Morris RG, Mayes AR (2004) Long-term spatial memory: introduc-
McNaughton BL, Morris RGM (1987) Hippocampal synaptic tion and guide to the special section. Neuropsychology
enhancement and information storage within a distributed 18:403404.
memory system. TINS 10:408415. Morris RGM (1981) Spatial localisation does not depend on the pres-
McNaughton BL, Barnes CA, Meltzer J, Sutherland RJ (1989) ence of local cues. Learn Motiv 12:239260.
Hippocampal granule cells are necessary for normal spatial Morris RGM (1983) Neural subsystems of exploration in rats. In:
learning but not for spatially-selective pyramidal cell discharge. Exploration in animals and humans (Archer J, Birke L, eds), pp
Exp Brain Res 76:485496. 117146. London: Van Nostrand Reinhold.
McNaughton BL, Chen LL, Markus EJ (1991) Dead reckoning, Morris RGM (1984) Developments of a watermaze procedure for
landmark learning, and the sense of direction: a neuro- studying spatial learning in the rat. J Neurosci Methods 11:4760.
physiological and computational hypothesis. J Cogn Neurosci Morris RGM (1991) Distinctive computations and relevant associa-
3:190202. tive processes: hippocampal role in processing, retrieval, but not
McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW, storage of allocentric spatial memory. Hippocampus 1:287290.
Knierim JJ, Kudrimoti H, Qin Y, Skaggs WE, Suster M, Weaver Morris RGM (2001) Episodic-like memory in animals: psychological
KL (1996) Deciphering the hippocampal polyglot: the hip- criteria, neural mechanisms and the value of episodic-like tasks
pocampus as a path integration system. J Exp Biol 199:173185. to investigate animal models of neurodegenerative disease.
McNaughton BL, Barnes CA, Battaglia FP, Bower MR, Cowen SL, Philos Trans R Soc Lond B Biol Sci 356:14531465.
Ekstrom AD, Gerrard JL, Hoffman KL, Houston FP, Karten Y, Morris RGM (2006) Elements of a neurobiological theory of the hip-
Lipa P, Pennartz CMA, Sutherland GR (2003) Off-line repro- pocampus: the role of activity-dependent synaptic plasticity in
cessing of recent memory and its role in memory consolidation: episodic-like memory. Eur J Neurosci 23:28292846.
a progress report. In: Sleep and synaptic plasticity (Smith C, Morris RGM, Frey U (1997) Hippocampal synaptic plasticity: role in
Maquet P, eds), pp 225246. New York: Oxford University Press. spatial learning or the automatic recording of attended experi-
Meunier M, Bachevalier J, Mishkin M, Murray EA (1993) Effects on ence? Philos Trans R Soc Lond B Biol Sci 352:14891503.
visual recognition of combined and separate ablations of the Morris RGM, Kennedy MB (1992) The pierian spring. Curr Biol
entorhinal and perirhinal cortex in rhesus monkeys. J Neurosci 2:511514.
13:54185432. Morris RGM, Garrud P, Rawlins JN, OKeefe J (1982) Place naviga-
Mishkin M (1978) Memory in monkeys severely impaired by com- tion impaired in rats with hippocampal lesions. Nature
bined but not by separate removal of amygdala and hippocam- 297:681683.
pus. Nature 273:297298. Morris RGM, Hagan JJ, Rawlins JN (1986a) Allocentric spatial learn-
Mishkin M (1982) A memory system in the monkey. Philos Trans R ing by hippocampectomised rats: a further test of the spatial
Soc Lond B Biol Sci 298:8395. mapping and working memory theories of hippocampal
Mishkin M, Delacour J (1975) An analysis of short-term visual mem- function. Q J Exp Psychol B 38:365395.
ory in the monkey. J Exp Psychol Anim Behav Process 1:326334. Morris RGM, Anderson E, Lynch GS, Baudry M (1986b) Selective
Mishkin M, Murray EA (1994) Stimulus recognition. Curr Opin impairment of learning and blockade of long-term potentiation
Neurobiol 4:200206. by an N-methyl-D-aspartate receptor antagonist, AP5. Nature
Mishkin M, Suzuki WA, Gadian DG, Vargha-Khadem F (1997) 319:774776.
Hierarchical organization of cognitive memory. Philos Trans Morris RGM, Schenk F, Tweedie F, Jarrard LE (1990) Ibotenate
Royal Soc Lond B Biol Sci 352:14611467. lesions of hippocampus and/or subiculum: dissociating compo-
Mishkin M, Vargha-Khadem F, Gadian DG (1998) Amnesia and nents of allocentric spatial learning. Eur J Neurosci 2:10161028.
the organization of the hippocampal system. Hippocampus Morris RGM, Martin SJ, Moser EI, Riedel G, Sandin J, Day M,
8:212216. OCarroll C (2003) Elements of a neurobiological theory of the
Mittelstaedt H, Mittelstaedt M (1982) Homing by path integration. hippocampus: the role of activity-dependent synaptic plasticity
In: Avian navigation (Papi F, Wallraff HG, eds), pp 290297. in memory. Philos Trans Roy Soc Lond B 358:773786.
Heidelberg: Springer-Verlag. Moscovitch M (1995) Recovered consciousness: a hypothesis
Miyashita Y (2004) Cognitive memory: cellular and network concerning modularity and episodic memory. J Clin Exp
machineries and their top-down control. Science 306:435440. Neuropsychol 17:276290.
Mizumori SJ, Miya DY, Ward KE (1994) Reversible inactivation of the Moscovitch M, Nadel L (1998) Consolidation and the hippocampal
lateral dorsal thalamus disrupts hippocampal place representa- complex revisited: in defense of the multiple-trace model. Curr
tion and impairs spatial learning. Brain Res 644:168174. Opin Neurobiol 8:297300.
Moita MA, Rosis S, Zhou Y, LeDoux JE, Blair HT (2003) Moser E, Moser MB, Anderson P (1993) Spatial learning impair-
Hippocampal place cells acquire location-specic responses to ment parallels the magnitude of dorsal hippocampal lesions,
the conditioned stimulus during auditory fear conditioning. but is hardly present following ventral lesions. J Neurosci
Neuron 37:485497. 13:39163925.
Morgan DG, Diamond DM, Gottschall PE, Ugen KE, Dickey C, Moser EI, Krobert KA, Moser MB, Morris RGM (1998) Impaired spa-
Hardy JD, Duff K, Jantzen P, DiCarlo G, Wilcock D, Connor K, tial learning after saturation of long-term potentiation. Science
Hatcher J, Hope C, Gordon MN, Arendash GW (2000) A beta 281:20382042.
706 The Hippocampus Book

Moser MB, Moser EI (1998a) Distributed encoding and retrieval of spatial scene learning and object reversals. Behav Neurosci
spatial memory in the hippocampus. J Neurosci 18:75357542. 112:12911303.
Moser MB, Moser EI (1998b) Functional differentiation in the hip- Murray EA, Graham KS, Gaffan D (2005) Perirhinal cortex and its
pocampus. Hippocampus 8:608619. neighbours in the medial temporal lobe: contributions to mem-
Moser MB, Moser EI, Forrest E, Andersen P, Morris RGM (1995) ory and perception. Q J Exp Psychol 58:378396.
Spatial learning with a minislab in the dorsal hippocampus. Proc Nadel L (1991) The hippocampus and space revisited. Hippocampus
Natl Acad Sci USA 92:96979701. 1:221229.
Moses SN, Ryan JD (2006) A comparison and evaluation of the pre- Nadel L (1995) The role of the hippocampus in declarative memory:
dictions of relational and conjunctive accounts of hippocampal a comment on Zola-Morgan, Squire, and Ramus (1994).
function. Hippocampus 16:4365. Hippocampus 5:232239.
Moss M, Mahut H, Zola-Morgan S (1981) Concurrent discrimina- Nadel L, Bohbot V (2001) Consolidation of memory. Hippocampus
tion learning of monkeys after hippocampal, entorhinal, or 11:5660.
fornix lesions. J Neurosci 1:227240. Nadel L, Hardt O (2004) The spatial brain. Neuropsychology
Moyer JR, Thompson LT, Disterhoft JF (1996) Trace eyeblink condi- 18:473476.
tioning increases CA1 excitability in a transient and learning- Nadel L, Moscovitch M (1997) Memory consolidation, retrograde
specic manner. J Neurosci 16:55365546. amnesia and the hippocampal complex. Curr Opin Neurobiol
Moyer JR, Power JM, Thompson LT, Disterhoft JF (2000) Increased 7:217227.
excitability of aged rabbit CA1 neurons after trace eyeblink con- Nadel L, Moscovitch M (1998) Hippocampal contributions to corti-
ditioning. J Neurosci 20:54765482. cal plasticity. Neuropharmacology 37:431439.
Muller R (1996) A quarter of a century of place cells. Neuron Nadel L, Willner J (1980) Context and conditioning: a place for space.
17:813822. Physiol Psychol 8:218228.
Mumby DG (2001) Perspectives on object-recognition memory fol- Nadel L, OKeefe J, Black AH (1975) Slam on the brakes: a critique of
lowing hippocampal damage: lessons from studies in rats. Behav Altman, Brunner and Bayers response-inhibition model of hip-
Brain Res 127:159181. pocampal function. Behav Biol 14:151162.
Mumby DG, Pinel JPJ, Wood ER (1990) Nonrecurring-items delayed Nadel L, Samsonovich A, Ryan L, Moscovitch M (2000) Multiple
nonmatching-to-sample in rats: a new paradigm for testing trace theory of human memory: computational, neuroimaging,
nonspatial working memory. Psychobiology 18:321326. and neuropsychological results. Hippocampus 10:352368.
Mumby DG, Wood ER, Pinel JP (1992a) Object-recognition memory Nader K (2003) Memory traces unbound. Trends Neurosci 26:6572.
is only mildly impaired in rats with lesions of the hippocampus Nakazawa K, Quirk MC, Chitwood RA, Watanabe M, Yeckel MF, Sun
and amygdala. Psychobiology 20:1827. LD, Kato A, Carr CA, Johnston D, Wilson MA, Tonegawa S
Mumby DG, Wood ER, Pinel JPJ (1992b) Object recognition mem- (2002) Requirement for hippocampal CA3 NMDA receptors in
ory in rats is only mildly impaired by lesions of the hippocam- associative memory recall. Science 297:211218.
pus and amygdala. Psychobiology 20:1827. Nakazawa K, Sun LD, Quirk MC, Rondi-Reig L, Wilson MA, Tone-
Mumby DG, Pinel JPJ, Kornecook TJ, Shen MJ, Redila VA (1995) gawa S (2003) Hippocampal CA3 NMDA receptors are crucial
Memory decits following lesions of hippocampus or amygdala for memory acquisition of one-time experience. Neuron
in rats: an assesment by an object-memory tets battery. 38:305315.
Psychobiology 23:2636. Nakazawa K, McHugh TJ, Wilson MA, Tonegawa S (2004) NMDA
Mumby DG, Astur RS, Weisend MP, Sutherland RJ (1999) Retrograde receptors, place cells and hippocampal spatial memory. Nat Rev
amnesia and selective damage to the hippocampal formation: Neurosci 5:361372.
memory for places and object discriminations. Behav Brain Res Nelson W (1988) Food restriction, circadian disorder and longevity
106:97107. of rats and mice. J Nutr 118:286289.
Murray EA, Bussey TJ (1999) Perceptual-mnemonic functions of the Nemanic S, Alvarado MC, Bachevalier J (2004) The hippocampal/
perirhinal cortex. Trends Cogn Sci 3:142151. parahippocampal regions and recognition memory: insights
Murray EA, Bussey TJ (2001) Consolidation and the medial tempo- from visual paired comparison versus object-delayed non-
ral lobe revisited: methodological considerations. Hippocampus matching in monkeys. J Neurosci 24:20132026.
11:17. Nunn JA, LePeillet E, Netto CA, Hodges H, Gray JA, Meldrum BS
Murray EA, Mishkin M (1983) Severe tactual memory decits in (1994) Global ischaemia: hippocampal pathology and spatial
monkeys after combined removal of the amygdala and hip- decits in the watermaze. Behav Brain Res 62:4154.
pocampus. Brain Res 270:340344. Ogden JA, Corkin S (1991) Memories of H.M. In: Memory mecha-
Murray EA, Mishkin M (1985) Amygdalectomy impairs crossmodal nisms: a tribute to G.V. Goddard (Abraham WC, Corballis M,
association in monkeys. Science 228:604605. White KG, eds), pp 195215. Hillsdale, NJ: Erlbaum,
Murray EA, Mishkin M (1998) Object recognition and location Ohno M, Frankland PW, Chen AP, Costa RM, Silva AJ (2001)
memory in monkeys with excitotoxic lesions of the amygdala Inducible, pharmacogenetic approaches to the study of learning
and hippocampus. J Neurosci 18:65686582. and memory. Nat Neurosci 4:12381243.
Murray EA, Wise SP (2004) What, if anything, is the medial tempo- OKeefe J (1976) Place units in the hippocampus of the freely mov-
ral lobe, and how can the amygdala be part of it if there is no ing rat. Exp Neurol 51:78109.
such thing? Neurobiol Learn Mem 82:178198. OKeefe J (1996) The spatial prepositions in English, vector grammar
Murray EA, Gaffan D, Mishkin M (1993) Neural substrates of visual and the cognitive map theory. In: Language and space (Bloom P,
stimulus-stimulus association in rhesus monkeys. J Neurosci Peterson MA, Nadel L, Garrett MF, eds), pp 277316.
13:45494561. Cambridge, MA: MIT Press.
Murray EA, Baxter MG, Gaffan D (1998) Monkeys with rhinal cortex OKeefe J (2005) Vector grammar, places, and the functional role of
damage or neurotoxic hippocampal lesions are impaired on the spatial prepositions in English. In: Representing direction in
 Theories of Hippocampal Function 707

language and space (van de Zee E, Slack J, eds), pp 6985. Parslow DM, Rose D, Brooks B, Fleminger S, Gray JA, Giampietro V,
Oxford, UK: Oxford University Press. Brammer MJ, Williams S, Gasston D, Andrew C, Vythelingum
OKeefe J, Burgess N (1996) Geometric determinants of the place GN, Loannou G, Simmons A, Morris RG (2004) Allocentric
elds of hippocampal neurons. Nature 381:425428. spatial memory activation of the hippocampal formation meas-
OKeefe J, Conway DH (1978) Hippocampal place units in the freely ured with fMRI. Neuropsychology 18:450461.
moving rat: why they re where they re. Exp Brain Res Patel SN, Clayton NS, Krebs JR (1997) Spatial learning induces neu-
31:573590. rogenesis in the avian brain. Behav Brain Res 89:115128.
OKeefe J, Dostrovsky J (1971) The hippocampus as a spatial map: Paulsen O, Moser EI (1998) A model of hippocampal memory
preliminary evidence from unit activity in the freely-moving encoding and retrieval: GABAergic control of synaptic plastic-
rat. Brain Res 34:171175. ity. Trends Neurosci 21:273278.
OKeefe J, Nadel L (1978) The hippocampus as a cognitive map. Pearce JM, Bouton ME (2001) Theories of associative learning in ani-
Oxford, UK: Clarendon Press. mals. Annu Rev Psychol 52:111139.
OKeefe J, Nadel L (1979) Precis of OKeefe & Nadels The Pearce JM, Hall G (1980) A model for pavlovian learning: variations
Hippocampus as a cognitive map. Behav Brain Sci 2:487533. in the effectiveness of conditioned but not of unconditioned
OKeefe J, Speakman A (1987) Single unit activity in the rat hip- stimuli. Psychol Rev 87:532552.
pocampus during a spatial memory task. Exp Brain Res 68:127. Pearce JM, Roberts AD, Good M (1998) Hippocampal lesions disrupt
OKeefe J, Nadel L, Keightley S, Kill D (1975) Fornix lesions selec- navigation based on cognitive maps but not heading vectors.
tively abolish place learning in the rat. Exp Neurol 48:152166. Nature 396:7577.
Olton DS (1977) The function of septo-hippocampal connections in Pearce JM, Good MA, Jones PM, McGregor A (2004) Transfer of
spatially organized behaviour. Ciba Found Symp 58:327349. spatial behavior between different environments: implications
Olton DS, Papas BC (1979) Spatial memory and hippocampal func- for theories of spatial learning and for the role of the hip-
tion. Neuropsychologia 17:669682. pocampus in spatial learning. J Exp Psychol Anim Behav Process
Olton DS, Samuelson RJ (1976) Remembrance of places passed: spa- 30:135147.
tial memory in rats. J Exp Psychol Anim Behav Process 2:97116. Phillips RG, LeDoux JE (1992) Differential contribution of amygdala
Olton DS, Becker JT, Handelmann GE (1979) Hippocampus, space, and hippocampus to cued and contextual for conditioning.
and memory. Brain Behav Sci 2:313365. Behav Neurosci 106:274285.
OReilly RC (1998) Six principles for biologically-based computa- Phillips RG, LeDoux JE (1994) Lesions of the dorsal hippocampal
tional models of cortical cognition. Trends Cogn Sci 2:455462. formation interfere with background but not foreground con-
OReilly RC, Norman KA (2002) Hippocampal and neocortical con- textual fear conditioning. Learn Mem 1:3444.
tributions to memory: advances in the complementary learning Pickens CL, Holland PC (2004) Conditioning and cognition.
systems framework. Trends Cogn Sci 6:505510. Neurosci Biobehav Rev 28:651661.
OReilly RC, Rudy JW (2001) Conjunctive representations in learning Pinker S (1994) The language instinct. New York: William Morrow.
and memory: principles of cortical and hippocampal function. Pitkanen A, Pikkarainen M, Nurminen N, Ylinen A (2000) Reciprocal
Psychol Rev 108:311345. connections between the amygdala and the hippocampal for-
Otto T, Eichenbaum H (1992) Complementary roles of the orbital mation, perirhinal cortex, and postrhinal cortex in rat: a review.
prefrontal cortex and the perirhinal-entorhinal cortices in an Ann N Y Acad Sci 911:369391.
odor-guided delayed-nonmatching-to-sample task. Behav Poldrack RA, Clark J, Pare-Blagoev EJ, Shohamy D, Creso Moyano J,
Neurosci 106:762775. Myers C, Gluck MA (2001) Interactive memory systems in the
Overman WH, Ormsby G, Mishkin M (1990) Picture recognition vs. human brain. Nature 414:546550.
picture discrimination learning in monkeys with medial tempo- Poucet B (1993) Spatial cognitive maps in animals: new hypotheses
ral removals. Exp Brain Res 79:1824. on their structure and neural mechanisms. Psychol Rev
Overton DA (1964) State-dependent or dissociation learning pro- 100:163182.
duced with pentobarbital. J Comp Physiol Psychol 57:312. Poucet B, Benhamou S (1997) The neuropsychology of spatial cogni-
Owen MJ, Butler SR (1981) Amnesia after transection of the fornix in tion in the rat. Crit Rev Neurobiol 11:101120.
monkeys: long-term memory impaired, short-term memory Poucet B, Lenck-Santini PP, Paz-Villagran V, Save E (2003) Place cells,
intact. Behav Brain Res 3:115123. neocortex and spatial navigation: a short review. J Physiol Paris
Packard MG, McGaugh JL (1996) Inactivation of hippocampus or 97:537546.
caudate nucleus with lidocaine differentially affects expression Pravosudov VV, Clayton NS (2002) A test of the adaptive specializa-
of place and response learning. Neurobiol Learn Mem 65:6572. tion hypothesis: population differences in caching, memory,
Papadimitriou A, Wynne C (1999) Preserved negative patterning and and the hippocampus in black-capped chickadees (Poecile atri-
impaired spatial learning in pigeons (Columba livia) with capilla). Behav Neurosci 116:515522.
lesions of the hippocampus. Behav Neurosci 113:683690. Prusky GT, West PW, Douglas RM (2000a) Behavioral assessment of
Papi F (1990) Olfactory navigation in birds. Experientia 46:352363. visual acuity in mice and rats. Vision Res 40:22012209.
Papini MR (2002) Pattern and process in the evolution of learning. Prusky GT, West PW, Douglas RM (2000b) Reduced visual acuity
Psychol Rev 109:186201. impairs place but not cued learning in the Morris water task.
Parkinson JK, Murray EA, Mishkin M (1988) A selective mnemonic Behav Brain Res 116:135140.
role for the hippocampus in monkeys: memory for the location Quirk GJ, Muller RU, Kubie JL, Ranck JB (1992) The positional ring
of objects. J Neurosci 8:41594167. properties of medial entorhinal neurons: description and com-
Parron C, Save E (2004) Comparison of the effects of entorhinal and parison with hippocampal place cells. J Neurosci 12:19451963.
retrosplenial cortical lesions on habituation, reaction to spatial Ramos JM (1998) Retrograde amnesia for spatial information: a dis-
and non-spatial changes during object exploration in the rat. sociation between intra- and extramaze cues following hip-
Neurobiol Learn Mem 82:111. pocampus lesions in rats. Eur J Neurosci 10:32953301.
708 The Hippocampus Book

Rawlins JN, Lyford GL, Seferiades A, Deacon RM, Cassaday HJ Rodrigo T (2002) Navigational strategies and models. Psicologica
(1993) Critical determinants of nonspatial working memory 23:232.
decits in rats with conventional lesions of the hippocampus or Rodrigo T, Chamizo VD, McLaren IPL, Mackintosh NJ (1997)
fornix. Behav Neurosci 107:420433. Blocking in the spatial domain. J Exp Psychol Anim Behav Process
Rawlins JNP (1985) Associations across time: the hippocampus as a 23:110118.
temporary memory store. Behav Brain Sci 8:479497. Rolls ET, Treves A (1998) Neural networks and brain function. Oxford,
Redish AD (1999) Beyond the cognitive map: from place cells to UK: Oxford University Press.
episodic memory. Cambridge, MA: MIT Press. Rosenbaum RS, Winocur G, Moscovitch M (2001) New views on old
Reisel D, Bannerman DM, Schmitt WB, Deacon RM, Flint J, memories: re-evaluating the role of the hippocampal complex.
Borchardt T, Seeburg PH, Rawlins JN (2002) Spatial memory Behav Brain Res 127:183197.
dissociations in mice lacking GluR1. Nat Neurosci 5:868873. Rosenzweig ES, Redish AD, McNaughton BL, Barnes CA (2003)
Remondes M, Schuman EM (2004) Role for a cortical input to hip- Hippocampal map realignment and spatial learning. Nat
pocampal area CA1 in the consolidation of a long-term mem- Neurosci 6:609615.
ory. Nature 431:699703. Ross RT, Orr WB, Holland PC, Berger TW (1984) Hippocampectomy
Rempel-Clower NL, Zola SM, Squire LR, Amaral DG (1996) Three disrupts acquisition and retention of learned conditional
cases of enduring memory impairment after bilateral damage responding. Behav Neurosci 98:211225.
limited to the hippocampal formation. J Neurosci 16:52335255. Rossier J, Kaminsky Y, Schenk F, Bures J (2000) The place
Renner MJ (1990) Neglected aspects of exploratory and investigatory preference task: a new tool for studying the relation between
behavior. Psychobiology 18:1622. behavior and place cell activity in rats. Behav Neurosci
Rescorla RA (1973) Effect of US habituation following conditioning. 114:273284.
J Comp Physiol Psychol 82:137143. Rotenberg A, Mayford M, Hawkins RD, Kandel ER, Muller RU (1996)
Rescorla RA (1988) Pavlovian conditioning: its not what you think it Mice expressing activated CaMKII lack low frequency LTP and
is. Am Psychol 43:151160. do not form stable place cells in the CA1 region of the hip-
Rescorla RA, Wagner AR (1972) A theory of pavlovian conditioning: pocampus. Cell 87:13511361.
the effectiveness of of reinforcement and nonreinforcement. In: Rotenberg A, Abel T, Hawkins RD, Kandel ER, Muller RU (2000)
Classical conditioning. II. Current research and theory (Black AH, Parallel instabilities of long-term potentiation, place cells, and
Prokasy WF, eds), pp 6499. New York: Appleton-Century- learning caused by decreased protein kinase A activity. J
Crofts. Neurosci 20:80968102.
Ressler KJ, Paschall G, Zhou X-L, Davis M (2002) Regulation of Rothblat LA, Hayes LL (1987) Short-term object recognition mem-
synaptic plasticity genes during consolidation of fear condition- ory in the rat: nonmatching with trial-unique junk stimuli.
ing. J Neurosci 22:78927902. Behav Neurosci 101:587590.
Richmond MA, Yee BK, Pouzet B, Veenman L, Rawlins JNP, Feldon J, Roullet P, Lassalle JM (1992) Behavioural strategies, sensorial
Bannerman DM (1999) Dissociating context and space within processes and hippocampal mossy bre distribution in radial
the hippocampus: effects of complete, dorsal, and ventral exci- maze performance in mice. Behav Brain Res 48:7785.
totoxic hippocampal leesions on conditioned freezing and spa- Rudy JW, Sutherland RJ (1995) Congural association theory and the
tial learning. Behav Neurosci 113:11891203. hippocampal formation: an appraisal and reconguration.
Riedel G, Micheau J, Lam AG, Roloff E, Martin SJ, Bridge H, Hoz L, Hippocampus 5:375389.
Poeschel B, McCulloch J, Morris RGM (1999) Reversible neural Rudy JW, OReilly RC (1999) Contextual fear conditioning, conjunc-
inactivation reveals hippocampal participation in several mem- tive representations, pattern completion, and the hippocampus.
ory processes. Nat Neurosci 2:898905. Behav Neurosci 113:867880.
Ringo JL (1991) Memory decays at the same rate in macaques with Rudy JW, Barrientos RM, OReilly RC (2002) Hippocampal forma-
and without brain lesions when expressed in d or arcsine terms. tion supports conditioning to memory of a context. Behav
Behav Brain Res 42:123134. Neurosci 116:530538.
Ringo JL (1993) Spared short-term memory in monkeys following Rumelhart DE, McClelland JL (1986) Parallel distributed processing:
medial temporal lobe lesions is not yet established: a reply to explorations in the microstructure of cognition. Cambridge, MA:
Alvarez-Royo, Zola-Morgan and Squire. Behav Brain Res Bradford Books.
59:6572. Rupniak NMJ, Gaffan D (1987) Monkey hippocampus and
Rizzolatti G, Craighero L (2004) The mirror-neuron system. Annu learning about spatially directed movements. J Neurosci
Rev Neurosci 27:169192. 7:23312337.
Robbins TW, James M, Owen AM, Sahakian BJ, McInnes L, Rabbitt P Ryle G (1949) The concept of mind. London: Hutchinson.
(1994) Cambridge Neuropsychological Test Automated Battery Sahgal A (1993) Behavioural neuroscience: a practical approach,.
(CANTAB): a factor analytic study of a large sample of normal Oxford: IRL Press.
elderly volunteers. Dementia 5:266281. Sakai K, Miyashita Y (1991) Neural organization for the long-term
Roberts WA (2000) Are animals stuck in time? Psychol Bull memory of paired associates. Nature 354:152155.
128:473489. Salmon DP, Zola-Morgan S, Squire LR (1987) Retrograde amnesia
Robertson RG, Rolls ET, Georges-Francois P (1998) Spatial view cells following combined hippocampus-amygdala lesions in mon-
in the primate hipocampus: effects of removal of view details. J keys. Psychobiology 15:3747.
Neurophysiol 79:11451156. Sandi C, Loscertales M, Guaza C (1997) Experience-dependent facil-
Robinson L, Bridge H, Riedel G (2001) Visual discrimination learn- itating effect of corticosterone on spatial memory formation in
ing in the watermaze: a novel test for visual acuity. Behav Brain the watermaze. Eur J Neurosci 9:637642.
Res 119:7784. Sapolsky RM (1985) A mechanism for glucocorticoid toxicity in the
 Theories of Hippocampal Function 709

hippocampus: increased neuronal vulnerability to metabolic Shettleworth SJ (1998) Cognition, evolution, and behavior. New York:
insults. J Neurosci 5:12281232. Oxford University Press.
Saunders RC, Weiskrantz L (1989) The effects of fornix transection Shettleworth SJ (2001) Animal cognition and animal behaviour.
and combined fornix transection, mammillary body lesions and Anim Behav 61:277286.
hippocampal ablations on object-pair association memory in Shettleworth SJ (2003) Memory and hippocampal specialization in
the rhesus monkey. Behav Brain Res 35:8594. food-storing birds: challenges for research on comparative cog-
Save E, Poucet B, Foreman N, Buhot MC (1992a) Object exploration nition. Brain Behav Evol 62:108116.
and reactions to spatial and nonspatial changes in hooded rats Shettleworth SJ, Krebs JR (1982) How marsh tits nd their hoards:
following damage to parietal cortex or hippocampal formation. the roles of site preference and spatial memory. J Exp Psychol
Behav Neurosci 106:447456. Anim Behav Process 8:354375.
Save E, Buhot MC, Foreman N, Thinus-Blanc C (1992b) Exploratory Shiett MW, Smulders TV, Benedict L, DeVoogd TJ (2003) Reversible
activity and response to a spatial change in rats with hippocam- inactivation of the hippocampal formation in food-storing
pal or posterior parietal cortical lesions. Behav Brain Res black-capped chickadees (Poecile atricapillus). Hippocampus
47:113127. 13:437444.
Schacter DE, Tulving E (1994) Memory systems, pp 269310. Shimamura AP, Squire LR (1987) A neuropsychological study of fact
Cambridge, MA: MIT Press. memory and source amnesia. J Exp Psychol 13:464473.
Schenk F, Morris RGM (1985) Dissociation between components of Siapas AG, Wilson MA (1998) Coordinated interactions between hip-
spatial memory in rats after recovery from the effects of retro- pocampal ripples and cortical spindals during slow-wave sleep.
hippocampal lesions. Exp Brain Res 58:1128. Neuron 21:11231128.
Schmajuk NA (1984) Psychological theories of hippocampal func- Silva AJ, Paylor R, Wehner JM, Tonegawa S (1992) Impaired spatial
tion. Physiol Psychol 12:166183. learning in alpha-calcium-calmodulin kinase II mutant mice.
Schmitt WB, Sprengel R, Mack V, Draft RW, Seeburg PH, Deacon Science 257:206211.
RM, Rawlins JN, Bannerman DM (2005) Restoration of spatial Sirota A, Csicsvari J, Buhl D, Buzsaki G (2003) Communication
working memory by genetic rescue of GluR-A-decient mice. between neocortex and hippocampus during sleep in rodents.
Nat Neurosci 8:270272. Proc Natl Acad Sci USA 100:20652069.
Schwegler H, Lipp HP (1981) Is there a correlation between hip- Skelton RW, McNamara RK (1992) Bilateral knife cuts to the per-
pocampal mossy ber distribution and two-way avoidance per- forant path disrupt spatial learning in the Morris watermaze.
formance in mice and rats? Neurosci Lett 23:2530. Hippocampus 2:7380.
Schwegler H, Lipp HP (1983) Hereditary covariations of neu- Smith ML, Milner B (1981) The role of the right hippocampus in the
ronal circuitry and behavior: correlations between the recall of spatial location. Neuropsychologia 19:781793.
proportions of hippocampal synaptic elds in the regio infe- Smulders TV, Sasson AD, DeVoogd TJ (1995) Seasonal variation in
rior and two-way avoidance in mice and rats. Behav Brain Res hippocampal volume in a food-storing bird, the black-capped
7:138. chickadee. J Neurobiol 27:1525.
Schwegler H, Crusio WE, Brust I (1990) Hippocampal mossy bers Smulders TV, Shiett MW, Sperling AJ, DeVoogd TJ (2000) Seasonal
and radial-maze learning in the mouse: a correlation with spa- changes in neuron numbers in the hippocampal formation of a
tial working memory but not with non-spatial reference mem- food-hoarding bird: the black-capped chickadee. J Neurobiol
ory. Neuroscience 34:293298. 44:414422.
Schwegler H, Crusio WE, Lipp HP, Brust I, Mueller GG (1991) Early Sole LM, Shettleworth SJ, Bennett PJ (2003) Uncertainty in pigeons.
postnatal hyperthyroidism alters hippocampal circuitry and Psychon Bull Rev 10:738745.
improves radial-maze learning in adult mice. J Neurosci Spooner RIW, Thomson A, Hall J, Morris RGM, Salter SH (1994) The
11:21022106. Atlantis Platform: A new design and further developments of
Selden NRW, Everitt BJ, Jarrard LE, Robbins TW (1991) Buresovas on-demand platform for the watermaze. Learn Mem
Complementary roles for the amygdala and hippocampus in 1:203211.
aversive conditioning to explicit and contextual cues. Squire LR (1992) Memory and the hippocampus: a synthesis from
Neuroscience 42:335350. findings with rats, monkeys, and humans. Psychol Rev
Seyfarth RM, Cheney DL, Marler P (1980) Monkey responses to three 99:195231.
different alarm calls: evidence of predator classication and Squire LR (1993) The hippocampus and spatial memory. TINS
semantic communication. Science 210:801803. 16:5657.
Shallice T (1988) From neuropsychology to mental structure. New Squire LR (2004) Memory systems of the brain: a brief history and
York: Cambridge University Press. current perspective. Neurobiol Learn Mem 82:171177.
Sherry DF, Schacter DL (1987) The evolution of multiple memory Squire LR, Alvarez P (1995) Retrograde amnesia and memory con-
systems. Psychol Rev 94:439454. solidation: a neurobiological perspective. Curr Opin Neurobiol
Sherry DF, Vaccarino AL (1989) Hippocampus and memory for food 5:169177.
caches in black-capped chickadees. Behav Neurosci Squire LR, Cohen NJ (1984) Human memory and amnesia.
103:308318. In: Neurobiology of learning and memory (Lynch G, McGaugh
Sherry DF, Vaccarino AL, Buckenham K, Herz RS (1989) The hip- JL, Weinberger NM, eds), pp 364. New York: Gilford
pocampal complex of food-storing birds. Brain Behav Evol Press.
34:308317. Squire LR, Zola SM (1998) Episodic memory, semantic memory and
Sherry DF, Jacobs LF, Gaulin SJ (1992) Spatial memory and amnesia. Hippocampus 8:205211.
adaptive specialization of the hippocampus. Trends Neurosci Squire LR, Zola-Morgan S (1983) The neurology of memory: the case
15:298303. for correspondence between the ndings for human and non-
710 The Hippocampus Book

human primate. In: The physiological basis of memory, 2nd ed. Suzuki S, Augerinos G, Black AH (1980) Stimulus control of spatial
(Deutsch JA, ed). San Diego: Academic Press. behavior on the eight-arm maze in rats. Learn Motiv 18.
Squire LR, Zola-Morgan S (1991) The medial temporal lobe memory Suzuki WA, Amaral DG (1994a) Perirhinal and parahippocampal
system. Science 253:13801386. cortices of the macaque monkey: cortical afferents. J Comp
Squire LR, Knowlton B, Musen G (1993) The structure and organi- Neurol 350:497533.
zation of memory. Annu Rev Psychol 44:453495. Suzuki WA, Amaral DG (1994b) Topographic organization of the
Squire LR, Clark RE, Knowlton BJ (2001) Retrograde amnesia. reciprocal connections between the monkey entorhinal cortex
Hippocampus 11:5055. and the perirhinal and parahippocampal cortices. J Neurosci
Squire LR, Stark CE, Clark RE (2004) The medial temporal lobe. 14:18561877.
Annu Rev Neurosci 27:279306. Suzuki WA, Clayton NS (2000) The hippocampus and memory: a
Staubli U, Ivy G, Lynch G (1984) Hippocampal denervation causes comparative and ethological perspective. Curr Opin Neurobiol
rapid forgetting of olfactory information in rats. Proc Natl Acad 10:768773.
Sci USA 81:58855887. Suzuki WA, Eichenbaum H (2000) The neurophysiology of memory.
Steele K, Rawlins JN (1993) The effects of hippocampectomy on per- Ann NY Acad Sci 911:175191.
formance by rats of a running recognition task using long lists Suzuki WA, Zola-Morgan S, Squire LR, Amaral DG (1993) Lesions of
of non-spatial items. Behav Brain Res 54:110. the perirhinal and parahippocampal cortices in the monkey
Steele RJ, Morris RGM (1999) Delay-dependent impairment of a produce long-lasting memory impairment in the visual and tac-
matching-to-place task with chronic and intrahippocampal tual modalities. J Neurosci 13:24302451.
infusion of the NMDA-antagonist D-AP5. Hippocampus Swanson LW (2000) Cerebral hemisphere regulation of motivated
9:118136. behavior. Brain Res 886:113164.
Stefanacci L, Buffalo EA, Schmolck H, Squire LR (2000) Profound Swanson LW (2004) Brain maps: structure of the rat brain, 3rd ed.
amnesia after damage to the medial temporal lobe: a neu- Amsterdam: Elsevier.
roanatomical and neuropsychological prole of patient E.P. J Sweatt JD (2003) Mechanisms of memory. London: Academic
Neurosci 20:70247036. Press.
Steffenach HA, Sloviter RS, Moser EI, Moser MB (2002) Impaired Tanner WP Jr, Swets JA (1954) A decision-making theory of visual
retention of spatial memory after transection of longitudinally detection. Psychol Rev 61:401409.
oriented axons of hippocampal CA3 pyramidal cells. Proc Natl Taube JS (1998) Head direction cells and the neurophysiological basis
Acad Sci USA 99:31943198. for a sense of direction. Prog Neurobiol 55:225256.
Stein DG, Finger S, Hart T (1983) Brain damage and recovery: prob- Taube JS (1999) Some thoughts on place cells and the hippocampus.
lems and perspectives. Behav Neural Biol 37:185222. Hippocampus 9:452457.
Steward O, Cotman CW, Lynch GS (1974) Growth of a new ber pro- Taube JS, Schwartzkroin PA (1987) Intracellular-recording from hip-
jection in the brain of adult rats: re-innervation of the dentate pocampal ca1 interneurons before and after development of
gyrus by the contralateral entorhinal cortex following ipsilateral long-term potentiation. Brain Res 419:3238.
entorhinal lesions. Exp Brain Res 20:4566. Taube JS, Muller RU, Ranck JB Jr (1990a) Head-direction cells
Strasser R, Ehrlinger JM, Bingman VP (2004) Transitive behavior in recorded from the postsubiculum in freely moving rats. I.
hippocampal-lesioned pigeons. Brain Behav Evol 63:181188. Description and quantitative analysis. J Neurosci 10:420435.
Suddendorf T, Busby J (2003) Mental time travel in animals? Trends Taube JS, Muller RU, Ranck JB Jr (1990b) Head-direction cells
Cogn Sci 7:391396. recorded from the postsubiculum in freely moving rats. II.
Suddendorf T, Corballis MC (1997) Mental time travel and the evo- Effects of environmental manipulations. J Neurosci 10:438447.
lution of the human mind. Genet Soc Gen Psychol Monogr Taube JS, Kesslak JP, Cotman CW (1992) Lesions of the rat post-
123:133167. subiculum impair performance on spatial tasks. Behav Neural
Sutherland RJ, Dyck RH (1984) Place navigation by rats in a swim- Biol 57:131143.
ming pool. Can J Psychol 38:322247. Teng E, Squire LR (1999) Memory for places learned long ago is
Sutherland RJ, Rudy JW (1989) Confugural association theory: the intact after hippocampal damage. Nature 400:675677.
role of the hippocampal formation in learning, memory, and Teng E, Stefanacci L, Squire LR, Zola SM (2000) Contrasting effects
amnesia. Psychobiology 17:129144. on discrimination learning after hippocampal lesions and con-
Sutherland RJ, Hamilton DA (2004) Rodent spatial navigation: at the joint hippocampal-caudate lesions in monkeys. J Neurosci
crossroads of cognition and movement. Neurosci Biobehav Rev 20:38533863.
28:687697. Teuber HL (1955) Physiological psychology. Annu Rev Psychol
Sutherland RJ, Whishaw IQ, Kolb B (1983) A behavioural analysis of 6:267296.
spatial localization following electrolytic, kainate- or colchicine- Teyler TJ, DiScenna P (1986) The hippocampal memory indexing
induced damage to the hippocampal formation in the rat. Behav theory. Behav Neurosci 100:147154.
Brain Res 7:133153. Teyler TJ, DiScenna P (1987) Long-term potentiation. Annu Rev
Sutherland RJ, Whishaw IQ, Kolb B (1988) Contributions of cingu- Neurosci 10:131161.
late cortex to two forms of spatial learning and memory. J Thinus-Blanc C, Bouzouba L, Chaix K, Chapuis N, Durup M, Poucet
Neurosci 8:18631872. B (1987) A study of spatial parameters encoded during explo-
Sutherland RJ, Weisend MP, Mumby D, Astur RS, Hanlon FM, ration in hamsters. J Exp Psychol Anim Behav Process
Koerner A, Thomas MJ, Wu Y, Moses SN, Cole C, Hamilton DA, 13:418427.
Hoesing JM (2001) Retrograde amnesia after hippocampal Thomas KG, Hsu M, Laurance HE, Nadel L, Jacobs WJ (2001) Place
damage: recent vs. remote memories in two tasks. Hippocampus learning in virtual space. III. Investigation of spatial naviga-
11:2742. tion training procedures and their application to fMRI and clin-
 Theories of Hippocampal Function 711

ical neuropsychology. Behav Res Methods Instrum Comput Van Essen DC, Drury HA (1997) Structural and functional analyses
33:2137. of human cerebral cortex using a surface based atlas. J Neurosci
Thompson LT, Best PJ (1990) Long-term stability of the place-eld 17:70797102.
activity of single units recorded from the dorsal hippocampus of Vann SD, Aggleton JP (2004) Testing the importance of the retros-
freely behaving rats. Brain Res 509:299308. plenial guidance system: effects of different sized retrosplenial
Thornton JA, Malkova L, Murray EA (1998) Rhinal cortex ablations cortex lesions on heading direction and spatial working mem-
fail to disrupt reinforcer devaluation effects in rhesus monkeys ory. Behav Brain Res 155:97108.
(Macaca mulatta). Behav Neurosci 112:10201025. Vann SD, Brown MW, Erichsen JT, Aggleton JP (2000) Fos imaging
Tischmeyer W, Grimm R (1999) Activation of immediate early genes reveals differential patterns of hippocampal and parahippocam-
and memory formation. Cell Mol Life Sci 55:564574. pal subeld activation in rats in response to different spatial
Tolman EC (1948) Cognitive maps in rats and men. Psychol Rev memory tests. J Neurosci 20:27112718.
55:189208. Van Praag H, Christie BR, Sejnowski TJ, Gage FH (1999) Running
Tomita H, Ohbayashi M, Nakahara K, Hasegawa I, Miyashita Y enhances neurogenesis, learning, and long-term potentiation in
(1999) Top-down signal from prefrontal cortex in executive mice. Proc Natl Acad Sci USA 96:1342713431.
control of memory retrieval. Nature 401:699703. Vargha-Khadem F, Gadian DG, Watkins KE, Connelly A, Van
Tommasi L, Thinus-Blanc C (2004) Generalization in place learning Paesschen W, Mishkin M (1997) Differential effects of early hip-
and geometry knowledge in rats. Learn Mem 11:153161. pocampal pathology on episodic and semantic memory. Science
Tonegawa S, Li Y, Erzurumlu RS, Jhaveri S, Chen C, Goda Y, Paylor R, 277:376380.
Silva AJ, Kim JJ, Wehner JM (1995) The gene knockout technol- Vazdarjanova A, Guzowski JF (2004) Differences in hippocampal
ogy for the analysis of learning and memory, and neural devel- neuronal population responses to modications of an environ-
opment. Prog Brain Res 105:314. mental context: evidence for distinct, yet complementary, func-
Tooby J, Cosmides L (2000) Toward mapping the evolved functional tions of CA3 and CA1 ensembles. J Neurosci 24:64896496.
organisation of mind and brain. In: The new cognitive neuro- Vazdarjanova A, McGaugh JL (1999) Basolateral amygdala is involved
sciences (Gazzaniga M, ed). Cambridge, MA: MIT Press. in modulating consolidation of memory for classical fear condi-
Treves A, Rolls ET (1994) Computational analysis of the role of the tioning. J Neurosci 19:66156622.
hippocampus in memory. Hippocampus 4:118. Vinogradova OS (1995) Expression, control, and probable functional
Trullier O, Wiener SI, Berthoz A, Meyer JA (1997) Biologically based signicance of the neuronal theta rhythm. Prog Neurobiol
artificial navigation systems: review and prospects. Prog 45:523583.
Neurobiol 51:483544. Viskontas IV, McAndrews MP, Moscovitch M (2000) Remote
Trullier O, Shibata R, Mulder AB, Wiener SI (1999) Hippocampal episodic memory decits in patients with unilateral temporal
neuronal position selectivity remains xed to room cues only in lobe epilepsy and excisions. J Neurosci 20:58535857.
rats alternating between place navigation and beacon approach Voermans NC, Petersson KM, Daudey L, Weber B, Van Spaendonck
tasks. Eur J Neurosci 11:43814388. KP, Kremer HP, Fernandez G (2004) Interaction between the
Tsien JZ, Chen DF, Gerber D, Tom C, Mercer EH, Anderson DJ, human hippocampus and the caudate nucleus during route
Mayford M, Kandel ER, Tonegawa S (1996) Subregion- and cell recognition. Neuron 43:427435.
type-restricted gene knockout in mouse brain. Cell Wan H, Aggleton JP, Brown MW (1999) Different contributions of
87:13171326. the hippocampus and perirhinal cortex to recognition memory.
Tulving E (1972) Episodic and semantic memory. In: Organisation of J Neurosci 19:11421148.
memory (Tulving E, Donaldson W, eds), pp 381403. San Diego: Wang R, Spelke E (2002) Human spatial representation: insights from
Academic Press. animals. Trends Cogn Sci 6:376.
Tulving E (1983) Elements of episodic memory. New York: Oxford Warburton EC, Baird A, Morgan A, Muir JL, Aggleton JP (2001) The
University Press. conjoint importance of the hippocampus and anterior thalamic
Tulving E (2004) Episodic memory and autonoesis: uniquely human? nuclei for allocentric spatial learning: evidence from a discon-
In: The missing link in cognition: evolution of self-knowing con- nection study in the rat. J Neurosci 21:73237330.
sciousness (Terrace H, Metcalfe J, eds). New York: Oxford Warburton EC, Koder T, Cho K, Massey PV, Duguid G, Barker GR,
University Press. Aggleton JP, Bashir ZI, Brown MW (2003) Cholinergic neuro-
Tulving E, Markowitsch HJ (1998) Episodic and declarative memory: transmission is essential for perirhinal cortical plasticity and
role of hippocampus. Hippocampus 8:198204. recognition memory. Neuron 38:987996.
Ungerleider LG, Mishkin M (1982) Two cortical visual systems. In: Warrington EK, Weizkrantz L (1968) A study of learning and reten-
Analysis of visual behavior (Ingle DJ, Goodale MA, Manseld tion in amnesic patients. Neuropsychologia 6:283291.
RJW, eds), pp 549586. Cambridge, MA: MIT Press. Wehner R, Lehrer M, Harvey WC (1996) Navigation: migration and
Vanderwolf CH, Cain DP (1994) The behavioral neurobiology of homing. J Exp Biol 199:1260.
learning and memory: a conceptual reorientation. Brain Res Weiskrantz L (1982) Comparative aspects of studies of amnesia.
Brain Res Rev 19:264297. Philos Trans R Soc Lond B Biol Sci 298:97109.
Vanderwolf CH, Bland BH, Whishaw IQ (1973) Diencephalic hip- Weiskrantz L (1997) Consciousness lost and found. Oxford, UK:
pocampal and neocortical mechanisms in voluntary movement. Oxford University Press.
In: Efferent organisation and the integration of behavior (Masser Whishaw IQ (1998) Place learning in hippocampal rats and the path
JD, ed), pp 229263. San Diego: Academic Press. integration hypothesis. Neurosci Biobehav Rev 22:209220.
Van Elzakker M, OReilly RC, Rudy JW (2003) Transitivity, exibility, Whishaw IQ (2004a) Posterior neocortical (visual cortex) lesions in
conjunctive representations, and the hippocampus. I. An empir- the rat impair matching-to-place navigation in a swimming
ical analysis. Hippocampus 13:334340. pool: a reevaluation of cortical contributions to spatial behavior
712 The Hippocampus Book

using a new assessment of spatial versus non-spatial behavior. Winograd T (1975) Frame representations and the proce-
Behav Brain Res 155:177184. dural/declarative controversy. In: Representation and under-
Whishaw IQ (2004b) The behavior of the laboratory rat: a handbook standing: studies in cognitive science (Bobrow DG, Collins A,
with tests. New York: Oxford University Press. eds), pp 185210. San Diego: Academic Press.
Whishaw IQ, Jarrard LE (1996) Evidence for extrahippocampal Winters BD, Bussey TJ (2005) Glutamate receptors in perirhinal cor-
involvement in place learning and hippocampal involvement in tex mediate encoding, retrieval, and consolidation of object
path integration. Hippocampus 6:513524. recognition memory. J Neurosci 25:42434251.
Whishaw IQ, Tomie JA (1991) Acquisition and retention by hip- Winters BD, Forwood SE, Cowell RA, Saksida LM, Bussey TJ (2004)
pocampal rats of simple, conditional, and congural tasks using Double dissociation between the effects of peri-postrhinal cor-
tactile and olfactory cues: implications for hippocampal func- tex and hippocampal lesions on tests of object recognition and
tion. Behav Neurosci 105:787797. spatial memory: heterogeneity of function within the temporal
Whishaw IQ, Tomie JA (1997) Perseveration on place reversals in lobe. J Neurosci 24:59015908.
spatial swimming pool tasks: further evidence for place learning Wirth S, Yanike M, Frank LM, Smith AC, Brown EN, Suzuki WA
in hippocampal rats. Hippocampus 7:361370. (2003) Single neurons in the monkey hippocampus and learn-
Whishaw IQ, Kolb B, Sutherland RJ (1983) The analysis of behavior ing of new associations. Science 300:15781581.
in the laboratory rat. In: Behavioral approaches to brain research Witter MP, Naber PA, van Haeften T, Machielsen WC, Rombouts SA,
(Robinson TE, ed), pp 141211. New York: Oxford University Barkhof F, Scheltens P, Lopes da Silva FH (2000) Cortico-
Press. hippocampal communication by way of parallel parahippocam-
Whishaw IQ, Cassel JC, Jarrard LE (1995) Rats with mbria-fornix pal-subicular pathways. Hippocampus 10:398410.
lesions display a place response in a swimming pool: a dissocia- Wixted JT, Squire LR (2004) Recall and recognition are equally
tion between getting there and knowing where. J Neurosci impaired in patients with selective hippocampal damage. Cogn
15:57795788. Affect Behav Neurosci 4:5866.
Whishaw IQ, Hines DJ, Wallace DG (2001) Dead reckoning (path Wood ER, Dudchenko PA, Eichenbaum H (1999) The global
integration) requires the hippocampal formation: evidence from record of memory in hippocampal neuronal activity. Nature
spontaneous exploration and spatial learning tasks in light (allo- 397:613616.
thetic) and dark (idiothetic) tests. Behav Brain Res 127:4969. Wood ER, Dudchenko PA, Robitsek RJ, Eichenbaum H (2000)
White NM, McDonald RJ (2002) Multiple parallel memory systems Hippocampal neurons encode information about different
in the brain of the rat. Neurobiol Learn Mem 77:125184. types of memory episodes occurring in the same location.
Wiener SI (1993) Spatial and behavioral correlates of striatal neurons Neuron 27:623633.
in rats performing a self-initiated navigation task. J Neurosci Worden RP (1992) Navigation by fragment tting: a theory of hip-
13:38023817. pocampal function. Hippocampus 2:165187.
Wigstrom H, Gustafsson B (1983) Facilitated induction of hip- Wulff P, Wisden W (2005) Dissecting neural circuitry by combining
pocampal long-lasting potentiation during blockade of inhibi- genetics and pharmacology. Trends Neurosci 28:4450.
tion. Nature 301:603604. Yanike M, Wirth S, Suzuki WA (2004) Representation of well-
Wiig KA, Bilkey DK (1995) Lesions of rat perirhinal cortex exacer- learned information in the monkey hippocampus. Neuron
bate the memory decit observed following damage to the 42:477487.
mbria-fornix. Behav Neurosci 109:620630. Yonelinas AP (2001) Components of episodic memory: the contri-
Willshaw D, Dayan P (1990) Optimal plasticity from matrix memo- bution of recollection and familiarity. Philos Trans R Soc Lond B
ries: what goes up must come down. Neural Commun 8593. Biol Sci 356:13631374.
Willshaw DJ, Buckingham JT (1990) An assessment of Marrs theory Yonelinas AP, Kroll NE, Quamme JR, Lazzara MM, Sauve MJ,
of the hippocampus as a temporary memory store. Philos Trans Widaman KF, Knight RT (2002) Effects of extensive temporal
R Soc Lond B Biol Sci 329:205215. lobe damage or mild hypoxia on recollection and familiarity.
Wilson A, Brooks DC, Bouton ME (1995) The role of the rat hip- Nat Neurosci 5:12361241.
pocampal system in several effects of context in extinction. Zamanillo D, Sprengel R, Hvalby O, Jensen V, Burnashev N, Rozov
Behav Neurosci 109:828836. A, Kaiser KM, Koster HJ, Borchardt T, Worley P, Lubke J,
Wilson BA (1999) Case studies in neuropsychological rehabilitation. Frotscher M, Kelly PH, Sommer B, Andersen P, Seeburg PH,
New York: Oxford University Press. Sakmann B (1999) Importance of AMPA receptors for hip-
Wilson MA, McNaughton BL (1993) Dynamics of the hippocampal pocampal synaptic plasticity but not for spatial learning. Science
ensemble code for space. Science 261:10551058. 284:18051811.
Wilson MA, Tonegawa S (1997) Synaptic plasticity, place cells and Zeman A (2002) Consciousness: a users guide. New Haven, CT: Yale
spatial memory: study with second generation knockouts. University Press.
Trends Neurosci 20:102106. Zhang WN, Bast T, Feldon J (2001) The ventral hippocampus and
Winocur G (1990) Anterograde and retrograde amnesia in rats with fear conditioning in rats: different anterograde amnesias of fear
dorsal hippocampal or dorsomedial thalamic lesions. Behav after infusion of N-methyl-D-aspartate or its noncompetitive
Brain Res 38:145154. antagonist MK-801 into the ventral hippocampus. Behav Brain
Winocur G, Rawlins JNP, Gray JA (1987) The hippocampus and con- Res 126:159174.
ditioning to contextual cues. Behav Neurosci 101:617625. Zhu XO, Brown MW, McCabe BJ, Aggleton JP (1995) Effects of the
Winocur G, Moscovitch M, Fogel S, Rosenbaum RS, Sekeres M novelty or familiarity of visual stimuli on the expression of the
(2005) Preserved spatial memory after hippocampal lesions: immediate early gene c-fos in rat brain. Neuroscience
effects of extensive experience in a complex environment. Nat 69:821829.
Neurosci 8:273275. Zinyuk L, Kubik S, Kaminsky Y, Fenton AA, Bures J (2000)
 Theories of Hippocampal Function 713

Understanding hippocampal activity by using purposeful tion: evidence for a time-limited role in memory storage.
behavior: place navigation induces place cell discharge in both Science 250:288290.
task-relevant and task-irrelevant spatial reference frames. Proc Zola-Morgan S, Squire LR, Mishkin M (1982) The neuroanatomy of
Natl Acad Sci USA 97:37713776. amnesia: amygdala-hippocampus versus temporal stem. Science
Zola SM, Squire LR (2001) Relationship between magnitude of dam- 218:13371339.
age to the hippocampus and impaired recognition memory in Zola-Morgan S, Squire LR, Amaral DG (1986) Human amnesia and
monkeys. Hippocampus 11:9298. the medial temporal region: enduring memory impairment fol-
Zola SM, Squire LR, Teng E, Stefanacci L, Buffalo EA, Clark RE (2000) lowing a bilateral lesion limited to eld CA1 of the hippocam-
Impaired recognition memory in monkeys after damage limited pus. J Neurosci 6:29502967.
to the hippocampal region. J Neurosci 20:451463. Zola-Morgan S, Squire LR, Amaral DG (1989) Lesions of the amyg-
Zola-Morgan S, Squire LR (1984) Preserved learning in monkeys dala that spare adjacent cortical regions do not impair memory
with medial temporal lesions: sparing of motor and cognitive or exacerbate the impairment following lesions of the hip-
skills. J Neurosci 4:1071085. pocampal formation. J Neurosci 9:19221936.
Zola-Morgan S, Squire LR (1985) Medial temporal lesions in mon- Zola-Morgan S, Squire LR, Ramus SJ (1994) Severity of memory
keys impair memory on a variety of tasks sensitive to human impairment in monkeys as a function of locus and extent of
amnesia. Behav Neurosci 99:2234. damage within the medial temporal lobe memory system.
Zola-Morgan S, Squire LR (1990) The primate hippocampal forma- Hippocampus 4:483495.
This page intentionally left blank
14 Neil Burgess
Computational Models of the Spatial
and Mnemonic Functions of the Hippocampus

data a vast body of knowledge has been collected regarding

14.1 Overview the neural representation of the spatial location and orienta-
tion of freely moving rats (see Chapter 11). Hypotheses
Some of the most striking data relating cognitive behavior to regarding the function of the hippocampus have traditionally
neuronal ring, damage, or metabolic activity in the brain been expressed in words. However, it is often possible to inter-
concerns the hippocampus. The use of computational models pret verbal descriptions in more than one way, or to change
has been invaluable for exploring the link between neurons their interpretation retrospectively to suit the facts. In addi-
and behavior, enabling hypothetical mechanisms to be dened tion, it can be difficult to tell whether the proposed explana-
precisely and examined quantitatively. Here I review many of tion would actually work as described and, if so, difficult to
these models, including models of spatial functions, models make unambiguous quantitative predictions that can be used
of more general associative mnemonic functions, models that to test it. These problems become even more acute when
stress feedforward processing through the hippocampal sys- hypotheses address the question of how a putative function
tem, and those stressing recurrent processing within it. I arises from the cooperative behavior of large numbers of neu-
review the spatial models rst, as they are most rmly rooted rons and synapses. One way around this is to express such a
in the known electrophysiology of the region. These models hypothesis in terms of equations or computer simulations,
cover both the representation of the animals spatial location referred to as a computational model. An advantage of this
and orientation and the use of this information in spatial nav- approach is that all of the parameter values and assumptions
igation. The models of mnemonic function, specically asso- necessary to generate the behavior concerned need to be made
ciative or episodic memory, follow from Marrs seminal 1971 explicit; another is that the operation of the model is unam-
model. I use this model as a generic framework in which to biguously specied. These advantages mean that computa-
consider the various subsequent developments to it. Finally, I tional modeling has an important role to play in the progress
review those models attempting to bring together the spatial of scientic understanding, most importantly in its interac-
and mnemonic functions of the hippocampus. tion with experimental investigation: by predicting critical
experiments, undergoing revision to reect their results, and
predicting new ones. They are not a panacea for all ills; and
interpretation of a model, the way it works, and the values of
14.2 Introduction its parameters can still be changed or disputed. At the most
basic level a computational model can serve as an existence of
There have been many attempts to understand and quantify proof of the behavior that can result from a proposed mecha-
the contribution of the hippocampus to cognitive behavior. In nism, but more generally it can serve to dene a theoretical
this chapter I focus on models of the link between the cogni- understanding and provide a powerful framework within
tive ability of the animal and the action of individual cells which the nature of a theory relating brain to behavior can be
and synapses. As reviewed in Chapter 13, lesion studies in a understood.
variety of mammals (including humans) have implicated the Examples of potential insights into the function of the
hippocampus in spatial navigation, whereas human neu- nervous system that would not have been fully understood,
ropsychology has most notably implicated it in episodic or and in some cases even suspected, without computational
declarative memory (see Chapter 12). In addition to these modeling include the relation of voltage-dependent ion chan-

715
716 The Hippocampus Book

nels to the propagation of action potentials (Hodgkin and space, drawing mostly on experimental data collected in rats.
Huxley, 1952); the possible relation of the pattern of ring of I next consider models of the use of spatial representations in
dopamine neurons during conditioning experiments to the guiding behavior. Together, models of the representation of
process of learning from trial and error (interpretable via the location and orientation from sensory input and models of
Rescorla-Wagner law and subsequent computational models how these representations are used to guide behavior provide
of reinforcement learning) (e.g., Shultz et al., 1997); the one of the best examples of quantitative understanding of the
relation of partially shifting gain eld responses in parietal links between perception, cognition, and action and between
cortex and optimal integration of multiple cues (made clear cells and systems and behavior. The second half of the chapter
by computational understanding of the way attractor net- concerns attempts to model the more general role of the
works can perform this function) (e.g., Deneve et al., 2001); human hippocampus in memory for personal experience. The
and the effect of genetic knockout of N-methyl-D-aspartate chapter concludes with consideration of the models that
(NMDA) receptors in hippocampal region CA3 on the attempt to reconcile these two streams of research (spatial and
robustness of both the mouses spatial memory and the ring mnemonic). As we shall see, the role of the recurrent collater-
of its place cells when a subset of environmental landmarks als in area CA3 of the hippocampus maintains a common
are removed (interpretable as pattern completion in a model point of contact between these models: In both the spatial and
of CA3 acting as an attractor network) (Marr, 1971; Nakazawa episodic memory frameworks they are assumed to perform an
et al., 2002) (see Section 14.3.3 and 14.5.3). associative memory function.
Computational modeling of the hippocampus has fol-
lowed two largely independent streams over the years: one
seeking to explain a general role in associative memory and
the other focusing on its role in spatial navigation. Models 14.3 Hippocampus and Spatial Representation
usually start from as detailed a biophysical level as is useful for
the level of the hypothesis they seek to investigate. Although This section addresses the representation of spatial location
many models reviewed here involve detailed simulation of cel- and orientation. Models of the representation of location
lular and synaptic electrophysiology, the aim of this chapter is embodied by the ring of place cells are considered rst. These
to explain the neural bases of spatial and mnemonic behav- models take two often equivalent forms: those relying pre-
iorsomething that is made easier by focusing on the sim- dominantly on feedforward connections to capture the data
plest level of description capturing the likely functional and those relying predominantly on the recurrent connections
consequences of cellular and synaptic events (e.g., whether an in CA3. It is not only the ring rates of place cells that encode
action potential was red). Thus, the activity (ring rate) of a location but also the time of ring relative to the ongoing theta
neuron is simply viewed as a monotonic function of the rhythm on the electroencephalogram (EEG) (OKeefe and
amount by which the net input to it exceeds some threshold Recce, 1993). Both of these aspects of ring have been the sub-
value. The net input to a neuron is the sum of the activity of jects of extensive computational modeling. The representation
each neuron connected to it weighted by the strength of the of the animals orientation embodied by the head-direction
connection (occasionally inhibitory inputs are modeled as a (HD) cells is equally striking; and the nature of this signal,
divisive term in the net input rather than a subtractive term, considered in the remainder of this section, has also been
see Figure 1410, later). Learning corresponds to modica- investigated by computational modeling in more recent years.
tion of the connection strengths. Most commonly learning is Finally, the discovery of grid cells (Hafting et al., 2005) has
of a Hebbian nature (Hebb, 1949) such that simultaneous provided a new focus, reviewed by McNaughton et al. (2006).
pre- and post-connection activity leads to increased connec-
tion strength and is often used in explicit analogy to synaptic 14.3.1 Representing Spatial Location
processes such as long-term potentiation (LTP) (see Chapter and Orientation: Data
10) (see Box 141). Other concepts are explained as and where
necessary. Readers interested in neural computation more A rich set of experimental data has been gathered on the neu-
generally should see the relevant literature (e.g., McClelland ral representation of spatial behavior found in and around the
and Rumelhart, 1986; Rumelhart and McClelland, 1986; Hertz hippocampus. Here I briey summarize the results with the
et al., 1990; Dayan and Abbott, 2002). greatest relevance to the models described below (see Chapter
Because the anatomy of the region is similar in rats, pri- 11 for more details). The ring of place cells in the hip-
mates, and humans, it seems sensible to start with the compu- pocampi of freely moving rats encodes the location of the ani-
tational functions of the hippocampal region in the rat and mal, each cell ring when the animal is within a particular
then consider how these functions might have been adapted portion of its environment (the place eld). Cells with sim-
during evolution. This has the advantage of applying the most ilar responses have also been observed in primates (Hori et al.,
detailed electrophysiological constraints at the outset. 2003; Ludvig et al., 2004) including humans (Ekstrom et al.,
Accordingly, in this chapter I concentrate initially on neural 2003). In open environments through which the rat can move
models starting from the reliable and well understood body of freely, ring rates are not inuenced by the animals orienta-
data regarding place cells and the neural representation of tion, whereas in environments in which movement direction
 Computational Models of Spatial and Mnemonic Functions 717

Box 141
Learning via Synaptic Modication: Hebbian Learning Rules

In the simplest type of neural network model, the ring rate, or activity, of a neuron (a) is
simply a function (the transfer function f) of the net current coming into the neuron, which
in turn is simply a weighted sum of the ring rates (u i) of the neurons connecting to it. That is:
a f(i w i u i), often written as: a  f(w.u), where w is the vector of connection weights
modeling the strengths (e.g., net synaptic efficacy) of connections from the input neurons, and
the dot is the vector dot product. With the simplest, linear, transfer function, the activation is
given by:
a w.u (1)
In such networks, learning corresponds to modication of the connection weights w. Below I
discuss some of the Hebbian learning rules mentioned in the rest of the chapter, and their
effects, following the discussion in Dayan and Abbott (2002), where further details can be
found.
A learning rule directly implementing Hebbs (1949) postulate of coincident ring leading to
increased coupling between neurons describes the change in connection weights in terms of the
product of pre- and post-synaptic ring rates:
dw dw
i  aui, or   au (2)
dt dt
where gives the rate of change of connection weights with time. When this rule is applied to a
training set of n example input patterns of activity u, each presented for an equal duration
over a total time T, we can integrate equation 2 to see the total change in w:
T
w wau (3)

where a  w.u from equation 1. If the connection weights are updated only after presenta-
tion of all of the input patterns, we can say:
T nT
w w(w.u)u  wQ.w (4)

where Q is the correlation matrix of the input patterns (Q  u u, where   denotes the
average over input patterns, and u u is the outer product of u with itself). Thus, simple
Hebbian learning rules are also known as correlation-based learning rules. Inspection of equa-
tion 4 indicates that the weight vector w, if plotted in the same space as the input vectors u,
eventually follows the principal eigenvector of the correlation matrix; that is, it will lie along
the direction from the origin to the mean input pattern (u) or, if u is at the origin,
along the rst principal component of the set of input patterns. However, this learning rule is
not stable: Large weights produce large output activations, which produce large increases in
weights and so on. More formally, it can be seen from the dot product of w in equation 2 that
the length of the weight vector increases whenever the output neuron is active:
d w 2 dw 2aw.u 2a2
  2w.    (5)
dt dt
One way to introduce balance into the learning rule is to allow for a connection weight to
increase or to decrease according to the levels of pre- and postconnection activity, by analogy
with long-term potentiation and depression (LTP and LTD) (see Chapter 10). In this way
equation 2 could become
dw dw
  (a
)u, or   a(u-) (6)
dt dt
where either a postsynaptic threshold or a set of presynaptic thresholds are applied to deter-
mine the sense and size of weight changes (respectively: is the level postsynaptic activity must
surpass for the connection to increase rather than decrease; or , which is the vector of activity
levels each input neuron must surpass). The most obvious choice of threshold for the pre- or
postsynaptic neuron is its average activity over the training set. In this case, following a similar
derivation to equation 4, both versions produce the same learning rule:
nT
w w C.w (7)

(Continued)
718 The Hippocampus Book

Box 141
Learning via Synaptic Modication: Hebbian Learning Rules (Continued)

where C is the covariance matrix of the input patterns: C  (u- u )2   u u - u

2  (u- u ) u . These learning rules are also known as covariance rules. Inspection of
equation 7 indicates that the weight vector will eventually follow the principal eigenvector of
the covariance matrix (i.e., it will lie along the direction of the rst principal component of the
set of input patterns). It should be noted that these rules are also not stable; in this case d w 2/dt
is proportional to the variance (a2
a2) of the output activity over the training set.
The BCM learning rule, derived from experimental investigation of visual cortical plasticity
(Bienenstock et al., 1982), proposes that:
dw
  au(a
(a)) (8)
dt
This requires both pre- and postsynaptic activity for modication of a connection weight
(unlike the rules in equation 6), and also involves a sliding postsynaptic threshold ((a)), which
varies with postsynaptic activity. So long as the postsynaptic threshold increases as a power of
a  1 (typically following a time-averaged estimate of a2), it can ensure stability of the learning
rule: effectively increasing the threshold to an overactive output neuron so connection weights
to it tend to be reduced.
The other common way in which Hebbian learning rules are stabilizede.g., in Rumelhart
and Zipsers (1986) competitive learning algorithmis to use divisive normalization: explic-
itly constraining the length of the weight vector to remain constant during learning by dividing
all weights by w . Although this is a nonlocal operation, because synaptic strengths must be
altered according to the state of other, distant synapses, a similar effect can be achieved by the
local learning rule of Oja (1982):
dw
  au
a2w (9)
dt
An analysis similar to equation 5 shows that w2 tends to a value 1/ under repeated applica-
tion of this rule.
As well as being stable, the BCM and normalized Hebbian learning rules involve competition
between connections: Increasing some of the connection weights onto a neuron leads to a de-
crease in the others. Under the BCM rule, this occurs because of increased activity leading to a
higher postsynaptic threshold and thus an increased incidence of LTD versus LTP. With normal-
ization it occurs directly owing to the increase in the length of the weight vector. These learning
rules tend to allow a neuron to become tuned to respond to specic patterns of input activation
(see text). There is at least some evidence for competitive interaction between synapses such
that increasing the strengths of one set of synapses leads to a decrease in the strengths of others
so as to normalize the total synaptic strength onto the neuron (Royer and Pare, 2003). Note
however, that recent evidence of the dependence of synaptic plasticity on the precise timing of
pre- and postsynaptic activity (see Chapter 10) changes the likely nature of learning rules based
on LTP and LTD, as in the effects of temporal asymmetry noted in Section 14.4.3.

is constrained (e.g., linear tracks, eight-arm mazes) ring is sal presubiculum (Taube et al., 1990). The HDCs code for
strongly modulated by the rats direction of motion (see head direction within an environment, each ring whenever
Chapter 11). The orientation of the place cell representation is the animals head points in a specic direction, independently
controlled by distal cues at or beyond the edge of the envi- of the animals location. The orientation of the head direction
ronment (OKeefe and Conway, 1978; Muller and Kubie, representation is controlled by distal visual cues in the same
1987) but not by those within it (Cressant et al., 1997). A place way as the place cell representation. The overall orientation of
cells spatially localized ring appears to be robust to the both place and HDC representations may be disrupted by dis-
removal of subsets of cues and indeed removal of all of the orientation (rotating the rat in a covered container); and
controlling visual cues while the rat remains in the environ- when both have been recorded simultaneously, both represen-
ment (Muller and Kubie, 1987; OKeefe and Speakman, 1987), tations have remained in register with each other (Knierim et
although remaining uncontrolled cues may be important in al., 1995). Interestingly, hints that an expanded representation
these cases (Save et al., 2000). of neurons responding to specic combinations of place and
The complementary representation of orientation inde- direction have been found in the presubiculum (Sharp, 1996;
pendent of location is found in head direction cells (HDCs) Cacucci et al., 2004). A third type of representation, grid
in the mammillary bodies, anterior thalamic nuclei, and dor- cells, has now been found in dorsomedial entorhinal cortex
 Computational Models of Spatial and Mnemonic Functions 719

(Hafting et al., 2005). Each grid cell res in a set of locations the environment but with the incorporation of an element of
which is laid out on a hexagonal grid. The intriguing proper- competitive learning (Rumelhart and Zipser, 1986). Briey,
ties of these cells are summarized at the end of Section 14.3.3. this involves neurons arranged in groups dominated by lateral
However, at the time of this writing they have not yet been inhibition such that only the neuron with the greatest input
fully incorporated into models of hippocampal function and can re. Normalized Hebbian learning is then applied (i.e.,
so are not a focus of this chapter. increasing the strengths of connections between simultane-
Beyond the clear encoding of spatial location and orienta- ously active neurons while decreasing the others so the overall
tion in the ring rates of these neurons, the picture becomes strength of connections to a neuron does not change (see Box
slightly more complex. Initial experiments in which place cells 141). This results in specic neurons coming to represent
were recorded in environments of different shape (Muller and specic patterns of input, with each neuron responding to a
Kubie, 1987) reported completely different patterns of ring in particular pattern or patterns similar to it. Sharps model
the two environments (remapping). A place cell active in one envisaged two types of sensory input regarding each distal
environment might re in an unrelated location in the second cue: one representing its distance from the rat and the other
environment or might be silent. In other experiments, perhaps representing both its distance from the rat and its direction
involving less complete changes to the environment or less relative to the rats heading. This sensory input passed forward
extensive pretraining in the various environments, parametric to a layer of entorhinal cortical cells and thence to a layer of
changes in the pattern of ring were observed (e.g., OKeefe place cells (Fig. 141). Competitive learning at each layer
and Burgess, 1996). Interestingly, although the ring rate of causes the entorhinal and place cells to learn to respond selec-
place cells can be shown to encode the animals location within tively to the pattern of sensory input present in a particular
a given environment (e.g., Wilson and McNaughton, 1993), portion of the environment, and it produces reasonable
the times at which they re relative to the theta rhythm robustness to cue removal. The successive layers of competi-
encodes additional information (see Chapter 11) (OKeefe and tion produce sharper tuning to position and greater robust-
Recce, 1993; Skaggs et al., 1996; Jensen and Lisman, 2000). ness to cue removal in place cells than entorhinal cells. Due to
the use of distance-related inputs, expansion of the environ-
14.3.2 Representing Spatial Location: ment produces results that are qualitatively similar to those of
Feedforward Models Muller and Kubies (1987) experiment. The most interesting
aspect of this model concerns the directional modulation of
Computational modeling of place cell ring began with place cell ring. In the model, place cell ring is initially direc-
Zipsers (1985) model. In this model, sensory details of the tionally modulated owing to the partially directional sensory
environment feedforward to landmark detectors and thence inputs. During random exploration through an open environ-
to place cells. Landmark detectors are neurons specic to a ment, competitive learning allows a given place cell to learn to
specic place cell and to a specic aspect of the sensory scene respond to the sensory inputs occurring for different orienta-
(a location parameter). The output of these detectors is pro- tions at the place eld, producing nondirectional ring. By
portional to the match between the stored state of a location contrast, this does not occur during constrained motion (i.e.,
parameter and its currently perceived state. A place cells activ- back and forth in a single direction). This provides a good,
ity corresponds to a thresholded sum of the strengths of the simple model of the directionality of place cell firing,
matches it receives from several landmark detectors. Interest- although a more detailed look at directionality data indicates
ingly, the most obvious location parameterdistance from a that, if anything, place elds in open environments are ini-
landmarkwas rejected in favor of measures that scale with tially nondirectional and become directional as a result of
environmental size, such as the retinal angle between two experience. For alternative models see Blum and Abbott
landmarks, on the basis of a misinterpretation of Muller and (1996), Brunel and Trullier (1998) and Kali and Dayan (2000),
Kubies (1987) experiment. In this experiment, a signicant discussed below.
but small number of place elds were shown to expand fol- In an attempt to derive the form of the sensory input to the
lowing a doubling in the linear size of the environment. place cells, OKeefe and Burgess (1996) systematically varied
However, this expansion corresponded to, at most, approxi- the shape and size of the rats environment while recording
mate doubling of the place eld in area rather than the quad- from the same cells. The patterns of ring across environ-
rupling predicted if everything scaled up proportionately. ments included place elds that stretched or became bimodal
Furthermore, in many cells the expansion often appeared to when the environment expanded. These patterns were not
be along only one environmental dimension rather than both. consistent with those obtained in previous models of place
The model captures some of the motility of place elds in the elds depending on the relative locations of discrete land-
presence of manipulations of environmental cues and some of marks from the rat (e.g., Zipser, 1985) but, rather, indicated
their robustness to the removal of subsets of cues and (incor- continuous dependence on environmental boundaries.
rectly) produces a place eld that scales up proportionately Specically, place elds were viewed as a thresholded linear
with environmental expansion. sum of inputs tuned to respond to the presence of a boundary
Sharp (1991) followed in the same vein of feedforward at a given distance along a given allocentric direction (i.e.,
modeling of the response of place cells to sensory input from independent of the orientation of the rat and probably
720 The Hippocampus Book

Figure 141. Sharps (1991) model of place cell ring in response in an open cylinder, place cell ring is only weakly modulated by
to sensory input. A. Two groups of neocortical cells respond to the the heading direction of the animal (overall place eld shown
distance or (egocentric) direction of specic cues as the rat moves in center, ring elds at each of eight heading direction shown
in a cylinder (shown in B, cues marked by letters). Competitive around the edge). D. If exploration is constrained to specic
learning in successive layers of entorhinal cells and place cells directions of movement, as on a radial arm maze, place cell ring
leads, after exploration, to spatially tuned ring that is robust to is strongly modulated by the heading direction. (Source: Adapted
removal of subsets of cues. C. If exploration is unconstrained, as from Sharp, 1991.)

determined relative to the head-direction system; see below) One phenomenon not addressed by the above models is
(Fig. 142). These hypothetical inputs were termed boundary the remapping of place cell representations across different
vector cells. By tting a place cells ring pattern across several environments. This remapping can be both partial and incre-
environmental shapes, the model can predict its ring pattern mental over time (e.g., Bostock et al., 1991; Skaggs and
in an environment of novel shape (Hartley et al., 2000). In an McNaughton, 1998; Lever et al., 2002), with the eventual cre-
experiment complementary to the variation of environmental ation of stable but distinct patterns of ring in the two envi-
shape, Fenton et al. (2000) parametrically varied the position ronments. The factors inuencing the speed and completeness
of two cue cards around the edge of a cylindrical environment. of remapping are not currently well understood, but one com-
Although each card alone can control the overall orientation mon change in an individual place cell is to continue to re in
of the recorded place elds, inconsistent movement of both the environment in which it res most strongly and to stop
(i.e., moving them closer together or apart) produces inho- ring in the other. This aspect of remapping was addressed by
mogeneous parametric movement of the place elds such that Fuhs and Touretzky (2000) in a model of learning in the per-
their movement depends on their location. The added com- forant path projection from entorhinal cortex to place cells in
plexity of the cue cards acting as both identiable boundaries CA3. They found that the usual learning rules relating synap-
and directional cues can be incorporated into the boundary tic modication to the product of pre- and postsynaptic activ-
vector cell model by assuming that inconsistent cue-card ity (i.e.. Hebbian learning) or to its covariance were unable to
movement warps the representation of head direction, and reproduce this behavior. In the case of Hebbian learning, a
boundary vector cells become sensitive to variations in texture place cell with strong ring in one environment and weak r-
after sufficient exposure (Burgess and Hartley, 2002). ing in the other strengthens its ring in both environments. In
 Computational Models of Spatial and Mnemonic Functions 721

b) c)

a) d) e)
Figure 142. Model of the geometrical inuence on place elds b. Place elds recorded from the same cell in four environments of
(Hartley et al., 2000), assuming a stable directional reference frame. different shape or orientation relative to distal cues. c. Simulation of
Place elds are composed from thresholded linear sums of the the place elds in b by the best tting set of four BVCs constrained
ring rates of boundary vector cells (BVCs). a. Above: Each BVC to be in orthogonal directions (BVCs shown on the left, simulated
has a Gaussian tuned response to the presence of a boundary at elds on the right). The simulated cell can now be used to predict
a given distance and bearing from the rat (independent of its ring in novel situations. Real and predicted data from three novel
orientation). Below: The sharpness of tuning of a BVC decreases environments are shown in d and e, respectively, showing good
as the distance to which it is tuned increases. The only free parame- qualitative agreement. (Source: Adapted from Burgess and Hartley,
ters of a BVC are the distance and direction of the peak response. 2002.)

the case of covariance learning, exposure to the second envi- of place cell representations across different environments has
ronment tends to lead to loss of the place cell representation also been ascribed to the effects of the recurrent connections
of the rst environment. By contrast, the BCM (Bienenstock- in CA3. Models stressing these connections are the subject of
Cooper-Munro) learning rule (Bienenstock et al., 1982) (see the next section.
Box 141), which explicitly makes the direction of synaptic
modication dependent on the strength of the postsynaptic 14.3.3 Representing Spatial Location
activity, did produce the desired result: strong ring remain- and Orientation: Feedback Models
ing stable and weak ring reducing with experience. This type
of learning can also capture the way place elds become more The long-range recurrent connections between pyramidal
coherent with time and the temporal dynamics of their cells in area CA3 of the hippocampus have long been inter-
response to the introduction of a barrier into the environment preted as enabling this region to work as an autoassociative
(Barry et al., 2006). neural network (e.g., Marr, 1971; Hopeld, 1982; Amit, 1989).
Evidence of experience-dependent change in place cell r- This type of network is most often used to provide a content-
ing also comes from experiments by Mehta et al. (1997, 2000). addressable memory, a subject explored in Section 14.5 (see
They found that, over the rst few runs through a CA1 place Box 142). In this section I consider the role played by recur-
eld on a linear track, the spatial distribution of ring changes rent collaterals in the spatial representations of place and HD
from roughly symmetrical to slightly asymmetrical, caused by cells. Interestingly, direct experimental evidence for an asso-
additional ring at low rates earlier on the track. They suggest ciative function for CA3 has emerged recently, with indica-
that this change results from the known temporal asymmetry tions that the NMDA receptors in this region are involved in
of LTP (which is greater when presynaptic activity precedes making both the place elds and the rats spatial memory
postsynaptic activity than vice versa) acting on the CA3 to robust to cue removal (Nakazawa et al., 2002). In parallel,
CA1 pathway (see Chapter 10). Other models have implicated attractor dynamics have been found in the place cell represen-
the recurrent connections in CA3 as responsible for this effect. tation of two environments of different shape after fast
Interestingly, the phase shift seen in place cell ring (see remapping caused by exposure to the two environmental
Section 14.3.1 for data and Section 14.3.5 for models) (see shapes made of different materials (Wills et al., 2005). In these
also Chapter 11) acts to increase this effect of temporal asym- representations, in contrast to those that have not fast-
metry in LTP, causing place cells with elds early on the path remapped (Leutgeb et al., 2005), the two shapes act as attrac-
to re before those with elds later along the path during each tors: all place cells in intermediate shaped environments
theta cycle. The apparently unrelated effect of the remapping coherently returning to one or other representation.
722 The Hippocampus Book

Box 142
Attractors in Memory, Neural Coding, and Path Integration

POINT ATTRACTORS AND MEMORY

A network of recurrently connected neurons can be arranged so a nite number of discrete

patterns of activation across the neurons are stable states, or attractors. This means that any
pattern of activation similar enough to one of these attractors will evolve into the attractor
pattern under the dynamics of the network. These patterns of activation are stored in the
network in the sense that they are retrieved from any similar enough initial pattern. Such
networks are also referred to as autoassociative networks, and are an example of a content-
addressable memory in that a pattern of activity is retrieved by a pattern of similar content,
rather than, say, an unrelated index term or the address of a storage location.
In one of the simplest models (Hopeld, 1982), the activity of neuron i is modeled as ai 
 1. Connections between neurons i and j are symmetrical, with synaptic weight wij  wji
The dynamics of the network are given by: ai(t1)  sign(j wij aj(t)), such that the energy
or Lyapunov function of the network:
E
ij wijaiaj
can only reduce. If connection weights undergo a form of Hebbian learning when the to-be-
stored patterns of activation (a , say) are present, such that wij aiaj , these representa-
tions become attractors so long as the number of stored patterns is not too large (less than
around 0.14N, where N is the number of neurons, in this case). That is, a similar enough pat-
tern of activation converges onto the stored pattern under the dynamics of the network (Cohen
and Grossberg, 1983). This situation is often visualized by imagining how the energy of the
network varies as a function of the networks state au  (a1, a2 , .. aN)the attractor states
being local minima of the energy surface to which nearby states evolve under the networks
dynamics (Fig. 143A). Similar behavior is also shown by more biologically realistic models
(Amit, 1992; Treves and Rolls, 1992; McClelland et al., 1995) (Fig. 1410).
LINE ATTRACTORS AND NEURAL CODING

The value of a continuous variable (or stimuluss) often seems to be represented in the ring
rates of a population of neurons, each of which is tuned to respond preferentially to a single
preferred value. For example, head-direction cells can be thought of in this way, with s repre-
senting the rats heading. The pattern of activation of the population is often visualized by
imagining the neurons arranged so their location reects their preferred values: showing a
smooth bump of activity across the neurons peaked at the actual value of the stimulus.
However, if the ring rates are noisy it is difficult to estimate the precise value of the stimulus.
The presence of recurrent connections between neurons, arranged so that the weight of the
connection between each pair is simply a decreasing function of the difference in their pre-
ferred values (or physical separation when arranged as above), can help by ensuring that the
ring pattern takes the shape of a smooth bump1 (Fig. 143B). With the appropriate choice
of recurrent connections, such a network can perform optimal decoding (Latham et al., 2003),
including the situation where the representation is formed from different, unreliable sources
of information (Deneve et al., 2001).
The patterns of activation comprising a smooth bump can be thought of as a line in the N
dimensional state space au  (a1, a2 , .. aN) of the network. Each point on the line corresponds to
a different estimate of s (referred to as s). Conversely, all of the possible noisy patterns of activa-
tion that end up producing the same s lie on an N-1 dimensional subspace within which the
action of the recurrent connections corresponds to convergence onto the line (Fig. 143C).
An important aspect of these networks is that, although the recurrent connections ensure that
patterns of activation move onto the line attractor, movement along it, corresponding to chang-
ing s, is not affected by the recurrent connections (because the connections between a pair of
neurons depends only on the difference in their preferred values, not what those preferred
values are).
LINE ATTRACTORS AND PATH INTEGRATION

Because the (symmetrical) recurrent connections provide no resistance to the motion of the
bump of activity along the line attractor, its position is easily moved by asymmetrical connec-
tions from each neuron to neighbors farther along the line (Skaggs et al., 1995; Zhang, 1996).
1 These patterns have low energy as activation is concentrated in nearby neurons, which have the
strongest interconnections.
 Computational Models of Spatial and Mnemonic Functions 723

The greater the size of the asymmetrical connectionswhich should correspond to the spatial
derivative of the symmetrical connections for the bump to move without changing shape
(Zhang, 1996)compared to the symmetrical ones, the faster the movement of the bump
(Figs. 143 to 145). Thus, if the strength of the asymmetrical connections is proportional to
angular velocity, the location of the bump of activity in a ring of head direction cells tracks the
head direction of the animalperforming angular path integration. For a more detailed
model, see Redish et al. (1996).
As noted by Zhang (1996) and McNaughton et al. (1996), the angular path integration mod-
els of head direction cell ring can be extended to path integration models of place cell ring.
In this case, the place cells are imagined as a two-dimensional array, so the location of each
neuron corresponds to the location of its place eld in the environment (Fig. 146). Again,
symmetrical connections decreasing in strength with the physical separation of the pre- and
postsynaptic neurons can ensure that neural activity forms a single-peaked bump over the
array. Asymmetrical connections from each neuron to its neighbors along a given direction
causes the bump to shift in that direction (Figs. 143 and 144). In this case, to perform path
integration of position, the strength of the asymmetrical connections between a pair of neu-
rons displaced in a given direction needs to be proportional to the velocity of the rat in that
direction. See Samsonovich and McNaughton (1997) for a more detailed model and Droulez
and Berthoz (1991) and Dominey and Arbib (1992) for the origins of this type of model.

Continuous Attractor Models input from sensory cells (visual cells in Fig. 145) and so
of Head Direction Cells can be associated with those sensory inputs appearing at a sta-
ble bearing during exploration of a new environment. These
The simplest examples of the use of continuous attractors (see sensory inputs subsequently prevent the cumulative errors that
Box 142) to model spatial representations come from mod- would otherwise occur in the integration of angular velocity.
els of the representation of head direction rather than loca- This basic model has been implemented and extended in
tion. In many respects the literature on head-direction cells various ways, developing in hand with our knowledge of the
(HDCs) is much more straightforward than that on place operation of the head-direction system. This is now thought
cells. The overall orientation of the head-direction representa- to involve a circuit from the mammillary bodies (MBs) to
tion can be controlled by sensory cues in a way similar to that anterior thalamic nuclei (ATN) to dorsal presubiculum (PS).
of the place cell representation. Unlike the place cells, how- In this circuit, cells in the MBs code for head direction further
ever, there have been no reports to date of HDCs changing in the future (6070 ms) than those in the ATN (2030 ms),
their preferred orientations relative to each other. Even when whereas those in the PS code for current or past head direc-
the rat is disoriented or is in a symmetrical environment with- tion (0 to 10 ms) (Blair and Sharp, 1995; Blair et al., 1998;
out polarizing cues, the preferred directions of simultaneously Taube, 1998; Taube and Muller, 1998) (see Chapter 11).
recorded HDCs remain in synchronyif they rotate, all rotate Various of the additional detailed properties of these systems,
together. For this reason all models of the head-direction sys- such as time advances and asymmetrical responses during
tem follow the same basic mechanism of a one-dimensional turning, have been modeled more recently (e.g., Touretzky
continuous attractor, or line attractor (Skaggs et al., 1995; and Redish, 1996; Blair et al., 1997; Goodridge and Touretzky,
Zhang, 1996) from which the two-dimensional continuous 2000). However, here I focus on the hippocampus, and return
attractor models of place cells developed (Figs. 143 to 146) to models of location.
(see Box 142).
If HDCs are imagined laid out in a ring with each cells Continuous Attractor Models of Place Cells
location corresponding to its preferred direction and each is
connected to its neighbors (see Box 142), activity is smoothly Samsonovich and McNaughton (1997) produced a detailed
peaked at the current heading direction. Skaggs et al.s model model of the place cell representation as a continuous attrac-
contains two more rings of cells, with each cell receiving con- tor, following Zhang (1996). In this model, the recurrent con-
nections from the corresponding HDC. One ring is composed nections in CA3 are preconfigured to provide several
of left rotation cells, which project back to the HDCs to the continuous attractor representations of location (termed
left (anticlockwise) of their location; the other ring is com- charts). Each chart involves a different set of place cells, the
posed of right rotation cells, which project back to HDCs to relative positions of whose place elds are predetermined. The
the right (clockwise) of their location. The left rotation cells strength of the recurrent connection between two cells in a
corresponding to the current heading direction are activated chart is set as a Gaussian function of the proximity of their
when the rat is turning left because of inputs from the vestibu- place elds. The place cells connect with a path integration
lar system as well as the HDCs, causing the HDC activation to (PI) system, thought to be in the subiculum, in which neurons
move leftward. In addition to these cells, the HDCs receive respond to combinations of the rats location and orientation
724 The Hippocampus Book

(s)

(s)

(s)
Figure 143. Point attractors and line attractors in neural systems preferred direction (top). If ring rates are noisy, it may be diffi-
(see Box 142). A. Point attractor is a pattern of activation a  (a1, cult to estimate the actual value of the variable (middle). Recurrent
a2, a3...) into which other nearby patterns evolve under the dynam- connections can be organized so all other patterns of activation
ics of the network (determined by the pattern of connection evolve into smooth bump-shaped patterns of activation (below).
strengths, update rule, and so on). In some cases a function can be This process can provide an optimal way of decoding the value of
dened that can only decrease under the dynamics (a Lyanpunov the variable from the population (the peak of the bump is the
function, or the energy (E) of a physical system), so that attractor dashed line). C. The set of smooth bump-shaped patterns of activa-
states lie at local minima of this function. B. Population encoding tion form a line attractor, a continuous set of patterns of activity
and decoding. Neural populations can encode the current value of a onto which other nearby patterns evolve but along which move-
variables, say in the pattern of activation across neurons, each of ment is unimpeded. Locations along the line attractor can be
which is tuned to respond to a preferred value. These are often thought of as estimates of the variable (S); all of the patterns of
imagined to be laid out so that the position of each neuron on the activity that end up at a given estimate (such as S 1) form a subspace
ordinate corresponds to its preferred value. For example, a set of a(S) within which the intersection with the line is a point attractor.
head-direction (HD) cells might each be tuned to a different (Source: Adapted from Latham et al., 2003.)

(Sharp, 1996; Cacucci et al., 2004). Specically, the place cells integration in supporting short-term continuity in a cognitive
connect to PI neurons representing similar locations, and the map.
return projections connect back to place cells representing The Samsonovich and McNaughton model is consistent
slightly different locations shifted along the rats direction of with data showing that the stability of the place cell represen-
motion. The gain of this return projection is modulated by tation is dependent on NMDA receptors (Kentros et al., 1998)
information relating to the rats speed of self-motion and that place eld locations remain consistent with each
(presumably carried by motor efference signals). This system, other but slowly drift in the absence of anchoring sensory cues
although demanding a highly specic set of hard-wired or if the rat is consistently disoriented before each trial
connections, provides a self-consistent continuous attractor (Knierim et al., 1995). In addition, the involvement of some
representation of location that moves automatically with self- form of path integration is suggested by the increased inu-
motion. Finally, the hippocampus also receives sensory input ence of the boundary the rat is running from compared to the
so that when the rat is placed in an environment for the one it is running toward (Gothard et al., 1996; OKeefe and
rst time associations between the sensory scene and the Burgess, 1996; Redish et al., 2000). Note that the important
internal representation of location can be formed, which can role played by path integration in this model does not, con-
then be used to reset the system periodically. Overall, the versely, imply that the hippocampus is necessary for guiding
model can be seen as a possible implementation of OKeefe behavior in tests of path integration (e.g., Alyan and
and Nadels (1978, pp. 220230) view of the role of path McNaughton, 1999). The role of the CA3 recurrent collaterals
 S

S S

Figure 144. Continuous attractor networks of place and head tion weights. The asymmetrical component is the spatial derivative
direction cells (HDCs). A. Emergence of a stable ring prole from of the symmetrical component along the direction of drift, and its
an arbitrary initial state in a network of HDCs arranged as a one- size () determines the speed of drift of the represented head direc-
dimensional continuous attractor. The cells are indexed by their pre- tion. D. Two-dimensional place cell network similar to the one-
ferred ring direction and connected by weights with a symmetrical dimensional HDC network, showing emergence of a stereotyped
distribution (i.e., an even function of the difference between cells stable ring prole from an arbitrary initial state, using symmetrical
tuning directionssee C, top row). B. Movement of the peak caused weight distribution (a Gaussian with constant inhibitory back-
by an asymmetrical component in the weight distribution (see C, ground). E. As with the one-dimensional network, the addition of
middle row). C. Distribution of connection weights in a one-dimen- an asymmetrical component to the connection weights causes the
sional attractor network (bottom row), showing a symmetrical com- represented location to drift (again, the asymmetrical component is
ponent (top row) and an additional asymmetrical component the spatial derivative along the direction of drift, and its size deter-
(middle row). Note the slight asymmetry in the combined connec- mines the speed of drift). (Source: Adapted from Zhang, 1996.)

725
726 The Hippocampus Book

Vestibular cell (right) Visual cell with other systems for path integration and for maintaining
Vestibular cell (left) orientation (Touretzky and Redish, 1996). However, no spe-
cic mechanism has been proposed for the path integration.
Because errors accumulate so rapidly in this system, it can be
expected to behave very differently from a perfect system. The
integration by place cells of visual inputs and inputs from a
recurrently connected parietal system, and its relation to
remapping, was explored further by Guazzelli et al. (2001).
Samsonovich and McNaughton suggested that remap-
ping reects the system switching between uncorrelated
charts. This is a reasonable model of the situation after fast-
remapping, which is consistent with each chart acting as an
Rotation cell (left) attractor (Wills et al., 2005), although in this case the charts
Rotation cell (right) would not be precongured but formed by the process of fast
remapping. However, the model is not consistent with the sit-
uations in which individual place elds can move relative to
each other in response to environmental change (e.g., OKeefe
and Burgess, 1996; Fenton et al., 2000). To t these data
Head direction cell
requires feedforward inputs to dominate, replacing the
models main feature. The model is also not consistent with
Figure 145. Skaggs et al.s (1995) model of HDCs, showing the lat- slow remapping (Lever et al., 2002) or partial remapping (e.g.,
eral connections among HDCs providing a continuous attractor
Skaggs and McNaughton, 1998). The experimental conditions
and connections from left or right rotation cells, visual inputs, and
and neural mechanisms resulting in fast versus slow remap-
vestibular inputs. The input from rotation cells serves the same pur-
pose as the asymmetrical component of lateral connections in
ping are currently not well understood (Knierim, 2003).
Figure 144. (Source: Adapted from Skaggs et al., 1995.) Recurrent networks have also been used to model place
eld directionality, as modeled in a feedforward manner
by Sharp (1991). In these models (Brunel and Trullier, 1998;
in self-localization; forming the most accurate representation Kali and Dayan, 2000), place cell ring is initially derived from
of location within a continuous attractor network given con- orientation-specic sensory input at each location (referred to
icting inputs was also stressed by Redish and Touretzky as the local view from that location), and the dynamics of
(1998). In addition, they elaborated the relation of this system the network are strongly dependent on the recurrent connec-

Figure 146. Place cell activity on a chart. The activity of a of the square and is moving to the left and toward the viewer.
population of 36 simultaneously recorded place cells shown The shape of the activity packet does not depend on velocity,
symbolically distributed in the box (in which the rat foraged for acceleration, future trajectory of motion, or theta frequency.
food) each cell being placed at the center of its place eld. Units on (Source: Samsonovich and McNaughton, 1997.)
horizontal axes are centimeters. The animal is located at the center
 Computational Models of Spatial and Mnemonic Functions 727

tions in CA3. If exploration is unconstrained and random, that place cell ring in open elds is initially nondirectional
Hebbian learning in the recurrent collaterals of these models but can become directional as a result of experience.
results in a continuous attractor of the sort hard-wired Finally, note that the recent discovery of grid cells in the
by Samsonovich and McNaughton and an orientation- dorsomedial entorhinal cortex (Hafting et al., 2005) are likely
independent place cell representation of space. One caveat to to have profound implications for ideas of hippocampal func-
this is that the Hebbian learning must be modulated by nov- tion. Each cell res when the rat occupies multiple spatial
elty to prevent inhomogeneous exploration from causing locations laid out on a startlingly regular hexagonal grid. The
highly nonuniform weight structures and unrealistic ring orientation and scale of these grids seems to be constant in
patterns (Kali and Dayan, 2000). Such novelty information different environments and between different nearby cells.
has been suggested as a function of the cholinergic septal Overall, grids get larger as the recording site moves ventrolat-
inputs to the hippocampus (Hasselmo et al., 1996; but see erally. Because the grids of neighboring cells have the same
Hasselmo and Fehlau, 2001 and Lisman and Grace, 2005). As orientation and scale and appear to have xed offsets, it is pos-
with Sharp (1991), in these models directionally constrained sible that the recurrent connections between local sets of grid
exploration results in orientation-dependent place elds. cells could be endlessly tuned to perform path integration,
Kali and Dayan (2000) also simulated the effects of chang- irrespective of the location of the animal, providing the con-
ing the environment on the pattern of place cell ring. The tinuous attractor envisaged in CA3 by Zhang (1996) and
model rst learns a representation in one environment, as Samsonovich and McNaughton (1997). The stable association
described above, and is then exposed to a second environ- of each grid to the environment might occur via connections
ment. If the second environment is sufficiently novel, the pat- to place cells that, by virtue of generally having only a single
tern of sensory input is assumed to be completely different ring location, could be associated with the environmental
and novelty-modulated learning is enabled. As a result, the stimuli at that location. If this is correct, place cell ring could
system successfully learns a second, unrelated, place cell repre- reect the superposition of multiple grids that all overlap at
sentation, corresponding to complete remapping. In simula- the place eld, and remapping could reect changes in offset
tions using two similar boxes (as in Skaggs and McNaughton, in these multiple inputs (OKeefe and Burgess, 2005, see also
1998), similar ring in the two boxes is imposed by similar Fuhs and Touretzky (2000) and McNaughton et al., (2006)).
sensory input, and the model is also shown to be capable of
storing and recalling partially overlapping representations. 14.3.4 Modeling Phase Coding in Place Cells
Finally, if the second environment differs from the rst only
geometrically, the same sensory inputs are assumed to be used The origin of the intriguing phenomenon of the phase coding
(but with different values, as the distances to boundaries have of place cell ring with respect to the concurrent theta rhythm
changed) and there is no new learning. This results in para- of the EEG (OKeefe and Recce, 1993) has been the subject of
metric changes to the pattern of ring. In this case, the effect several computational models. Before considering these mod-
of the recurrent collaterals is to preserve the relative spatial els, I briey review the relevant experimental ndings (see
relation of the place elds and so acts to maintain the posi- Chapter 11 for more details). The theta rhythm is a large-
tions of elds in terms of the ratio of distances between amplitude oscillation of around 6 to 10 Hz of the EEG and
boundaries (effectively scaling up the representation). By con- is present whenever the rat is moving its head through the
trast, the sensory inputs are equivalent to the boundary vector environment. As the rat runs through a place eld, the corre-
cell model (OKeefe and Burgess, 1996; Hartley et al., 2000) sponding place cell tends to re spikes with a systematic phase
and so act to maintain the position of elds in terms of their relation to the theta rhythm. On entering the eld, spikes
absolute distance from boundaries. Thus, this model can are red at a late phase; as the rat passes through the eld,
account for both types of behavior. To t the data in OKeefe spikes are red at successively earlier phases so that on exiting
and Burgess (1996), the model should be dominated by the the eld the phase of ring may be up to 360 earlier (corre-
sensory input rather than the recurrent connections, as place sponding to an early phase). This effect is most pronounced
elds tend to maintain a xed distance to environmental when the rat runs repetitively in a directionally constrained
boundaries in this experiment. By contrast, the apparently manner (e.g., on a linear track). Interestingly, the phase of r-
similar experiment of Muller and Kubie (1987) results in ing correlates better with the location of the rat in the place
complete remapping. It is not yet clear whether models such eld than with other variables such as the time spent in the
as that of Kali and Dayan (2000), which depend on assigning eld or the instantaneous ring rate of the cell (Huxter et
different amounts of similarity to two environments, can cap- al., 2003).
ture the complexities of the data concerning remapping. For The apparently tight coupling between location and phase
example, the differences between fast and slow remapping implies that the phase of ring might depend on the sensory
referred to above imply some learning in the feedforward con- input driving the place cell in a simple feedforward way. Thus,
nections (to model slow remapping, see, e.g., Barry et al., OKeefe (1990, 1991) proposed that each place cell might act
2005; Fuhs and Touretzky, 2000), as well as in the recurrent as a phasor representing location relative to the centroid and
connections (to model fast-remapping). Like Sharps (1991) eccentricity, or slope, of a subset of the sensory cues in an
model, these recurrent models are also inconsistent with hints environment: the ring rate encoding proximity to the cen-
728 The Hippocampus Book

troid and the phase relative to theta encoding the bearing to stronger than that between phase and rate; and on the linear
the centroid relative to the direction dened by the eccentric- track at least, the weaker correlation is a side effect of the
ity, or slope, of the subset of cues. The advantages of the pha- stronger one (OKeefe and Burgess, 2005).
sor representation for simple vector calculations were stressed As with models of place cell ring, a second type of model
in this model, but its easy relation to the subsequent experi- of the phase shift stresses the recurrent connections in contrast
mental data on phase is at least as signicant. Burgess et al. to the feedforward connections. Indeed, simulations of the
(1993) suggested a direct interpretation of phase of ring: that Samsonovich and McNaughton model, in which net activa-
it could be used to separate place cells with elds ahead of the tion is made to oscillate at the theta frequency, shows some-
rat from those with elds behind the rat, which can be useful thing qualitatively similar to the phase shift owing to path
for navigation (see Section 14.4.2). This was subsequently ver- integration occurring within each cycle. That is, initially the
ied experimentally (Burgess et al., 1994; Skaggs et al., 1996; set of active place cells settles to just those place cells with
Samsonovich and McNaughton, 1997). They also demon- elds centered on the rat and then expands to include those
strated that a phase shift in each individual cell is consistent with elds centered ahead of the rat. The rst quantitative
with theta modulation of the net activity of the population of model of the phase shift was proposed by Tsodyks et al. (1996).
cells. In a simple feedforward model of place cell ring, In this model, the recurrent connections between place cells in
Burgess et al. (1994) simulated the phase of ring as a func- CA3 are asymmetrically arranged so that each place cell proj-
tion of the rats position relative to the sensory cues that drove ects to place cells farther along a learned path (see also Blum
the cell. The input to entorhinal comes from pairs of sen- and Abbott, 1996, discussed below). External input to a CA3
sory cells, each tuned to respond at a given distance from a place cell arrives at a xed (early) phase of theta, causing place
particular sensory cue such that the maximum input ampli- cell activity at this phase, which in turn propagates through
tude occurs at the centroid of the two cues. The phase of r- the recurrent connections to place cells with elds farther
ing of entorhinal cells was assumed to reect the angle of the along the path and causes them to re. Overall activity is
cue-centroid from the rat: ring at a late phase when driven by inhibited at the end of each theta cycle, preventing further
cues ahead of the rat and at an early phase when driven by propagation of activity into the next cycle. Thus, when the rat
cues behind the rat. This model is broadly consistent with the enters a place eld, the corresponding cell starts to re at a late
synchrony and oscillations seen in some sensory circuits (e.g., phase owing to propagated activity from cells with elds ear-
Nicolelis et al., 1995) but remains more qualitative than quan- lier on the path; and it res earlier during each cycle as the rat
titative. Bose and Recce (2001) propose an alternative model advances owing to activity having to propagate through fewer
that also accounts for the lack of phase shift seen in place cells cells, until nally ring at the early phase is due solely to exter-
when a rat runs in a running wheel (Hirase et al., 1999), nal inputs. Several similar mechanisms that depend on the
involving the dynamics of the interaction of place cells and association of place cells ring earlier along a learned path to
interneurons and the assumption that the frequency of the those ring later along it have now been proposed (Jensen and
theta rhythm depends on running speed on the linear track Lisman, 1996; Touretzky and Redish, 1996; Wallenstein and
but not in the running wheel. However, this picture became Hasselmo, 1997). The Jensen and Lisman (1996) model makes
more complicated with the subsequent claim that phase shift- interesting suggestions for the gamma rhythm, in separating
ing does occur in the running wheel at high ring rates the ring of cells corresponding to the current and succes-
(Harris et al., 2002). sively farther advanced locations, and for the dynamics of
An appealingly simple alternative feedforward model is NMDA channels, in separating each route retrieval into suc-
possible (Mehta et al., 2000; Harris et al., 2002): The excitatory cessive cycles of the theta rhythm. Wallenstein and Hasselmo
synaptic input to a place cell might increase as the rat runs (1997) emphasized the role of GABAB receptors in varying the
through the place eld, and the theta rhythm might reect a relative inuence of the inputs to CA1 from CA3 compared to
sawtooth-shaped inhibitory input (i.e., inhibition decreasing those directly from entorhinal cortex (EC) over the theta cycle:
through each cycle). In this model, the ring phase would allowing sensory (EC) input to dominate early and predictive
advance simply because the increasing excitatory input man- input from CA3 to dominate late in the cycle.
ages to overcome the inhibitory input successively earlier in These models t nicely with the observations of Mehta et
each cycle. The cause of the increasing excitatory input might al. (1997, 2000): They produce a phase shift limited to 360
reect an exaggerated form of the asymmetry reported by that is more strongly correlated with position than time, and
Mehta et al. (1997) or an increasing then decreasing input but that is greater for well learned paths than for random explo-
with lack of ring on the decreasing portion due to effects ration. Interestingly, Mehta et al. (2000) suggested a similar
such as habituation (Harris et al., 2002). These models capture model based on learned asymmetry in the connections from
the observation that the phase shift becomes more reliable CA3 to CA1, the main testable difference here being that the
over the rst few runs of a trial, as does the asymmetry of phase shift should not be observed in CA3. Other aspects of
place elds (Mehta et al., 2002), and allows phase to be ana- these models seem less likely. First, because the initial ring
lyzed in terms of ring rate during nontranslational behaviors of a place cell depends on both the externally driven activity
such as dreaming or wheel running (Harris et al., 2002). of other cells and its propagation through the network, it
However, the correlation between phase and location is seems likely that on cell-by-cell and run-by-run bases the
 Computational Models of Spatial and Mnemonic Functions 729

initial phase of ring should be more variable than the (exter- (see Chapter 13). This is comparable to behavior in smaller
nally driven) nal phase of ring. This is not the case in the scale spaces and over shorter durations, such as visually guided
data (Skaggs et al., 1996; Huxter et al., 2003). Second, one reaching, which are most commonly associated with the pos-
mechanism for producing the required asymmetrical connec- terior parietal lobe (e.g., Burgess et al., 1999). As with models
tions is that suggested by Mehta et al., (2000). However, it has of spatial representation, these models can be approximately
since been found that, although the development of asymme- divided into those stressing the role of feedforward connec-
try in place elds over the rst few runs of a trial is prevented tions and those stressing the role of recurrent connections.
by blockade of NMDA receptors, the phase shift phenomenon
is unaffected by this manipulation (Ekstrom et al., 2001). 14.4.1 Spatial Navigation: Data
A third type of model stresses the inherent oscillatory
nature of some cellular processes, as did Jensen and Lisman Behavioral data indicate that rats learn about the spatial lay-
(1996) for different reasons. Thus, OKeefe and Recce (1993) out of their environment during exploration in the absence of
pointed out that the phase and amplitude characteristics of explicit goals or rewards (e.g., Tolman, 1948). They can also
place cell ring could be modeled as the interference pattern prot from being placed at the goal location without having
between an 11-Hz external input to the cell (perhaps the sen- explored the rest of the environment (Keith and McVety,
sory input) and a 9-Hz external or internal oscillation corre- 1988). These processes are referred to as latent learning. Rats
sponding to theta (perhaps driven by the septal input). This also appear to be able to perform short cuts and detours.
produces an oscillation of 10 Hz, corresponding to ring that These abilities contributed to the idea that rats form a cogni-
shifts in phase relative to theta and a 1-Hz envelope, one-half tive map of their environment rather than simply learning to
cycle of which corresponds to the place eld. This model has associate individual stimuli with responses (Tolman, 1948;
since been extended (Lengyel et al., 2003), identifying the rst OKeefe and Nadel, 1978). Of course, the learning of stimulus-
input as a voltage-controlled oscillation of the membrane response associations also plays an important role in spatial
potential (e.g., Hoppensteadt, 1986) in the dendrites and the navigation, such as when the goal is directly visible or when a
second as an inhibitory input to the soma of xed frequency. well learned turn or sequence of turns is to be performed.
The frequency of the dendritic oscillation is assumed to However, there seems to be good evidence that these types of
increase above that of the somatic oscillation proportionally behavior are less dependent on the hippocampus than those
with the strength of the dendritic input, which is assumed to associated with cognitive mapping (OKeefe and Nadel, 1978;
be zero outside the place eld and proportional to the rats Morris et al., 1982; Packard and McGaugh, 1996). These issues
running speed within it (McNaughton et al., 1983; Czurko et are dealt with in more detail in Chapter 13.
al., 1999; Ekstrom et al., 2001; Huxter et al., 2003). Thus, the The other main spur to the association of the hippocampus
two oscillations destructively interfere outside of the place with a cognitive map of the rats environment was the discov-
eld, whereas phase of ring relative to the somatic input ery of place cells whose response is not easily described in
within the eld can shift more rapidly as the rat runs faster, terms of a simple response to a single stimulus. However, a gap
preserving the relation between phase and location. In addi- remains between the properties of place cells and the proper-
tion, the dendritic oscillation needs to be weakly driven in ties required of a system for spatial navigation. Two features of
antiphase to the somatic input so in the absence of any den- place cell ring are particularly problematic. First, information
dritic input it returns to being in phase with it to ensure com- about a place in an environment (i.e., ring of the correspon-
plete destructive interference. Recent corroborative evidence ding place cells) can only be accessed locally (by actually visit-
for interference models comes from the observation that the ing that place). Second, place elds appear to be no more
increase in place eld size along the dorsoventral axis of the affected by the location of the goal than by the location of any
hippocampus parallels a corresponding decrease in the intrin- other cue. That is, place cells tell you only where you are cur-
sic ring frequency of place cells, reducing toward the theta rently and not where to go to get to your goal (Speakman and
frequency in more ventral regions (Maurer et al., 2005). OKeefe, 1990). A caveat to the rst point may be indicated by
Intriguingly, the full, repeating, interference pattern is the recording of spatial view cells in the macaque hippocam-
expressed in the recently discovered entorhinal grid cells pus (Rolls et al., 1997), which re as a function of where the
(McNaughton et al., 2006) as predicted by an inference model monkey is looking rather than where it is physically located.
(OKeefe and Burgess, 2005).
14.4.2 Spatial Navigation: Feedforward Models

Zipsers (1986) view eld model was built on the observa-

14.4 Hippocampus and Spatial Navigation tion that in some circumstances place cell ring is modu-
lated by the orientation of the rat. In this model, a set of
In this section we consider the contribution of hippocampal orientation-dependent place cells, or view-field units,
spatial representations to guiding behavior. I focus on large- become associated with a set of goal units, which encode
scale spatial navigation, the spatial behavior most commonly the direction to the goal relative to the current heading direc-
associated with the hippocampus and medial temporal lobes tion. So long as the appropriate cells become associated, the
730 The Hippocampus Book

population vector of directions represented by the goal units the goal would be strongly affected if stereotyped routes were
guide the rat to the goal, as goal units driven by place cells rep- used during learning.
resenting the current location re the most strongly. However, A related way to think about spatial navigation is to imag-
it is unclear how the directions to the goal would become ine dening a surface over the environment such that gradient
associated with the place units if the goal were not visible. In ascent on it leads to the goal. The simplest model of this sort
a second model (the beta coefficient model), Zipser sug- is for the place cells to be connected to a goal cell via reward-
gested that the location of the goal relative to subsets of land- modulated Hebb-modiable connections such that encoun-
marks is calculated and stored (as the coefficient of the linear tering a goal causes strengthening of connections to it from
sum of landmark locations that is equal to the goal location). the concurrently active place cells (Burgess and OKeefe,
In this model, learning at the goal location is sufficient, 1996). The subsequent activity of the goal cell then increases
although the neural mechanisms required to implement the with the proximity of the goal, as the net activity of those
desired calculations are not explained. One possible route to place cells with strengthened connections increases with the
enabling place cells to perform this type of computation was proximity to the goal (Fig. 147). The task for the rat in nd-
provided by the phasor model of OKeefe (1991) (see above). ing the goal is then to move in the direction that increases the
A simplistic version of this type of model was simulated by rate of ring of the goal cell representing the desired goal (Fig.
Wilkie and Palfrey (1987), in which the distance of the goal 147B). This type of model qualitatively captures the rapid
from two landmarks is stored so that navigation back to the nature of learning a goal location once place cell ring has
goal can be effected by moving so as to match these landmark become established and the ability to learn simply by being at
distances. the goal location rather than having to nd it many times.
Several models have followed Zipsers view eld model in However, nding the goal would involve the rat hunting
associating places to movements or local views. See Trullier et around to determine the best direction in which to move. This
al. (1997) for a wider review of biologically based articial behavior, known as vicarious trial and error, can be observed
navigation systems. McNaughton and Nadel (1990) suggested at choice points but is not common. A second problem raised
that routes might be learned as a chain of associations from a by this model is the range over which spatial information
local view to an action and thence to the next local view, and is accessible. If there are no place cells that re at both the
so on. This model was not actually simulated, and simply stor- goal location and the current location of the rat, there is
ing routes is insufficient to enable spatial navigation (see no gradient in the ring rate of the goal cell (being locally
Section 14.6). Even if a given route can be correctly selected in zero). Hence, this type of model requires the population of
terms of the locations to which it leads, navigational abilities place cells to include some cells that have nonzero ring rates
such as generating novel shortcuts and detours are beyond a at any two points in the environment, however far apart.
simple route-based system. The task of accumulating route- Consideration of the population of place eld shapes and sizes
independent spatial information faces several problems, in the environments used so far (i.e., not more than about 2 m
including the credit assignment problem: deciding which in diameter) indicate that this is quite probable (Hartley et al.,
actions along a route are critical for determining whether it 2000). However, whether this is true of larger environments is
eventually leads to the goal. not known.
Brown and Sharp (1995) provided a more sophisticated A more sophisticated model of navigation (Burgess et al.,
model for associating locations with actions. In their model 1994) was proposed to make two improvements on the simple
the possible actions in a place are represented by a left-turn model above. The rst was to be able to calculate the direction
cell and a right-turn cell (in nucleus accumbens) driven by to the goal from any subsequent location after a single visit to
each place cell. These turn cells receive modiable connec- the goal location without having to hunt around. The model
tions from head-direction cells (HDCs), which support the made use of the fact that rats placed at the goal location like to
rats spatial learning. When the rat reaches the goal, connec- rear up and look around and the supposition that place cells
tions between HDCs and turn cells are modied according to ring at a late phase relative to the theta rhythm have elds
a recency-weighted index of their simultaneous activity. Thus, peaked ahead of the rat (see Section 14.3.4). The model posits
if turning left in a particular place when facing north leads a set of goal cells for each goal, such that each cell in a set is
immediately to the goal, the HDC representing north becomes associated with a different head direction. The connections
more strongly associated with the left-turn cell driven by the from place cells to a north goal cell are modied when the
corresponding place cell. The recency weighting of connection rat is at the goal and facing north and similarly for the goal
modication is designed to provide an approximate solution cells associated with other directions. Crucially, connections
to the credit assignment problem (when applied over many are modied at late phases of the theta cycle so the active
trials) by effectively dividing credit according to the number place cells (i.e., those from which connections are strength-
of steps within which an action leads to nding the goal, see ened) are cells with elds ahead of the rat. Therefore, after
reinforcement learning (e.g., Dayan and Abbott, 2002) for a spending at least one theta cycle at the goal facing in each
more principled approach. This model successfully simulates direction, a north goal cell has strong connections from place
learning in the Morris watermaze but would not show latent cells with elds peaked to the north of the goaland similarly
learning; performance would not be affected by whether the for south, east, and west goal cells. As a consequence, a north
rat can look around from the goal location; and navigation to goal cells ring rate forms a surface over the environment that
 Computational Models of Spatial and Mnemonic Functions 731

Figure 147. Simple model of place cells and navigation. A. A the proximity of the goal (G). C. Firing rate maps of four goal cells
goal cell stores a goals location by taking a snapshot of place cell whose population vector codes for the goal location (G). Each is
activity via Hebbian synaptic modication when the goal cell is associated with the allocentric direction ui in which the location of
excited by the attributes of a particular goal location. Solid circles its peak ring rate is displaced from G; thus, the vector sum of the
are active place cells; open circles are inactive place cells; and solid directions ui weighted by the instantaneous ring rates fi of the goal
squares mark potentiated synapses between place cell axons and cells (i.e., i fiui/i fi) codes for the direction of the rat from the goal,
goal cell dendrites. B. Firing rate map of the goal cell (roughly an and the net ring rate of the goal cells (i.e., i fi) codes for the goals
inverted cone) during subsequent movements of the rat codes for proximity. (Source: Adapted from Burgess and OKeefe, 1996.)

is peaked to the north of the goal locationand similarly for (Sharp and Green, 1994), and larger place elds have since
south, east, and west goal cells (Fig. 147C). This is useful for been found in the ventral hippocampus (Jung et al., 1994) that
subsequent navigation, as the rat can now use the relative r- might also serve this purpose.
ing rates of the various goal cells to indicate the direction of To model hippocampal navigation Foster et al. (2000; see
the goal. If the rat is to the north of the goal, the north goal also Dayan, 1991) used temporal difference learning in
cell has a higher ring rate than the south goal cell. More pre- which the state corresponds to the place cell representation
cisely, the population vector (Georgopoulos et al., 1986) of a of the rats current location, and its value reects the expected
set of goal cells successfully indicated the direction of the rat number of steps to reach the goal (one unit of reward is
from the goal in simulations, and different sets of goal cells received on reaching the goal). Temporal difference learning
can be used simultaneously to indicate different locations of (Sutton and Barto, 1988) can be implemented by simply con-
interest. Conversion to egocentric (e.g., left or right) move- necting place cells to actor and critic units with connec-
ment directions was envisaged to take place in the parietal tions that are adjusted according to a modied Hebbian rule.
cortex or basal ganglia, given knowledge of the current head Activation of the critic unit is the evaluation of the current
direction (Burgess et al., 2001a). state (the expected future reward from that state, discounted
The second improvement was to ameliorate the problems by distance into the futuree.g., Dayan and Abbott, 2002), a
of the range over which spatial information is available from more principled analogue of the simple goal cell above. The set
place cell ring. The proposed solution was to interpose a set of action units, only one of which can be active at a given time,
of subicular cells between the place cells and goal cells. Given represent movements North, South, and so on. At each step the
weaker inhibitory competition between subicular cells than connection weight from a place cell to the critic unit or to the
between place cells, competitive learning in the connections active actor unit is adjusted by an amount proportional to the
from place to subicular cells during exploration causes the place cells activity times the amount by which the reward
subicular cells to build up larger ring elds, each effectively exceeds that expected from the change in the activity of the
composed of several place elds. This learning is goal- critic unit (Fig. 148). This type of learning is consistent with
independent and corresponds to latent learning in preparing a role for dopaminergic modulation of LTP (e.g., Montague et
the ground for effective one-shot learning of the location of al., 1996; Schultz et al., 1997). Over many routes to the goal,
any goals should they be encountered. Large spatial ring ideally involving performing all actions at all locations many
elds in subicular cells are consistent with experimental data times, this rule causes (1) the critic to provide an accurate esti-
732 The Hippocampus Book

Figure 148. Learning in the actor-critic system in a watermaze. and tortuous path is taken to the platform. Trial 7: The critics value
The plots for each trial show the critics value function C(p) (above), function having peaked in the northeast quadrant of the pool, the
the preferred actions at various locations (below left; the length of preferred actions are correct for locations close to the platform but
each arrow is related to the probability that the particular action not for locations further away. Trial 22: The critics value function
shown is taken), and a sample path (below right). Trial 2: After a has spread across the whole pool and the preferred actions are close
timed-out rst trial, the critics value function remains zero every- to correct in most locations, so the actor takes a direct route to the
where, the actions point randomly in different directions, and a long platform. (Source: Foster et al., 2000.)

mation of value and (2) the appropriate actions to be associ- 14.4.3 Spatial Navigation: Feedback Models
ated with each state (Fig. 148). For a given conguration of
goal and environment, this can provide the optimal strategy, As well as being linked with spatial representation and asso-
which is not the case with the approximate recency weighting ciative memory (see Box 142), the CA3 recurrent collaterals
(see Brown and Sharp, 1995, above; Blum and Abbott, 1996, have also been proposed to play a role in spatial navigation.
below) or, if obstacles are present, with goal cells simply indi- The rst model to formalize such a role was suggested by
cating the physical proximity of the goal (Burgess et al., 1994). Muller et al. (1991, 1996) and focused on the Hebb-associative
However, learning is both experience-dependent and goal- effects of LTP on the recurrent collaterals. If pre- and postsy-
dependent (e.g., to know to head north from a given place naptic ring within a short time interval leads to a small
results from this combination having previously led to the increase in synaptic strength, the ring of place cells as the rat
goal) and so perhaps does not capture the characteristics of the moves around an environment leads to the strength of a con-
hippocampal contribution to spatial learning. nection between two place cells depending on the proximity
To simulate goal independent learning over many trials in of their place elds. This occurs simply because the greater the
which the location of the goal changes, Foster et al. (2000) overlap between place elds the more often they re near-
proposed that a second system learns to form a coordinate coincidentally during random exploration. Muller et al.
representation of the rats position by using the rats locally (1996) showed that after extensive exploration the synaptic
accurate ability to estimate self-motion. The place cells are strengths represent a cognitive graph, each approximately
connected to two units that learn to estimate the x and y coor- representing the minimum path length between the centers of
dinates of the rat, again using temporal difference learning to the place elds of the cells it connects. Their model proposes
adjust connection weights. In this system, the change in a con- that the rat navigates by moving through the place elds of the
nection weight to the x unit is proportional to the place cell cells most strongly connected to the cells with elds at the cur-
activation times the amount by which the change in x, as esti- rent and destination positions. This mechanism, reminiscent
mated by self-motion, exceeds that estimated by the change in of a resistive grid (Connelly et al., 1990), works well but relies
the activity of the x unit. The explicit representation of x and on a graph search. It is not easy to imagine how such a process
y coordinates enables accurate navigation after one exposure could be implemented in a biologically plausible fashion,
to the goal and so corresponds well with latent learning. given the apparent lack of inuence of the goal location on the
However, it is not clear if such a representation actually exists ring of place cells (Speakman and OKeefe, 1990). One sug-
in the brain, or if it would necessarily be more useful than a gestion (Gorchetchnikov and Hasselmo, 2002) is that activa-
place eld-like representation that covered the appropriate tion corresponding to the goal location occurs in entorhinal
length scales (e.g., Burgess et al., 1994). cortex while activation corresponding to the current location
 Computational Models of Spatial and Mnemonic Functions 733

occurs in CA3. Activation in each region spreads along the graph in which locations are represented as nodes connected
available paths (although more slowly in CA3 than entorhinal by (asymmetrical) edges that represent the movement neces-
cortex) until a commonly activated location is detected in sary to get from one node to the next. In terms of the naviga-
CA1. This location represents the next immediate destination tion of autonomous agents, Scholkopf and Mallot (1995)
for the rat, although how this information is interpreted in described a view graph in which the local view or sensory
terms of whether to turn left or right is not described. perception at given locations form the nodes and the actions
In a related model, Blum and Abbott (1996) make use of required to get from one view to the next form the edges. See
the temporal asymmetry of LTP to strengthen recurrent con- also McNaughton and Nadel (1990) for a description of how
nections from CA3 place cells that re early on the rats path a view graph might be implemented in the hippocampus.
to those that re later along it. This causes the activation of Mallot and Gillner (2000) argued that such view graphs are a
place cells to spread backward along the path (see also Mehta good model for human navigation despite their simplicity.
et al., 1997, 2000 and models of spatial representation, above). One key requirement for building a world graph is to be able
If the ring of place cells is interpreted by systems down- to decide whether to assign a new node to a location. This can
stream of CA3 as representing the location of the rat, this shift be done on the basis of its familiarity (see Touretzky and
in ring (backward along the path) would be interpreted as a Redish, 1996; Kali and Dayan, 2000, for models relating to
shift in the location of the rat forward along the path. Blum this) or possibly on the basis of the sequence of actions that
and Abbott suggested that navigation along a previously per- lead back to a location. Lieblich and Arbib further suggested
formed route could be undertaken by moving from the cur- that the nodes of a world graph might also represent a loca-
rent location (e.g., read from CA1) to the shifted location tions motivational valence and thus become a general model
represented in CA3, although how this could be implemented for goal-directed behavior, even though its creation might
was not described. To enable navigation to a goal location, the correspond to latent learning.
rules for synaptic change were modied to be proportional to
the amount of pre- and postsynaptic activation weighted by
how recently it occurred prior encountering the goal (i.e.,
modication similar to that suggested by Brown and Sharp, 14.5 Hippocampus and Associative
1995, above). or Episodic Memory
It is interesting to note that the symmetrical pattern of
connection strengths learned in Muller et al.s model resem- In contrast to the vast amount of animal work linking hip-
bles that used in continuous attractor models of place cell r- pocampal damage to decits in spatial memory, the major
ing and so also serves to produce a consistent pattern of impairment noted in humans following bilateral damage to
activity to represent each location (see above). By contrast, the the hippocampus is amnesia: a much more general impair-
additional asymmetry in connection strengths learned in ment in memory. The extent of this impairment into various
Blum and Abbotts model serve to shift the represented loca- subdivisions of memory and into information acquired prior
tion along the learned route. Indeed, the latter property has to the damage is a contentious issue (see Chapter 12). Here I
been shown to allow the represented location to move briey review the data on human hippocampal function in
smoothly along the route over time, suggested as a model for memory, introduce the generic model for it derived from
mental replay during sleep (Redish and Touretzky, 1998). The Marrs seminal paper in 1971, and discuss various details and
goal-independent encoding of spatial proximity (as in the developments of this model over the years.
symmetrical connections of Muller et al.s cognitive graph) or
of previously traveled routes (as in the asymmetrical connec- 14.5.1 Hippocampus and Memory: Data
tion of the later models) correspond to latent learning.
However, the way these models incorporate a new goal loca- Substantial bilateral damage to the hippocampus and medial
tion necessarily requires many trips to the goal and thus prob- temporal lobes almost invariably leads to amnesia, character-
ably falls short of a rats abilities. Another drawback associated ized as a drop in the memory component of the intelligence
with these models is that none makes clear the details of how quotient (MIQ) of at least 20 points relative to full-scale IQ.
the rats brain might deduce the direction it should move in or Because only a relatively small number of cases of damage
if it would be able to generate a short-cut or detour. They also restricted to the hippocampus have been studied, it is difficult
assume place elds of xed relative location but might still to draw general conclusions regarding its role in memory, in
make some interesting predictions regarding the locus of contrast to the roles of the surrounding cortical areas.
search in environments that had changed in shape or size. To However, some general points can be made (see Spiers et al.,
build up a true distance metric in complex environments 2001 and Chapter 12 for details). They include a ubiquitous
would take a long time, in common with the reinforcement decit in memory for personally experienced events that
learning approaches (see Foster et al., 2000, above). occur after the lesion (i.e., an anterograde deficit in
These models can be viewed in the context of a more gen- episodic memory) and spared procedural and working
eral set of higher-level algorithms for navigation based on memory. The extent of retrograde amnesia (loss of memory
directed graphs. Lieblich and Arbib (1982) described a world for information acquired prior to the lesion) appears to vary
734 The Hippocampus Book

across patients and possibly across types of information neocortical representation of an event into a simple repre-
(Nadel and Moscovitch, 1997). Memory loss can extend over sentation in hippocampus, with modiable connections to
the entire lifetime or can be restricted to shorter periods prior and from the hippocampus storing the mappings between the
to the damage; but it does seem to be relatively limited in the full representation and the simple representation. The CA3
case of lesions to the fornix. More controversial ndings recurrent collaterals are modied to store the simple repre-
include relative sparing of semantic memory (memory for sentation as an associative memoryso if a simple represen-
facts) and of familiarity-based recognition in some cases of tation is incompletely activated, the collateral effect results
focal hippocampal damage. Relative sparing of recognition in the full representation being recovered. Thus, partial acti-
memory is consistent with ndings in monkeys showing that vation of the neocortical representation of an event can lead
this type of memory is perhaps more strongly dependent on to complete reactivation of its simple representation in the
nearby cortical areas (e.g., Gaffan, 1994; Zola-Morgan et al., hippocampus, which in turn can reactivate the entire neocor-
1994; Aggleton and Brown, 1999; Baxter and Murray, 2001). tical representation. In Marrs view, the capacity of the hip-
Finally, it should be noted that the human hippocampus, par- pocampal system should be enough to store a days events so
ticularly in the right hemisphere, is also involved in spatial the process of categorization and long-term storage in neo-
navigation (Burgess et al., 2002). cortex can take place during the following nights sleep. Note
that this now seems at odds with the much larger extent of ret-
14.5.2 Marrs Hippocampo-neocortical rograde amnesia following hippocampal lesions, and the pos-
Model of Long-term Memory sibility that some types of information (e.g., episodic, in
contrast to semantic) remain forever dependent on the hip-
Most modeling work on the role of the hippocampus in mem- pocampus (Nadel and Moscovitch, 1997). Marr further sug-
ory can be considered part of a long tradition reaching back to gested that the simple representations need reect only those
Marr (1971). In this section I sketch the main components of parts of the event through which they are addressed, and that
Marrs model, which provide a common framework for all of they should be sparsely encoded to reduce possible interfer-
the subsequent models, and indicate how these components ence between representations. The sparseness of a representa-
correspond to various aspects of the data on human memory tion refers to the fraction of neurons that are active: If this
(see also Willshaw and Buckingham, 1990; Burgess et al., fraction is low, the chance of the same neuron being active in
2001a). I refer to this as the generic hippocampo-neocortical the representation of different events is small, and interference
model (Fig. 149). between representations is minimized.
In Marrs (1971) model, events in the outside world are It should be noted that the proposed capacity of the hip-
represented by patterns of activity in neocortical areas. The pocampus has varied widely in subsequent models, as memo-
role of the hippocampus is to store these representations over ries remain hippocampus-dependent for more than 1 day and
the short term so relevant events can be categorized and possibly never become fully independent of it (see below).
stored for the long term in neocortex (see Marrs 1970 model The issue of capacity itself is also confounded by the need to
of the cerebral neocortex). This is achieved by mapping the specify how to divide continuous experience into discrete
binary patterns. Another confusing issue is the correspon-
dence between Marrs rather abstract generic model of mem-
Figure 149. Generic hippocampo-neocortical model of long-term ory and the processes of retrieval in human memory. For
memory. Strong connections and active cells are shown in black. example, it is not clear whether the model refers to recogni-
Relatively dense recurrent connections and sparse representations tion or recollection, where the incomplete retrieval cue comes
in the hippocampus enable efficient pattern completion. Connec- from, or whether the model applies more to memory for some
tions between neocortex and hippocampus allow the hippocampal kinds of information than others. With regard to retrieval
representation of an event to be associated with its sensory details,
cues, the prefrontal cortex probably plays an important role in
including reactivation of the representations in various neocortical
the strategic organization of retrieval, taking this issue beyond
areas dealing with different sensory modalities. Abstracted seman-
tic representations may also be learned over time in the neocortex. the scope of this chapter (e.g., Roberts et al., 1998). By the end
The recurrent connections in each neocortical area allow unimodal of Section 14.5 some of the other issues are resolved, and oth-
recognition. (Source: Burgess et al., 2001a.) ers are discussed further in Section 14.6.

hippocampus 14.5.3 Associative Memory

and the Hippocampus

Much of the development and analysis of the generic model

has followed from basic research on the associative properties
of feedforward networks (e.g., Willshaw et al., 1969) and
recurrent networks (Kohonen, 1972; Gardner-Medwin, 1976;
Hopeld, 1982) based on Hebbian learning (Fig. 1410).
Much of the initial development of the model concentrated on
neocortex matching the major anatomical properties of the hippocam-
 Computational Models of Spatial and Mnemonic Functions 735

Heteroassociation Autoassociation
0 1 0
0 0 0
1 1 0 B
0 0 1
1 1 1
1 0 1
x3 x2 x1 1 0 1 0 1
1 0 0 0 0
0 1 0 1 1

A 1
0
0
0
1
1
1
1
0
0
1
1
1
0
0
y1 y2 y3 x2 x1
axon
dendrite unpotentiated synapse
input neuron potentiated synapse
output neuron detonator synapse
inhibitory interneuron inhibitory synapse

y inputs

C y inputs
1 0 0 1 1 0 y3 D 1
1
1
0
0 0 0
0 1 1
1
0
y4
y3 E 0 0 1 0 1 1 x2
1 0 1 1 0 0 x1
0 0 1 0 1 1 y2 0 0 1 0 1 1 y2
1 1 0 1 0 0 y1 1 1 0 1 0 0 y1 0 1 1 0 1 1 0 0
0 0 0 0 0 0 0 0

x inputs
0 1 0 0 0 1 0 1 1 0 0 1 0 0 0 1 0 1 1 1 1 1 0 1 1 1 1
0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 0 0 0 1 0 1 1 0 1 1 0 0
x inputs
x inputs

1 1 0 1 0 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 0 0 1 0 1 1
0 0 1 1 1 0 1 0 0 1 0 0 1 1 1 0 1 0 1 1 0 0 0 1 0 1 1
1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 x2 x1
1 0 1 1 1 0 1 1 0 0 1 0 1 1 1 0 1 1 0
x3 x2 x1 x4 x3 x2 x1 PATTERN COMPLETION
(0 0 1 0 0 1) is part of x2
CORRECT RECALL SATURATION (0 0 1 0 0 1) C = (1 0 2 1 2 2)
x3 = (0 0 1 0 1 1) x4 = (0 1 1 1 0 0) (1 0 2 1 2 2)/ 2 = (0 0 1 0 1 1) = x2
x3 C = (3 2 2 3 3 2) x4 C = (3 3 1 2 1 3)
(3 2 2 3 3 2)/ 3 = (1 0 0 1 1 0) = y3 (3 3 1 2 1 3)/ 3 = (1 1 0 0 0 1) = y4 ERROR CORRECTION
BUT (0 0 0 1 1 1) is a corrupted x2
PATTERN COMPLETION x3 C = (3 3 2 3 3 2) (0 0 0 1 1 1) C = (1 0 3 1 2 2)
(0 0 1 0 0 1) is part of x3 (3 3 2 3 3 2)/ 3 = (1 1 0 1 1 0) =/ y3 (1 0 3 1 2 2)/ 3 = (0 0 1 0 0 0) =/ x2
(0 0 1 0 0 1) C = (2 1 1 2 2 1) BUT
(2 1 1 2 2 1)/ 2 = (1 0 0 1 1 0) = y3 (1 0 3 1 2 2)/ 2 = (0 0 1 0 1 1) = x2
Figure 1410. Biological implementation of associative memory in binary vectors (e.g., x1, y1) are presented to the system for storage.
the hippocampus. A. Heteroassociative network associates pattern A Hebbian learning rule is used: A synaptic connection is strength-
of activity y1 with input x1, and y2 with x2, and so on to form an ened (set to 1 from 0) given pre- and postsynaptic activity (i.e., both
associative memory. (The matrix of connection weights shown in A inputs set to 1). Because the net input to a cell is the sum of inputs
is the result of successive presentations of input pattern x1 and out- multiplied by the strength of the connection mediating it, the net
put pattern y1, x2 and y2, x3 and y3 to an initially blank matrix.) input to a cell corresponds to the number of active inputs arriving
This enables input x1 to reproduce activation y1. Even an incom- via strengthened connections. To re, the net input to a cell must
plete or corrupted version of x1 can reproduce y1 (see C). Storing equal the number of currently active inputs. Thus, retrieval of a vec-
too much information leads to saturation of the system and tor given its corresponding paired associate is achieved by matrix-
retrieval failures (see D). B. Autoassociative network associates pat- vector multiplication (e.g., pattern y3 in A is extracted by multiply-
tern x1 with itself, x2 with itself, and so on to form an autoassocia- ing the matrix rows by corresponding elements of vector x3 and
tive memory. This enables an incomplete or corrupted pattern to be summing the columns) followed by integer division by the number
cleaned up (see E). The essential functional components of these of active bits in the input vector. Provided not too many patterns
networks are a set of powerful detonator synapses that can impose have been stored, any unique subset of an x vector can recall the cor-
the pattern of activity to be stored, a set of extensively connected rect y vector. The autoassociative cases (B and E) work as the
inputs with modiable synapses, and a set of inhibitory interneu- heteroassociative cases (A, C, D) with output y1  input x1,
rons whose role is to set a divisive threshold for the activity of the y2  x2, and so on. (Source: Adapted from McNaughton and
cells. CE. Details of associative memory using the correlation Nadel, 1990.)
matrix formalism (Willshaw et al., 1969; Kohonen, 1972). Pairs of
736 The Hippocampus Book

pus (see Chapter 3), with constraints on the representations

(a) Hippocampal anatomy
and learning mechanisms indicated by functional analysis
of associative memory (see Figs. 1410 to 1412). The rst CA1
major attempt of this sort (McNaughton and Morris, 1987) 4
CA3
highlighted the potential contribution of the dentate gyrus 3
(DG). 5 2
First, the much larger number of projection cells in the DG 1 Dentate gyrus
(around one million granule cells in the rat) than in either the
entorhinal cortex (EC) (around 200,000 layer II cells project
into the hippocampus) or region CA3 (around 300,000
pyramidal cells) indicate that it could be used for pattern sep- 6
aration (Amaral et al., 1990). This means that distinct (i.e., V III II
Entorhinal
nonoverlapping) patterns of activation are created in the DG cortex
despite similarity in patterns of EC activation representing
similar events. This process is also referred to as orthogonal- (b) Computational model
ization. Thus pattern completion in CA3, so important for Medial septum
retrieving the representation of a familiar event, does not lead Regulation of
to a similar novel event also causing retrieval of the represen- learning dynamics
tation of the familiar event. Of course, one cannot have per- 7 ACh 8 3
fect pattern completion of the representation of an event from Region CA3
Region CA1 4
partial cues and perfect pattern separation of it from the rep- Hetero- Autoassociative
Comparison associative recall
resentations of similar events in the same system. One mech- recall
2
anism must fail once the partial cue is less similar to the old
Self-organization5 6 Dentate gyrus
event than the new event. Self-organization
Second, the specic nature of the various synaptic inputs Hippocampus
1
to CA3 pyramidal cells suggests different functions for them.
Entorhinal cortex
The input from DG comes from a small number (around 46) Afferent input
of very large synapses around the soma. A much large number Current Opinion in Neurobiology
of connections are received farther up the dendrites from
within CA3 (up to 12,000), and up to 3750 connections are Figure 1412. Hippocampal anatomy (a) and a computational
received from the EC onto the distal apical dendrites. It was model of hippocampal episodic memory function (b). Numbers
suggested (McNaughton and Morris, 1987) that the powerful label synaptic connections mediating various functions in the model.
input from DG (via detonator synapses) serve to impose a 1: Synapses of the perforant path bers projecting from entorhinal
new pattern of activity to be learned. A strong input is cortex layer II to the dentate gyrus undergo sequential self-organiza-
required to impose a new pattern of activity in CA3 in the face tion to form sparse, less overlapping representations of entorhinal
of the interference due to feedback via the recurrent connec- activity patterns. 2: Mossy bers projecting from dentate gyrus to
tions, which tend to cause the system to return to a previously region CA3 transfer dentate gyrus activity to CA3. 3: Excitatory
stored pattern of activity. Once the representation of a new recurrent connections in region CA3 mediate autoassociative encod-
ing and retrieval of the features of episodic memories. 4: Schaffer
event has been imposed, the Hebbian modication of both
collaterals from region CA3 to CA1 encode and retrieve associations
the recurrent connections and the connections from EC can
between CA3 activity and activity patterns induced by entorhinal
input to region CA1. 5: Perforant path input to region CA1 under-
Figure 1411. Anatomy of the inputs to CA3 pyramidal cells, goes self-organization, forming new representations of entorhinal
showing the approximate number of synapses onto each cell in cortex input for comparison with recall from CA3. 6: Projections
the rat. (Adapted from Treves and Rolls, 1992.) from region CA1 to deep layers of the entorhinal cortex store associ-
ations between CA1 activity and entorhinal cortex activity, allowing
CA1 representations in CA1 to activate the associated patterns in entorhi-
nal cortex. 7: Output from region CA1 to the medial septum regu-
Perforant path lates cholinergic modulation. 8: Cholinergic modulation from the
<
~ 3750 synapses medial septum sets appropriate dynamics for encoding in hip-
Recurrent collaterals pocampus. (Source: Hasselmo and McClelland, 1999.)
< 12000 synapses
~
Mossy fibres occur. The large number of recurrent synapses per cell allow
~ 46 synapses
large autoassociative memory capacity, and the large number
of synapses in the input from EC allow large heteroassociative
memory capacity in associating the EC representation to the
CA3 CA3 representation (Treves and Rolls, 1992) (Fig. 1411).
 Computational Models of Spatial and Mnemonic Functions 737

Third, the requirements of Hebbian learning in the CA3 tate gyrus, the hippocampal representations are independent
recurrent connections and the connections from EC to CA3, of the details of the events and can be thought of as simply an
but not in the connections from the DG, are consistent with index for them (Teyler and DiScenna, 1986). However, as the
the physiology of these various connections. The synapses in hippocampal representation must initially be activated by the
the former two pathways are thought to be capable of NMDA neocortical representation, the overall memory system is still
receptor-dependent LTP, whereas the mossy ber connections content addressable and its behavior (e.g., pattern separa-
from DG show only non-Hebbian modication (see Chapter tion or pattern completion) depends on how different aspects
10). Equally, the divisive normalization required by associative of an event and its context contribute to the activation of its
nets (Willshaw et al., 1969) (see Box 142) is consistent with hippocampal representation. This relates to Marrs observa-
the action of interneurons providing inhibition by opening tion that the hippocampal representation should include
ion channels near the soma-to shunt input current in the den- those aspects of an event used for its retrieval.
drites (Fig. 1410). Further analysis of autoassociative mem- In a simple associative memory, all elements of the repre-
ory indicates that progessive recall improves performance sentation of the event are equally associated with all other ele-
(Gardner-Medwin, 1976). With this model, inhibition is ments. However, as implied by Marr, it seems plausible that
slowly reduced during retrieval so the rst few cells that some aspects of an event are better able to cue associative
become active are the most likely to be correct, and feedback retrieval than others, and other aspects of an event can be
from their activation decreases the chances of subsequent associatively retrieved more easily than others. Thus, the sim-
erroneous activation. Such periodic uctuation of inhibition ple representations envisaged for the hippocampus would
(or, equivalently, the cells ring threshold) may provide a reect some aspects more than others. A related suggestion is
functional interpretation for the theta rhythm. that the hippocampal representation reects efficient com-
In Marrs model an incomplete or incorrect neocortical pression of the neocortical representationsextracting the
representation feeds into the hippocampus, where the correct distinguishing features of each event (Gluck and Myers, 1996).
hippocampal representation is retrieved and feeds back to For example, the name of someone you met only once is often
complete or correct the neocortical representation. This sim- a good cue to recalling the meeting but can be difficut to
ple picture has been elaborated to include separate input retrieve, whereas the location of the meeting is often both a
and output representations in the entorhinal cortex (in the good cue and relatively easy to retrieve. Similar to someones
supercial layers and deep layers, respectively). The processes name, the sequential position of an item in a list is easier to
of storage and retrieval obviously do not start and end at the use as a cue than it is to retrieve (Jones, 1976). Thus, even the
entorhinal cortex. The processes of feedforward cueing of simplest associative model of memory should include asym-
a representation, its pattern completion and feedback can metrical associations between the elements of an event.
be modeled as occurring sequentially in the cortical areas One of the distinguishing features of episodic memory is
between sensory cortex and entorhinal cortex (Rolls, 1996). the ability to retrieve the ongoing context within which the
The principal difference between the processes in the hip- event occurs (Gardiner and Java, 1993; Tulving, 1993;
pocampus and EC and those in lower cortical areas are Knowlton and Squire, 1995), and one suggestion for the role
the additional pattern separation provided by the DG and the of the human hippocampus is that it provides the spatiotem-
greater autoassociative power of the much longer-range recur- poral context for episodic memory (OKeefe and Nadel,
rent collaterals in CA3 compared to cortex. The long-range 1978). Theoretical analyses of associative memory have also
recurrent collaterals in CA3 also provide the best opportunity made the distinction between the content of the event and its
to associate information from different sensory modalities. context (e.g., Raaijmakers and Shiffrin, 1981) or between the
Many of the above ideas are reviewed or developed further record of the event and the header or index term used to ref-
in the literature (e.g., Amit, 1989; McClelland et al., 1995; erence it (e.g., Morton et al., 1985). One possibility is that the
Rolls and Treves, 1997; Hasselmo and McClelland, 1999; hippocampus serves to associate the content of the event with
Redish, 1999). its context (e.g., Wallenstein and Hasselmo, 1997) (see Section
14.6). A related possibility is that the role of the hippocampus
14.5.4 Hippocampal Representation, is to generate a representation of temporal context itself (e.g.,
Context, and Novelty slowly varying patterns of activity generated by its own recur-
rent dynamics) (Levy, 1996). This type of model coincides
The above considerations of sparseness and pattern separa- with a considerable psychological literature of models of
tion regarding the hippocampal representation of an event memory in which retrieval of stored items is held to depend,
raise the issue of how these representations relate to the vari- at least in part, on their association with a representation of
ous elements of its content and context. First note that the context that changes slowly with time or experience. In some
conicting processes of pattern separation and pattern com- of these models the context representation changes inde-
pletion serve to dene the similarity space of retrieval (i.e., pendently of the to-be-remembered items (Mensink and
which dimensions a retrieval cue can vary along but still Raaijmakers, 1988; Burgess and Hitch, 1992; Davelaar et al.,
retrieve the event and which dimensions serve to discriminate 2005), whereas in others the context representation is derived
events). In the limit of complete orthogonalization in the den- from the items themselves (Howard and Kahana, 2001).
738 The Hippocampus Book

Retrieval corresponds to nding the item most strongly asso- neuropsychological patients (e.g., Spiers et al., 2001) (see
ciated with a re-presented context. Chapter 12). Second, blocking the ACh receptors prior to
I briey describe the operation of one of these models, the encoding (by injecting scopolamine) seems to impair recall
temporal context model, as I return to it in the section on more strongly than recognition (Hasselmo and Wyble, 1997).
models attempting to draw links between the medial tempo- This is also the case in the model, assuming that CA3 serves to
ral roles in spatial and episodic memory. As the with other associate events and their contexts because recall is more
models of episodic memory described above, item representa- reliant on these associations than recognition. However, dis-
tions are associated with context representations such that a ruption of the hippocampus might also impair recall more
given item is retrieved according to the similarity between the than recognition in many other models (e.g., due to disrupt-
context at retrieval and that associated with the item. ing associations with context) (see Section 14.6). In addition,
However, in this model the context representation is derived more recent work has begun to include the role of dopamine
from the presented or retrieved items themselves, becoming a in novelty and encoding (Lisman and Grace, 2005).
recency-weighted sum of the context arising from each item.
After presentation of the list, the context vector is most simi- 14.5.5 Consolidation and Cross-modal
lar to the most recent items, producing the well known Binding of Events in Memory
recency effect. In addition, when an item is presented, it affects
the context vector with which subsequent items are associ- Ever since the initial reports of dense anterograde amnesia but
ated. Thus, recall of a given item changes the context vector so weaker or temporally graded retrograde amnesia after bilat-
that retrieval of immediately subsequent items in the list is eral medial temporal lobectomy (Scoville and Milner, 1957)
more likely (by making the context vector more similar to how (see Chapter 12 and the caveats in Section 14.5.2), researchers
it was when those items were presented). This leads to an have considered how the hippocampus contributes to the
asymmetry such that forward associations in the list are long-term consolidation of memories. One mechanism for
stronger than backward associations, as is often found with consolidation, consistent with the preservation of informa-
free recall. tion acquired prior to hippocampal damage, is that the hip-
Consideration of the different requirements of encoding pocampus enables information to be stored elsewhere in the
and retrieval also raises some interesting questions. brain after which it is no longer needed for retrieval. Marrs
Notwithstanding McNaughton and Morriss suggestion that (1971) model follows this view, suggesting that the days
detonator synapses from the dentate gyrus can impose a events are stored in the hippocampus and that this informa-
new pattern of activation on CA3 despite the retrieval-related tion is used to allow long-term categorization and storage in
feedback from recurrent connections, there must be some the neocortex. Note that processes of temporal consolidation
mechanism to determine whether the system should be in might also occur in the hippocampus (i.e., without transfer of
encoding or retrieval mode (McNaughton and Morris, 1987). information from one region to another).
One proposal is that the supply of acetylcholine (ACh) from The experimental data regarding the gradient of retrograde
the medial septum switches the hippocampus between encod- amnesia (i.e., the sparing of memories acquired sufficiently
ing and retrieval modes (Hasselmo et al., 1995; Murre, 1996; long enough before the damage) is inconsistent and remains
Wallenstein and Hasselmo, 1997). In this model, increased controversial in animals (see Chapter 13) and humans (see
ACh increases the rate of synaptic modication of the recur- Chapter is 12) (Spiers et al., 2001). One problem in the human
rent connections in CA3 and suppresses the synaptic trans- data is that the amnesia often extends back to childhood,
mission of intrinsic activity (within CA3 and from CA3 to implying that the hippocampus stores several decades worth
CA1), enabling new patterns of activity to be stored. The level of information. Several computational arguments have also
of delivery of ACh is determined by the novelty of the neo- been put forward for the transfer of information out of the
cortical input, as represented by direct activation of CA1 from hippocampus for consolidation in the neocortex. Marr (1971)
EC, compared to the most similar previously stored event, as suggested that a short-term buffer is needed to store informa-
represented by the input to CA1 from CA3 after it settles to a tion online (i.e., the hippocampus), whereas only information
stored state. Specically, if both CA3 and entorhinal cortex deemed relevant to the animals future needs to be incorpo-
inputs to CA1 are matching (as with a familiar stimulus), rated into the animals long-term store of knowledge and
strong activation of CA1 activates interneurons in the medial must be rst appropriately categorized with respect to it. The
septum, which decrease the activity of the cholinergic cells hippocampal store of unprocessed experience necessarily has
projecting to the hippocampus (Fig. 1412). a limited capacity, and the process of abstracting relevant
Two types of evidence support this model. First, by empha- information from this experience to expand a long-term data-
sizing the connections between the hippocampus and the base requires extensive offline processing, perhaps during
medial septum, the model begins to address data showing the sleep (see also McClelland et al., 1995).
importance of the fornix, the large ber bundle connecting A second argument concerns the anatomical convergence
the hippocampus with the medial septum and other subcorti- of information from difference sensory modalities at the hip-
cal structures. In the model, sectioning the fornix prevents the pocampus. Thus, in the absence of long-range connections
learning of new memories due to lack of ACh, which corre- between different sensory cortical areas, associations between
sponds well to reviews of the effects of damage to the fornix in the elements of an event, such as its sight, sound, and smell,
 Computational Models of Spatial and Mnemonic Functions 739

cannot be formed in the lower-level cortices. Damasio (1989) A reinterpretation of the experimental data regarding the
suggested that convergence zones must exist where these gradient of retrograde amnesia associated with medial tempo-
associations could be formed. Several models have extended ral lobe damage led Nadel and Moscovitch (1997) to a differ-
this idea to include rapid learning of a hippocampal, or ent conclusion regarding consolidation. Noticing that in many
medial temporal lobe, representation reciprocally connected instances retrograde amnesia extends back over a much longer
to all the unimodal cortical representations of the event which time than that envisaged by Marr, they proposed that the hip-
allows them to be associated with each other (Alvarez and pocampus remains necessary for the retrieval of detailed
Squire, 1994; Murre, 1996; Moll and Miikkulainen, 1997). episodic or spatial information (but for evidence that early
These models also suggest that, after multiple rehearsals of a spatial memories become hippocampus-independent see
memory driven by the hippocampus, long-range associations Teng and Squire, 1999; Rosenbaum et al., 2000). They pro-
can be learned directly between the unimodal representations, posed a new model to account for both the common occur-
nally making the stored information independent of the rence of a temporal gradient in memory for other types of
hippocampus. information and the possibility of partial damage to the hip-
The arguments regarding data abstraction and anatomical pocampus in some cases. In this model, whenever a memory
convergence are represented in a model by Kali and Dayan is rehearsed or reactivated, a new hippocampal representation
(2004). In this model the hippocampus serves to aid the learn- is formed, again connected to all of the neocortical represen-
ing of a hierarchical semantic system by being able to reinstate tations. The result is that although a complete lesion of the
the top-level (i.e., entorhinal/ perirhinal/ parahippocampal) medial temporal lobe impairs retrieval of all memories, the
representation of events during sleep. The semantic system older the memory, the more robust it is to partial damage by
contains reciprocal connections between higher cortical areas virtue of being represented in multiple locations. Some evi-
and those immediately below them in the hierarchy but no dence for the reactivation of specic memories comes from
direct connections between areas at the same level. Aided by the recently revived study of reconsolidation (Nader et al.,
the hippocampus, the neocortical system learns to form an 2000). In these experiments re-presentation of the context of
associative representation of patterns of activation in lower an event (usually application of an electric shock) appears
cortical areas. This is achieved by the higher level representa- to render the memory for the event labile in the sense
tions learning a generative model of representations lower in that protein synthesis is again required for long-term storage
the hierarchy (e.g., Hinton and Sejnowski, 1999). Thus of the memory, as is the case when the event is rst experi-
higher-level representations can be cued by input from a sub- enced.
set of lower cortical areas and can then cause pattern comple-
tion in all lower areas via the reciprocal connections. During 14.5.6 Hippocampal Contributions to Various
further simulations, Kali and Dayan (2004) noted that regular Types of Memory and Retrieval
reactivation of episodic hippocampal representations is
required to maintain them in register with the slowly chang- A distinction has been made between retrieval of episodic in-
ing semantic representations. formation, retrieval of semantic information, and familiarity-
A third argument for consolidation refers to the effects of based recognition (see Chapter 12). Episodic retrieval is
interference. In some learning systemssuch as using a xed characterized by the ability to retrieve detailed information
feedforward structure and the error back-propagation learn- about the event and its context. Semantic retrieval is charac-
ing rule (Rumelhart et al., 1986) to form associations consis- terized by factual knowledge without retrieval of the indi-
tent with a set of examplesnew information can interfere vidual events and contexts in which it was acquired.
with previously stored information, sometimes catastrophi- Familiarity-based recognition depends purely on an unattrib-
cally such that all information is lost. This problem can also utable feeling of familiarity associated with a stimulus in the
occur with associative memories (Fig. 1410). McClelland et absence of detailed information about the event and context in
al. (1995) proposed a solution to this in which long-term neo- which it was encountered. Of these three processes, the hip-
cortical learning involves random interleaved re-presentation pocampus is claimed to be primarily involved in episodic
of all previous knowledge along with newly acquired knowl- retrieval (e.g., Aggleton and Brown, 1999) (see Chapter 12).
edge to produce a single integrated neocortical representation The hippocampo-neocortical model suggests that semantic
of semantic knowledge. Note, however, that this mechanism knowledge is abstracted from combinations of hippocampal
requires the temporary store to have capacity for the entire memories of unique events, but eventually becomes inde-
data set. Other solutions include associative memories with pendent of the hippocampus. This is consistent with memory
continuous but bounded connection weights (Hopeld, for the unique content and context of a specic event (episodic
1982) and constructive algorithms in which the addition of memory) depending on the hippocampus but semantic mem-
new information is accompanied by the addition of new pro- ory depending on other areas of the temporal lobe (e.g.,
cessing units (Gallant, 1986; Mezard and Nadal, 1989; Graham and Hodges, 1997). It is also consistent with the sug-
Fahlman and Lebiere, 1990; Frean, 1990). The latter algo- gestion of Nadel and Moscovich (1997) that semantic memo-
rithms may relate to the recent observation of neurogenesis in ries show a temporal gradient of retrograde amnesia, whereas
the adult dentate gyrus that is related to learning (Shors et al., detailed episodic memories do not. The model is less obvi-
2001). ously consistent with the observation that semantic informa-
740 The Hippocampus Book

tion can be acquired despite early bilateral hippocampal The hippocampal contribution to recognition memory
pathology (Vargha-Khadem et al., 1997). However, the possi- (via recollection) can be modeled by using the stimulus-
bility of partial sparing of the hippocampus (Squire and driven medial temporal neocortical representation as the cue
Zola, 1998) or the use of external rehearsal of information to retrieval of a stored pattern of activation in CA3 and com-
(Baddeley et al., 2001) might provide explanations for these paring the retrieved activation to the stimulus-driven activa-
developmental cases within the framework of the model. tion in the entorhinal cortex (Norman and OReilly, 2003). By
The distinction between episodic retrieval and familiarity- explicitly retrieving an entire stored pattern, the process is all-
based recognition also has interesting parallels in the or-nothing, and even foils that resemble presented patterns
hippocampo-neocortical model. It has been argued that the and are not falsely recognized because the retrieval product
hippocampus is required to provide associations between rep- still differs from the stimulus-driven activation. The use of
resentations in disparate cortical areas, whereas associations sparse hippocampal representations, orthogonalized via the
within each area can be formed locally. Thus, effects mediated dentate gyrus, serves to prevent interference between different
by the familiarity of single stimuli might be supported by the stored events. By contrast, familiarity-based recognition is
association of elements within each of the neocortical areas, modeled in terms of sharpening of the neocortical represen-
independent of the hippocampus. By contrast, correct recogni- tation. With this model, although a new item is represented by
tion of a pair of cross-modal associates among equally familiar weak activation of a large number of neocortical neurons,
distractors would require the hippocampus. Evidence indicates repeated presentation of the item results in strong activation
that simple recognition memory does not depend on the hip- of a smaller number of neocortical neurons via a competitive
pocampus but on nearby neocortical areas (Zhu et al., 1996; learning mechanism. Activation of the neocortical neurons
Baxendale et al., 1997; Vargha-Khadem et al., 1997; Murray and that re in response to a given stimulus can then be used as a
Mishkin, 1998; Aggleton and Brown, 1999; Wan et al., 1999; measure of familiarity-based recognition. One advantage of
Holdstock et al., 2000; Baddeley et al., 2001: but see also Manns this scheme is that repeated presentations of some items in a
and Squire, 1999; Zola et al., 2000). In contrast, there is some list does not impair recognition of the other items in the list,
evidence that recognition of cross-modal associations is as seen experimentally (Ratcliff, 1990), which is not true of
impaired by bilateral damage restricted to the hippocampus some other psychological models of recognition memory
(Vargha-Khadem et al., 1997; Holdstock et al., 2000). More (Norman, 2003). A characteristic of this model is that hip-
extensive unilateral damage may also impair the binding of ele- pocampal damage specically impairs recognition memory
ments within the same modality (Kroll et al., 1996). The logi- when related lures (novel items that resemble previously seen
cal extension of this idea is that episodic memory requires full items) are included in the test. The familiarity-based recogni-
recollection of an event and its context in all of its multimodal tion mechanism would produce false alarms to these stimuli,
detail and so requires an intact hippocampus. whereas the hippocampal recollection mechanism would not.
Much of the analysis of the differential role of the hip- Some evidence consistent with this hypothesis has been found
pocampus in episodic retrieval (or recollection), and (Holdstock et al., 2002).
familiarity-based recognition has focused on the idea that
these two processes contribute independently to the perform-
ance of recognition memory tests (e.g., Yonelinas, 2002). In
forced choice and yesno recognition paradigms, the recollec- 14.6 Reconciling the Hippocampal
tive component is assumed to be all or nothing and high- Roles in Memory and Space
threshold. That is, the stimulus is recalled in great detail or
not at all, and a novel foil is never falsely recalled. By contrast, In this section I discuss some of the models that have
the familiarity-based process is more like a signal detection attempted to draw together the common strands present in the
problem: The subject guesses whether the item is familiar, plethora of hippocampal models reviewed above. One feature
informed by a noisy measure of familiarity (see Chapter 13). common to most of the spatial and nonspatial models is use of
The hippocampus has been associated with the recollective the autoassociative properties of the recurrent network in
component (Aggleton and Brown, 1999; Yonelinas et al., 2002) CA3. Even feedforward models of spatial representation (see
and so should provide a high-threshold, all-or-nothing mech- Section 14.3.2) usually note that the various sets of cells used
anism in recognition memory tests. One method for detecting to represent different environments might be stored in this
this component involves the receiver-operator characteristic way (e.g., McNaughton and Nadel, 1990; Burgess et al., 1994;
(ROC) curve: the plot of the proportion of previously seen Kali and Dayan, 2000).
items that are correctly identied (hits) versus the propor- Many other models concern the common functional
tion of new items incorrectly identied (false alarms) as a requirement of storing and retrieving a sequence of patterns of
function of the subjects confidence in their response. activation. For example, Redish and Touretzky (1998) used
Recollection is marked by high-condence hits in the absence replay of a spatial route as a demonstration of episodic recall.
of high-condence false alarms, producing a positive intercept In Levys (1996) model, the slowly varying patterns of activity
in the ROC curve. See Chapter 13 and Rugg and Yonelinas (resulting from the presence of asymmetrical connections in
(2003) for a review. CA3) (see Section 14.3.3) might be used as a context repre-
 Computational Models of Spatial and Mnemonic Functions 741

sentation, associations to which could store sequences of (TCM) of free recall (Howard and Kahana, 2001), described
memories. In a rat moving through its environment, it is above, to model the noisy place-specic ring of some cells
argued that they might also resemble place elds, although it in entorhinal cortex (Quirk et al., 1992; Frank et al., 2000;
is not clear that this would be the case for a trajectory that Fyhn et al., 2004). To model free recall, the TCM forms a
crosses itself. Jensen and Lismans (1996, 2005) model of long- context representation derived from a recency-weighted sum
term memory and working memory, involving association of of the contributions to context from the nature of the list
one item to the next and preservation of serial order during items themselves. Howard et al. proposed that, in terms of
each theta cycle, is also applied to model the storage of routes spatial memory, these context contributions might be
as sequences of place elds. encoded by neurons with ring rates reecting running speed
However, spatial navigation is not just a subset of episodic in particular head directions. If each context neuron per-
memory; simply storing routes and sequences of places would formed a recency-weighted sum of one of these inputs, they
not enable accurate navigation, detours, or shortcuts. These would be effectively supporting an approximate route mem-
require a metric or at least a representation of proximity (see ory, or path integration. In simulations, these cells show
Section 14.4.2). Put another way, specic constraints apply to noisy place-specic ring, rather like some cells in the
spatial information that are not necessarily required of a more entorhinal cortex, as they re most following a series of
general memory system. What then is the relation between the movements along their preferred direction (e.g., a cell with a
constraints involved in spatial and episodic memory such that west head direction input res most along the west edge of the
a unique structure, the hippocampus, might support both? environment). They also show retrospective coding: that
OKeefe and Nadel (1978) argued that the spatial function is, modulation of ring by the recent history of movements, as
seen in the rat hippocampus is augmented by a linear sense of found by Frank et al. (2000). In addition, cells whose ring
time in humans, which allows the human hippocampus to rate reects the time integral of their input have been found
supply the spatiotemporal context of events required by in entorhinal cortex (Egorov et al., 2002), and systematic
episodic memory. This idea has been taken up by two types of errors in path integration have been found that are consistent
model: one investigating the provision of temporal context with recency weighting (Etienne et al., 1996). However, the
and the other the provision of spatial context. model predicts that entorhinal cell ring is directionally selec-
Following the models of Levy (1996) and Jensen and tive and always peaks at the edge of the environment, which
Lisman (1996) described above, Wallenstein et al. (1998) may not be the case (Fyhn et al., 2004). This idea of entorhi-
investigated the idea that the hippocampus could provide a nal processing may also have to be reviewed in the light of the
signal representing temporal context as a solution to the need discovery of grid cells (Hafting et al., 2005) (see Section
to form associations between temporarily discontiguous 14.3.3).
events. Associating events that occur much farther apart in The idea of using spatial context to index memory was
time than the time scale required for pre- and postsynaptic combined with Marrs hippocampo-neocortical model to
activity to induce LTP (i.e., around 100 ms) (see Chapter 10) produce a simple navigation system (Recce and Harris, 1996).
obviously requires some kind of bridging mechanism. One In this model, perceptual representations of the environment
indication that the hippocampus is involved in this process are stored in parietal cortex and retrieved via an index repre-
comes from trace conditioning, in which the animal must sentation in the hippocampus corresponding to the location
learn to respond at a xed delay after the disappearance of a of the animal. This allows retrieval of the correct perceptual
cue, because the timing of the response is disrupted by hip- map for a given location. Navigation additionally requires that
pocampal lesions. This contrasts with delay conditioning in the location of an encountered goal be added to the perceptual
which the cue does not disappear, for which there is no hip- maps (the mechanism for this is not specied). A more
pocampal lesion effect (see Chapter 13). In Wallenstein et al. detailed model of long-term memory and retrieval of a spatial
(1998), as with Levy (1996), unsupervised learning in the hip- environment (Becker and Burgess, 2001) also serves as a
pocampus in the presence of temporally correlated inputs cre- model of memory for the spatial context of an event (Burgess
ates patterns of activity that vary more slowly in time than the et al., 2001a) (Fig. 1414). This model explicitly makes use of
original stimuli (Fig. 1413). Associations between the various the constraints associated with spatial information, and our
states of this temporal context signal can allow the original detailed knowledge of how it is represented in the brain. The
stimuli to be reproduced in order. Indeed, this type of mech- representations of spatial layout in long-term memory are
anism provides a good model for short-term serial recall assumed to be allocentric (e.g., independent of the orientation
(Burgess and Hitch, 2005). This model can be extended to of the person) in contrast to the egocentric short-term repre-
spatial navigation in terms of associations between spatially sentations involved in perception, action, and working mem-
discontiguous events, but note that, as with storing temporal ory (Goodale and Milner, 1992; Burgess et al., 1999; Milner et
sequences to aid navigation, temporal contiguity alone is al., 1999). By extension from spatial models of the hippocam-
insufficient to generate metric information. pus (Hartley et al., 2000), the geometry of the environment is
Howard et al. (2005) attempted to draw parallels between encoded in (parahippocampal) boundary vector cells bidirec-
the roles of the hippocampal region in both spatial and tionally connected to (hippocampal) place cells. The place
episodic memory by extending the temporal context model cells are connected up to form a continuous attractor. This
742 The Hippocampus Book

Figure 1413. Context eld development. Each rectangle shows a synaptic potentiation between cells ring in the background and
subset of 50 simulated pyramidal cells ring across time, each cells that encode sequence items owing to a simple Hebbian learn-
action potential represented by a vertical line. A unique group of ing rule. B. After the fourth learning trial, this repeated potentiation
ve cells ring at the same time encodes a single item in a 13-item leads to a condition where background cells begin to respond to the
sequence (see the top portion of each rectangle). This can be appearance of contiguous segments of the entire sequence. The por-
thought of as afferent activation of CA3 pyramidal cells due to a tion of the full sequence to which the cell responds is called the
specic pattern of sensory events. A. Note the background ring context eld of the cell. Because the context elds overlap, the
during the rst learning trial that does not encode sequence items entire sequence can be reconstructed by interdigitating them in the
directly. This stems from activity at recurrent excitatory synapses. proper order. (Source: Wallenstein et al., 1998.)
Repeated exposure to the same sequence can lead to enhanced

acts as a spatial memory system in that partial activation of feedforward connections) and retrieval of its spatial context
boundary vector cells (representing the occurrence of partic- (via the feedback connections).
ular large objects at particular distances and bearings) causes Retrieval of allocentric spatial information into a format
activation of the corresponding place cells, which then acti- appropriate for visual imagery requires that it be translated
vate the boundary vector cells representing the remaining into an egocentric reference frame. Accordingly, in the model,
landmarks in the environment. The location of a specic the parahippocampal boundary vector cell representation
event is represented by a goal cell, as in the simple model of (arranged in terms of distances and bearings) is translated
navigation (Burgess and OKeefe, 1996) (see Section 14.4.2). successively into trunk-centered then head-centered represen-
Bidirectional connections between place cells and these tations by networks designed for this purpose to model poste-
event cells allows for both navigation to the location (via the rior parietal cortex (e.g., Pouget and Sejnowski, 1997).
 Computational Models of Spatial and Mnemonic Functions 743

Figure 1414. A. Functional architecture of the model for encoding northwest corner of the square (hippocampal cell activity shown
and retrieving the spatial context of an event. Neurons shown in as the brightness of the pixel corresponding to the location of each
gray illustrate a possible pattern of activation corresponding to an cells place eld). iii: The parahippocampus correctly retrieves the
imagined location with extended landmarks nearby to the west, far locations of the other buildings (parahippocampal cell activity
away to the north, and at intermediate distances east and south; the shown as the brightness of the pixel for the location encoded by
imagined westerly heading direction means that these landmarks each cell, relative to the subject at the center). The line indicates that
are imagined as far to the right, near straight ahead, and at an inter- the imagined head direction is south. iv: Medial parietal (MP) cell
mediate distance to the left (and behind, but this is not in view). activity. The parahippocampal map has been correctly rotated given
(Source: Burgess et al., 2001a.) B. Simulation of retrieving spatial head direction south (straight ahead is up); stars indicate a direc-
information in the Milan square experiment of Bisiach and Luzzatti tion of inspection to the left, circles to the right. v: Perirhinal (PR)
(1978). i: Training consists of simulated exploration of the square cell activations correctly showing building 5 to the left and building
(shaded area, north is up). The system is cued to imagine being near 7 to the right. vi: Effect of a right parietal lesion on the medial pari-
the cathedral (i.e., the perirhinal cell for the texture of building 1 etal representation (note lack of activation on the left) and perirhi-
and parahippocampal cell for a building at a short distance north nal activations (vii). Note the decrease in activation of building 5
are activated), and the hippocampal-parahippocampal-perirhinal when inspection is to the left. (Source: Adapted from Becker and
system settles. ii: The hippocampus settles to a location in the Burgess, 2001.)
744 The Hippocampus Book

Likewise, forming an allocentric representation from egocen- Alyan S, McNaughton BL (1999) Hippocampectomized rats are
tric perception requires translation in the reverse direction. capable of homing by path integration. Behav Neurosci 113:
Visual imagery occurs in medial parietal areas using the nal 1931.
head-centered representation with the intermediate represen- Amaral DG, Ishizuka N, Claiborne B (1990) Neurons, numbers and
the hippocampal network. Prog Brain Res 83:111.
tations buffered in retrosplenial cortex. This model is consis-
Amit DH (1989) Modelling brain function. Cambridge, UK:
tent with the observation of hemispatial neglect in imagery
Cambridge University Press.
following parietal damage (e.g., Bisiach and Luzzatti, 1978) Amit DJ (1992) Modeling brain function: the world of attractor neural
and functional imaging of the retrieval of spatial context networks. Cambridge, UK: Cambridge University Press.
(Burgess et al., 2001b). It also provides an explanation for the Baddeley AD, Vargha-Khadem F, Mishkin M (2001) Preserved recog-
involvement of Papezs circuit (see Chapter 3) in supporting nition in a case of developmental amnesia: implications for the
both episodic memory (see Chapter 12) and the representa- acquisition of semantic memory? J Cogn Neurosci 13:357369.
tion of head direction (see Chapter 11). Finally it implies a Barry C, Lever C, Hayman R, Hartley T, Burton S, OKeefe J, Jeffery
strong link between episodic recollection (and its neural KJ, Burgess N (2006) The boundary vector cell model of place
bases) and mental imagery. cell ring and spatial memory. Rev Neurosci 17:7197.
Baxendale SA, Van Paesschen W, Thompson PJ, Duncan JS, Shorvon
SD, Connelly A (1997) The relation between quantitative MRI
measures of hippocampal structure and the intracarotid amo-
barbital test. Epilepsia 38:9981007.
14.7 Conclusions Baxter MG, Murray EA (2001) Opposite relationship of hippocampal
and rhinal cortex damage to delayed nonmatching-to-sample
As in other areas of neuroscience, the ability to specify a pro- decits in monkeys. Hippocampus 11:6171.
posed mechanism of hippocampal function in terms of a Becker S, Burgess N (2001) A model of spatial recall, mental imagery
computational model has been invaluable in many ways. First, and neglect. Adv Neural Inform Proc Syst 13:96102.
it reduces ambiguity and hypothesis drift. Second, it enables Bienenstock EL, Cooper LN, Munro PW (1982) Theory for the devel-
quantitative simulation of the resulting behavior. At the very opment of neuron selectivity: orientation specicity and binoc-
least this can serve as a demonstration of the viability of a ular interaction in visual cortex. J Neurosci 2:3248.
Bisiach E, Luzzatti C (1978) Unilateral neglect of representational
given proposal. Beyond this, it enables quantitative predic-
space. Cortex 14:129133.
tions to be made at the levels of cells, systems, and behavior
Blair HT, Sharp PE (1995) Anticipatory head direction signals in
regarding the effects of experimental manipulations. Some of anterior thalamus: evidence for a thalamocortical circuit that
the models reviewed here have been used to make such pre- integrates angular head motion to compute head direction. J
dictions and have also contributed to the design of experi- Neurosci 15:62606270.
ments to test them. Many questions for future research are Blair HT, Lipscomb BW, Sharp PE (1997) Anticipatory time intervals
prompted by these models, not least the possibility of com- of head-direction cells in the anterior thalamus of the rat: impli-
paring or combining the inuences of spatial and temporal cations for path integration in the head-direction circuit. J
contexts as retrieval cues. Because any worthwhile theory Neurophysiol 78:145159.
must make potentially falsiable predictions, computational Blair HT, Cho J, Sharp PE (1998) Role of the lateral mammillary
modeling has a crucial role to play in the development of the nucleus in the rat head direction circuit: a combined single unit
recording and lesion study. Neuron 21:13871397.
eld. Investigating the link between the properties of cells
Blum KI, Abbott LF (1996) A model of spatial map formation in the
interacting in a complex system such as the brain to the result-
hippocampus of the rat. Neural Comput 8:8593.
ing behavior of the animal would become almost impossible Bose A, Recce M (2001) Phase precession and phase locking of hip-
without the aid of computational models. pocampal pyramidal cells. Hippocampus 11:204215.
Bostock E, Muller RU, Kubie JL (1991) Experience-dependent modi-
fications of hippocampal place cell firing. Hippocampus
ACKNOWLEDGMENTS 1:193205.
Brown MA, Sharp PE (1995) Simulation of spatial learning in the
I am pleased to acknowledge useful discussions with Peter Latham Morris watermaze by a neural network model of the hippocam-
and John OKeefe, and the support of the Medical Research Council, pal formation and nucleus accumbens. Hippocampus 5:171188.
UK. Brunel N, Trullier O (1998) Plasticity of directional place elds in a
model of rodent CA3. Hippocampus 8:651665.
Burgess N, Hartley T (2002) Orientational and geometric deter-
REFERENCES minants of place and head-direction. In: Neural informa-
tion processing systems 14, pp 165172. Cambridge, MA: MIT
Aggleton JP, Brown MW (1999) Episodic memory, amnesia, and Press.
the hippocampal-anterior thalamic axis. Behav Brain Sci 22: Burgess N, Hitch GJ (1992) Toward a network model of the articula-
425490. tory loop. J Mem Language 31:429460.
Alvarez P, Squire LR (1994) Memory consolidation and the medial Burgess N, Hitch GJ (2005) Computational models of working mem-
temporal lobe: a simple network model. Proc Natl Acad Sci USA ory: putting long term memory into context. Trends Cogn Sci
91:70417045. 9:535541.
 Computational Models of Spatial and Mnemonic Functions 745

Burgess N, OKeefe J (1996) Neuronal computations underlying the Ekstrom AD, Meltzer J, McNaughton BL, Barnes CA (2001) NMDA
ring of place cells and their role in navigation. Hippocampus receptor antagonism blocks experience-dependent expansion of
6:749762. hippocampal place elds. Neuron 31:631638.
Burgess N, OKeefe J, Recce M (1993) Using hippocampal place cells Ekstrom AD, Kahana MJ, Caplan JB, Fields TA, Isham EA, Newman
for navigation, exploiting phase coding. In: Neural information EL, Fried I (2003) Cellular networks underlying human spatial
processing systems 5 (Hanson SE, Cowan JD, Giles CL, eds), pp navigation. Nature 425:184188.
929936. San Francisco: Morgan Kaufmann. Etienne AS, Maurer R, Seguinot V (1996) Path integration in mam-
Burgess N, Recce M, OKeefe J (1994) A model of hippocampal func- mals and its interaction with visual landmarks. J Exp Biol
tion. Neural Netw 7:10651081. 199:201209.
Burgess N, Jeffery KJ, OKeefe J (1999) Integrating hippocampal and Fahlman S, Lebiere C (1990) The cascade-correlation learning archi-
parietal functions: a spatial point of view. In: The hippocampal tecture. In: Neural information processing systems 2 (Touretzky
and parietal foundations of spatial cognition (Burgess N, Jeffery DS, ed), pp 524532. San Francisco: Morgan Kaufmann.
KJ, OKeefe J, eds), pp 329. Oxford, UK: Oxford University Fenton AA, Csizmadia G, Muller RU (2000) Conjoint control of hip-
Press. pocampal place cell ring by two visual stimuli. I. The effects of
Burgess N, Maguire E, OKeefe J (2002) The human hippocampus moving the stimuli on ring eld positions. J Gen Physiol
and spatial and episodic memory. Neuron 35:625641. 116:191209.
Burgess N, Becker S, King JA, OKeefe J (2001a) Memory for events Foster DJ, Morris RG, Dayan P (2000) A model of hippocampally
and their spatial context: models and experiments. Philos Trans dependent navigation, using the temporal difference learning
R Soc Lond B Biol Sci 356:14931503. rule. Hippocampus 10:116.
Burgess N, Maguire EA, Spiers HJ, OKeefe J (2001b) A temporopari- Frank LM, Brown EN, Wilson M (2000) Trajectory encoding in the
etal and prefrontal network for retrieving the spatial context of hippocampus and entorhinal cortex. Neuron 27:169178.
lifelike events. Neuroimage 14:439453. Frean MR (1990) The Upstart algorithm: a method for constructing
Cacucci F, Lever C, Wills TJ, Burgess N, OKeefe J (2004) Theta- and training feedforward neural networks. Neural Comput
modulated place-by-direction cells in the hippocampal forma- 2:198209.
tion in the rat. J Neurosci 24:82658277. Fuhs MC, Touretzky DS (2000) Synaptic learning models of map sep-
Cohen MA, Grossberg S (1983) Absolute stability of global pattern aration in the hippocampus. Neurocomputing 32:379384.
formation and parallel memory storage by competitive neural Fuhs MC, Touretzky DS (2006) A spin glass model of path integra-
networks. IEEE Trans Syst Man Cybern 13:815821. tion in rat medial entorhinal cortex. J Neurosci 26:42664276.
Connelly CI, Burns JB, Weiss R (1990) Path planning using Laplaces Fyhn M, Molden S, Witter MP, Moser EI, Moser MB (2004) Spatial
equation. In: Proceedings of the 1990 IEEE international confer- representation in the entorhinal cortex. Science 305:12581264.
ence on robotics and automation, pp 21022106. Gaffan D (1994) Dissociated effects of perirhinal cortex ablation,
Cressant A, Muller RU, Poucet B (1997) Failure of centrally placed fornix transection and amygdalectomy: evidence for multiple
objects to control the ring elds of hippocampal place cells. J memory systems in the primate temporal lobe. Exp Brain Res
Neurosci 17:25312542. 99:411422.
Czurko A, Hirase H, Csicsvari J, Buzsaki G (1999) Sustained activa- Gallant SI (1986) Three constructive algorithms for network learn-
tion of hippocampal pyramidal cells by space clamping in a ing. In: Proceedings of the 8th annual conference of the Cognitive
running wheel. Eur J Neurosci 11:344352. Science Society, pp 652660.
Damasio AR (1989) The brain binds entities and events by multire- Gardiner JM, Java RI (1993) In: Theories of memory (Collins A,
gional activation from convergence zones. Neural Comput Gathercole S, Morris P, eds), pp 168188. Hillsdale, NJ: Erlbaum.
1:123132. Gardner-Medwin AR (1976) The recall of events through the learn-
Davelaar EJ, Goshen-Gottstein Y, Ashkenazi A, Usher M (2005) A ing of associations between their parts. Proc R Soc Lond B Biol
context activation model of list memory: dissociating short- Sci 194:375402.
term from long-term recency effects. Psychol Rev 112:3453. Georgopoulos AP, Schwartz AB, Kettner RE (1986) Neuronal popula-
Dayan P (1991) Navigating through temporal difference. In: Neural tion coding of movement direction. Science 233:14161419.
information processing systems 3 (Lippman RP, Moody JE, Gluck MA, Myers CE (1996) Integrating behavioral and physio-
Touretzky DS, eds), pp 464470. San Mateo, CA: Morgan logical models of hippocampal function. Hippocampus 6:
Kaufmann. 643653.
Dayan P, Abbott LF (2002) Computational neuroscience. Cambridge, Goodale MA, Milner AD (1992) Separate visual pathways for percep-
MA: MIT Press. tion and action. Trends Neurosci 15:2025.
Deneve S, Latham PE, Pouget A (2001) Efficient computation and Goodridge JP, Touretzky DS (2000) Modeling attractor deformation
cue integration with noisy population codes. Nat Neurosci in the rodent head-direction system. J Neurophysiol 83:
4:826831. 34023410.
Dominey PF, Arbib MA (1992) A cortico-subcortical model for gen- Gorchetchnikov A, Hasselmo ME (2002) A model of hippocampal
eration of spatially accurate sequential saccades. Cereb Cortex circuitry mediating goal-driven navigation in a familiar envi-
2:153175. ronment. Neurocomputing 4446:423427.
Droulez J, Berthoz A (1991) A neural network model of sensoritopic Gothard KM, Skaggs WE, McNaughton BL (1996) Dynamics of mis-
maps with predictive short-term memory properties. Proc Natl match correction in the hippocampal ensemble code for space:
Acad Sci USA 88:96539657. interaction between path integration and environmental cues. J
Egorov AV, Hamam BN, Fransen E, Hasselmo ME, Alonso AA (2002) Neurosci 16:80278040.
Graded persistent activity in entorhinal cortex neurons. Nature Graham KS, Hodges JR (1997) Differentiating the roles of the hip-
420:173178. pocampus complex and the neocortex in long-term memory
746 The Hippocampus Book

storage: evidence from the study of semantic dementia and Howard MW, Fotedar MS, Datey AV, Hasselmo ME (2005) The tem-
Alzheimers disease. Neuropsychology 11:7789. poral context model in spatial navigation and relational learn-
Guazzelli A, Bota M, Arbib MA (2001) Competitive Hebbian learn- ing: toward a common explanation of medial temporal lobe
ing and the hippocampal place cell system: modeling the inter- function across domains. Psychol Rev 112:75116.
action of visual and path integration cues. Hippocampus Huxter J, Burgess N, O Keefe J (2003) Independent rate and tempo-
11:216239. ral coding in hippocampal pyramidal cells. Nature 425:828832.
Hafting T, Fyhn M, Molden S, Moser MB, Moser EI (2005) Jensen O, Lisman JE (1996) Hippocampal CA3 region predicts mem-
Microstructure of a spatial map in the entorhinal cortex. Nature ory sequences: accounting for the phase precession of place
436:801806. cells. Learn Mem 3:279287.
Harris KD, Henze DA, Hirase H, Leinekugel X, Dragoi G, Czurko A, Jensen O, Lisman JE (2000) Position reconstruction from an ensem-
Buzsaki G (2002) Spike train dynamics predicts theta-related ble of hippocampal place cells: contribution of theta phase cod-
phase precession in hippocampal pyramidal cells. Nature ing. J Neurophysiol 83:26022609.
417:738741. Jensen O, Lisman JE (2005) Hippocampal sequence-encoding
Hartley T, Burgess N, Lever C, Cacucci F, OKeefe J (2000) Modeling driven by a cortical multi-item working memory buffer. Trends
place elds in terms of the cortical inputs to the hippocampus. Neurosci 28:6772.
Hippocampus 10:369379. Jones VJ (1976) A fragmentation hypothesis of memory: cued recall
Hasselmo ME, Fehlau BP (2001) Differences in time course of ACh of pictures and of sequential position. J Exp Psychol Gen
and GABA modulation of excitatory synaptic potentials in slices 105:277293.
of rat hippocampus. J Neurophysiol 86:17921802. Jung MW, Wiener SI, McNaughton BL (1994) Comparison of spatial
Hasselmo ME, McClelland JL (1999) Neural models of memory. Curr ring characteristics of units in dorsal and ventral hippocampus
Opin Neurobiol 9:184188. of the rat. J Neurosci 14:73477356.
Hasselmo ME, Schnell E, Barkai E (1995) Dynamics of learning and Kali S, Dayan P (2000) The involvement of recurrent connections in
recall at excitatory recurrent synapses and cholinergic modula- area CA3 in establishing the properties of place elds: a model.
tion in rat hippocampal region CA3. J Neurosci 15:52495262. J Neurosci 20:74637477.
Hasselmo ME, Wyble BP (1997) Free recall and recognition in a net- Kali S, Dayan P (2004) Off-line replay maintains declarative memo-
work model of the hippocampus: simulating effects of scopo- ries in a model of hippocampal-neocortical interactions. Nat
lamine on human memory function. Behav Brain Res 89:134. Neurosci 7:286294.
Hasselmo ME, Wyble BP, Wallenstein GV (1996) Encoding and Keith JR, McVety KM (1988) Latent place learning in a novel envi-
retrieval of episodic memories: role of cholinergic and ronment and the influence of prior training in rats.
GABAergic modulation in the hippocampus. Hippocampus Psychobiology 16:146151.
6:693708. Kentros C, Hargreaves E, Hawkins RD, Kandel ER, Shapiro M, Muller
Hebb DO (1949) The organisation of behavior. New York: Wiley. RV (1998) Abolition of long-term stability of new hippocampal
Hertz J, Krogh A, Palmer R (1990) Introduction to the theory of neural place cell maps by NMDA receptor blockade. Science
computation. New York: Perseus Books. 280:21212126.
Hinton GE, Sejnowski TJ (1999) Unsupervised learning. Cambridge, Knierim JJ (2003) Hippocampal remapping: implications for
MA: MIT Press. spatial learning and navigation. In: The neurobiology of spatial
Hirase H, Czurko HH, Csicsvari J, Buzsaki G (1999) Firing rate and behaviour (Jeffery KJ, ed), pp 226239. Oxford, UK: Oxford
theta-phase coding by hippocampal pyramidal neurons during University Press.
space clamping. Eur J Neurosci 11:43734380. Knierim JJ, Kudrimoti HS, McNaughton BL (1995) Place cells, head
Hodgkin AL, Huxley AF (1952) A quantitative description of mem- direction cells, and the learning of landmark stability. J Neurosci
brane current and its application to conduction and excitation 15:16481659.
in nerve. J Physiol 117:500544. Knowlton BJ, Squire LR (1995) Remembering and knowing: two dif-
Holdstock JS, Mayes AR, Cezayirli E, Isaac CL, Aggleton JP, Roberts ferent expressions of declarative memory. J Exp Psychol Learn
N (2000) A comparison of egocentric and allocentric spatial Mem Cogn 21:699710.
memory in a patient with selective hippocampal damage. Kohonen T (1972) Correlation matrix memories. IEEE Trans Comp
Neuropsychologia 38:410425. C-21:353359.
Holdstock JS, Mayes AR, Roberts N, Cezayirli E, Isaac CL, OReilly Kroll NE, Knight RT, Metcalfe J, Wolf ES, Tulving E (1996) Cohesion
RC, Norman KA (2002) Under what conditions is recognition failure as a source of memory illusions. J Mem Language
spared relative to recall after selective hippocampal damage in 35:176196.
humans? Hippocampus 12:341351. Latham PE, Deneve S, Pouget A (2003) Optimal computation with
Hopeld JJ (1982) Neural networks and physical systems with emer- attractor networks. J Physiol Paris 97:683694.
gent collective computational abilities. Proc Natl Acad Sci USA Lengyel M, Szatmary Z, Erdi P (2003) Dynamically detuned oscilla-
79:25542558. tions account for the coupled rate and temporal code of place
Hoppensteadt FC (1986) An introduction to the mathematical proper- cell ring. Hippocampus 13:700714.
ties of neurons. Cambridge, UK: Cambridge University Press. Leutgeb JK, Leutgeb S, Treves A, Meyer R, Barnes CA, McNaughton
Hori E, Tabuchi E, Matsumura N, Tamura R, Eifuku S, Endo S, BL, Moser MB, Moser EI (2005) Progressive transformation of
Nishijo H, Ono T (2003) Representation of place by monkey hippocampal neuronal representations in morphed environ-
hippocampal neurons in real and virtual translocation. ments. Neuron 48:345358.
Hippocampus 13:190196. Lever C, Wills T, Cacucci F, Burgess N, OKeefe J (2002) Long-term
Howard M, Kahana MJ (2001) A distributed representation of tem- plasticity in the hippocampal place cell representation of envi-
poral context. J Math Psychol 46:269299. ronmental geometry. Nature 416:9094.
 Computational Models of Spatial and Mnemonic Functions 747

Levy WB (1996) A sequence predicting CA3 is a exible associator Mezard M, Nadal J-P (1989) Learning in feedforward layered net-
that learns and uses context to solve hippocampal-like tasks. works: the Tiling algorithm. J Physics A 22:21912203.
Hippocampus 6:579590. Milner AD, Dijkerman HC, Carey DP (1999) Visuospatial processing
Lieblich I, Arbib MA (1982) Multiple representations of space iun- in a case of visual form agnosia. In: The hippocampal and parietal
derlying behavior. Behav Brain Sci 5:627659. foundations of spatial cognition (Burgess N, Jeffery KJ, OKeefe J,
Lisman JE, Grace AA (2005) The hippocampal-VIA Loop: control- eds), pp 443466. Oxford, UK: Oxford University Press.
ling the entry of information into long-term memory. Neuron Moll M, Miikkulainen R (1997) Convergence-zone episodic memory:
46:703713. analysis and simulations. Neural Networks 10:10171036.
Ludvig N, Tang HM, Gohil BC, Botero JM (2004) Detecting location- Montague PR, Dayan P, Sejnowski TJ (1996) A framework for mes-
specic neuronal ring rate increases in the hippocampus of encephalic dopamine systems based on predictive Hebbian
freely-moving monkeys. Brain Res 1014:97109. learning. J Neurosci 16:19361947.
Mallot HA, Gillner S (2000) Route navigating without place recogni- Morris RGM, Garrud P, Rawlins JN, OKeefe J (1982) Place naviga-
tion: what is recognised in recognition-triggered responses? tion impaired in rats with hippocampal lesions. Nature
Perception 29:4355. 297:681683.
Manns JR, Squire LR (1999) Impaired recognition memory on the Morton J, Hammersley RH, Bekerian DA (1985) Headed records: a
doors and people test after damage limited to the hippocampal model for memory and its failure. Cognition 20:123.
region. Hippocampus 9:495499. Muller RU, Kubie JL (1987) The effects of changes in the environ-
Marr D (1970) A theory for cerebral cortex. Proc R Soc Lond B Biol ment on the spatial ring of hippocampal complex-spike cells. J
Sci 176:161234. Neurosci 7:19511968.
Marr D (1971) Simple memory: a theory for archicortex. Philos Trans Muller RU, Kubie JL, Saypoff R (1991) The hippocampus as a cogni-
R Soc Lond B Biol Sci 262:2381. tive graph (abridged version). Hippocampus 1:243246.
Maurer AP, Van Rhoads SR, Sutherland GR, Lipa P, McNaughton Muller RU, Stead M, Pach J (1996) The hippocampus as a cognitive
BL (2005) Self-motion and the origin of differential spatial graph. J Gen Physiol 107:663694.
scaling along the septo-temporal axis of the hippocampus. Murray EA, Mishkin M (1998) Object recognition and location
Hippocampus 15:841852. memory in monkeys with excitotoxic lesions of the amygdala
McClelland JL, McNaughton BL, OReilly RC (1995) Why there are and hippocampus. J Neurosci 18:65686582.
complementary learning systems in the hippocampus and neo- Murre JM (1996) TraceLink: a model of amnesia and consolidation
cortex: insights from the successes and failures of connectionist of memory. Hippocampus 6:675684.
models of learning and memory. Psychol Rev 102:419457. Nadel L, Moscovitch M (1997) Memory consolidation, retrograde
McClelland JL, Rumelhart DE (1986) Parallel distributed pro- amnesia and the hippocampal complex. Curr Opin Neurobiol
cessing: explorations in the microstructure of cognition. Vol 2. 7:217227.
Psychological and biological models. Cambridge, MA: MIT Nader K, Schafe GE, LeDoux JE (2000) The labile nature of consoli-
Press. dation theory. Nat Rev Neurosci 1:216219.
McNaughton BL, Barnes CA, OKeefe J (1983) The contributions of Nakazawa K, Quirk MC, Chitwood RA, Watanabe M, Yeckel MF, Sun
position, direction, and velocity to single unit activity in the LD, Kato A, Carr CA, Johnston D, Wilson MA, Tonegawa S
hippocampus of freely-moving rats. Exp Brain Res 52:4149. (2002) Requirement for hippocampal CA3 NMDA receptors in
McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW, associative memory recall. Science 297:211218.
Knierim JJ, Kudrimoti H, Qin Y, Skaggs WE, Suster M, Weaver Nicolelis MA, Baccala LA, Lin RC, Chapin JK (1995) Sensorimotor
KL (1996) Deciphering the hippocampal polyglot: the hip- encoding by synchronous neural ensemble activity at multiple
pocampus as a path integration system. J Exp Biol 199:173185. levels of the somatosensory system. Science 268:13531358.
McNaughton BL, Battaglia FP, Jensen O, Moser EI, Moser MB (2006). Norman KA (2003) Episodic memory, computational models of. In:
Path integration and the neural basis of the cognitive map. Nat Encyclopedia of cognitive science (Nadel L, ed), pp 1523.
Rev Neurosci 7:663678. London: Nature Publishing.
McNaughton BL, Morris RG (1987) Hippocampal synaptic enhance- Norman KA, OReilly RC (2003) Modeling hippocampal and
ment and information storage within a distributed memory sys- neocortical contributions to recognition memory: a comple-
tem. Trends Neurosci 10:408415. mentary-learning-systems approach. Psychol Rev 110:611646.
McNaughton BL, Nadel L (1990) Hebb-Marr networks and the neu- OKeefe J (1990) A computational theory of the hippocampal cogni-
robiological representation of action in space. In: Neuroscience tive map. Prog Brain Res 83:301312.
and connectionist theory (Gluck MA, Rumelhart DE, eds), pp OKeefe J (1991) The hippocampal cognitive map and navigational
163. Hillsdale, NJ: Lawrence Erlbaum. strategies. In: Brain and Space (Paillard J, ed), pp 273295.
Mehta MR, Barnes CA, McNaughton BL (1997) Experience- Oxford, UK: Oxford University Press.
dependent, asymmetric expansion of hippocampal place elds. OKeefe J, Burgess N (1996) Geometric determinants of the place
Proc Natl Acad Sci USA 94:89188921. elds of hippocampal neurons. Nature 381:425428.
Mehta MR, Quirk MC, Wilson MA (2000) Experience-dependent OKeefe J, Burgess N (2005) Dual phase and rate coding in hip-
asymmetric shape of hippocampal receptive elds. Neuron pocampal place cells: theoretical signicance and relationship to
25:707715. entorhinal grid cells. Hippocampus 15:853866.
Mehta MR, Lee AK, Wilson MA (2002) Role of experience and oscil- OKeefe J, Conway DH (1978) Hippocampal place units in the freely
lations in transforming a rate code into a temporal code. Nature moving rat: why they re where they re. Exp Brain Res
417:741746. 31:573590.
Mensink GJ, Raaijmakers JG (1988) A model for interference and for- OKeefe J, Nadel L (1978) The hippocampus as a cognitive map.
getting. Psychol Rev 95:434455. Oxford, UK: Oxford University Press.
748 The Hippocampus Book

OKeefe J, Recce ML (1993) Phase relationship between hippocam- Foundations (Rumelhart DE, McClelland JL, eds), pp 151193.
pal place units and the EEG theta rhythm. Hippocampus MIT Press.
3:317330. Samsonovich A, McNaughton BL (1997) Path integration and cogni-
OKeefe J, Speakman A (1987) Single unit activity in the rat hip- tive mapping in a continuous attractor neural network model. J
pocampus during a spatial memory task. Exp Brain Res 68:127. Neurosci 17:59005920.
Oja E (1982) A simplied neuron model as a principal component Save E, Nerad L, Poucet B (2000) Contribution of multiple sensory
analyzer. J Math Biol 15:267273. information to place eld stability in hippocampal place cells.
Packard MG, McGaugh JL (1996) Inactivation of hippocampus Hippocampus 10:6476.
or caudate nucleus with lidocaine differentially affects expres- Scholkopf B, Mallot HA (1995) View-based cognitive mapping and
sion of place and response learning. Neurobiol Learn Mem path planning. Adaptive Behav 3:311348.
65:6572. Schultz W, Dayan P, Montague PR (1997) A neural substrate of pre-
Pouget A, Sejnowski TJ (1997) A new view of hemineglect based on diction and reward. Science 275:15931599.
the response properties of parietal neurones. Philos Trans R Soc Scoville WB, Milner B (1957) Loss of recent memory after bilateral
Lond B Biol Sci 352:14491459. hippocampal lesions. J Neurol Neurosurg Psychiatry 20:1121.
Quirk GJ, Muller RU, Kubie JL, Ranck JB Jr (1992) The positional r- Sharp PE (1991) Computer simulation of hippocampal place cells.
ing properties of medial entorhinal neurons: description and Psychobiology 19:103115.
comparison with hippocampal place cells. J Neurosci Sharp PE (1996) Multiple spatial/behavioral correlates for cells in the
12:19451963. rat postsubiculum: multiple regression analysis and comparison
Raaijmakers JG, Shiffrin RM (1981) Search of associative memory. to other hippocampal areas. Cereb Cortex 6:238259.
Psychol Rev 88:93134. Sharp PE, Green C (1994) Spatial correlates of ring patterns of sin-
Ratcliff R (1990) Connectionist models of recognition memory: con- gle cells in the subiculum of the freely moving rat. J Neurosci
straints imposed by learning and forgetting functions. Psychol 14:23392356.
Rev 97:285308. Shors TJ, Miesagaes G, Beylin A, Zhao M, Rydel T, Gould E (2001)
Recce M, Harris KD (1996) Memory for places: a navigational model Neurogenesis in the adult is involved in the formation of trace
in support of Marrs theory of hippocampal function. memories. Nature 410:372376.
Hippocampus 6:735748. Skaggs WE, Knierim JJ, Kudrimoti H, McNaughton BL (1995) A
Redish AD (1999) Beyond the cognitive map: from place cells to model of the neural basis of the rats sense of direction. In:
episodic memory. Cambridge, MA: MIT Press. Neural information processing systems 7 (Hanson SJ, Cowan JD,
Redish AD, Elga AN, Touretzky DS (1996) A coupled attractor model Giles CL, eds), pp 173180. Cambridge, MA: MIT Press.
of the rodent head direction system. Network 7:671685. Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta
Redish AD, Rosenzweig ES, Bohanick JD, McNaughton BL, Barnes phase precession in hippocampal neuronal populations and the
CA (2000) Dynamics of hippocampal ensemble activity realign- compression of temporal sequences. Hippocampus 6:149172.
ment: time versus space. J Neurosci 20:92989309. Skaggs WE, McNaughton BL (1998) Spatial ring properties of hip-
Redish AD, Touretzky DS (1998) The role of the hippocampus in pocampal CA1 populations in an environment containing two
solving the Morris watermaze. Neural Comput 10:73111. visually identical regions. J Neurosci 18:84558466.
Roberts AC, Robbins TW, Weiskrantz L (1998) The prefrontal cortex. Speakman A, OKeefe J (1990) Hippocampal complex spike cells do
Oxford, UK: Oxford University Press. not change their place elds if the goal is moved within a cue
Rolls ET (1996) A theory of hippocampal function in memory. controlled environment. Eur J Neurosci 7:544555.
Hippocampus 6:601620. Spiers HJ, Maguire EA, Burgess N (2001) Hippocampal amnesia.
Rolls ET, Treves A (1997) Neural networks and brain function. Oxford, Neurocase 7:357382.
UK: Oxford University Press. Squire LR, Zola SM (1998) Episodic memory, semantic memory, and
Rolls ET, Robertson RG, Georges-Francois P (1997) Spatial view cells amnesia. Hippocampus 8:205211.
in the primate hippocampus. Eur J Neurosci 9:17891794. Sutton RS, Barto AG (1988) Reinforcement learning: an introduction.
Rosenbaum RS, Priselac S, Kohler S, Black SE, Gao F, Nadel L, Cambridge, MA: MIT Press.
Moscovitch M (2000) Remote spatial memory in an amnesic Taube JS (1998) Head direction cells and the neuropsychological
person with extensive bilateral hippocampal lesions. Nat basis for a sense of direction. Prog Neurobiol 55:225256.
Neurosci 3:10441048. Taube JS, Muller RU (1998) Comparisons of head direction cell activ-
Royer S, Pare D (2003) Conservation of total synaptic weight through ity in the postsubiculum and anterior thalamus of freely mov-
balanced synaptic depression and potentiation. Nature ing rats. Hippocampus 8:87108.
422:518522. Taube JS, Muller RU, Ranck JB Jr (1990) Head-direction cells
Rugg MD, Yonelinas AP (2003) Human recognition memory: a cog- recorded from the postsubiculum in freely moving rats. II.
nitive neuroscience perspective. Trends Cogn Sci 7:313319. Effects of environmental manipulations. J Neurosci 10:436447.
Rumelhart DE, McClelland JL (1986) Parallel distributed processing: Teng E, Squire LR (1999) Memory for places learned long ago is
explorations in the microstructure of cognition. Vol 1. intact after hippocampal damage. Nature 400:675677.
Foundations. Cambridge, MA: MIT Press. Teyler TJ, DiScenna P (1986) The hippocampal memory indexing
Rumelhart DE, Hinton GE, Williams RJ (1986) Learning internal theory. Behav Neurosci 100:147154.
representations by error propagation. In: Parallel distributed Tolman EC (1948) Cognitive maps in rats and men. Psychol Rev
programming. Vol 1. Foundations (Rumelhart DE, McClelland 55:189208.
JL, eds), pp 318364. Cambridge, MA: MIT Press. Touretzky DS, Redish AD (1996) Theory of rodent navigation
Rumelhart DE, Zipser D (1986) Feature discovery by competitive based on interacting representations of space. Hippocampus
learning. In: Parallel distributed programming Vol. 1. 6:247270.
 Computational Models of Spatial and Mnemonic Functions 749

Treves A, Rolls ET (1992) Computational constraints suggest the Willshaw DJ, Buckingham JT (1990) An assessment of Marrs theory
need for two distinct input systems to the hippocampal CA3 of the hippocampus as a temporary memory store. Philos Trans
network. Hippocampus 2:189199. R Soc Lond B Biol Sci 329:205215.
Trullier O, Wiener SI, Berthoz A, Meyer JA (1997) Biologically based Willshaw DJ, Buneman OP, Longuet-Higgins HC (1969) Non-holo-
artificial navigation systems: review and prospects. Prog graphic associative memory. Nature 222:960962.
Neurobiol 51:483544. Wilson MA, McNaughton BL (1993) Dynamics of the hippocampal
Tsodyks MV, Skaggs WE, Sejnowski TJ, McNaughton BL (1996) ensemble code for space. Science 261:10551058.
Population dynamics and theta rhythm phase precession of hip- Yonelinas AP (2002) The nature of recollection and familiarity: a
pocampal place cell firing: a spiking neuron model. review of 30 years of research. J Mem Language 46:441517.
Hippocampus 6:271280. Yonelinas AP, Kroll NE, Quamme JR, Lazzara MM, Sauve MJ,
Tulving E (1993) What is episodic memory? Curr Perspect Psychol Sci Widaman KF, Knight RT (2002) Effects of extensive temporal
2:6770. lobe damage or mild hypoxia on recollection and familiarity.
Vargha-Khadem F, Gadian DG, Watkins KE, Connelly A, Van Nat Neurosci 5:12361241.
Paesschen W, Mishkin M (1997) Differential effects of early hip- Zhang K (1996) Representation of spatial orientation by the intrinsic
pocampal pathology on episodic and semantic memory. Science dynamics of the head-direction cell ensemble: a theory. J
277:376380. Neurosci 16:21122126.
Wallenstein GV, Eichenbaum H, Hasselmo ME (1998) The hip- Zhu XO, McCabe BJ, Aggleton JP, Brown MW (1996) Mapping visual
pocampus as an associator of discontiguous events. Trends recognition memory through expression of the immediate early
Neurosci 21:317323. gene c-fos. Neuroreport 7:18711875.
Wallenstein GV, Hasselmo ME (1997) GABAergic modulation of Zipser D (1985) A computational model of hippocampal place elds.
hippocampal population activity: sequence learning, place eld Behav Neurosci 99:10061018.
development, and the phase precession effect. J Neurophysiol Zipser D (1986) Place recognition. In: Parallel distributed program-
78:393408. ming. Vol 2. Psychological and biological models (McClelland
Wan H, Aggleton JP, Brown MW (1999) Different contributions of JL, Rumelhart DE, eds), pp 432470. Cambridge, MA: MIT
the hippocampus and perirhinal cortex to recognition memory. Press.
J Neurosci 19:11421148. Zola SM, Squire LR, Teng E, Stefanacci L, Buffalo EA, Clark RE (2000)
Wilkie DM, Palfrey R (1987) A computer simulation model of rats Impaired recognition memory in monkeys after damage limited
place navigation in the Morris watermaze. Behav Res Meth to the hippocampal region. J Neurosci 20:451463.
Instrum Comput 19:400403. Zola-Morgan S, Squire LR, Ramus SJ (1994) Severity of memory
Wills T, Lever C, Cacucci F, Burgess N, OKeefe J (2005) Attractor impairment in monkeys as a function of locus and extent of
dynamics in the hippocampal representation of the local envi- damage within the medial temporal lobe memory system.
ronment. Science 308:873876. Hippocampus 4:483495.
This page intentionally left blank
15 Richard Morris

Stress and the Hippocampus

formation is a part, others also have suggested that inhibitory

15.1 Overview processing should not be ignored (as it tends to be in the
memory theories discussed in Chapter 13) because hip-
Anxiety, fear and stress are, on the face of it, very different pocampal lesions also cause decits in various forms of
entities from memory, and it may therefore come as a surprise nonspatial-context extinction, negative-occasion settings, and
that the hippocampus has also been implicated in these discrimination reversal (Chan et al., 2001; Davidson and
processes. This chapter focuses primarily on the role of the Jarrard, 2004). Acceptance of the earliest version of the septo-
hippocampal formation and other higher structures, such as hippocampal theory of anxiety suffered from exclusion of the
the amygdala and prefrontal cortex, in the negative feedback amygdala from the BIS despite growing evidence of the role of
regulation of the hypothalamic-pituitary-adrenal (HPA) the amygdala in learned fear and the modulation of memory
axisthe major system involved in orchestrating the bodys in emotional situations. This exclusion has been remedied in
reactions to both acute and chronic stress. There are two main the updated theory that now extends the envelope of the puta-
aspects to consider. The rst concerns how hippocampal tive BIS to a number of brain structures that are held to medi-
memory processing and synaptic plasticity is modulated by ate the neural processing involved in coordinating behavior in
stress hormones, with particular reference to an inverted U- conict situations, such as those involving fear and anxiety.
shaped function relating stress to optimal performance. The McNaughton and Corr (2004) have taken this further by
second is how the hippocampus interconnects with structures incorporating the Blanchards idea of a categorical distinction
that, in turn, innervate the paraventricular nucleus of the between fear and anxiety and by extending their intriguing
hypothalamus to complete the loop of HPA axis regulation. concept of defensive distance (Blanchard et al., 1997). Fear,
The larger part of the chapter is concerned with the rst of according to this view, is a set of behaviors elicited by a pred-
these topics, about which much more is known, but we also ator that includes ght, ight, and freezing. What behavior is
touch on the second. chosen depends, in part, on the perceived distance of the
Before outlining these subjects, it is important to note that prey from the predator. Anxiety, in contrast, is a process of risk
an early proponent of the idea that fear and anxiety were assessment associated with the potential presence of a preda-
mediated by the hippocampus was Gray (1982) with his tor that is expressed as the inhibition of whatever behavior in
septo-hippocampal theory. It was based on the notion of which the prey animal is then engaged. Anxietybut not
hippocampal processing of signals of reward and punishment fearis thought to be sensitive to anxiolytic drugs and only
and the specic role the hippocampal theta rhythm, at learned anxiety mediated by the amygdala/septo-hippocampal
frequencies of around 7 Hz, might play in mediating the system. Fear is in the province of the amygdala. The main
inhibition of prepotent behavioral responses that lead to pun- empirical basis for implicating the hippocampus in anxiety is
ishment or frustrating nonreward One experimental phe- the perceived similarity between the pattern of effects of a
nomenenonincreased resistance to extinction after partial wide range of anxiolytic drugs on hippocampus-dependent
reinforcement for running in a simple runwaywas a corner- tasks and that of the frank lesions that dene such tasks (Gray,
stone of the original theory. Over the years modications 2000). Certain studies pointing to differential functions of the
became necessary, and it was been updated and elaborated by dorsal (septal) and ventral (temporal) regions of the hip-
Gray in 2000. Not alone in maintaining that the brain has a pocampal formation provide partial support for this idea, as
behavioral inhibition system (BIS) of which the hippocampal lesions of the ventral hippocampus do impair tasks involving

751
752 The Hippocampus Book

both unconditioned fear (Kjelstrup et al., 2002) and some

anxiety-related behaviors (Bannerman et al., 2004). However, Box 151
Evidence Linking the Hippocampus and Stress
the theorys treatment of amnesia as hypermnesia rather
than true memory loss does not command much support. 1. The presence of mineralocorticoid and glucocorticoid
An ostensibly related idea, but of distinct provenance, is receptors in the animal and human hippocampus.
that the hippocampus has a key role in the regulation of the 2. An inverted U-shaped function between level of acute
HPA axis. The discovery by McEwen in 1968 that the hip- stress and memory.
pocampus contains a high density of corticosteroid receptors 3. Stress modulation of intrinsic hippocampus excitability
opened up a eld of research that had hitherto focused largely and activity-dependent synaptic plasticity associated with
on the HPA axis alone (McEwen et al., 1968). This eld has learning and memory.
enabled classic concepts, such as the general adaptation syn- 4. Chronic exposure to high levels of stress or stress hor-
drome of Seyle (1936) and the inverted U-shaped function mones is associated with structural changes in area CA3
of the hippocampus and, during pregnancy, fetal repro-
relating stress to alterations in many aspects of performance,
gramming of the HPA axis that lasts into adulthood.
to be tackled at a more mechanistic level. Modern molecular
5. Stress or stress hormones can impair neurogenesis in the
and cell-biological research coupled to behavioral studies is dentate gyrus.
also helping us toward an understanding of why low levels of
stress are good for cognitive exibility and memory formation
but higher levels are not. Indeed, sustained high levels of stress axis activity itselfhas multiple effects, including the cascade
cause deleterious structural changes to specic regions of the of corticotrophin-releasing hormone (CRH) release from the
hippocampal formation. On this view, two strands of thinking paraventricular nucleus (PVN) of the hypothalamus, leading
about the hippocampus are linked: its role in stress and in to ACTH secretion in the anterior pituitary and consequently
memory. Thinking about this issue is also a preliminary to the glucocorticoid (GC) release from the adrenal cortex into the
topic of the nal chapter of this book, which deals with the circulatory system. The cortisol (humans) or corticosterone
hippocampus and neurological disease (see Chapter 16). (rodents) so released then feeds back onto both hypothalamic
Following his pioneering work, further studies by McEwen and higher brain structures, including the hippocampus. The
and his colleagues revealed that the highest concentrations HPA axis also functions in constitutive housekeeping mode
of mineralocorticoid receptors in the brain are found in the by regulating the secretion of GCs in response to the diurnal
hippocampus (McEwen et al., 1980). As increased release of cycle of rest and activity, primarily through reexive feed-
corticosterone from the adrenals occurs during stressful situ- back routes rather than via higher brain structures mediating
ations, these observations linked the hippocampus directly planning and memory. Under basal conditions, GCs exhibit a
with bodily responses to stress, a concept that is extremely dif- 24-hour circadian rhythm during which maximum steroid
cult to dene precisely because it can refer to both distress concentrations (cortisol in humans, corticosterone in rats and
and exhilaration (Goldstein and McEwen, 2002). McEwens mice) are associated with awakening and minimum levels
seminal observations on corticosterone led to the birth of the with nocturnal sleeping periods. A rhythm is also observed in
stresshippocampus link (Lupien and Lepage, 2001). There nocturnal species but with a different pattern of entrainment.
has since been a great deal of research evaluating the partici- The term allostasis was introduced by Sterling and Eyer
pation of the hippocampus in the many facets of Selyes gen- (1988) to characterize how blood pressure and heart rate
eral adaptation syndrome. Five key lines of evidence implicate responses vary with experience and throughout the day and
the hippocampus in stress (see Box 151). how the set-point for these can be changed. McEwen (2002)
On perception of a stressor, be it physical or cognitive, the argued for a broader denition in which allostasis refers to a
central nervous system (CNS) mediates the synergistic activa- process, including activation of the SAM and HPA axes, whose
tion of the neuroendocrine, immune, and autonomic nervous essential function is to actively maintain homeostasis. Allosta-
systems. In so doing, it optimizes physiological parameters sis is thus a process of stability through change (somewhat
that help subserve short-term coping needs (e.g., gluconeoge- analogous to the shock absorbers and suspension of a car),
nesis, enhanced oxygen delivery, increased heart rate, effective and it applies not only to protective internal physiological
vigilance). The most commonly studied physiological sys- responses but also to overt, ostensibly paradoxical alterations
temsthe sympathetic-adrenal medullary system (SAM) and in cognition or behavior that favor future survival (McEwen,
the HPA axiscoordinate a two-phase cascade of events. 2002). It may, for example, include enhanced memory for sig-
First, emotional arousal triggers rapid sympathetic adrenergic nicant aspects of a stressful experience coupled with poorer
activity, inducing the release of catecholamines from the adre- consolidation of other unrelated information. This shift in the
nal medulla into the bloodstream. This SAM activation medi- pattern of information processing is to be distinguished from
ates the rapid phases of the ght or ight response to a merely asserting that, for example, memory has gotten better
threat, with adrenaline modulating the memory for events or worse. Stress has a further downside if it is sustained. The
triggering this arousal via an action within the amygdala separate concept of allostatic load (McEwen, 2000) refers to
(McGaugh, 1973; Davis et al., 1993). The second waveHPA the pathological cost of prolonged or overactive allostatic
 Stress and the Hippocampus 753

stress that can, over time, induce tissue damage and maladap- is not always obtained. Whereas neurotoxic lesions of area CA3
tive functioning. of the hippocampus do cause stress-induced hypersecretion of
This chapter considers ndings relevant to the stress corticosterone in rats (Roozendaal et al., 2001), large selective
hippocampus link beginning with a model of HPA axis regu- neurotoxic lesions of the entire hippocampus have, paradoxi-
lation; it then turns to allostatic modulation of hippocampal cally, been observed to cause no changes in corticosterone lev-
processing and, next, to the problems that arise with allostatic els at rest or under stress (Tuvnes et al., 2003). Moreover, work
load. This is followed by discussion of the interaction between during the 1970s revealed inconsistencies in the patterns of
the hippocampus and other brain structures in regulating regional retention of naturally occurring corticosterone and
stress and, nally, the mechanistic issue of how the hip- synthetic corticosteroids, raising the possibility of there being
pocampus orchestrates certain cognitive sequelae of arousing more than one adrenal steroid recognition system (de Kloet et
aversive experiences. al., 1975; McEwen et al., 1976). These questions are now
resolved in favor of models such as that shown in Figure 151
in which the HPA axis itself (shown in gray) is regulated reex-
ively by input from, for example, the subfornical organ and
15.2 Glucocorticoid Receptors more cognitively by a composite of higher brainstructures.
and Hippocampal Function Following the development of selective pure glucocorti-
coid compounds (Veldhuis et al., 1982) conrmed that corti-
15.2.1 Glucocorticoid Receptors Are Present costerone binds to two separate glucocorticoid receptor types
in the Animal and Human Hippocampus in the rat hippocampus. These receptors are transcription fac-
tors and now commonly referred to as the high-affinity min-
The discovery that glucocorticoid receptors are highly eralocorticoid (MR), or type I, receptor and the lower-affinity
expressed in the rodent hippocampus led to the conjecture glucocorticoid (GR), or type II, receptor. They display differ-
that they might regulate aspects of the HPA axis. A number of ential binding properties, distribution, and intracellular
early studies led to the supposition that the hippocampus has mechanisms (Joels, 2001). MRs are highly expressed in the
an inhibitory action (i.e., negative feedback) on circulating hippocampus and amygdala, and they have an approximately
levels of glucocorticoids (Feldman and Conforti, 1980; 10-fold higher affinity for corticosterone than GRs. In con-
Fischette et al., 1980; Wilson et al., 1980; Sapolsky et al., 1984, trast, GRs are widely expressed in most brain tissues and are
1990; Lupien and Lepage, 2001). The primary observation was detected, among other regions, in limbic structures, the cere-
a sustained corticosteroid response to various stressors in ani- bral cortex, and brain stem monoaminergic nuclei (Reul and
mals given hippocampus or fornix lesions. However, this result de Kloet, 1985; de Kloet et al., 1999; Helm et al., 2002; de

Figure 151. Neural afferents, corticosteroids, and

hypothalamo-pituitary-adrenal (HPA) control. The
HPA axis (gray shading) consists of the paraventricular
nuclei (PVN) of the hypothalamus that is activated ()
and inhibited (-) via descending pathways from the
prefrontal cortex, the hippocampus and lateral septum,
and the amygdala. It is also regulated reexively by cor-
ticosteroid signals entrained to the diurnal circadian
rhythm and by neural input from the subfornical organ
and area postrema. The PVN releases cortico-releasing
hormone (CRH) and vasopression (AVP) via the portal
blood vessels to activate the pituitary, which, in turn,
releases adrenocorticotropic hormone (ACTH) into the
bloodstream, which activates the adrenal gland. The
corticosterone then released into the bloodstream nds
its way to the higher brain structures involved in its
feedback regulation. Binding of corticosteroid to min-
eralocorticoid (type I) receptors (MRs) and glucocorti-
coid (type II) receptors (GRs) triggers genomic and
nongenomic mechanisms with one eventual result
being negative feedback (-) from the hippocampus
onto the PVN. (Source: After Welberg and Seckl, 2001
and Herman et al., 2003.)
754 The Hippocampus Book

Kloet, 2004). It is now recognized that corticosteroid actions 15.2.2 There Is an Inverted U-Shape Function
in the brain are, at least in part, mediated by both MRs and Between Level of Stress and Memory
GRs, with the ratio of receptor activation at these two sites
being critical to the modulatory effects observed on cellular It is commonplace for parents and schoolteachers to assert
activity, neural excitability, and network function (de Kloet et that a small amount of stress is good for you while also recog-
al., 1999; Joels, 1999). We shall see that other parameters, such nizing that too much can be harmful. This folk wisdom is now
as the context in which a stressor is perceived, also inuence on a rmer scientic footing (Fig. 152), with a large scientic
the dynamics of the HPA axis response. literature attesting to an inverted U-shaped function between
The various routes by which these higher brain structures severity of acute stress and cognitive function in animalsthe
might affect the PVN are gradually being unraveled (Herman so-called Yerkes-Dodson law. This now classic idea emerged
et al., 2003). A key idea is the contrast between a reexive from studies showing the initially synergistic but later delete-
and an anticipatory route by which the PVN is activated rious effects of increasing stress on learning in mice (Yerkes
or inhibited. With the former, sensory information reect- and Dodson, 1908). Experimentally, this is often investigated
ing actual homeostatic challenges (e.g., nociception) promotes with surrogate indices, such as stress being administration of
the activity of the PVN and thus the appropriate release of a particular dose of adrenal steroids and cognitive function
stress hormones. With the latter, signals associated with plan- being performance in a watermaze or other learning/memory
ning and decision-making (prefrontal lobe), memory of task. For example, mild stress and low concentrations of
stressful or potentially stressful experiences (hippocampus), administered steroids can facilitate spatial memory (Sandi et
and learned fear (amygdala) can all activate pathways leading al., 1997; Akirav et al., 2004), passive avoidance learning in
to hypothalamic nuclei in the neighborhood of the PVN. In chicks (Sandi and Rose, 1994, 1997; Liu et al., 1999), contex-
this way, stress hormone release anticipates the circumstances tual fear conditioning (Cordero et al., 2003), and classic eye-
in which it may be needed. For example, lesions of the ventral blink conditioning (Shors, 2001). Conversely, moderate to
subiculum reduce the corticosterone response to novelty high levels of stress or exogenous steroid have been reported
(Herman et al., 1998). Conversely, binding of corticosterone to to impair spatial memory (Diamond et al., 1996; de Quervain
MRs and GRs in the hippocampus can also inhibit PVN activ- et al., 1998; Stillman et al., 1998; Conrad et al., 1999; Diamond
ity upon termination of the stressor or the risk associated with and Park, 2000), recognition memory (Baker and Kim, 2002),
it and so limit the duration of a stress response. As we have and contextual fear conditioning (Pugh et al., 1997; Rudy et
seen, it was the extended duration of the stress response in al., 1999). These data fall neatly across the putative inverted U-
hippocampus-lesioned animals that indicated that this struc- shaped function linking levels of stress to learning ability, the
ture participated in regulation of the HPA axis. That this rst part of the function being the improvements that mild
response is not always seen suggests that multiple brain sys- stress can bring about, and the later parts of the curve being
tems are capable of exerting independent tonic inhibition of the impairments (Fig. 152). Although the criticism can be
the HPA axis.
Risold and Swanson (1996) suggested that there may be
Figure 152. Yerkes-Dodson law. Cognitive function, including
two routes by which the hippocampus mediates such mem-
learning, has a biphasic relation to levels of stress. As stress
ory-associated effects. One is via the orderly CA3 output to
increases, learning rst improves and then declines to levels below
the lateral septum (LS) and thence, via the precommissural baseline.
fornix, as a tripartite projection to the hypothalamus. The
other route is via the postcommissural fornix, which arises in
the subiculum and ends in the mammillary body and anterior
thalamus. Interestingly, there do not appear to be any direct,
monosynaptic inputs from the hippocampal formation to the +
PVN. The sophisticated summary by Herman et al. (2003) of
the relevant neuroanatomy suggests that central stress regula-
learning

tion relies heavily on hierarchical rather than dichotomous

pathways (p. 169) with the overall output of the stress
response ( or
) being dependent on the overall set of both
reexive and anticipatory inputs. Forebrain inuences on the _
PVN are polysynaptic, with the hippocampus having interac-
tions with the bed nucleus of the stria terminalis (BST) and
the hypothalamus, whereas the prefrontal lobe projects to the
BST and nucleus of the solitary tract. Notwithstanding
numerous important anatomical studies, it is still unclear how
Stress
activity in the hippocampus can both trigger corticosterone
release and limit its duration.
 Stress and the Hippocampus 755

made that the full inverted-U function has been witnessed in four trials per day in a hippocampus-dependent radial-arm
few if any behavioral studies, this interpretation of the data is watermaze (with the water at 24C), which contained either
still reasonably secure. four or six arms, and then given a retention trial 30 minutes
Two illustrative examples of behavioral studies of the later as their fth exposure to the maze. Training continued
inverted U-shaped function are the work of Sandi et al. across days with a new start and goal location (hidden plat-
(1997), in which mild stress improved the performance of rats forms) on each day. This is a working-memory task in which
in a conventional watermaze, and that of Diamond et al. the animals must acquire new information daily and exibly
(1999) in which exposure to a predator impaired performance retrieve information acquired that day, rather than a previous
in a working-memory version of a radial-arm watermaze. In day, during the retention trial. As shown in Figure 153B,
Sandi et al.s (1997) study, rats trained over 3 days with the exposure of the animals to a cat predator during the memory
water temperature at 19C showed a faster rate of acquisition delay interval between training and retention, which demon-
and better retention 1 week later than those trained at 25C strably increased stress in the rats exposed to the cat, resulted
(Fig. 153A). When trained at 25C and given injections in an elevation of the number of errors in the six-arm maze
of either vehicle or corticosterone immediately after each and in one of two training protocols for the four-arm maze.
training session, retention in the corticosterone-treated These data reect the right-hand side of the inverted U-
group was comparable to that of the 19C trained animals shaped function, with the stress-induced impairment being
from the rst experiment. In the third study of the series, the more prominent on a more a difficult learning task.
corticosterone-induced improvement did not occur when it In keeping with this body of work, the performance of
was given to animals being trained in 19C water. These data adrenalectomized (ADX) rats is impaired in a variety of
reect the left-hand side of the inverted U-shaped function. In behavioral tasks but is restored by administration of steroids
Diamond et al.s study (Fig. 153B), the rats were trained for in a dose-dependent manner (Vaher et al., 1994; Conrad et al.,

Figure 153. The inverted U-shaped impact of stress on spatial costerone at the higher but not the lower water temperature during
learning. A. Mean escape latencies during a retention test 1 week training. (Source: After Sandi et al., 1997.) B. Exposure to a cat pred-
after training in a wate rmaze at two temperatures. Training at the ator after daily training trials caused disruption of performance
colder temperature produced better performance (lower escape (increased errors) in a difficult radial-arm watermaze but not an
latency). This response was mimicked after administration of corti- easier one. HC, home cage. (Source: After Diamond et al., 1999.)

A. Mild stress improves retention of B. More severe stress impairs retention

spatial memory of spatial memory

40 1.5 4-arm maze-center entry

25C
19C
30
1.0
*
20
0.5
10

0 0.0
mean time (sec) to find the platform

40 1.5 4-arm maze-arm entry

error on the retention trial

vehicle
corticosterone
30 *
1.0
20 *
0.5
10

0 0.0
6-arm maze-arm entry
40 1.5 *
19 + vehicle *
19 + corticosterone
30 *
1.0
20
0.5
10

0 0.0
retention day HC 1-2 3-4 5-6
days of cat exposure
756 The Hippocampus Book

1997; McCormick et al., 1997). In a striking report, it has been cedures in which rodents are exposed to long periods of
claimed that steroids can even reverse the decit in memory in enforced immobility and unavoidable electric shock. That
a watermaze task that is induced by neurotoxic CA3 lesions, these procedures work is not in doubt, but it is unclear that
though it is unclear how they can compensate for the inter- such extreme conditions are really necessary to investigate the
rupted intrahippocampal communication that would neces- biological mechanisms of stress. Opinions change over time,
sarily occur with such lesions (Roozendaal et al., 2001). and the techniques used by Seyle (1936), for example, would
Indeed, with respect to the freezing response that reects con- not be regarded as ethical today. Mindful of the neuroetho-
ditioned fear and is often used in studies of contextual fear logical strategy discussed in relation to the cognitive map
conditioning, the reestablishment of control levels of per- theory (see Chapter 13), the assumption that the delivery of
formance by steroid administration in ADX rats is reported to electric shock is a biologically valid stressor is less clear than
occur only when the hippocampus is intact (Takahashi, 1995). researchers on stress may appreciate. Appearances may, how-
There are several outstanding puzzles with respect to these ever, be deceptive as exposure to the rst day of swimming in
ideas and data. First, the denition of severity of stress dif- the more ecologically natural situation of a watermaze can
fers across research groups and behavioral paradigms. For yield changes in corticosterone levels comparable to those
example, lowering water temperature during watermaze seen after delivery of electric shock. This habituates rapidly
training by 6C has been described as moderate to high stress such that a watermaze probe test as early as day 3 of training
(Akirav et al., 2004), whereas repeated foot shocks given over is associated with minimal corticosterone release (Tuvnes et
5 days has been reported as mild (Xiong et al., 2003). The basis al., 2003). The use of predator stress, introduced by Diamond
of these assignations is unclear. There is also circularity of the (Diamond et al., 1999; Mesches et al., 1999), appears to be a
attempts to identify severity because comparisons of expo- useful step forward and one that avoids using a stimulus that,
sure to foot shock, placement in a watermaze, or confronta- like shock, can cause overt tissue damage.
tion with a potential predator are not easily made in Research on the inverted U-shaped function in humans
procedural terms but can be quantied by resorting to the has, for the most part, substantiated the dose-dependent effect
surrogate measure of circulating corticosteroid levels. This is of GCs on cognitive function. Supraphysiological stress (or
probably reasonable but clearly assumes the very relation that administration of cortisol) impairs declarative memory
the experiments are in part intended to help establish. (Kirschbaum et al., 1996; Newcomer et al., 1999; de Quervain
Although many studies have shown a correlation between pre- et al., 2000). In contrast, acute low doses of cortisol adminis-
sumed severity of stress and circulating corticosterone levels, tered 1 hour before memory encoding has been shown to facil-
elevations in serum concentrations also accompany nonaver- itate word recall performance (Becker and Olton, 1980). It also
sive arousing experiences such as the presentation of a sexu- improves the long-term recall of emotionally arousing pictures
ally receptive female to a male rat. Spatial working memory compared to neutral stimuli (Buchanan and Lovallo, 2001).
was unimpaired by the exposure to a sexually receptive female Not all such effects are likely to be due to the actions of GCs
rat, a stimulus that was novel and arousing but not aversive. alone, however, with a well known study by (Cahill et al., 1994)
There was, nonetheless, an equivalent increase in serum corti- showing that the enhanced memory of an emotionally laden
costerone levels in the male rats exposed to a cat or to the prose passage compared to that of a more neutral story can be
female rat, but only the cat-exposed rats exhibited a signi- blocked by the -adrenergic receptor antagonist propranolol.
cant correlation between corticosterone levels and impaired When focusing on hippocampal regulation of the HPA axis, it
memory (Woodson et al., 2003). is vital not to ignore the key role that the SAM plays in imme-
Second, dening stress for animals used in experimental diate reactions to anxiety- and stress-provoking stimuli.
work can also be problematic. In humans, specic arousing or Discrepancies in published ndings also exist and have been
aversive situations may be perceived and then declared stress- explained as being a consequence of differential time of steroid
ful by some but not others, an opportunity for distinguishing administration during the natural circadian rhythm (Lupien
differing potential stressors that we do not have with animals. and Lepage, 2001). These disquieting items aside, a body of
Even in humans, stress really is an interaction between dispo- work using a plethora of methodologies and behavioral para-
sition or personality on the one hand and a range of different digms in both humans and animals has provided solid evi-
physical or cognitive potential stressors on the other. As ani- dence that varying corticosteroid levels may account for many
mals cannot make overt declarations of what is stressful to aspects of the inverted U-shaped function with respect to the
them, there is no alternative but to resort to assumptions or hippocampus and, indeed, other brain structures.
surrogate markers. The likely cognitive outcome of stress after
a training experience, such as facilitated or impaired memory 15.2.3 Stress Modulates Intrinsic Hippocampal
consolidation, seems to depend on the aversive qualities of the Excitability and Activity-dependent Synaptic
stress experience and stressor-specic pathways activated (i.e., Plasticity Associated with Learning and Memory
sensory or psychological) and even the sex of the animal (de
Kloet, 2003; Shors, 2004). When thinking about the mechanisms by which stress has its
A third concern about some research on stress using ani- impact on hippocampal function, the time course is critical.
mals is that it sometimes seems needlessly cruel, such as pro- Some effects are fast. Immediate effects of a stressor, such as
 Stress and the Hippocampus 757

those mediated by the sympathetic system, may occur via the de Kloet et al., 1999; Kim and Diamond, 2002). The parallel
action of adrenaline on the amygdala, which then affects the effects of varying levels of stress on synaptic plasticity and on
hippocampus through their mutual anatomical connections. certain forms of memory is tantalizing and can be taken as
Corticosteroid receptor activation can also induce immediate indirect support for the synaptic plasticity and memory
responses through reaggregation and molecular remodeling hypothesis outlined in Chapter 10.
of receptor molecules in the cytoplasm (de Kloet, 2003). The How stress engenders such dose-dependent biphasic effects
modulatory effects on glutamate and -aminobutyric acid on hippocampal plasticity and memory performance has been
(GABA)-mediated transmission, apparent after as short a explained in terms of differential concomitant corticosteroid
period as 20 minutes, are two examples of responses now receptor mediated actiona theory known as the binary
thought to be mediated by fast nongenomic mechanisms. For hormone response (Evans and Arriza, 1989) or the MR/GR
example, Karst et al. (2005) have shown very rapid increases in balance (Oitzl et al., 1995) hypothesis. Owing to the high
the frequency of miniature excitatory postsynaptic currents affinity of MRs to bind GCs, MRs are thought to be predom-
(EPSCs) in CA1 pyramidal cells in vitro that are blocked by an inantly, though not entirely, occupied at basal circulatory lev-
MR antagonist and absent in MR knockout mice. They sug- els of corticosteroids (observed during the circadian trough
gested that such effects are likely due to MRs that have shut- and nonstress periods). As corticosterone levels increase, GRs
tled from the nucleus/cytoplasm to cell membranes. become progressively occupied until both MRs and GRs are
However, the delayed effects of corticosteroids are now extensively activated. The action of GRs has a higher capacity
widely agreed to involve gene transcription. Intracellular stud- to be modulated through this progressive increase in occu-
ies in vitro by Joels and her colleagues have examined the con- pancy. In contrast, MR actions can be regulated through a
sequences of delayed GC transcriptional regulation (Joels, change in receptor number more than by ligand occupation
2001). Although no striking effects on hippocampal electrical per se. For example, acute swim stress, which leads to upregu-
activity are observed following MR occupation, there are a lation of MR expression in the hippocampus (Gesing et al.,
number of clear effects: reduced turnover of dentate granule 2001), can result in prolongation of LTP in the dentate gyrus
cells, reduced after-hyperpolarization (AHP), weaker res- when given after the induction of early LTP (Korz and Frey,
ponses to serotonin (5-HT), and smaller Ca2 currents. More- 2003). Similarly, rats treated with the antidepressant amitri-
over, induced alterations in membrane potential have revealed ptyline resulted in increased MR mRNA expression in the hip-
effects of GR activation on Ca2 currents through voltage- pocampus, and this increase correlated positively with their
dependent Ca2 channels. The increased inux of Ca2 may, improved spatial memory performance in a watermaze (Yau
for example, activate Ca2-dependent K currents that help et al., 1995). These effects are shown in Figure 155. MR acti-
limit excitability. As this then limits the fast transfer of excita- vation and partial GR occupancy is associated, through cell-
tory information, it causes subtle alterations in hippocampal biological mechanisms that are still unclear, with stable or
network activity that may be the basis of the impaired mem- even enhanced glutamatergic receptor transmission. This pro-
ory seen in the presence of a strong stressor. motes the maintenance of hippocampal information ow and
As discussed in Chapter 10, long-term potentiation (LTP) a tonic inuence on the HPA system (de Kloet et al., 1999;
and long-term depression (LTD) are physiological models of Joels, 2001). In contrast, strong stress and consequent GR sat-
activity-dependent synaptic plasticity and memory forma- uration attenuate excitatory input, a state thought to underlie
tion. Extensive observations obtained using in vitro and in the negative effects on plasticity mechanisms. These opposing
vivo electrophysiological techniques and a variety of physical effects are also schematically represented in the model linking
and psychosocial stressors have revealed that high levels of synaptic plasticity to the classic inverted U-shaped function
stress impair both LTP and primed burst potentiation (PBP), shown in Figure 154B.
a lower threshold form of LTP, whereas they enhance LTD In parallel with these differential effects on neural plastic-
induction in the dentate gyrus and CA1 regions of the hip- ity, MR and GR activation are thought to modulate distinct
pocampal formation (Foy et al., 1987; Diamond et al., 1992; components of information processing. Pharmacological
Kim et al., 1996; Garcia et al., 1997; Xu et al., 1997; Mesches et intervention using selective receptor antagonists administered
al., 1999; Diamond and Park, 2000). Furthermore, low corti- at specic stages of information processing indicate that MR
costeroid serum levels and acute, low doses of glucocorticoid activation can inuence processes involved in the attention to
receptor agonists are positively correlated with facilitated and interpretation of environmental stimuli and the conse-
induction of both PBP and LTP (Bennett et al., 1991; Pavlides quent selection of appropriate behavioral responses in ani-
et al., 1994). PBP appears to be particularly sensitive mals (Oitzl and de Kloet, 1992; Sandi and Rose, 1994), a
(Diamond et al., 1992). These ndings indicate that adrenal process akin to selective attention in humans. For example,
steroids and acute stress reversibly and biphasically modulate Oitzl and de Kloet (1992) showed that blocking MR receptors
synaptic plasticity in the hippocampal formation (Fig. can alter the search strategies of rats in a watermaze. In con-
154A), leading to the suggestion that stress and associated trast, activation of GRs is thought to regulate the acquisition
steroids might also induce metaplasticity (i.e., alter the phys- and consolidation processes of memory (Oitzl and de Kloet,
iological range of endogenous plasticity without necessarily 1992; Conrad et al., 1999). For example, in an elegant study
inducing either LTP or LTD directly) (Abraham, 1996, 2004; using a Y maze, Conrad et al. (1999) investigated the effects of
758 The Hippocampus Book

A. Primed burst potentiation with increasing corticosterone B. MR/GR regulation of synaptic plasticity

context of stressor unrelated to

learning paradigm learning paradigm
300
*
population spike amplitude (% change)

250

200 Threshold variable

depending on history
150 * +
of synaptic plasticity

synaptic strengh
and transcriptional
100
LTP activation
50 *

0 LTD
_
-50
*
-100
0 10 20 30 40 50 60 70 80 90 100

occupancy
relative
serum corticosterone (g/dl) GR

MR

Figure 154. The inverted U-shaped relation between stress, synap- As it increases, it differentially occupies MRs and GRs. The y-axis
tic plasticity, and cognitive function. A. The magnitude of primed depicts the potential for altering synaptic plasticity ( or
) and/
burst potentiation (PBP) as a function of corticosterone levels. or cognitive performance. The diagram depicts a model of how
(Source: After Diamond et al., 1992). B. Stress is plotted on the the pattern of activation of the hippocampal corticosteroid recep-
x-axis as the endogenous corticosteroid level induced by a stressor tors could affect both plasticity and memory. (Source: After de
or as the surrogate of an articial exogenous corticosteroid dose. Kloet et al., 1999.)

MR/GR agonists and antagonists on both ADX and unoper- U-shaped function, whereas the positive action of MRs and
ated rats. They found that spatial memory was impaired dur- GRs acting in concert on attentional and consolidation
ing both complete blockade of GRs (left-hand side of inverted processes may account for the peak in the inverted curve.
U) and high occupation (right-hand side) but was normal for The molecular mechanisms underlying the delayed modu-
all other treatments. In a similar protocol, Oitzl and de Kloet latory effects of corticosterone on hippocampal cell properties
(1992) had earlier shown that GR, but not MR, blockade are known to be transcription-dependent (de Kloet et al.,
interfered with the consolidation of spatial information. 1998; Joels, 2001). Current understanding of the mode by
Taken together, this suggests that GRs may, paradoxically, be which transcription is regulated is predominantly based on
responsible for both the upward and downward slopes of the investigations of GR processes in CA1 pyramidal and dentate

Figure 155. Parallels between the persistence of long-term poten- LTP over 24 hours is sensitive to an MR antagonist but not a GR
tiation (LTP) and spatial memory induced by exposure to stress. antagonist. (Source: After Korz and Frey, 2003.) B. Rats treated with
A. Induction of LTP by a weak tetanus results in decaying E-LTP the antidepressant amitriptyline showed an increase in MR mRNA.
unless followed by a 2-minute period of swimming in cold water. The extent of the increase correlated positively with spatial memory.
This exposure increases serum corticosterone. The persistence of (Source: After Yau et al., 1995.)

A. Post-induction stress can transform E-LTP into L-LTP B. Mineralocorticoid mRNA and
spatial learning
300 300 saline
swimming antagonist-injection
control vehicle-injection amitriptyline
*
PS amplitude (% Change)

* 80
% time in platform quadrant

200 200
* * 70
* 60
*
50
100 100 *
40

30 r= 0.85; p < 0.01

no treatment
tetanus
0 or swimming 0
20
-1 0 1 2 3 4 5 6 7 8 24 GR-ant MR-ant GR/MR-ant 100 110 120 130 140 150 160
time (h) MR mRNA expression (grains/ neuron)
 Stress and the Hippocampus 759

granule cells of the hippocampal formation. However, there is 1986), a theory relating the effects of chronic stress to hip-
some indication that these molecular processes may differ pocampal damage and dysfunction. As McEwen described it:
between receptor types. The long-established view was that The hypothesis states that glucocorticoids participate in a
GR complexes translocate to the nucleus, where they bind as feed-forward cascade of effects on the brain and body, in which
homodimers to glucocorticoid response elements (GRE) in progressive GC-induced damage to the hippocampus pro-
DNA of target genes, such as N-methyl-D-aspartate (NMDA) motes progressive elevations of adrenal steroids and dysregula-
receptors (Weiland et al., 1997) and sgk (rat serum- and tion of the HPA axis (McEwen, 1999). Subsequent research
glucocorticoid-inducible kinase), one of the genes proposed to has generally provided support for this hypothesis, the key
be involved in memory consolidation of spatial learning in rats observation being that prolonged stress and chronically ele-
(Tsai et al., 2002). In addition, however, GR complexes have vated HPA activity results in microscopic and even gross struc-
also been reported to regulate gene expression through trans- tural changes in the hippocampus in a variety of species.
activation or proteinprotein interaction with transcription There are several lines of evidence for different types of
factors such as AP1, NFKB, and CREB (Auphan et al., 1995). structural change in the hippocampal formation (Fig 156).
This nding indicates that steroid receptor-mediated and First, moderate durations of stress and high GC exposure in
transcription-dependent events in the hippocampus poten- rats (and tree shrews) cause reversible atrophy of apical den-
tially operate through the same signaling pathways. Given this, trites of CA3 pyramidal cells and dentate gyrus neurons
if cross-talk between signaling cascades were to occur, the tim- (Woolley et al., 1990; Magarinos and McEwen, 1995;
ing and nature of the inputs would be critical for determining Magarinos et al., 1996) and alteration in synaptic terminal
whether gene transcription was enhanced or repressed. It is structure (Magarinos et al., 1997). Not surprisingly, such den-
important to emphasize, however, that MR/GR gene-mediated dritic remodeling is accompanied by reversible impairment of
processes that alter neuronal activity do so conditionally, such the initial learning of a spatial reference memory task (Luine
as only when the membrane potential is shifted from its rest- et al., 1994). Second, chronic longer-term physical and psy-
ing level (de Kloet, 2003). Thus, there is considerable potential chosocial stress was reported to result in overt loss of hip-
for dynamic interaction between the strictly neural and strictly pocampal CA1 and CA3 neurons in rats and primates
neuroendocrine modes of cellular communication. (Sapolsky et al., 1985; Kerr et al., 1991; Mizoguchi et al., 1992).
Glucocorticoid receptor action is only part of the story: However, these ndings have not been replicated using
Stress-induced alterations in electrophysiological function are improved stereological methodology (West and Gundersen,
likely to involve multiple mechanisms and pathways over 1990; West, 1993), and whether cell death occurs in response
various time courses. Several endogenous neurotrans- to stress is now an issue of debate. Third, MRI studies of indi-
mitter/modulators (e.g., catecholamines, GABA, serotonin, viduals suffering from Cushings diseasea syndrome associ-
opioids, and tissue mediators such as glucose and neu- ated with chronic endogenously induced elevated levels in
rotrophins to name but a few) can modulate intrinsic proper- cortisolshow atrophy of the hippocampus. Specically, sig-
ties and cellular activity in the hippocampus and elsewhere in nicant correlations have been found between the severity of
the forebrain (McEwen and Sapolsky, 1995; McEwen, 2000; the hypercortisolemia, the extent of atrophy, and the magni-
Joels, 2001). Whether these mechanisms act independently in tude of cognitive impairment in a verbal learning and recall
parallel or are downstream of adrenal steroid action is yet to test (Starkman et al., 1992).
be elucidated. Nevertheless, progress in this eld has provided Stress-induced changes in hippocampal morphology
valuable insight into the neurobiological processes underlying can be attributed to reduced dendritic branching (in CA3) or
acute allostatic responses as well as the chronic maladaptive possibly to apoptosis (in the dentate gyrus). However, a third
stress consequences that may contribute to conditions such as factor has recently been considered as an additional con-
depression and neurodegenerative diseases such as Alzheimers tender: suppression of adult neurogenesis in the dentate
that have been associated with aberrant HPA axis activity and gyrus.
hippocampus-dependent cognitive decit (Holsboer and
Barden, 1996; Welberg and Seckl, 2001; Muller et al., 2004). 15.3.2 Stress or Stress Hormones Can Impair
Neurogenesis in the Hippocampus

Neurogenesis, the generation of new neurons in the adult

15.3 Stress and Hippocampal Structure brain (see Chapter 9), is now known to occur in (at least) two
regions of the CNS: the lateral ventricle and the subgranular
15.3.1 Chronic Exposure to High Levels of Stress zone in the dentate gyrus (DG) of the hippocampus. It has
or Stress Hormones Is Associated with been reported in a range of species, including rodents, mon-
Structural Changes in the Hippocampus keys and humans, as discussed in detail in Chapter 9.
Extensive studies using both acute and chronic stress/steroid
Evidence that adrenal steroids have a role in neuronal aging in protocols have also revealed that corticosteroids can inhibit
the hippocampus was rst provided by Landeld et al. (1978) postnatal and adult granule cell proliferation (Cameron and
and Sapolsky et al. (1984). This led Sapolsky to formulate the Gould, 1994; Gould et al., 1997). As granule cell precursors do
glucocorticosteroid cascade hypothesis (Sapolsky et al., not express MR and GR, this implies an indirect glucocorti-
760 The Hippocampus Book

Inhibition of neurogenesis Dendritic atrophy

entorinal
cortex
+

dentate MR
PreS R NMDA R +
gyrus + & GR inter +
-neuron
+
CA3
PreS R
+
Figure 156. Chronic stress causes struc-
glucocorticoid serotonin tural changes in the hippocampal formation.
A model of the mechanisms by which stress
alters neurogenesis and cell death (left), and
the integrity of the dendrites of area CA3 of
the hippocampal formation. (Source: After
STRESS
McEwen, 1999.)

coid mechanism, which Kim and Diamond (2002) suggested 15.3.3 Fetal Programming of GC Regulation
may occur through NMDA receptor activation.
McEwen argued that short-term structural plasticity (den- Adverse events during early life can inuence prenatal devel-
dritic pruning in CA3) and inhibition of neurogenesis (Fig. opment and may cause structural and functional changes in
156) are examples of adaptive protective mechanisms the brain that persist for the life-span, a phenomenon known
(allostasis) that could reduce the impact of glutamate and glu- as early-life programming (Seckl and Meaney, 2004). Fetal
cocorticoids causing more permanent damage. However, exposure to GCs has been proposed as a likely programming
stress is also known to exacerbate the effects of neurological factor. Although GCs are clearly necessary for normal develop-
insults such as hypoxia, ischemia, and seizures, increasing the ment, excess exposure has deleterious effects: inhibiting fetal
susceptibility of hippocampal cell death putatively through growth and altering the trajectory of tissue maturation. For
both GC-induced inhibition of glucose transport and example, exposure of pregnant rats to exogenous or endoge-
enhanced Ca2 inux. Why such a maladaptive response nous GCs not only reduces the birth weight of their progeny, it
occurs after neurological injury is puzzling. An intriguing idea produces permanent hypertension, hyperglycemia, and hyper-
proposed by Sapolsky, supported by evidence that the HPA insulinemia in the offspring, continuing through to adult life.
glucocorticosteroid cascade is an evolutionary conserved phe-
nomenon, is that the body simply has not evolved the capac- Figure 157. Stress and hippocampal size in humans. A longitudi-
ity or tendency to not secrete glucocorticoids during a crisis nal study of the correlation between sustained cortisol levels and
(Sapolsky, 2004). In effect, evolution has only gotten so far. hippocampal volume in humans. Higher cortisol levels are associ-
With this view, during protracted periods of intermittent ated with a smaller volume, providing evidence for the atrophy
stress the hippocampus may become impeded in its role in predicted in the model. (Source: Lupien et al., 1998.)
shutting off HPA axis stress activity. This results in 0.46
increased secretion of GCs and, in a positive feedback cycle
with negative consequences, ends up damaging the hip- 0.44
Hippocampal volume

pocampus itself, thereby further reducing the ability of the 0.42

hippocampus to regulate the HPA axis. The observations by
Lupien et al. (1998) of reduced hippocampal volume and 0.40
poorer memory in humans in association with chronically
0.38
high levels of cortisol is an apparently striking example of
such an effect (Fig. 157). Whether hippocampal atrophy and 0.36
inhibition of neurogenesis are the causes or consequences of
elevated GCs is therefore a chicken or egg question. What is
0.34
apparent is that this brain structure is both centrally involved 0.32
in the neural reaction to aspects of prolonged stress and itself -0.6 -0.4 -0.2 -0.0 0.2 0.4
a target of chronic stress. Cortisol slope (g/dl/year)
 Stress and the Hippocampus 761

Given that MRs and GRs are highly expressed in the devel- receptors, increase cAMP generation, and the induction of
oping brain (Fuxe et al., 1985; Kitraki et al., 1997), it is not specic transcription factors, most notably NGFI-A and AP-2.
surprising that the brain is particularly sensitive to early envi- These then bind to the GR gene promoter, inducing a specic
ronmental manipulations. The fetus is normally protected GR transcript in the hippocampus (McCormick et al., 2000).
from the relatively high levels of maternal physiological GCs However, a key question that remains is how perinatal
by an enzyme called 11 -HSD2, which catalyzes the conver- environmental events (e.g., stress) can permanently alter
sion of active GCs to inert forms; this enzyme is highly gene expression. Some evidence is emerging for methylation/
expressed in the placenta. There is also abundant 11 -HSD2 demethylation of specic promoters of the GR gene. This may
in the CNS at mid-gestation that presumably protects vul- change histone structure (acetylation/deacetylation), perma-
nerable developing cells from the premature action of GCs nently opening or hiding specic transcription features for the
(Brown et al., 1996). Expression of this enzyme is then dra- life-span. Preliminary data suggest that the putative NGFI-A
matically switched off at the end of mid-gestation in the rat site on the GR promoter (around exon 17one of the six
and mouse brain, and similarly in the human fetal brain alternate rst exons/promoters of the GR gene that is utilized
between gestational weeks 19 and 26 (Stewart et al., 1994), in the hippocampus) is subject to differential and permanent
thereby allowing exposure of the developing brain regions to methylation and/or demethylation in association with varia-
circulating GCs during the later stages of pregnancy. tions in maternal care (Weaver et al., 2004).
In animal studies, stress and late prenatal GC exposure per- To summarize this section, chronic exposure to GCs can
manently increase HPA axis activity and impair cognitive cause structural changes in the hippocampus, can alter neuro-
function in adults. This is likely, in part, because of reduced genesis, and during pregnancy can program cardiovascular,
MR and GR expression in the adult hippocampus. Interest- metabolic, and neuroendocrine disturbances that persist into
ingly, postnatal events such as neonatal handling or adop- adult life. The molecular mechanisms may reect permanent
tion can reverse the effects of prenatal stress (Maccari et al., changes in the expression of specic transcription factors
1995; Vallee et al., 1999). Thus, there appear to be several dis- including GR itself.
tinct time windows during which GC exposure has minimal
(early) or maximal (late) effects in rodents. The human HPA
axis also appears to be programmed by the early life environ-
ment. Maternal stress during pregnancy in humans reduces 15.4 Other Higher Brain Structures
birth weight, just as it does in rats. Mutations of the 11 - Implicated in Stress and Their
HSD2 gene that alter its effectiveness to protect the develop- Interaction with the Hippocampus
ing fetus also cause low birth weight, and low birth weight in
humans has been shown to correlate with increased basal and In a book focusing on the hippocampus, it is important to rec-
stimulated cortisol levels during adult life. Interestingly, the ognize once again that it cannot be considered alone and,
glucocorticoid agonist dexamethasone (DEX) is often given in specically, to appreciate the synergistic role that other brain
obstetric practice, such as when preterm labor threatens dur- structures play in hippocampus-mediated aspects of stress.
ing the last trimester. However, DEX readily crosses the pla- For example, electrophysiological and lesion studies provide
centa, but its effects on cognition and behavior in humans strong evidence that activation of the amygdala plays a critical
have not yet been investigated in detail. This point is impor- role in the rapid mediation of the effects of emotionally
tant to investigate as human neonates exposed to DEX may be arousing events on memory consolidation (McGaugh, 2000;
at risk of later disturbances of the HPA axis. McGaugh and Roozendaal, 2002). Adrenaline, like glucocorti-
Insight into the molecular mechanisms by which early life coid, is known to modulate long-term memory consolidation
environmental factors may program offspring physiology for fear-related or emotionally arousing events. Gold and
have come from studies dissecting the processes that underpin Van Buskirk (1975) were the rst to demonstrate that system-
postnatal environmental programming of the HPA axis in the atic injections of adrenaline (epinephrine) administered after
neonatal handling paradigm (Meaney et al., 2000). A short inhibitory avoidance training produced dose-dependent
period of daily handling of rat pups during the rst 2 weeks of enhancement of long-term retention. Subsequent research
life (around 15 min/day) results in enhanced maternal care- has indicated dose- and time-dependent effects of epineph-
related behaviors and permanently increased hippocampus rine the consolidation of a variety of aversion- and appetite-
GR levels. This potentiates the HPA axis sensitivity to GC- motivated tasks (Cahill and McGaugh, 1998). Many effects of
negative feedback and lowers plasma GC levels throughout amygdala activation are entirely separate from hippocampal
lifea programmatic resetting of the axis. The daily han- function, but not all.
dling model is of physiological relevance because natural An accumulating body of evidence suggests that the amyg-
variation in such maternal behavior correlates similarly with dala, particularly the basolateral nucleus, is involved in mod-
offspring HPA physiology and hippocampal GR expression ulating memory storage processes in the hippocampus. For
(Liu et al., 1997). The altered maternal care-related behaviors instance, the integrity of the basolateral nucleus is essential for
elevate thyroid hormones, which then induces 5-HT to the memory-modulatory effects induced by intrahippocam-
increase GR gene expression in the hippocampus via 5-HT7 pal infusion of GR agonists and antagonists (Roozendaal,
762 The Hippocampus Book

2002). Amygdala lesions also prevent the stress-induced frontal cortex but are detected only weakly in the hippocam-
suppression of hippocampal LTP (Kim et al., 2001, 2005), and pal formation (Sanchez et al., 2000). Lupien and Lepage
amygdala stimulation potentiates the long-term persistence (2001) have also drawn attention to studies indicating that
of hippocampal LTP (Frey et al., 2001). Moreover, intra- frontal lobe processing, (e.g., executive and working memory
basolateral amygdala (BLA) infusions of a GR antagonist tasks) can be more sensitive than hippocampus-dependent
impair retention of spatial memory in the watermaze processing to acute and short-term increases in corticos-
(Roozendaal and McGaugh, 1997). Such ndings have been teroids. In short, although research of stress-induced frontal
interpreted as support for the concept of emotional tagging, lobe dysfunction is in its relative infancy, it is now recognized
which refers to the idea that, emotional arousing events mark that other higher brain structures than the hippocampus play
an experience as important (Richter-Levin and Akirav, 2003). a role in engendering changes in cognitive function associated
More specically they suggest that full expression of the with stress states. Stress is thus partially emancipated from
effects of stress on the persistence of hippocampus-dependent the grip of pituitary control.
memory requires the co-activation of the amygdala, acting via There is a wider context in which to understand the chang-
reciprocal hippocampus-amygdala circuitry (Majak and ing role of higher brain structures, such as the hippocampus
Pitkanen, 2003). Furthermore, in a study using event-related and prefrontal cortex, over the course of evolution. As dis-
functional magnetic resonance imaging (MRI) to measure cussed by Keverne et al. (1996), much behavior in small-
neural activity in young healthy female adults during the brained mammals is part of physiological homeostasis and
retrieval of emotional and neutral pictures after a 1-year strongly determined by endocrine state. Thus, maternal
retention period, it was found that emotion selectively behavior never occurs outside the context of pregnancy and
enhanced activity that indexed retrieval success in both the parturition; sexual behavior is inevitably reproductive and
amygdala and the hippocampus (Dolcos et al., 2005). As generally restricted to periods of fertility; and forgaing and
Dolcos put it: One way of explaining their co-activation dur- feeding frequently occur in response to signals from the gut
ing emotional recollection is that emotion enhances recollec- and fat reserves. In contrast, in large-brained mammals such
tion-related activity in the hippocampus, whereas recollection as primates and humans, most sexual activity is nonreproduc-
enhances emotion-related activity in the amygdala. The two tive; feeding occurs to avoid, rather than react to, hunger; and
brain structures are acting in synergy. Emotional tagging and females do not necessarily have to undertake pregnancy and
the related idea of synaptic tagging (see Chapter 10) both pro- parturition to be maternal. Decision-making has switched
vide ways of thinking about how activity in an afferent brain from hormonal to cognitive determinants (Keverne, 2004).
structure can modulate the persistence of synaptic potentia- The mediation of stress by the HPA axis can be considered
tion in the hippocampus. A detailed discussion of the impact another example of how the evolution of the mammalian
of adrenal stress hormones on the amygdala and its contribu- brain has provided increased executive control of behavior
tion to the modulation of memory is beyond the scope of this and a degree of emancipation from hormonal determinants
book but is discussed by McGaugh and Roozendaal (2002). (Herman et al., 2003). In the case of human stress, cortisol is
Another brain area implicated in regulation of the HPA still released from the adrenals, but the entire HPA axis is reg-
axis is the prefrontal cortex (Diorio et al., 1993). In humans, ulated, both positively and negatively, by cognitive processing
an individuals perception of an experience, intricately tied to in areas of the brain involved in planning, decision-making,
past experience of similar events, strongly inuences whether and memory.
a specic arousing event has a positive or negative effect on
performance. For example, the impact of a specic physical
stressor on animals may depend on whether they can perform
any adaptable response to escape it or must merely endure it. 15.5 How the Hippocampus
Such psychologically driven factors are thought to involve Orchestrates Behavioral Responses
planning and decision-making in higher brain structures (e.g., to Arousing Aversive Experiences
frontal cortex) that then positively triggers top down stres-
sor specic pathways resulting in decreased HPA axis activity To conclude, several lines of evidence support the idea of an
(de Kloet, 2003). In this and other ways, the frontal cortex has allostasishippocampus link. They point to likely mechanisms
now been elevated to the status of being part of the regulatory by which stress steroids generate changes in hippocampal
HPA feedback circuitry (Lupien and Lepage, 2001), an addi- excitability and synaptic plasticity and how the hippocampus
tional anatomical component, as shown in Figure 151. (with other higher brain structures) is involved in stress feed-
The role of the frontal cortex during short-term and back circuitry. The nal topic for discussion is the nature, or
chronic stress has also received attention. Interest has arisen, character, of hippocampal processing during stress and the
in part, by the discovery that the GR distribution in humans impact that cognitive changes induced by it may have on
and nonhuman primates differs from rodent expression behavior during such experiences.
patterns (Sarrieau et al., 1986; Patel et al., 2000; Sanchez et First, acute stress induces an adaptive response that not
al., 2000). For instance, in monkeys GRs are reported to be only involves internal physiological mechanisms serving
preferentially distributed in cortical areas particularly the pre- short-term coping needs but also changes in behavior that
 Stress and the Hippocampus 763

favor future survival. Which action is takenght, ight, the death of President John Kennedy (November 1963) or the
freezeis determined in part by information acquired during unfolding events of the 9/11 tragedy (September 2001).
the stress episode or recalled from previous episodes. Focused Importantly, it is not only the stressor itself that is marked as
attention to and encoding of salient features of an individual important material to be remembered but also certain trivial
stress experience is therefore critical to the adaptiveness of details of the day on which it happened. Simultaneous hip-
the response. That emotionally arousing experiences tend to pocampal and amygdala memory processing is likely to be
be well and long remembered could be one selection pressure involved in the encoding, storage, and consolidation of any
that may have helped the various body and brain mechanisms potentially relevant information that may then aid adaptive
mediating stress to evolve (Cahill and McGaugh, 1996; Schafe action another time (McGaugh et al., 1993; Nadel and Jacobs,
et al., 2001; McGaugh and Roozendaal, 2002). 1996; Frey and Morris, 1998).
Why does stress have both positive and negative effects on Another connection between HPA housekeeping activity
hippocampus-dependent memory? Severity of stress is one and the hippocampus is in relation to the processing of novel
factor inuencing the outcomethe system may have evolved contextual information. Humans with unilateral or bilateral
to cope constructively with mild stress, but the mechanisms hippocampal damage do not show the normal peak cortisol
involved play the dangerous price of interacting with impor- response to awakening even though the remainder of the
tant entities such as Ca2 channels. Another possibility is that diurnal cycle is unaffected (Buchanan et al., 2004; Wolf et al.,
retrograde amnesia induced by unrelated (i.e., out of con- 2004). The psychological signicance of the cortisol awaken-
text) stress, whereas ostensibly a negative outcome is actu- ing response is unclear. However, one explanation is that it is
ally a positive adaptive response that helps to ensure greater part of the process of detecting change such as the transition
priority to the storage of information relevant to enhancing from the sleeping state to the waking state. It would be inter-
survival. Spatial memory in rats is impaired when learning esting to know if this response is larger when a person awak-
is followed by a variety of unrelated stressors, including ens in an unusual or unexpected place (e.g., a different place
inhibitory avoidance training, contextual fear conditioning, from that in which they went to sleep).
and forced exposure to an unfamiliar environment or preda- Finally, the experience of unique, acute, life-threatening
tor (Diamond et al., 2004). Diamond suggested that unrelated traumatic events (e.g., being unexpectedly involved in military
stressors cause alterations in synaptic weight (owing to LTP- action) can give rise to a persistent clinical condition called
like plasticity in the hippocampus caused by stress), which posttraumatic stress disorder (PTSD). Evidence from extensive
results in neural representations of the stressful experience studies investigating the neurobiology and cognitive character-
itself having the effect of overwriting any original and unre- istics of PTSD implicate GR receptor mechanisms in the hip-
lated learning going on at the time. pocampus. Three lines of evidence support this assertion. First,
Emotionally charged experiences within the context of a blood samples from PTSD subjects indicate lymphocyte GC
learning task have been the focus of much research by Sandi. receptors are more numerous and have increased sensitivity
With others, she has recognized that strong context-related (Yehuda et al., 1995). Second, enhanced cortisol suppression in
aversive stimuli can also give rise to poor memory of very response to administration of exogenous corticosteroids (i.e.,
demanding learning tasks. However, her data also indicate sensitization of the HPA axis) is consistent with the clinical
that moderate cold-water stress elicited during training can presentation of hyperreactivity (Yehuda, 2001). Third, brain
result in better acquisition even though the corticosteroid lev- imaging studies point to hippocampal atrophy and lack of the
els then induced are actually equivalent to those seen in stress- activation sometimes seen in the hippocampus during declar-
ful situations that are deleterious to memory (Sandi et al., ative memory tasks (Bremner et al., 1995; Gurvits et al., 1996;
1997; Akirav et al., 2001, 2004). A possible explanation relates Bremner, 2001; Villarreal et al., 2002), and decits in attention
to the in-context nature of the stressor and an apparent right- and short-term verbal memory are commonly reported
ward shift in the inverted U-shaped function; that is, high (Elzinga and Bremner, 2002; Sala et al., 2004).
corticosterone levels are now within the optimum perform- From a psychological perspective, individuals suffering
ance zone. from PTSD display repetitive reliving of the stress experience
Facilitation of memory induced by mild within-context and evident amplication of memory for the traumatic stim-
stress has also been explained in terms of the enhanced atten- ulus with, in contrast to ashbulb memory, poor memory for
tion and reactivity to stimuli and improved consolidation of the surrounding contextual material (Layton and Krikorian,
information (de Kloet et al., 1999; Akirav et al., 2004; Joels et 2002). This failure to consolidate material proximate to the
al., 2006). These factors may also contribute to a striking trauma could be due to strong activation of the amygdala and
human memory phenomenonflashbulb memoryin a negative impact of GCs on hippocampal function. This
which the vivid and persistent memories for the context of an would then be expected to give rise to the memory of the
emotionally arousing public event are also associated with trauma lacking any spatial and temporal frame of reference
surprisingly good memory for otherwise innocuous details of (Nadel and Jacobs, 1996). Instead of the events being rmly
the circumstances in which they were witnessed. Different bound in the past, the memory takes on a quality of the here
generations have different exemplars of this phenomenon, and now, leading individuals to believe they are literally reex-
such as remembering when and where one heard the news of periencing the event.
764 The Hippocampus Book

The themes of this chapter have been that stress mediates with combat-related posttraumatic stress disorder. Am J
certain of its effects via the hippocampus, which provides Psychiatry 152:973981.
feedback regulation of the HPA axis. Acute stress has a bipha- Bremner JD (2001) Hypotheses and controversies related to effects of
sic effect on synaptic plasticity and cognition. Chronic stress stress on the hippocampus: an argument for stress-induced
damage to the hippocampus in patients with posttraumatic
is, however, maladaptive and destructiveone of a number of
stress disorder. Hippocampus 11:7584.
ways in which the hippocampal formation can become dam-
Brown RW, Diaz R, Robson AC, Kotelevtsev YV, Mullins JJ, Kaufman
aged through life. Other pathological changes that affect the MH, Seckl JR (1996) The ontogeny of 11 beta-hydroxysteroid
hippocampal formation include the onset of specic neuro- dehydrogenase type 2 and mineralocorticoid receptor gene
logical diseases, such as epilepsy and neurodegenerative con- expression reveal intricate control of glucocorticoid action in
ditions such as Alzheimers disease. It is to these topics that we development. Endocrinology 137:794797.
turn in Chapter 16. Buchanan TW, Lovallo WR (2001) Enhanced memory for emotional
material following stress-level cortisol treatment in humans.
Psychoneuroendocrinology 26:307317.
ACKNOWLEDGMENTS Buchanan TW, Kern S, Allen JS, Tranel D, Kirschbaum C (2004)
Circadian regulation of cortisol after hippocampal damage in
R.G.M.M. acknowledges the advice and help of Helen Knight, Marie humans. Biol Psychiatry 56:651656.
Pezze, Carmen Sandi, and Joyce Yau on preparation of the chapter Cahill L, McGaugh JL (1996) Modulation of memory storage. Curr
and the critical comments of Marian Joels and Joyce Yau on the text. Opin Neurobiol 6:237242.
Cahill L, McGaugh JL (1998) Mechanisms of emotional arousal and
lasting declarative memory. Trends Neurosci 21:294299.
REFERENCES Cahill L, Prins B, Weber M, McGaugh JL (1994) Beta-adrenergic acti-
vation and memory for emotional events. Nature 371:702704.
Abraham WC (1996) Activity-dependent regulation of synaptic plas- Cameron HA, Gould E (1994) Adult neurogenesis is regulated by
ticity (metaplasticity) in the hippocampus. In: The hippocam- adrenal steroids in the dentate gyrus. Neuroscience 61:203209.
pus: functions and clinical relevance (Kato N, ed), pp 1526. Chan KH, Morell JR, Jarrard LE, Davidson TL (2001) Reconsideration
Amsterdam: Elsevier Science. of the role of the hippocampus in learned inhibition. Behav
Abraham WC (2004) Stress-related phenomena. Hippocampus Brain Res 119:111130.
14:675676. Conrad CD, Lupien SJ, Thanasoulis LC, McEwen BS (1997) The
Akirav I, Sandi C, Richter-Levin G (2001) Differential activation of effects of type I and type II corticosteroid receptor agonists on
hippocampus and amygdala following spatial learning under exploratory behavior and spatial memory in the Y-maze. Brain
stress. Eur J Neurosci 14:719725. Res 759:7683.
Akirav I, Kozenicky M, Tal D, Sandi C, Venero C, Richter-Levin G Conrad CD, Lupien SJ, McEwen BS (1999) Support for a bimodal
(2004) A facilitative role for corticosterone in the acquisition of role for type II adrenal steroid receptors in spatial memory.
a spatial task under moderate stress. Learn Mem 11:188195. Neurobiol Learn Mem 72:3946.
Auphan N, DiDonato JA, Rosette C, Helmberg A, Karin M (1995) Cordero MI, Venero C, Kruyt ND, Sandi C (2003) Prior exposure to
Immunosuppression by glucocorticoids: inhibition of NF- a single stress session facilitates subsequent contextual fear con-
kappa B activity through induction of I kappa B synthesis. ditioning in rats: evidence for a role of corticosterone. Horm
Science 270:286290. Behav 44:338345.
Baker KB, Kim JJ (2002) Effects of stress and hippocampal NMDA Davidson TL, Jarrard LE (2004) The hippocampus and inhibitory
receptor antagonism on recognition memory in rats. Learn learning: a gray area? Neurosci Biobehav Rev 28:261271.
Mem 9:5865. Davis M, Falls WA, Campeau S, Kim M (1993) Fear-potentiated star-
Bannerman DM, Rawlins JN, McHugh SB, Deacon RM, Yee BK, Bast tle: a neural and pharmacological analysis. Behav Brain Res
T, Zhang WN, Pothuizen HH, Feldon J (2004) Regional dissoci- 58:175198.
ations within the hippocampusmemory and anxiety. Neurosci De Kloet ER (2003) Hormones, brain and stress. Endocr Regul
Biobehav Rev 28:273283. 37:5168.
Becker JT, Olton DS (1980) Object discrimination by rats: the role of De Kloet ER (2004) Hormones and the stressed brain. Ann N Y Acad
frontal and hippocampal systems in retention and reversal. Sci 1018:115.
Physiol Behav 24:3338. De Kloet R, Wallach G, McEwen BS (1975) Differences in corticos-
Bennett MC, Diamond DM, Fleshner M, Rose GM (1991) Serum terone and dexamethasone binding to rat brain and pituitary.
corticosterone level predicts the magnitude of hippocam- Endocrinology 96:598609.
pal primed burst potentiation and depression in urethane- De Kloet ER, Vreugdenhil E, Oitzl MS, Joels M (1998) Brain corti-
anesthetized rats. Psychobiology 19:301307. costeroid receptor balance in health and disease. Endocr Rev
Blanchard RJ, Griebel G, Henrie JA, Blanchard DC (1997) 19:269301.
Differentiation of anxiolytic and panicolytic drugs by effects on De Kloet ER, Oitzl MS, Joels M (1999) Stress and cognition: are cor-
rat and mouse defense test batteries. Neurosci Biobehav Rev ticosteroids good or bad guys? Trends Neurosci 22:422426.
21:783789. De Quervain DJ-F, Roozendaal B, McGaugh JL (1998) Stress and glu-
Bremner JD, Randall P, Scott TM, Bronen RA, Seibyl JP, Southwick cocorticoids impair retrieval of long-term spatial memory.
SM, Delaney RC, McCarthy G, Charney DS, Innis RB (1995) Nature 394:787790.
MRI-based measurement of hippocampal volume in patients De Quervain DJ, Roozendaal B, Nitsch RM, McGaugh JL, Hock C
 Stress and the Hippocampus 765

(2000) Acute cortisone administration impairs retrieval of long- Gesing A, Bilang-Bleuel A, Droste SK, Linthorst AC, Holsboer F, Reul
term declarative memory in humans. Nat Neurosci 3:313314. JM (2001) Psychological stress increases hippocampal miner-
Diamond DM, Park CR (2000) Predator exposure produces retro- alocorticoid receptor levels: involvement of corticotropin-
grade amnesia and blocks synaptic plasticity: progress toward releasing hormone. J Neurosci 21:48224829.
understanding how the hippocampus is affected by stress. Ann Gold PE, Van Buskirk RB (1975) Facilitation of time-dependent
N Y Acad Sci 911:453455. memory processes with posttrial epinephrine injections. Behav
Diamond DM, Bennett MC, Fleshner M, Rose GM (1992) Inverted- Biol 13:145153.
U relationship between the level of peripheral corticosterone Goldstein DS, McEwen B (2002) Allostasis, homeostats, and the
and the magnitude of hippocampal primed burst potentiation. nature of stress. Stress 5:5558.
Hippocampus 2:421430. Gould E, McEwen BS, Tanapat P, Galea LA, Fuchs E (1997)
Diamond DM, Fleshner M, Ingersoll N, Rose GM (1996) Psychologi- Neurogenesis in the dentate gyrus of the adult tree shrew is reg-
cal stress impairs spatial working memory: relevance to electro- ulated by psychosocial stress and NMDA receptor activation. J
physiological studies of hippocampal function. Behav Neurosci Neurosci 17:24922498.
110:661672. Gray JA (1982) The neuropsychology of anxiety: an enquiry into the
Diamond DM, Park CR, Heman KL, Rose GM (1999) Exposing rats functions of the septo-hippocampal system. Oxford, UK: Oxford
to a predator impairs spatial working memory in the radial arm University Press.
watermaze. Hippocampus 9:542552. Gray JA (2000) The neuropsychology of anxiety: an enquiry into the
Diamond DM, Park CR, Woodson JC (2004) Stress generates emo- functions of the septo-hippocampal system, 2nd ed. Oxford, UK:
tional memories and retrograde amnesia by inducing an Oxford University Press.
endogenous form of hippocampal LTP. Hippocampus Gurvits TV, Shenton ME, Hokama H, Ohta H, Lasko NB, Gilbertson
14:281291. MW, Orr SP, Kikinis R, Jolesz FA, McCarley RW, Pitman RK
Diorio D, Viau V, Meaney MJ (1993) The role of the medial prefron- (1996) Magnetic resonance imaging study of hippocampal vol-
tal cortex (cingulate gyrus) in the regulation of hypothalamic- ume in chronic, combat-related posttraumatic stress disorder.
pituitary-adrenal responses to stress. J Neurosci 13:38393847. Biol Psychiatry 40:10911099.
Dolcos F, LaBar KS, Cabeza R (2005) Remembering one year later: Helm KA, Han JS, Gallagher M (2002) Effects of cholinergic lesions
role of the amygdala and the medial temporal lobe memory sys- produced by infusions of 192 IgG-saporin on glucocorticoid
tem in retrieving emotional memories. Proc Natl Acad Sci USA receptor mRNA expression in hippocampus and medial pre-
102:26262631. frontal cortex of the rat. Neuroscience 115:765774.
Elzinga BM, Bremner JD (2002) Are the neural substrates of memory Herman JP, Dolgas CM, Carlson SL (1998) Ventral subiculum regu-
the nal common pathway in posttraumatic stress disorder lates hypothalamo-pituitary-adrenocortical and behavioural
(PTSD)? J Affect Disord 70:117. responses to cognitive stressors. Neuroscience 86:449459.
Evans RM, Arriza JL (1989) A molecular framework for the actions of Herman JP, Figueiredo H, Mueller NK, Ulrich-Lai Y, Ostrander MM,
glucocorticoid hormones in the nervous system. Neuron Choi DC, Cullinan WE (2003) Central mechanisms of stress
2:11051112. integration: hierarchical circuitry controlling hypothalamo-
Feldman S, Conforti N (1980) Participation of the dorsal hippocam- pituitary-adrenocortical responsiveness. Front Neuroendocrinol
pus in the glucocorticoid feedback effect on adrenocortical 24:151180.
activity. Neuroendocrinology 30:5255. Holsboer F, Barden N (1996) Antidepressants and hypothalamic-
Fischette CT, Komisaruk BR, Edinger HM, Feder HH, Siegel A (1980) pituitary-adrenocortical regulation. Endocr Rev 17:187205.
Differential fornix ablations and the circadian rhythmicity of Joels M (1999) Effects of corticosteroid hormones in the hippocam-
adrenal corticosteroid secretion. Brain Res 195:373387. pus. Acta Physiol Scand 167:A3.
Foy MR, Stanton ME, Levine S, Thompson RF (1987) Behavioral Joels M (2001) Corticosteroid actions in the hippocampus. J
stress impairs long-term potentiation in rodent hippocampus. Neuroendocrinol 13:657669.
Behav Neural Biol 48:138149. Joels M, Pu Z, Wiegert O, Oitzl MS, Krugers HJ (2006) Learning
Frey S, Bergado-Rosado J, Seidenbecher T, Pape HC, Frey JU (2001) under stress: how does it work? Trends Cogn Sci 10:152158.
Reinforcement of early long-term potentiation (early-LTP) in Karst H, Berger S, Turiault M, Tronche F, Schutz G, Joels M (2005)
dentate gyrus by stimulation of the basolateral amygdala: het- Mineralocorticoid receptors are indispensable for nongenomic
erosynaptic induction mechanisms of late-LTP. J Neurosci modulation of hippocampal glutamate nongenomic modula-
21:36973703. tion of hippocampal glutamate transmission by corticosterone.
Frey U, Morris RGM (1998b) Synaptic tagging: implications for late Proc Natl Acad Sci USA 102:1920419207.
maintenance of hippocampal long-term potentiation. Trends Kerr DS, Campbell LW, Applegate MD, Brodish A, Landeld PW
Neurosci 21:181188. (1991) Chronic stress-induced acceleration of electrophysio-
Fuxe K, Wikstrom AC, Okret S, Agnati LF, Harfstrand A, Yu ZY, logic and morphometric biomarkers of hippocampal aging. J
Granholm L, Zoli M, Vale W, Gustafsson JA (1985) Mapping Neurosci 11:13161324.
of glucocorticoid receptor immunoreactive neurons in the Keverne EB, Martel FL, Nevison CM (1996) Primate brain evolution:
rat tel- and diencephalon using a monoclonal antibody genetic and functional considerations. Proc Biol Sci 263:689696.
against rat liver glucocorticoid receptor. Endocrinology 117: Keverne EB (2004) Understanding well-being in the evolutionary
18031812. context of brain development. Philos Trans R Soc Lond B Biol Sci
Garcia R, Musleh W, Tocco G, Thompson RF, Baudry M (1997) 359:13491358.
Time-dependent blockade of STP and LTP in hippocampal Kim JJ, Diamond DM (2002) The stressed hippocampus, synaptic
slices following acute stress in mice. Neurosci Lett 233:4144. plasticity and lost memories. Nat Rev Neurosci 3:453462.
766 The Hippocampus Book

Kim JJ, Foy MR, Thompson RF (1996) Behavioral stress modies Majak K, Pitkanen A (2003) Projections from the periamygdaloid
hippocampal plasticity through N-methyl-D-aspartate receptor cortex to the amygdaloid complex, the hippocampal formation,
activation. Proc Natl Acad Sci USA 93:47504753. and the parahippocampal region: a PHA-L study in the rat.
Kim JJ, Lee HJ, Han JS, Packard MG (2001) Amygdala is critical for Hippocampus 13:922942. 929.
stress-induced modulation of hippocampal long-term potenti- McCormick CM, McNamara M, Mukhopadhyay S, Kelsey JE (1997)
ation and learning. J Neurosci 21:52225228. Acute corticosterone replacement reinstates performance on
Kim JJ, Koo JW, Lee HJ, Han JS (2005) Amygdalar inactivation blocks spatial and nonspatial memory tasks 3 months after adrenalec-
stress-induced impairments in hippocampal long-term potenti- tomy despite degeneration in the dentate gyrus. Behav Neurosci
ation and spatial memory. J Neurosci 25:15321539. 111:518531.
Kirschbaum C, Wolf OT, May M, Wippich W, Hellhammer DH McCormick JA, Lyons V, Jacobson MD, Noble J, Diorio J, Nyirenda
(1996) Stress- and treatment-induced elevations of cortisol lev- M, Weaver S, Ester W, Yau JL, Meaney MJ, Seckl JR, Chapman
els associated with impaired declarative memory in healthy KE (2000) 5-Heterogeneity of glucocorticoid receptor messen-
adults. Life Sci 58:14751483. ger RNA is tissue specic: differential regulation of variant tran-
Kitraki E, Kittas C, Stylianopoulou F (1997) Glucocorticoid receptor scripts by early-life events. Mol Endocrinol 14:506517.
gene expression during rat embryogenesis: an in situ hybridiza- McEwen BS (1999) Stress and the aging hippocampus. Front
tion study. Differentiation 62:2131. Neuroendocrinol 20:4970.
Kjelstrup KG, Tuvnes FA, Steffenach HA, Murison R, Moser EI, McEwen BS (2000) The neurobiology of stress: from serendipity to
Moser MB (2002) Reduced fear expression after lesions of the clinical relevance. Brain Res 886:172189.
ventral hippocampus. Proc Natl Acad Sci USA 99:1082510830. McEwen BS (2002) Sex, stress and the hippocampus: allostasis, allo-
Korz V, Frey JU (2003) Stress-related modulation of hippocampal static load and the aging process. Neurobiol Aging 23:921939.
long-term potentiation in rats: involvement of adrenal steroid McEwen BS, Sapolsky RM (1995) Stress and cognitive function. Curr
receptors. J Neurosci 23:72817287. Opin Biol 5:205216.
Landeld PW, Waymire JC, Lynch G (1978) Hippocampal aging McEwen BS, Weiss JM, Schwartz LS (1968) Selective retention of
and adrenocorticoids: quantitative correlations. Science 202: corticosterone by limbic structures in rat brain. Nature
10981102. 220:911912.
Layton B, Krikorian R (2002) Memory mechanisms in posttraumatic McEwen BS, de Kloet R, Wallach G (1976) Interactions in vivo and in
stress disorder. J Neuropsychiatry Clin Neurosci 14:254261. vitro of corticoids and progesterone with cell nuclei and soluble
Liu D, Diorio J, Tannenbaum B, Caldji C, Francis D, Freedman A, macromolecules from rat brain regions and pituitary. Brain Res
Sharma S, Pearson D, Plotsky PM, Meaney MJ (1997) Maternal 105:129136.
care, hippocampal glucocorticoid receptors, and hypothalamic- McEwen BS, Stephenson BS, Krey LC (1980) Radioimmunoassay of
pituitary-adrenal responses to stress. Science 277:16591662. brian tissue and cell nuclear corticosterone. J Neurosci Methods
Liu L, Tsuji M, Takeda H, Takada K, Matsumiya T (1999) Adrenocor- 3:5765.
tical suppression blocks the enhancement of memory storage McGaugh JL (1973) Drug facilitation of learning and memory. Annu
produced by exposure to psychological stress in rats. Brain Res Rev Pharmacol 13:229241.
821:134140. McGaugh JL (2000) Memorya century of consolidation. Science
Luine V, Villegas M, Martinez C, McEwen BS (1994) Repeated stress 287:258251.
causes reversible impairments of spatial memory performance. McGaugh JL, Roozendaal B (2002) Role of adrenal stress hormones
Brain Res 639:167170. in forming lasting memories in the brain. Curr Opin Neurobiol
Lupien SJ, Lepage M (2001) Stress, memory, and the hippocampus: 12:205210.
cant live with it, cant live without it. Behav Brain Res 127: McGaugh JL, Introini-Collison IB, Cahill LF, Castellano C, Dalmaz C,
137158. Parent MB, Williams CL (1993) Neuromodulatory systems and
Lupien SJ, de Leon M, de Santi S, Convit A, Tarshish C, Nair NP, memory storage: role of the amygdala. Behav Brain Res
Thakur M, McEwen BS, Hauger RL, Meaney MJ (1998) Cortisol 58:8190.
levels during human aging predict hippocampal atrophy and McNaughton N, Corr PJ (2004) A two-dimensional neuropsychology
memory decits. Nat Neurosci 1:6973. of defense: fear/anxiety and defensive distance. Neurosci
Maccari S, Piazza PV, Kabbaj M, Barbazanges A, Simon H, Le Moal M Biobehav Rev 28:285305.
(1995) Adoption reverses the long-term impairment in gluco- Meaney MJ, Diorio J, Francis D, Weaver S, Yau J, Chapman K, Seckl
corticoid feedback induced by prenatal stress. J Neurosci JR (2000) Postnatal handling increases the expression of cAMP-
15:110116. inducible transcription factors in the rat hippocampus: the
Magarinos AM, McEwen BS (1995) Stress-induced atrophy of apical effects of thyroid hormones and serotonin. J Neurosci 20:
dendrites of hippocampal CA3c neurons: involvement of gluco- 39263935.
corticoid secretion and excitatory amino acid receptors. Mesches MH, Fleshner M, Heman KL, Rose GM, Diamond DM
Neuroscience 69:8998. (1999) Exposing rats to a predator blocks primed burst potenti-
Magarinos AM, McEwen BS, Flugge G, Fuchs E (1996) Chronic psy- ation in the hippocampus in vitro. J Neurosci 19:RC18.
chosocial stress causes apical dendritic atrophy of hippocampal Mizoguchi K, Kunishita T, Chui DH, Tabira T (1992) Stress induces
CA3 pyramidal neurons in subordinate tree shrews. J Neurosci neuronal death in the hippocampus of castrated rats. Neurosci
16:35343540. Lett 138:157160.
Magarinos AM, Verdugo JM, McEwen BS (1997) Chronic stress alters Muller MB, Uhr M, Holsboer F, Keck ME (2004) Hypothalamic-
synaptic terminal structure in hippocampus. Proc Natl Acad Sci pituitary-adrenocortical system and mood disorders: highlights
USA 94:1400214008. from mutant mice. Neuroendocrinology 79:112.
 Stress and the Hippocampus 767

Nadel L, Jacobs WJ (1996) The role of the hippocampus in PTSD, Sandi C, Loscertales M, Guaza C (1997) Experience-dependent facil-
panic and phobia. In: The hippocampus: functions and clinical itating effect of corticosterone on spatial memory formation in
relevance (Kato N, ed). Amsterdam: Elsevier. the watermaze. Eur J Neurosci 9:637642.
Newcomer JW, Selke G, Melson AK, Hershey T, Craft S, Richards K, Sapolsky RM (2004) Stressed-out memories. Sci Am Mind 14:2833.
Alderson AL (1999) Decreased memory performance in healthy Sapolsky RM, Krey LC, McEwen BS (1984) Glucocorticoid-sensitive
humans induced by stress-level cortisol treatment. Arch Gen hippocampal neurons are involved in terminating the adreno-
Psychiatry 56:527533. cortical stress response. Proc Natl Acad Sci USA 81:61746177.
Oitzl MS, de Kloet ER (1992) Selective corticosteroid antagonists Sapolsky RM, Krey LC, McEwen BS (1985) Prolonged glucocorticoid
modulate specic aspects of spatial orientation learning. Behav exposure reduces hippocampal neuron number: implications
Neurosci 106:6271. for aging. J Neurosci 5:12221227.
Oitzl MS, van Haarst AD, Sutanto W, de Kloet ER (1995) Sapolsky RM, Krey LC, McEwen BS (1986) The neuroendocrinology
Corticosterone, brain mineralocorticoid receptors (MRs) and of stress and aging: the glucocorticoid cascade hypothesis.
the activity of the hypothalamic-pituitary-adrenal (HPA) axis: Endocr Rev 7:284301.
the Lewis rat as an example of increased central MR capacity Sapolsky RM, Armanini MP, Packan DR, Sutton SW, Plotsky PM
and a hyporesponsive HPA axis. Psychoneuroendocrinology (1990) Glucocorticoid feedback inhibition of adreno-
20:655675. corticotropic hormone secretagogue release: relationship to
Patel PD, Lopez JF, Lyons DM, Burke S, Wallace M, Schatzberg AF corticosteroid receptor occupancy in various limbic sites.
(2000) Glucocorticoid and mineralocorticoid receptor mRNA Neuroendocrinology 51:328336.
expression in squirrel monkey brain. J Psychiatr Res 34:383392. Sarrieau A, Dussaillant M, Agid F, Philibert D, Agid Y, Rostene W
Pavlides C, Kimura A, Magarinos AM, McEwen BS (1994) Type I (1986) Autoradiographic localization of glucocorticosteroid
adrenal steroid receptors prolong hippocampal long-term and progesterone binding sites in the human post-mortem
potentiation. Neuroreport 5:26732677. brain. J Steroid Biochem 25:717721.
Pugh CR, Tremblay D, Fleshner M, Rudy JW (1997) A selective role Schafe GE, Nader K, Blair HT, LeDoux JE (2001) Memory consolida-
for corticosterone in contextual-fear conditioning. Behav tion of Pavlovian fear conditioning: a cellular and molecular
Neurosci 111:503511. perspective. Trends Neurosci 24:540546.
Reul JM, de Kloet ER (1985) Two receptor systems for corticosterone Seckl JR, Meaney MJ (2004) Glucocorticoid programming. Ann N Y
in rat brain: microdistribution and differential occupation. Acad Sci 1032:6384.
Endocrinology 117:25052511. Seyle H (1936) A syndrome produced by diverse nocuous agents.
Richter-Levin G, Akirav I (2003) Emotional tagging of memory for- Nature 138:32.
mationin the search for neural mechanisms. Brain Res Brain Shors TJ (2001) Acute stress rapidly and persistently enhances
Res Rev 43:247256. memory formation in the male rat. Neurobiol Learn Mem
Risold PY, Swanson LW (1996) Structural evidence for functional 75:1029.
domains in the rat hippocampus. Science 272:14841486. Shors TJ (2004) Learning during stressful times. Learn Mem
Roozendaal B (2002) Stress and memory: opposing effects of gluco- 11:137144.
corticoids on memory consolidation and memory retrieval. Starkman MN, Gebarski SS, Berent S, Schteingart DE (1992)
Neurobiol Learn Mem 78:578595. Hippocampal formation volume, memory dysfunction, and
Roozendaal B, McGaugh JL (1997) Basolateral amygdala lesions cortisol levels in patients with Cushings syndrome. Biol
block the memory-enhancing effect of glucocorticoid adminis- Psychiatry 32:756765.
tration in the dorsal hippocampus of rats. Eur J Neurosci Sterling P, Eyer J (1988) Allostasis: a new paradigm to explain arousal
9:7683. pathology. In: Handbook of life stress, cognition and health
Roozendaal B, Phillips RG, Power AE, Brooke SM, Sapolsky RM, (Fisher S, Reason J, eds), pp 629649. New York: Wiley.
McGaugh JL (2001) Memory retrieval impairment induced by Stewart PM, Murry BA, Mason JI (1994) Type 2 11 beta-hydroxys-
hippocampal CA3 lesions is blocked by adrenocortical suppres- teroid dehydrogenase in human fetal tissues. J Clin Endocrinol
sion. Nat Neurosci 4:11691171. Metab 78:15291532.
Rudy JW, Kuwagama K, Pugh CR (1999) Isolation reduces contextual Stillman MJ, Shukitt-Hale B, Levy A, Lieberman HR (1998) Spatial
but not auditory-cue fear conditioning: a role for endogenous memory under acute cold and restraint stress. Physiol Behav
opioids. Behav Neurosci 113:316323. 64:605609.
Sala M, Perez J, Soloff P, Ucelli di Nemi S, Caverzasi E, Soares JC, Takahashi LK (1995) Glucocorticoids, the hippocampus, and
Brambilla P (2004) Stress and hippocampal abnormalities in behavioral inhibition in the preweanling rat. J Neurosci 15:
psychiatric disorders. Eur Neuropsychopharmacol 14:393405. 60236034.
Sanchez MM, Young LJ, Plotsky PM, Insel TR (2000) Distribution of Tsai KJ, Chen SK, Ma YL, Hsu WL, Lee EH (2002) sgk, a primary
corticosteroid receptors in the rhesus brain: relative absence of glucocorticoid-induced gene, facilitates memory consolida-
glucocorticoid receptors in the hippocampal formation. J tion of spatial learning in rats. Proc Natl Acad Sci USA 99:
Neurosci 20:46574668. 39903995.
Sandi C, Rose SP (1994) Corticosterone enhances long-term reten- Tuvnes FA, Steffenach HA, Murison R, Moser MB, Moser EI (2003)
tion in one-day-old chicks trained in a weak passive avoidance Selective hippocampal lesions do not increase adrenocortical
learning paradigm. Brain Res 647:106112. activity. J Neurosci 23:43454354.
Sandi C, Rose SP (1997) Training-dependent biphasic effects of cor- Vaher PR, Luine VN, Gould E, McEwen BS (1994) Effects of adrena-
ticosterone in memory formation for a passive avoidance task in lectomy on spatial memory performance and dentate gyrus
chicks. Psychopharmacology (Berl) 133:152160. morphology. Brain Res 656:7178.
768 The Hippocampus Book

Vallee M, MacCari S, Dellu F, Simon H, Le Moal M, Mayo W (1999) Wolf OT, Kuhlmann S, Buss C, Hellhammer DH, Kirschbaum C
Long-term effects of prenatal stress and postnatal handling on (2004) Cortisol and memory retrieval in humans: inuence of
age-related glucocorticoid secretion and cognitive performance: emotional valence. Ann N Y Acad Sci 1032:195197.
a longitudinal study in the rat. Eur J Neurosci 11:29062916. Woodson JC, Macintosh D, Fleshner M, Diamond DM (2003)
Veldhuis HD, Van Koppen C, Van Ittersum M, De Kloet ER (1982) Emotion-induced amnesia in rats: working memory-specic
Specicity of the adrenal steroid receptor system in rat hip- impairment, corticosteronememory correlation, and fear ver-
pocampus. Endocrinology 110:20442051. sus arousal effects on memory. Learn Mem 10:326336.
Villarreal G, Hamilton DA, Petropoulos H, Driscoll I, Rowland LM, Woolley CS, Gould E, McEwen BS (1990) Exposure to excess gluco-
Griego JA, Kodituwakku PW, Hart BL, Escalona R, Brooks WM corticoids alters dendritic morphology of adult hippocampal
(2002) Reduced hippocampal volume and total white matter pyramidal neurons. Brain Res 531:225231.
volume in posttraumatic stress disorder. Biol Psychiatry 52: Xiong W, Yang Y, Cao J, Wei H, Liang C, Yang S, Xu L (2003) The stress
119125. experience dependent long-term depression disassociated with
Weaver IC, Cervoni N, Champagne FA, DAlessio AC, Sharma stress effect on spatial memory task. Neurosci Res 46:415421.
S, Seckl JR, Dymov S, Szyf M, Meaney MJ (2004) Epigenetic Xu L, Anwyl R, Rowan MJ (1997) Behavioural stress facilitates the
programming by maternal behavior. Nat Neurosci 7:847854. induction of long-term depression in the hippocampus. Nature
Weiland NG, Orchinik M, Tanapat P (1997) Chronic corticosterone 387:497500.
treatment induces parallel changes in N-methyl-D-aspartate Yau JL, Olsson T, Morris RG, Meaney MJ, Seckl JR (1995)
receptor subunit messenger RNA levels and antagonist binding Glucocorticoids, hippocampal corticosteroid receptor gene
sites in the hippocampus. Neuroscience 78:653662. expression and antidepressant treatment: relationship with spa-
Welberg LA, Seckl JR (2001) Prenatal stress, glucocorticoids and the tial learning in young and aged rats. Neuroscience 66:571581.
programming of the brain. J Neuroendocrinol 13:113128. Yehuda R (2001) Biology of posttraumatic stress disorder. J Clin
West MJ (1993) New stereological methods for counting neurons. Psychiatry 62(Suppl 17):4146.
Neurobiol Aging 14:275285. Yehuda R, Boisoneau D, Lowy MT, Giller EL Jr (1995) Dose-response
West MJ, Gundersen HJG (1990) Unbiased stereological estimation changes in plasma cortisol and lymphocyte glucocorticoid
of the number of neurons in the human hippocampus. J Comp receptors following dexamethasone administration in combat
Neurol 296:122. veterans with and without posttraumatic stress disorder. Arch
Wilson MM, Greer SE, Greer MA, Roberts L (1980) Hippocampal Gen Psychiatry 52:583593.
inhibition of pituitary-adrenocortical function in female rats. Yerkes RM, Dodson JD (1908) The relation of strength of stimulus to
Brain Res 197:433441. rapidity of habit-formation. J Comp Neurol Psychol 18:459482.
16 Matthew Walker, Dennis Chan, and Maria Thom

Hippocampus and Human Disease

is also often coincidental involvement of the brain stem and

16.1 Overview cerebellum. Limbic encephalitis can occur in response to cer-
tain cancers such as small-cell lung carcinomas, lymphomas,
The hippocampus is involved in many disparate disease pro- thymomas, and testicular tumors, as an immune-mediated
cesses, but only in rare instances is the hippocampus the sole syndrome (one of a number of so-called paraneoplastic syn-
site of pathological damage. It is subject to the same patholo- dromes). A similar syndrome has, however, been described in
gies that can affect other cortical areas, such as tumors, vascu- association with Wernickes encephalopathy, systemic lupus
lar malformations, and cortical dysgenesis; but in addition the erythematosus, and herpes simplex encephalitis. In these
hippocampus is also notable for its particular vulnerability to cases, there is a strong association with anti-neuronal anti-
damage as a consequence of ischemia/hypoxia, trauma, and bodies directed against intracellular antigens, but the patho-
hypoglycemia. There are also instances in which involvement logical role of these antibodies remains uncertain. Treatment
of the hippocampal formation is critical to the manifestation of the underlying malignancy can alleviate the symptoms.
of the disease; foremost among them are Alzheimers disease Hippocampal sclerosis has been observed in a proportion
and temporal lobe epilepsy, representing approximately 60% of elderly patients presenting with cognitive impairment. In
of all partial epilepsies. Alzheimers disease and epilepsy are one study (Dickson et al., 1994), hippocampal sclerosis was
among the most prevalent of all neurological diseases, with 20 observed in 26% of demented patients over the age of 80 years
million and 50 million people affected, respectively. Damage and in 16% of all patients aged over 80. In all cases there was
to the hippocampus is also the central component of a variety neuronal loss and gliosis affecting CA1, the subiculum, and
of rare conditions, such as limbic encephalitis and dementia dentate granule cells, with additional neuronal loss in the
with isolated hippocampal sclerosis. In addition, involvement entohinal cortex in a proportion of cases. However, concomi-
of the hippocampus is being increasingly recognized in schiz- tant pathology, such as ischemic vascular damage or Alzheimer
ophrenia, another common neuropsychiatric disorder. pathology, was noted in most of the cases in this study; pure
Acute encephalitis due to herpes simplex virus shows a hippocampal sclerosis is much rarer, affecting only 0.4% of
predilection for limbic structures, and infection can result in patients with dementia (Ala et al., 2000). These rare instances
selective damage to the hippocampus, amygdala, and associ- of pure hippocampal sclerosis are not associated with any
ated structures, resulting in acute limbic encephalitis. increase in risk factors for cerebrovascular disease, and in none
Subacute limbic encephalitis has also been described in which of the cases was there a history of a hypoxic episode preceding
the pathology more specically affects the limbic system the onset of cognitive impairment. The relation between hip-
(Corsellis et al., 1968). The clinical presentation is character- pocampal sclerosis, as a rare cause of dementia in the elderly,
ized by behavioral and psychiatric problems (usually aggres- and mesial temporal sclerosis, as a substrate for epilepsy affect-
sion and depression), disorientation, short-term memory ing a younger age group, remains undetermined, but it is pos-
deficits, hallucinations, seizures, and sleep disturbances sible that the two diseases arise as a consequence of differing
(Corsellis et al., 1968; Gultekin et al., 2000). The pathological etiologies, with pure hippocampal sclerosis occurring as a con-
nding is aggregation of lymphocytes around blood vessels sequence of a primary degenerative process rather than sec-
(perivascular lymphocytic cuffing), neuronal cell loss, and ondary to a systemic insult such as hypoxia or fever.
gliosis particularly affecting the hippocampus, dentate gyrus, Schizophrenia is thought to involve primarily the prefrontal
amygdala, cingulate, and parahippocampal structures. There cortex (Grossberg, 2000), but there is accumulating evidence

769
770 The Hippocampus Book

for involvement of mesial temporal lobe structures in its The characterization of hippocampal involvement in
pathophysiology. Indeed, there is evidence accumulating that human disease has been of great value to neuroscientists and
schizophrenia may be a result of fronto-hippocampal integra- clinicians. For instance, the unilateral nature of hippocampal
tion. Some patients with temporal lobe epilepsy exhibit a sclerosis in temporal lobe epilepsy has provided the opportu-
psychosis indistinguishable from schizophrenia. Structural nity to analyze the individual functions of the left and right
neuroimaging studies have shown a subtle but denite reduc- hippocampi. Also, the identication of hippocampal sclerosis
tion in hippocampal volume (Nelson et al., 1998), which in by high-resolution magnetic resonance imaging (MRI) in
some cases is present early in the disease; and functional imag- epilepsy patients has stimulated the use of curative epilepsy
ing studies using magnetic resonance spectroscopy have neurosurgery. The hippocampal atrophy that occurs as a
documented a reduction in levels of the metabolite N-acetyl- result of the pathological changes of Alzheimers disease can
D-aspartate, a marker of neuronal viability, in patients with be detected and quantied using in vivo neuroimaging tech-
schizophrenia (Maier et al., 1995). Neuropathological studies niques. As a biomarker of disease, the presence of hippocam-
indicate that the loss of hippocampal volume correlates with a pal atrophy provides important corroborative information at
reduction in the size of hippocampal neurons rather than the time of the clinical diagnosis, and the demonstration of
neuronal loss (Arnold et al., 1995). A reduction in neuronal progressive hippocampal volume loss is valuable for tracking
density in certain hippocampal regions, with the CA2 disease progression. Finally, case studies documenting the
interneurons particularly affected, is also observed in schizo- nature of cognitive impairments in those rare patients with
phrenia, as well as in manic depression (Benes et al., 1998). selective hippocampal pathology have provided important
Loss of synaptic proteins in the hippocampus (Eastwood and insights into the functions of the human hippocampus (see
Harrison, 1995) and abnormal MAP2 expression in subicular Chapter 13).
neuron dendrites also indicate abnormalities of connectivity in This chapter focuses on two disorders in which the role
patients with schizophrenia (Cotter et al., 2000). The lack of of the hippocampus has been extensively investigated:
demonstrable gliosis in those with schizophrenia has been Alzheimers disease and temporal lobe epilepsy. Although in
argued to support a neurodevelopmental rather than degener- Alzheimers disease the disease process results eventually
ative disease process; the demonstration in schizophrenia of in widespread destruction of the cerebral cortex, the damage
cytoarchitectural abnormalities of pre-alpha cells in the in the earliest stages of disease is restricted to the entorhinal
entorhinal cortex and of abnormal orientation of hippocam- cortex and the hippocampus, and the memory impair-
pal pyramidal neurons (Conrad et al., 1991) also support a ment that results from this disruption of the hippocampal
maldevelopmental disorder. formation represents one of the common characteristics of
The human hippocampus is the beneciary of a generous Alzheimers disease. In temporal lobe epilepsy, the pathologi-
arterial supply originating from a number of major arteries, cal damage is often restricted to the hippocampus in the form
including the anterior choroidal artery and branches of the of hippocampal sclerosis. However, unlike Alzheimers disease,
posterior cerebral artery (Erdem et al., 1993). Despite this, the in which the hippocampal damage is secondary to the under-
vascular supply to the hippocampus may be interrupted as a lying pathological process, the hippocampus in temporal lobe
result of embolic disease, as well as by prolonged anoxic epilepsy is not only sensitive to damage by seizure activity but
insults. Ischemic damage to the hippocampus can occur in can also act as the substrate for epileptic seizure generation.
isolation or as part of a more widespread cerebrovascular dis-
ease process. In addition, animal experiments have shown that
hippocampal damage, particularly affecting the CA1 subeld,
can occur as a result of chronic nonembolic vascular insuffi- 16.2 Mesial Temporal Lobe
ciency (de la Torre et al., 1992). Despite the known vulnera- Epilepsy and Hippocampal Sclerosis
bility of CA1 to ischemic damage (an observation dating back
to the observations made by Sommer in 1898) and evidence 16.2.1 Introduction
from animal studies that the hippocampus is particularly sen-
sitive to ischemic insult (Schmidt-Kastner and Freund, 1991), Epilepsy is the propensity to have seizures and is one of the
there exist in the clinical literature very few cases in which most common serious neurological conditions, affecting 0.4%
selective damage to the hippocampus has been observed. to 1.0% of the worlds population (Sander and Shorvon,
Hypoglycemia results in a pattern of damage different from 1996). There are approximately 20 to 70/100,000 new cases
that of ischemia, as it causes necrosis of predominantly the per year, and the lifetime chance of seizures is 3% to 5%
dentate granule cells (Auer and Siesjo, 1988); the clinical sig- (Sander and Shorvon, 1996). Seizure types can be divided into
nicance of such damage is not clear. Lastly, mesial temporal partial seizures, arising from one part of the brain, and gener-
structures are particularly vulnerable to traumatic brain alized seizures, arising simultaneously throughout the cortex;
injury; this is partly because of their location in the middle respectively, these constitute approximately 40% and 50% of
cranial fossa, leaving them susceptible to contusion and vas- seizures in newly diagnosed epilepsy (10% of seizures are
cular injury, but it may also be due to direct excitotoxic effects unclassiable) (Sander and Shorvon, 1996). Epilepsy itself can
(Tate and Bigler, 2000). be divided into a number of syndromes determined by seizure
 Hippocampus in Human Disease 771

type, electroencephalographic (EEG) abnormalities and con- increased co-morbidity including psychiatric problems
comitant neurological decits. Although all epilepsies are the (depression, psychosis), increased mortality and neuropsy-
result of an underlying brain abnormality (e.g., tumor or scar chological decits that relate to the side of the hippocampal
tissue in partial epilepsies, and a metabolic or genetic basis in sclerosis: verbal memory decits with dominant (usually left)
generalized epilepsies), a convincing cause is identied in only temporal lobe involvement and nonverbal memory decits
approximately 30% of patients with epilepsy (Sander and with nondominant lobe involvement.
Shorvon, 1996). The clinical manifestation of a seizure
depends not only on where the seizure starts but also on the Seizure Semiology
speed and pattern of seizure spread. Differing epilepsy syn-
dromes have different pathophysiologies and mechanisms; in Mesial temporal lobe seizures usually take the form of com-
this chapter we are concerned solely with temporal lobe plex partial seizures, in which consciousness is disturbed, and
epilepsy. less commonly simple partial seizures, in which consciousness
Temporal lobe epilepsy represents approximately 60% of is preserved (Walker and Shorvon, 1997). The seizure usually
all partial epilepsies. The commonest neuropathological has a gradual evolution over 1 to 2 minutes (substantially
lesion identied in temporal lobectomy series in patients with longer than extratemporal seizures) and lasts longer (210
mesial temporal lobe epilepsy (TLE) is hippocampal sclerosis, minutes) than complex partial seizures originating in
or Ammons horn sclerosis, which is seen in approximately extratemporal sites. The commonest warning (often termed
half of the cases (Bruton, 1988). Other major pathologies can aura, literally breeze) is that of a rising sensation from the
be grouped under lesion-associated TLE and include vascu- stomach. Other gastrointestinal auras can occur, especially
lar malformations, malformations of cortical development, nausea, stomach rumbling, and belching. Auras can also con-
and glioneuronal tumors (Wolf et al., 1993). Of those patients sist of olfactory-gustatory hallucinations, autonomic symp-
with drug-resistant epilepsy, hippocampal sclerosis is the toms, affective symptoms, disturbances of memory, or visual
commonest aetiology. hallucinations and illusions (especially with seizures involving
In 1825, Bouchet and Cazauvielh presented their ndings the temporal neocortex). Autonomic symptoms include
on 18 autopsied patients in a thesis that attempted to estab- changes in heart rate and blood pressure, pallor or ushing of
lish the relation between epilepsy, lpilepsie, and insanity, the face, pupillary dilatation, and piloerection. Affective
lalination mentale (Bouchet and Cazauvielh, 1825). They symptoms typically take the form of fear (the most common
noted that in ve cases where there were changes in the cornu and often extremely intense), depression, anger, and irritabil-
ammonis four were characterized by induration and one had ity. Euphoria and erotic thoughts have also been described.
softening. Sommer (1880) further described in detail the neu- Dreamy states and feelings of depersonalization commonly
ropathological nding of hippocampal sclerosis in the brains occur. Dj vu, dj entendu, and other abnormalities of
of patients with chronic epilepsy. He noted gliosis and pyram- memory such as recollections of childhood or even former
idal cell loss in predominantly the CA1 region of the hip- lives can also be present with this form of epilepsy.
pocampus, and he proposed that these lesions were the cause After the aura and in the early stages, motor arrest and
of the epilepsy. That same year, Peger described hemorrhagic absence are prominent. Typically, this is followed by marked
lesions in the mesial temporal lobe of a patient dying in status automatisms. The automatisms of mesiobasal TLE can be
epilepticus and concluded that neuronal necrosis was the prolonged and are characteristically oroalimentary (e.g., lips-
result of impaired blood ow or metabolic disturbances that macking, chewing) and/or gestural (e.g., dgeting, undress-
occurred during the seizure (Peger, 1880). Since that time ing, walking). Typically, the automatisms are more marked
the debate as to whether hippocampal sclerosis is the cause or ipsilaterally and may be associated with contralateral postur-
result of epilepsy has continued. ing. There may be some apparent responsiveness, and con-
Three lines of evidence indicate that seizures originate in scious behavior can occur during nondominant temporal
the sclerosed hippocampus. First, hippocampal sclerosis is lobe seizures. During the seizure, speech with recognizable
closely associated with a particular seizure semiology, the words lateralizes the focus to the nondominant temporal lobe.
psychomotor seizurea seizure type rst recognized by John Secondary generalization is less common than in extratempo-
Hughlings Jackson. Second, EEG evidence points to seizure ral lobe epilepsy. Postictal confusion is typical, and postictal
onset in the sclerosed hippocampus. Lastly, surgical resection dysphasia can occur following dominant temporal lobe
of the sclerosed hippocampus results in seizure remission. seizures. Postictal headache and postictal psychosis have also
been described. There is profound amnesia for the absence
16.2.2 Clinical Features and automatism (Walker and Shorvon, 1997).

The typical history of a patient with hippocampal sclerosis is EEG Characteristics

contained in Figure 161, see color insert. There is often an
antecedent history of an insult (usually febrile seizures) fol- Scalp EEG recordings usually demonstrate interictal epilepti-
lowed by a gap before seizures begin many years later. These form abnormalities (spikes, sharp waves) over the mid/ante-
seizures often prove resistant to treatment. There is an rior temporal region, but it is common for these epileptiform
 A A 32-year old man presented with refractory partial epilepsy. He had a prolonged febrile seizure
at the age of 18 months, and spontaneous seizures began at the age of 8 years. These seizures
have continued despite trying all available antiepileptic drugs. The seizures take the form of sim-
ple partial seizures in which he perceives a sense of fear rising from his stomach, and complex
partial seizures in which he has an aura as above and then loses touch with his surroundings.
During the complex partial seizure he is described as making chewing movements, his right
hand is postured and he ddles with his clothes with his left hand. The seizure lasts a couple
of minutes, and he is then confused and dysphasic afterward for approximately 5 minutes.

Figure 161. Clinical features of mesial temporal lobe epilepsy. A. Typical history. B. Magnetic
resonance imaging (MRI) ndings of left hippocampal sclerosis with high T2 signal in
shrunken hippocampus (arrow). C. Intracranial recordings from left (LH) and right (RH) hip-
pocampus and left (LA) and right (RA) amydala demonstrate well localized 1- to 2-Hz spikes
in left hippocampus and amygdala that evolve to low-amplitude fast activity at seizure onset.
772
 Hippocampus in Human Disease 773

Figure 161. (Continued) D. Histology of left hippocampal sclerosis. i. Loss of cells in hilus (H)
and CA1 (arrows) with preservation of CA2 (star) and subiculum (S); ii. Dynorphin staining
demonstrating many ber sprouting in granule cell layer (GCL) and molecular layer (ML); iii.
Mossy ber sprouting illustrated with Timms stain; iv. Hippocampus with minimal mossy
ber sprouting for comparison with iii.

abnormalities to occur bilaterally or independently over both localized or commonly spreads to involve a wider eld includ-
temporal regions (reasons why this may be so are discussed ing the contralateral temporal lobe.
below) (Williamson et al., 1993). Ictal scalp recordings usually Depth electrode studies have further conrmed the electro-
demonstrate a build-up of 5- to 10-Hz sharp activity localized graphic origin of these seizures in the hippocampal formation
to the mid/anterior temporal region. This activity may remain (King and Spencer, 1995). Although interictal spikes may
774 The Hippocampus Book

occur independently from either the hippocampus or pocampal sclerosis (French et al., 1993). The injury hypoth-
extrahippocampal sites (see below), the ictal discharges are esis implies that this insult irreversibly damages or alters the
usually relatively well localized. Preictal abnormalities can hippocampus, resulting in a template for the progression to
occur with well localized 1- to 2-Hz spikes that recur over sec- hippocampal sclerosis following a latent interval. There
onds or minutes (Fig. 161C, see color insert). With clinical appears to be age-specic sensitivity for this injury, with more
seizure onset, there is a 10- to 15-Hz low-amplitude discharge severe neuronal loss demonstrated with earlier onset of
that is initially conned to hippocampal electrodes (Fig.16 epilepsy (Davies et al., 1996). The most direct evidence of the
1C, see color insert) but grows in amplitude and then spreads association is the observation with serial neuroimaging that
to other regions (King and Spencer, 1995). A second pattern hippocampal sclerosis occurs following prolonged febrile con-
has also been described in which the seizure begins as a low- vulsions (Van Landingham et al., 1998). Febrile seizures have
amplitude, high-frequency discharge without the preictal spik- been modeled in animals by inducing hyperthermic seizures
ing. These two types of onset can occur in the same patient. in rats by blasts of hot air or water. The similarities between
The exact location for seizure onset can vary not only from the animal model and the human condition are that seizures
patient to patient but also within the same patient. This sug- occur in response to high body temperature and that increas-
gests that seizure onset and generation is not from a single area ing age confers resistance to these seizures (Baram et al., 1997;
in mesial temporal structures but from a distributed network. Walker and Kullmann, 1999). Although fever in humans is
associated with other physiological changes, reducing the
Hippocampal Resection body temperature is an effective way to reduce the likelihood
of seizures; thus, hyperpyrexia is probably the main trigger.
Surgery has provided the most compelling evidence of hip- There are, however, major differences between hyperthermic
pocampal sclerosis as the substrate for the epilepsy. Surgical seizures in rats and febrile convulsions in humans. Inducing
outcome for intractable TLE is most successful when mesial hyperthermia in young Sprague-Dawley rats apparently
temporal structures are included in the resection. In patients results in seizures in most of these animals (Baram et al.,
with drug-resistant epilepsy in whom there is concordance 1997), but convulsions are relatively rare in children with
between neuroimaging, electroclinical characteristics of the fever. In experimental models, prolonged hyperthermic
seizure, and neuropsychological tests, there is a better than seizures in immature rats did not cause spontaneous seizures
80% chance of curing the epilepsy with temporal lobe resec- during adulthood but did increase seizure susceptibility fol-
tion (Arruda et al., 1996). Furthermore, over 80% of patients lowing administration of a convulsant (a second hit) (Dube
without tumors rendered seizure-free by temporal lobectomy et al., 2000). However only 2% to 7% of children with a his-
have hippocampal sclerosis as their main pathology (French et tory of febrile convulsions go on to develop epilepsy (i.e.,
al., 1993). In these patients depth electrode recordings also unprovoked seizures) later in life (Annegers et al., 1987).
localized the seizure onset to the sclerosed hippocampus. Because many children with febrile convulsions may have a
There is thus a strong correlation between resection of a scle- predisposing susceptibility to seizures, the low incidence of
rosed hippocampus and cure of the epilepsy. Temporal lobe subsequent epilepsy could be explained by a protective effect
surgery, however, also involves removal of or damage to struc- of febrile seizures. Alternatively, febrile seizures alone are not
tures outside the hippocampus, including the amygdala and sufficient to result in development of epilepsy. (Walker and
parahippocampal structures and temporal neocortex. Further- Kullmann, 1999).
more, many patients, despite successful surgery, remain Other insults can result in hippocampal sclerosis including
dependent on antiepileptic drugs. These observations argue neonatal hypoxia and head injuries. In rat models, uid per-
that structures beyond the hippocampus are involved in the cussion injury to the dura results in hilar interneuron loss in
epileptic network. the hippocampus (Lowenstein et al., 1992). The mechanism
by which this occurs is unknown. The neuronal loss is accom-
16.2.3 Etiology panied by enhanced excitability of the hippocampus but again
no spontaneous seizures (Lowenstein et al., 1992). These
Pathogenesis of Hippocampal experimental and human studies do not, however, address two
Sclerosis and Developmental Aspects fundamental questions: (1) Why is hippocampal sclerosis pre-
dominantly a unilateral disease process in humans (see below)
There are predictable patterns of cell loss and alterations to following a global cerebral insult? (2) What is the nature of
the intrinsic circuitry of hippocampal sclerosis. However, the the second hit that results in the expression of epilepsy?
factors critical to the initiation of the cell loss and hippocam- The second hit does not necessarily have to be environ-
pal reorganization are still debated, and the precise etiology of mental but could be the coexistence of various genetic factors
hippocampal sclerosis remains elusive. or concomitant developmental abnormalities. Temporal lobe
A signicant cerebral insult (or initial precipitating injury) epilepsy is generally regarded as an acquired disorder with
occurring early in life, such as a febrile or prolonged seizure, only a small genetic contribution. There are familial cases
is often reported (3050% of casesbut up to 80% in one of febrile seizures, which are associated with ion channel
surgical series) in retrospective studies of patients with hip- mutations: sodium channel subunit and -aminobutyric acid
 Hippocampus in Human Disease 775

ionotropic receptor family A (GABAA) subunit mutations) pocampal sclerosis specimens (Blumcke et al., 1999b). Cajal
(Table 161) (Kullmann, 2002). However, these families usu- Retzius cells, through secretion of the reelin protein, play a
ally present with a hetererogeneous group of epilepsies that critical role in neuronal organization in the developing brain.
are distinct from the typical history of hippocampal epilepsy. Higher numbers of Cajal-Retzius cells were particularly
More recently, the leucine-rich, glioma-inactivated 1 gene has prominent in patients with hippocampal sclerosis and a his-
been associated with familial neocortical temporal lobe tory of febrile seizures. It is plausible that such an injury
epilepsies, although the mechanisms by which this mutation occurring early in life disrupts normal hippocampal develop-
results in epilepsy are unknown (Kullmann, 2002). We are at ment and maturation (one manifestation of which is an excess
present ignorant of the genetic mutations underlying most of of Cajal-Retzius cells), which in turn predisposes to hip-
the genetically determined epilepsies, let alone those that con- pocampal sclerosis. As it has been suggested that reelin in the
tribute to other epilepsies. Genetic predisposition to some adult cortex has a role in plasticity and axonal remodeling, an
forms of temporal lobe epilepsy and febrile convulsions have increased number of these cells may also be important for the
been described, and there are familial cases of febrile convul- reorganization of circuitry occurring in hippocampal sclerosis
sions and TLE but without hippocampal sclerosis (Baulac et (described below).
al., 2001). The nal argument supporting a maldevelopmental basis
More recent attention has focused on an underlying for hippocampal sclerosis comes from the observation that
maldevelopment of the hippocampus as a primary abnormal- hippocampal sclerosis is often observed in association with
ity predisposing to hippocampal sclerosis and to febrile subtle cytoarchitactural malformations in the neocortex, also
seizures. In an MRI study of families with familial febrile con- termed microdysgenesis (Hardiman et al., 1988). This may be
vulsions, a subtle preexisting hippocampal abnormality was indicative of a more widespread maldevelopmental process
detected (Fernandez et al., 1998), and hippocampal sclerosis involving both mesial and lateral temporal lobe structures.
has also been reported in patients in association with isolated One cytoarchitectural feature observed in microdysgenesis is
malformations of the hippocampus (Baulac et al., 1998). In also an excess of Cajal Retzius cells in the molecular layer
addition, an abnormal persistence of calretinin positive Cajal- (Garbelli et al., 2001), which interestingly seems to parallel
Retzius cells in the hippocampus has been reported in hip- ndings in hippocampal sclerosis.

Table 161.
Monogenic Epilepsies and Ion Channels Implicated in Human Epilepsy

Syndrome Gene Locus Protein

Autosomal Dominant Inheritance

Autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) CHRNA4 20q13.2-q13.3 Nicotinic ACh receptor subunit
CHRNB2 1p21 Nicotinic ACh receptor subunit
Benign familial neonatal convulsions (BFNC)* KCNQ2 20q13.3 Potassium channel subunit
KCNQ3 8q24 Potassium channel subunit
Generalized epilepsy with febrile seizures plus (GEFS) SCN1A 2q24 Sodium channel subunit
SCN1B 19q13.1 Sodium channel subunit
GABRG2 5q33.1-q33.1 GABAA receptor subunit
Juvenile myoclonic epilepsy (JME) GABRA1 5q34 GABAA receptor subunit
Autosomal dominant partial epilepsy with auditory features LGI1 10q24 Leucine-rich, glioma-inactivated
[ADPEAF] 1 [unknown functionnot an
lon channel]
Sporadic
Severe myoclonic epilepsy of Infancy (SMEI) SCN1A 2q24 Sodium channel subunit

Other Ion Channel Genes Implicated in Epilepsy

Febrile and afebrile seizures SCN2A 2q23-q24.3 Sodium channel subunit
Episodic ataxia type 1 with partial epilepsy KCNA1 12p13 Potassium channel subunit
Episodic ataxia type 2 with spike-wave epilepsy CACNA1A 19p13 Calcium channel subunit
JME CACNB4 2q22-23 Calcium channel subunit
Source: After Kullmann (2002), with permission.
ACh, acetylcholine; GABAA, -aminobutyric acid ionotropic receptor family A; GABAB, GABA ionotropic receptor family B.
*A 15% risk of epilepsy continuing beyond infancy.
Also associated with febrile seizures and absences.
Also known as autosomal dominant lateral temporal epilepsy.
Mutations identied in small number of patients, so role in epilepsy is less clear.
Also associated with generalized epilepsy with praxis-induced seizures.
776 The Hippocampus Book

Hippocampal sclerosis is also well recognized to occur in glutamatergic synaptic transmission may contribute to kin-
association with more severe cortical malformations, vascular dling by increasing the excitatory synaptic drive and the like-
malformations, and low-grade glioneuronal tumors (Cendes lihood of evoking after-discharges but is alone insufficient to
et al., 1995; Li et al., 1999). It possible that in these cases the explain the cellular mechanisms of kindling (Cain, 1989; Cain
epileptogenic extrahippocampal lesion kindles the hip- et al., 1992).
pocampal neuronal loss (i.e., the hippocampal sclerosis in Kindling alone is unlikely to explain the occurrence of hip-
these cases is a secondary event) (see below). It has been pocampal sclerosis in association with other pathology
shown, however, that in patients with dual pathologies because kindling itself usually results in no or minimal hip-
removal of both the lesion and the abnormal hippocampus pocampal damage and sclerosis (Tuunnen and Pitknen,
has the best outcome in terms of seizure control (Li et al., 2000). Kindling could, however, explain the progression of
1999), emphasizing the role of the hippocampus in temporal mesial temporal epilepsy. Eventually spontaneous seizures in
lobe seizures even when there is a second pathology. the kindling model result in progressive neuronal loss in the
hippocampus (Cavazos et al., 1994). Indeed, even following
Animal Models of Mesial single seizures there is evidence of both apoptotic cell death
Temporal Lobe Epilepsy and neurogenesis in the dentate granule cell layer (Bengzon et
al., 1997). This suggests that recurrent seizures may cause fur-
The interpretation of many of the pathological ndings and ther structural and functional changes in the hippocampus.
the electrophysiologic studies in human postsurgical speci- Human evidence for this has mainly been indirect. Epilepsy
mens is confounded by: (1) the inuence of treatment; (2) the duration correlates with hippocampal volume loss and pro-
difficulty differentiating cause from effect (i.e., it is possible gressive neuronal loss and dysfunction (Theodore et al.,
that the changes are the result, not the cause, of the seizures); 1999). There has also been a case reported of hippocampal
and (3) the lack of adequate control tissue for comparison. To volumes decreasing with time in hippocampal sclerosis (Van
overcome these handicaps, animal models of mesial TLE are Paesschen et al., 1998) and the appearance of hippocampal
used. The two most studied are the kindling model and the sclerosis de novo following secondary generalized brief tonic-
poststatus epilepticus model. Intrahippocampal injection of clonic seizures (Briellmann et al., 2001).
tetanus toxin also results in spontaneous seizures even after
clearance of the toxin, and this model has also contributed to Poststatus Epilepticus. Seizures are usually self-terminating
our understanding of the pathophysiology of mesial TLE and brief. Occasionally seizures persist unabated, or repeated
(Mellanby et al., 1977). This model does not result in hip- seizures can occur without recovery; this situation is termed
pocampal sclerosis (Jefferys et al., 1992), and the seizures usu- status epilepticus. Although status epilepticus may occur in
ally abate, in contrast to the human condition. We discuss the individuals with preexisting epilepsy, more than half of
kindling and the poststatus epilepticus models in more detail, patients who present with status epilepticus have no history of
as these models possibly have human correlates. seizures (DeLorenzo et al., 1996). In these patients, the status
epilepticus is often acutely precipitated by a central nervous
Kindling. Kindling is the repetition of tetanic (trains of) stim- system (CNS) infection, cerebral vascular accident, hypoxia,
uli that initially evoke after-discharges but not seizures or alcohol. The probability of then developing epilepsy
(Goddard, 1967; McNamara et al., 1993). Repetition of the (unprecipitated seizures) is 41% within 2 years compared
same trains of stimuli results in gradual lengthening of the with 13% for those with acute symptomatic seizures but no
after-discharges, eventually leading to progressively more status epilepticus (Hesdorffer et al., 1998). This suggests a
severe seizures. Once an animal has been kindled, the height- relation between the prolonged seizures of status epilepticus
ened response to the stimulus seems to be permanent, and and subsequent epileptogenesis, although a relation between
spontaneous seizures can occur (McNamara et al., 1993). The the length of the seizure and the nature and severity of the
hippocampus and amygdala are easily kindled, resulting in a precipitant cannot be discounted. In humans, status epilepti-
well described progression of limbic seizures. Kindling shares cus has been shown to result in hippocampal damage and
several characteristics with NMDA-dependent long-term subsequent hippocampal sclerosis. The hippocampus thus has
potentiation (LTP) of excitatory synaptic transmission. This a dichotomous role: as the substrate for epilepsy and as the
has led to the suggestion that kindling and LTP have similar structure susceptible to damage by prolonged seizures. Animal
underlying mechanisms. In support of this, the rate at which models of generalized convulsive as well as limbic status
kindling occurs is retarded in rodents treated with NMDA epilepticus have supported these ndings. Limbic status
receptor antagonists. There are, however, several differences epilepticus has been induced by the systemic or local admin-
between kindling and LTP. Although NMDA receptor antago- istration of kainic acid, systemic administration of pilo-
nists can completely block the induction of LTP, they are carpine (a muscarinic receptor agonist), or protocols using
unable to block kindling completely (Cain et al., 1992). electrical stimulation of limbic areas (Walker et al., 2002).
Perhaps a more fundamental difference is that the kindling Status epilepticus in these models in adult animals results in
process requires after-discharges; the repeated induction of hippocampal damage similar to that observed in humans.
LTP without after-discharges does not induce kindling. LTP of Following these acute episodes of limbic status epilepticus,
 Hippocampus in Human Disease 777

many of the animals go on to develop spontaneous limbic persistent sodium currents (Su et al., 2001). The effect of a
seizures after a latent period lasting days to weeks (Walker et burst of action potentials is to increase synaptic reliability;
al., 2002). within the excitatory network of the CA3 pyramidal cells,
burst ring in a single CA3 pyramidal cell can generate a syn-
16.2.4 Pathophysiology chronized burst throughout the whole network (Miles and
Wong, 1983). Because of the propensity for the CA3 pyrami-
One of the major points of confusion in understanding the dal cells to generate this synchronized burst, this region has
pathophysiology of epilepsy is the differentiation of a seizure often been considered the pacemaker for seizure activity.
(ictus) from interictal discharges and, indeed, from epilepsy Synchronized bursts can, however, also occur in the CA1 sub-
itself. Although obviously linked, they are separate entities. An eld (Karnup and Stelzer, 2001). In some situations the syn-
epileptic seizure is a transient paroxysm of excessive dis- chronization of CA1 pyramidal cells is secondary to a CA3
charges of neurons in the cerebral cortex causing a clinically generated burst, but synchronization can also occur through a
discernible event. Brief synchronous activity of a group of combination of nonsynaptic mechanisms including gap junc-
neurons leads to the interictal spike, and as we discuss, this tions, ephaptic transmission, and changes in the extracellular
shares some mechanisms with seizure generation; spikes milieu. The importance of these nonsynaptic mechanisms in
should, however, be recognized as a distinct phenomenon (de neuronal synchronization has been emphasized by the zero
Curtis and Avanzini, 2001). Epilepsy, on the other hand, is the calcium model of ictal discharges, in which reducing extracel-
propensity to have seizures; and epileptogenesis is the devel- lular calcium in a hippocampal slice preparation below that
opment of a neuronal network in which spontaneous seizures necessary for synaptic transmission results in synchronized
occur. epileptiform discharges due to increased axonal excitability
and ephaptic transmission (Jefferys, 1995). Furthermore,
Interictal Spike decreasing extracellular space (indirectly increasing ephaptic
transmission) can promote bursting (Roper et al., 1992),
Epileptiform interictal EEG abnormalities include spikes, whereas intracellular acidication with sodium propionate
which are fast electrographic transients lasting less than 80 indirectly decreasing electrotonic coupling (Perez and Carlen,
ms, and sharp waves, which last 80 to 120 ms (de Curtis and 2000)inhibits epileptiform bursts (Xiong et al., 2000).
Avanzini, 2001). That these abnormalities are pathological is Synchronization of principal cells can occur secondary to
supported by their rare occurrence ( 1%) in healthy indi- oscillations in the inhibitory interneuron network; indeed,
viduals (Gregory et al., 1993) and their strong association with single basket cells have been shown to synchronize the dis-
epilepsy (Ajmone-Marsan and Zivin, 1970). Spikes and sharp charges of pyramidal cells through synchronized somatic
waves are often followed by a slow wave lasting hundreds of inhibition (Cobb et al., 1995). Although the precise mecha-
milliseconds. As discussed below, this slow wave probably rep- nisms of neuronal synchronization in the hippocampus are
resents a period of relative refractoriness. It has been estab- still unclear, the observation of high-frequency oscillations
lished from concomitant eld potential and intracellular superimposed on spike discharges has led to the hypothesis
recordings that the intracellular correlate of the interictal that the same physiological mechanisms that subtend fast
spike is the paroxysmal depolarizing shift (PDS) (Matsumoto oscillations in the hippocampus are also responsible for
and Ajmone-Marsan, 1964), a slow depolarizing potential pathological synchronization (Perez and Carlen, 2000;Traub
with a high-frequency ( 200 Hz) burst of action potentials. et al., 2001).
In hippocampal slices from healthy animals, PDSs can be The interictal spike is terminated by activation of hyperpo-
observed if GABAA inhibition is reduced or if excitability is larizing GABAA and GABAB receptor-mediated currents and
increased by increasing potassium, reducing magnesium, calcium-dependent potassium currents (Alger and Nicoll,
reducing calcium, or blocking potassium channels with 4- 1980; Domann et al., 1994; Scanziani et al., 1994). There is also
aminopyridine (de Curtis and Avanzini, 2001). The PDS is some evidence of a contribution by other potassium currents,
characterized by an early phase that is maintained by intrinsic such as the sodium-dependent potassium current (Schwindt
properties of the neuron followed by a later phase that is sec- et al., 1989). Blocking the after-hyperpolarization, however,
ondary to recurrent excitation. Thus, the generation of inter- only results in moderate prolongation of the burst in CA3;
ictal spikes is dependent on two phenomena: the intrinsic and exhaustion of the immediately releasable pool of gluta-
burst properties of neurons and the synchronization of neu- mate has also been proposed to be a critical process in burst
ronal populations. Within the hippocampus, pyramidal cells termination (Staley et al., 1998). Furthermore, large depolar-
in area CA3 and some in area CA1 demonstrate burst proper- izations (rather than hyperpolarizations) herald the termina-
ties (see Chapter 5). The mechanisms underlying this are dif- tion of brief epileptic after-discharges (Bragin et al., 1997).
ferent for neurons from these two subelds. The bursting in This depolarization can be replicated by focal microinjection
CA3 pyramidal cells appears to be dependent on regenerative of potassium, and it has been hypothesized that potassium
dendritic potentials secondary to activation of calcium and ions released by discharging neurons result in propagating
sodium channels (Traub and Jefferys, 1994), whereas the burst waves of depolarization, which block spike generation in neu-
properties of some CA1 pyramidal cells is probably due to rons akin to spreading depression (Bragin et al., 1997).
778 The Hippocampus Book

Nevertheless, interictal spikes activate hyperpolarizing cur- hippocampal slice preparations, in which there is partial
rents resulting in a postspike refractory period, during which preservation of the trisynaptic loop, have conrmed the
neuronal activity is inhibited (de Curtis and Avanzini, 2001). antiepileptic potential of spikes. Spike discharges generated in
The effective activation of these currents by the interictal spike CA3 inhibited epileptic activity in the entorhinal cortex, so
thus raises the possibility that spikes can be anti-ictogenic. sectioning of the Schaffer collaterals led to potentiation of
There is evidence that this may be the case or at least that entorhinal cortex seizure activity (Fig. 162) (Barbarosie and
spikes are intrinsically different from a seizure. Depth EEG Avoli, 1997).
recordings in humans suggest that the interictal spike can Most of the ictal activity described thus far in the hip-
originate from a much wider eld than the ictal zone (see pocampal slice preparation has been brief and difficult to
above). Therefore, it is not uncommon to nd spikes originat- relate to seizures in vivo that last tens of seconds. Can such
ing in either hippocampus, whereas seizure activity is con- prolonged activity be mimicked in the slice, and does it differ
ned to one hippocampus. from briefer discharges? Prolonged ictal activity (seizure-like
A seizure is not the evolution of spike discharges but can activity) has been induced in the slice with high extracellular
begin as a distinct high-frequency rhythm (see above). Spike potassium (Traynelis and Dingledine, 1988). In this prepara-
discharges can precede the seizure with progressively less tion, the CA3 subeld generates regular interictal spikes,
effective after-hyperpolarizations, but ictal activity remains a which drive the generation of prolonged rhythmic seizure
distinct phenomenon. Furthermore, activation of interictal activity in the CA1 region; interestingly, the CA3 region in this
spikes occurs after the seizure, raising the possibility that this preparation is resistant to generating this ictal activity
is a compensatory antiepileptic response (de Curtis and (Traynelis and Dingledine, 1988). Inducing seizure-like activ-
Avanzini, 2001). Critical experiments in entorhinal cortex- ity in brain slices by other means (e.g., the lowering magne-

Figure 162. Interictal spikes inhibit

seizure activity. Spontaneous epilepti-
form activity induced by Mg2-free
articial cerebrospinal uid (ACSF)
before and after a Schaffer collateral
cut. Before the lesion (top), synchro-
nized interictal discharges are recorded
in the CA3, entorhinal cortex (EC),
and dentate gyrus (DG). Sectioning
the Schaffer collaterals (bottom) abol-
ishes interictal discharges in the EC
and discloses ictal epileptiform activity
that is recorded in the three areas.
Expanded traces of the experiment
shown in the top and bottom are illus-
trated in the middle. Note that before
the Schaffer collateral cut Mg2-free-
induced interictal discharges consist of
multiple components, whereas after
the cut (Interictal) they are markedly
reduced in duration and number of
events. (Source: Adapted from
Barbarosie and Avoli, 1997.)
 Hippocampus in Human Disease 779

sium level or GABAA receptor blockade) can result in the the EEG becomes less chaotic in large areas of cortex at the
generation of such activity in other regions in the hippocam- start of an ictus, suggesting that widespread synchronization
pal/parahippocampal formation such as the subiculum (Behr is occurring (Martinerie et al., 1998). To understand how a
and Heinemann, 1996), area CA3 (Borck and Jefferys, 1999), network that usually maintains oscillatory behavior becomes
and the entorhinal cortex (Jones, 1989). That small areas can epileptic, it is paramount to consider the changes that occur
generate seizure-like activity in the slice supports the hypoth- in the hippocampus during epileptogenesis.
esis that a network of only a few thousand neurons is neces- One problem with much that has been described is that it
sary to sustain seizure activity (Borck and Jefferys, 1999). That is associational (i.e., changes are observed and are assumed
the maintenance of seizure-like activity is different from the to contribute to the epileptogenic process). This may not
mechanisms underlying briefer epileptiform discharges is hold true for many of the changes, which could be compensa-
suggested by their differing pharmacology; NMDA receptor tory (i.e., protecting against the epileptogenic process).
antagonists can terminate brief epileptiform discharge but are Unfortunately, it has been difficult to distinguish between
ineffective during the maintenance phase of seizure-like these possibilities, and we are far from a comprehensive model
activity (Borck and Jefferys, 1999). of epileptogenesis. The changes that occur can be divided into
So how do spontaneous seizures (epilepsy) occur in vivo? structural change (neuronal loss, reorganization, synaptic
The brain slice studies can give us an insight into specic reorganization, changes in glia and extracellular space),
questions concerning the generation of epileptiform dis- changes in neurotransmission, and lastly changes in neuronal
charges. The interpretation and the in vivo extrapolation of properties. We discuss how each of these changes may con-
such studies are, however, complicated by certain observa- tribute to the epileptogenic process (Fig. 163).
tions: Seizure-like activity can be generated in vitro by quite
disparate means, and the mechanisms and structures involved Structural Change
in the generation of such activity can differ from study to
study. Seizure activity relies on oscillatory synchronization; Neuronal Loss. Hippocampal sclerosis is typically a unilateral
thus, mechanisms similar to those described in Chapter 8 process, affecting either hemisphere equally, with involvement
undoubtedly contribute to the emergence of in vivo seizure of the whole length of the hippocampus. In some cases more
activity. Furthermore, it is likely that mechanisms similar (but focal damage may be observed (Babb et al., 1984). and in oth-
not necessarily identical) to those that underlie spike dis- ers there is bilateral sclerosis (Van-Paesschen et al., 1997a). In
charges and longer epileptic after-discharges promote in so-called classic hippocampal sclerosis, selective loss of
vivo seizure activity. The transition from normal, physiologi- pyramidal cells is seen in the CA1 subeld and in the hilar
cal oscillatory behavior to epileptiform behavior is likely to be region with accompanying astrocytic gliosis. In some patients
due to greater spread and neuronal recruitment secondary to neuronal loss is restricted to the CA1 subeld (de Lanerolle et
enhanced connectivity, enhanced excitatory transmission, or a al., 2003). Pyramidal cells of CA2 and dentate granule cells
failure of inhibitory mechanisms. Indeed, in human studies, appear more resistant (Bruton, 1988). In severe hippocampal

Figure 163. Structural changes, changes in neurotransmission, and changes in the intrinsic
properties of neurons all contribute to the development of epilepsy.
780 The Hippocampus Book

sclerosis almost total neuronal loss is seen in all hippocampal This and the observation that the entorhinal cortex can alone
subelds and may be accompanied by marked deposition of maintain seizure-like activity (Jones, 1989) perhaps implicate
corpora amylacea. In the pattern of hippocampal sclerosis a more specic role for this region in seizure generation.
termed end folium gliosis, encountered in 3% to 4% of sur- The extent of any temporal neocortex neuronal loss does
gical cases (Bruton, 1988), the neuronal cell loss appears con- seem to correlate with the severity of hippocampal damage
ned to the hilus and includes loss of both principal cells and (Bruton, 1988). Neocortical neuronal loss also appears to be
interneurons. This pattern of hippocampal atrophy is less eas- layer-specic, with cortical layers II and III more severely
ily detected on preoperative MRI and is associated with a later affected.
onset of epilepsy than classic hippocampal sclerosis and a
worse postoperative seizure outcome (Van Paesschen et al., Mechanisms of Neuronal Death in Status Epilepticus. The pat-
1997b). tern of neuronal death in hippocampal sclerosis is mirrored by
Quantitative histological studies have been carried out in neuronal death seen in postmortem specimens following sta-
hippocampal sclerosis series, and pathological grading sys- tus epilepticus (DeGiorgio et al., 1992). Imaging studies have
tems have been proposed to categorize the severity of neu- demonstrated the occurrence of hippocampal sclerosis follow-
ronal loss in hippocampal sclerosis; for example, in one ing status epilepticus and prolonged seizures, further empha-
system, grade I hippocampal sclerosis correlates with less than sizing the vulnerability of the hippocampus to neuronal
10% of neuronal dropout in CA1 up to grade IV hippocampal damage. Insights into the mechanisms underlying this damage
sclerosis, which shows more than 50% neuronal loss in all have largely been derived from animal models of status epilep-
subelds (Wyler et al., 1992). This is based on a semiquantita- ticus (Meldrum, 1991). They have shown that although a cer-
tive assessment of neuronal loss in histological sections. Such tain amount of neuronal damage is secondary to physiological
analyses have proved useful as they allow pathological correla- compromise that occurs during status epilepticus (e.g.,
tion with clinical parameters (e.g., the age of the patient at hypoxia, hypoglycemia, hypotension) a large proportion of
epilepsy onset and the duration of seizures) (Davies et al., the damage is independent of these factors. This neuronal
1996) and with neuroimaging features. These grades may also damage is due to excitotoxicity in which the presence of
reect a progressive evolution of hippocampal sclerosis from epileptic activity mediates neuronal death through the activa-
grades I to IV, mirroring ongoing hippocampal atrophy that tion of glutamate receptors. Excessive inux of calcium (and
has been occasionally reported in sequential neuroimaging zinc at mossy ber synapses) through primarily NMDA recep-
studies. tors, but also through -amino-3-hydroxy-5-methyl-4-isoxa-
Marked cytological alterations have been observed in sur- zolepropionate (AMPA) receptors lacking the GluR2 subunit,
viving neurons in hippocampal sclerosis using immunohisto- results in a cascade of reactions leading to cell death (Weiss et
chemistry, electron microscopy, and confocal imaging al., 2000;Lipton and Rosenberg, 1994;Tanaka et al., 2000).
techniques (Blumcke et al., 1999a). These changes include
enlargement and accumulation of neurolaments in end folial Specic Neuronal Vulnerability in Hippocampal Sclerosis. Loss
neurons, abnormal dendritic nodular swellings, and ramica- of the principal pyramidal cells in hippocampal sclerosis is
tions of these cells. These features more likely represent sec- established, but it is difficult conceptually to conceive how
ondary or adaptive cellular changes due to the altered removal of principal excitatory neurons can contribute to a
connectivity in the reorganized hippocampus rather than a state of hyperexcitability. Undoubtedly, neuronal loss may
primary cellular abnormality. contribute to synaptic rearrangements and perhaps increased
Neuronal loss and gliosis may also be present in adjacent connectivity, but more important perhaps is the vulnerability
limbic structures, including the amygdala (Yilmazer-Hanke et of specic subsets of interneurons in the hippocampal forma-
al., 2000) and parahippocampal gyrus, which along with hip- tion, which may inuence the intrinsic circuitry of the hip-
pocampal sclerosis are collectively referred to as mesial tem- pocampus and seizure propagation. Most interneurons
poral sclerosis. Neuronal loss may also involve the entorhinal contain the neurotransmitter GABA but can be further subdi-
cortex, and volume loss has been demonstrated in the entorhi- vided according to their connectivity, calcium-binding pro-
nal cortex with quantitative MRI studies (Salmenpera et al., tein content, and neurotransmitter receptor status (Freund
2000). Neuropathological studies of this cortical region in and Buzsaki, 1996).
patients with hippocampal sclerosis have suggested signicant Neuropeptide Y (NPY)- and somatostatin-containing
loss of layer III cells (Du et al., 1993), although other studies inhibitory interneurons are normally numerous in the hilus
have suggested a more variable pattern of cell loss involving all and form a dense plexus of bers in the outer molecular layer
layers, including loss of pre-alpha cells (Yilmazer-Hanke et al., of the dentate gyrus which co-localizes with glutamic acid
2000). The observed loss of entorhinal neurons could indicate decarboxylase (GAD) (Amaral and Campbell, 1986). Loss of
either primary or secondary involvement of this region in the these interneuronal subtypes in the hilus was noted in hip-
pathophysiology of temporal lobe seizures. An important pocampal sclerosis (deLanerolle et al., 1989;Mathern et al.,
observation, however, has been the demonstration of entorhi- 1995a). NPY-containing axons also appeared to be reorgan-
nal cortex neuronal loss in the absence of hippocampal scle- ized in the dentate molecular layer in hippocampal sclerosis,
rosis (Yilmazer-Hanke et al., 2000; Bernasconi et al., 2001). and ectopic expression of NPY in granule cells has been
 Hippocampus in Human Disease 781

observed following seizures (Vezzani et al., 1999). This is likely mice, the absence of these proteins does not appear to affect
to represent plasticity in NPY inhibitory mechanisms in the the numbers of interneurons or excitotoxic-mediated cell loss
epileptogenic hippocampus. A more recent quantitative study in epilepsy (Bouilleret et al., 2000). Furthermore, in hip-
using in situ hybridization however, has suggested that NPY pocampal sclerosis there is loss of calbindin expression by
and somatostatin neurons in the fascia dentata are lost in pro- granule cells (Magloczky et al., 2000). Granule cells are typi-
portion to the overall cell loss and are not specically tar- cally more resistant to damage in hippocampal sclerosis than
geted in the disease process (Sundstrom et al., 2001). other principal neurons and it has controversially been pro-
The calcium-binding proteins calbindin D-28-K, parval- posed that the calbindin loss actually protects these cells from
bumin, and calretinin label different, nonoverlapping subsets Ca2-mediated neuronal damage (Nagerl et al., 2000). The
of inhibitory hippocampal interneurons, and the resistance or proposed mechanism underlying this proposal is that lack of
susceptibility of these cell populations in hippocampal sclero- calbindin results in larger intracellular free calcium transients
sis may directly affect hippocampal epileptogenesis. The nor- that inactivate calcium channels, thus limiting intracellular
mal distribution of calbindin is not restricted to interneurons calcium accumulation; on the other hand, buffering of free
but is also present in the dentate gyrus granule cells, mossy calcium transients with calcium-binding proteins permits
bers, and CA2 pyramidal cells; parvalbumin and calretinin greater accumulation of intracellular calcium. Thus, although
are present only in interneurons. Calbindin-positive interneu- these calcium-binding proteins may identify neurons that are
rons are mainly involved in the inhibition of principal cells in resistant to damage, the calcium-binding protein itself is
the dendritic region, whereas calretinin-positive interneurons probably not neuroprotective.
probably selectively innervate other interneurons (Magloczky Selective loss of hilar mossy cells, an excitatory interneuron
et al., 2000). An early study had suggested preferential survival with distinctive dendritic arborizations, has been described in
of calbindin and parvalbumin immunoreactive neurons in hippocampal sclerosis cases compared to patients with gener-
hippocampal sclerosis (Sloviter, 1991). Furthermore, alized seizures (Blumcke et al., 2000b). These excitatory
increased complexity of the terminal processes of powerful interneurons project to inhibitory basket cells, and their loss
inhibitory interneurons, the chandelier cell (which may be may result in reduced feedforward granule cell inhibition sup-
parvalbumin- or calbindin-positive) has also been demon- porting the experimental dormant basket cell hypothesis
strated in hippocampal sclerosis (Arellano et al 2004). More (Sloviter, 1991). However it is recognized in animal models
recent quantitative studies, however, have shown selective loss that basket cells also receive direct excitatory input from the
of parvalbumin-immunoreactive neurons in the hilus dispro- granule cells and perforant path bers, thus bypassing the
portionate to the overall cell loss (Zhu et al., 1997). The dis- mossy cells (Kneisler and Dingledine, 1995).
tribution of calbindin-positive interneurons in the dentate
gyrus in hippocampal sclerosis was shown not to differ from Pathophysiological Role of Neuronal Damage. Are epileptoge-
controls in one study but striking enlargement of their cell nesis and neuronal damage directly related? Following status
bodies with enhanced expression of calbindin and modica- epilepticus, those animals that develop spontaneous seizures
tion of the dendritic trees and synapses of these cells was have greater hilar interneuronal loss, perhaps resulting in
noted (Magloczky et al., 2000). Marked plasticity and reor- decreased inhibitory drive (Gorter et al., 2001). It has also been
ganization of calbindin-positive interneurons in the CA1 sub- suggested that damage to CA3 and the Schaffer collaterals may
eld in the hippocampus has also been shown, which may prevent spikes generated in CA3 from inhibiting epileptic
predate the pyramidal cell loss (Wittner et al., 2002). The nd- activity in the limbic system (see above). Selective neuronal
ings support a complex set of changes in interneural anatomy death can also result in a change in the nature of inhibition.
with changes in interneuron targets. Calretinin cells do not Dendritically expressed inhibitory postsynaptic potentials
appear to show abnormal distribution in hippocampal sclero- (IPSPs) modify the transmission of excitatory postsynaptic
sis (Blumcke et al., 1996); but increased numbers of a subset potentials (EPSPs) to the soma, whereas somatic and periso-
of calretinin-positive neurons, the Cajal-Retzius cells, occurs matic IPSPs depress the excitability of the neuron (Cossart et
in hippocampal sclerosis in some patients (Blumcke et al., al., 2001). Basket cells, which make multiple perisomatic and
1999b). Studies have demonstrated an expansion of calre- somatic synapses, have extensive axonal arborizsations leading
tinin-positive axonal networks in the molecular layer of the to connection of one basket cell with many pyramidal cells. By
dentate gyrus in hippocampal sclerosis (Blumcke et al., 1996, synchronously modulating the excitability of a group of
1999b; Magloczky et al., 2000). These bers are likely to rep- pyramidal neurons, one basket cell can synchronize pyramidal
resent those of the excitatory supramamillary pathway termi- cell activity (Cobb et al., 1995). The loss of oriens/lacunosum-
nating on granule cells rather than local axons. This moleculare interneurons in CA1 results in loss of distal den-
observation may indicate enhanced excitation of granule cells dritic inhibition, whereas the preservation and increased
by this pathway (Magloczky et al., 2000). connectivity of basket cells may result in greater pyramidal cell
The observed relative resistance of certain calbindin-con- synchronization (Fig. 164) (Cossart et al., 2001). The selective
taining neurons to neuronal damage has led to the suggestion loss of certain interneuronal populations in hippocampal scle-
that calbindin itself may be neuroprotective (Leranth and rosis is thus probably pro-epileptogenic; but is neuronal loss
Ribak, 1991). However, in calcium-binding protein knockout necessary for epileptogenesis? That there is a distinction
782 The Hippocampus Book

Figure 164. Inhibition is decreased in dendrites but increased at tic tissue. C, D. Somatic recordings from CA1 pyramidal cells. C.
the soma. A, B. Recordings from the apical dendrites of CA1 pyram- The cumulative probability plot of the amplitudes of spontaneous
idal cells. A. The cumulative probability plot of the amplitudes of IPSCs was shifted to the right, indicating an increased frequency
miniature events was shifted to the left, indicating decreased fre- of large-amplitude events in experimental animals (n  6 cells) as
quency of amplitudes larger than 15 pA in epileptic animals. The compared to controls (n  6 cells). D. Whole-cell recordings of
kinetics of miniature events (normalized averages) did not seem spontaneous (control) and miniature (TTX, 1 M) inhibitory post-
modied (inset). B. Whole-cell recordings of spontaneous (control) synaptic currents obtained at the reversal potential of glutamatergic
and miniature (TTX, 1 M) inhibitory postsynaptic currents events. The frequency of spontaneous events was similar in control
obtained at the reversal potential of glutamatergic events in a con- and epileptic neurons, but the number of large-amplitude events
trol and a kainate-treated animal. Continuous 4-s recordings are seemed to be increased in the epileptic neuron. (Source: Adapted
shown. The frequency of spontaneous events is lower in the epilep- from Cossart et al., 2001, with permission.)

between neuronal damage and epileptogenesis is indicated by compared to 100 m in control subjects (Houser, 1990). The
kindling, in which epileptogenesis occurs in the setting of no dispersed cells often appear elongated or fusiform in shape,
or minimal neuronal damage, and similarly in the intrahip- reminiscent of migrating neurons. Less often a bilaminar
pocampal tetanus toxin model (see above). Indeed, kindling arrangement of granule cells is observed (Houser et al., 1992;
may protect against neuronal damage induced by kainic acid Thom et al., 2001) or nests of GCs are present in the hilus
(Kelly and McIntyre, 1994), raising the intriguing possibility (Houser, 1990; Thom et al., 2001). The incidence of granule
that epilepsy itself can be neuroprotective. cell dispersion in hippocampal sclerosis surgical series is of the
order of 40% (Houser, 1990; Thom et al., 2001).
Granule Cell Dispersion. The observation of disorganization It has been suggested that granule cell dispersion repre-
or dispersion of granule cells into the molecular layer of the sents a primary abnormality of neuronal migration or an
dentate gyrus in hippocampal sclerosis was rst described in underlying hippocampal malformation (Houser et al., 1992).
detail by Houser (Houser, 1990; Houser et al., 1992). There are occasional reports of granule cell dispersion in asso-
Dispersed granule cells appear separated from the normally ciation with cortical malformations in the absence of a history
compact cell layer, which gives the impression of an undulated of seizures and with bilateral hippocampal involvement
irregular border with the molecular layer. In some cases the (Harding and Thom, 2001). Disorganization of the granule
deep (hilar border) of the granule cell layer is also ill-dened. cell layer and ectopic localization of neurons have also been
As a result, in hippocampal sclerosis the cell layer appears noted in several animal models with cortical malformations
broadened with a mean width of 180 m in TLE patients such as the reeler, and p35 mutant mice. The presence of gran-
 Hippocampus in Human Disease 783

ule cell dispersion in human hippocampal sclerosis has also have already demonstrated the existence of distinct popula-
been correlated with epileptic events occurring early in life, tions of granule cells, with one group showing abnormal
including febrile seizures (Houser, 1990), suggesting a vulner- excitability (Dietrich et al., 1999). We also know from animal
ability of these neurons during this time period. It has further studies that there is considerable potential for adaptability and
been shown that the presence of granule cell dispersion corre- plasticity of granule cells, such as increased basal expression of
lates well with the severity of hippocampal neuronal loss GAD (Sloviter et al., 1996), NPY induction (Vezzani et al.,
(Thom et al., 2001). This suggests that granule cell dispersion 1999), loss of calbindin expression (Magloczky et al., 1997),
represents an epiphenomenon of hippocampal sclerosis and altered ionotropic and metabotropic glutamate and
rather than a primary abnormality, the migration of granule GABA neurotransmitter receptor proles on granule cells
cells perhaps being inuenced by neurotrophin secretion dur- (Loup et al., 2000). It is plausible that such plasticity could be
ing seizures or other cellular signals. Interestingly, an inverse enhanced in newly generated granule cells and that it could
correlation between the levels of reelin protein and the extent contribute to seizure propensity.
of granule cell dispersion in the hippocampus has been
shown, suggesting a possible functional role for Cajal Retzius Mossy Fiber Sprouting. In 1974, using Golgi techniques,
cells in this process (Frotscher et al., 2003). Scheibel and colleagues identied aberrant axons from gran-
In animal models of epilepsy, such as the pilocarpine ule cell neurons ascending into the molecular layer of the den-
model, there is evidence to suggest that abnormally migrated tate gyrus in hippocampal specimens from patients with
granule cells are newly generated cells, neurogenesis being epilepsy (Scheibel et al., 1974). It has long been considered
stimulated by the seizures (Parent et al., 1997) (see Chapter 9). that reorganization of the excitatory glutamatergic mossy
Rapid dispersion of granule cells has been demonstrated fol- ber pathway is a key event in the development of chronic
lowing injury (Omar et al., 2000) and it has been shown that seizures (Sutula et al., 1989). Mossy ber sprouting in human
newly generated cells can migrate as far as CA3 and integrate hippocampal sclerosis specimens results in aberrant innerva-
into the CA3 neuronal network (Scharfman et al., 2000). tions of other granule cells and also of CA3 pyramidal neu-
Abnormal connections formed by new cells may contribute to rons, resulting in both feedback and feedforward excitation
seizure development (Parent et al., 1997), although experi- (Babb et al., 1991; Mathern et al., 1995b). In addition, aber-
mental inhibition of neurogenesis does not prevent granule rant mossy bers in animal models innervate interneurons,
cell axon reorganization (see below) in epilepsy models suggesting that new inhibitory circuits are established (Kotti
(Parent et al., 1999). Studies have conrmed that neurogene- et al., 1997).
sis also occurs in adult human dentate gyrus (Eriksson et al., Mossy ber sprouting in the supragranular layer of the
1998), and neuronal progenitor cells have been isolated from dentate gyrus can be demonstrated using the Timms histo-
the dentate gyrus (Roy et al., 2000). This pool of precursor chemical method, which highlights the zinc-rich mossy ber
cells may have important physiological roles, but it is conceiv- synaptic terminals (Babb et al., 1991), or with dynorphin
able that in human epilepsy, stimulated by seizures, an immunohistochemistry (Houser et al., 1990). Increased
increased rate of granule cell neurogenesis occurs, leading to expression of growth-associated protein GAP-43 in the supra-
the abnormal cell localization and reorganization observed in granular layer is thought to be indicative of active mossy ber
hippocampal sclerosis. There is also evidence emerging that sprouting in hippocampal sclerosis specimens (Proper et al.,
radial glial cells in this region could act as neuronal precur- 2000). Similarly increased syanaptogenesis in this region has
sors, and neurogenesis may be stimulated by NPY (Howell et been demonstrated by studying the distribution of 5nucleoti-
al., 2003). dase activity, which localizes in regions with more active
Another study has demonstrated the stem cell intermediate synaptic turnover (Lie et al., 1999). Overall reorganization of
lament protein nestin in granule cell neuronal precursors in synaptic terminals in hippocampal sclerosis has also been
young patients with mesial TLE and surgery before the age of demonstrated in human specimens using immunohistochem-
2 years (Blumcke et al., 2001). Similar cells were not found in istry for synaptic antigens such as synaptophysin, which shows
adults with hippocampal sclerosis; whether the nestin-positive a loss in the hilus and increased labeling in the dentate gyrus
cells represent newly generated cells or a delay in hippocam- molecular layer (Honer et al., 1994; Davies et al., 1998; Proper
pal development in these younger patients is not clear. Studies et al., 2000). Similarly, prominent immunolabeling for chro-
of cell cycle proteins, including Ki67, showed low expression mogranins (neuropeptide precursors that can be co-released
in the dentate gyrus subgranular layer in adult hippocampi with catecholamines and peptides) in the inner molecular
from patients with epilepsy (Del Bigio, 1999). Although this is layer of the dentate gyrus has been shown to correspond with
an insensitive technique for measuring cells with a low reorganized mossy fibers in patients with epilepsy
turnover rate, it does imply that neurogensis in hippocampal (Kandlhofer et al., 2000). In parallel with increased synapto-
sclerosis is a rare event and likely to be dependent on age. genesis, elaboration and increased complexity of granule cells
Even if migrated granule cells do represent newly gener- dendrites in the inner molecular layer has been demonstrated
ated cells, it is unknown whether there are any differences in in hippocampal sclerosis patients (von Campe et al., 1997).
the physiological properties of these less mature cells. The hypothesis that mossy ber collaterals form granule
Electrophysiological studies in human hippocampal sclerosis cellgranule cell synapses has been conrmed by visualizing
784 The Hippocampus Book

dentate granule cells and their mossy bers after terminal ing sprouting of CA1 pyramidal cell axons, resulting in an
uptake and retrograde transport of biocytin in epileptic rats increase in interconnectivity that results in increased excitabil-
secondary to status epilepticus (Okazaki et al., 1995). ity of the CA1 subeld (Esclapez et al., 1999). Sprouting of
Sprouting of mossy bers is thought to result from excitatory axons leading to increased interconnectivity may be
epilepsy-induced loss of target cells (e.g., hilar mossy cells). a powerful means of generating hyperexcitable circuits that
However, in animal models it may be an early event, occurring can maintain and propagate epileptic activity.
within 4 weeks following the start of kindling (Elmer et al.,
1996) and independent of hippocampal cell loss, possibly reg- Dormant Basket Cell Hypothesis. Immediately following 24
ulated by neurotrophic factors (Adams et al., 1997). hours of perforant path stimulation, there is loss of paired-
Preliminary studies also indicate that mossy ber sprouting is pulse inhibition to perforant path stimulation in the dentate
likely to be independent of any granule cell neurogenesis granule cell layer (Sloviter, 1987). The mechanism underlying
(Parent et al., 1997). this change was proposed to be the loss of excitatory input
onto basket cells due to excitotoxic loss of mossy cells
Pathological Role of Mossy Fiber Sprouting. It has long been (Sloviter, 1987). However, application of NBQX, an AMPA
considered that reorganization of the excitatory glutamatergic receptor antagonist, was shown to inhibit the loss of mossy
mossy ber pathway is a key event in the development of cells but to have no effect on the loss of paired-pulse inhibi-
chronic seizures. The dentate granule cells of the hippocam- tion (Penix and Wasterlain, 1994). Similarly, in the tetanus
pus probably act as a brake against seizure propagation toxin model, disinhibition occurs in the absence of hilar cell
through limbic circuitry (Perlin et al., 1992; Lothman and loss (Whittington and Jefferys, 1994). Other factors are there-
Bertram, 1993). This is mediated by the relative inexcitability fore likely to be responsible for the loss of paired-pulse inhi-
of dentate granule cells through strong tonic GABAergic inhi- bition, such as loss of affinity and activity of GABAA receptors
bition and relatively hyperpolarized membrane potentials. in the hippocampus or a shift in the chloride reversal poten-
Dentate granule cells do not show the burst properties char- tial to more positive values (Kapur et al., 1994). These changes
acteristic of hippocampal pyramidal cells in response to may be due in part to altered phosphorylation of GABAA
reduced GABAergic inhibition. Furthermore, dentate granule receptors (Kapur et al., 1994) but could also be due to altered
cell synchronization is difficult to achieve because of the low GABAA receptor subunit composition and expression (Sperk
rate of connectivity between the granule cells. et al., 1998). Perhaps one of the main mechanisms is decreased
Epileptogenesis may change the properties of dentate gran- recruitment of basket cells by excitatory inputs from dentate
ule cell receptors (see below); but, importantly, mossy ber granule cells, the perforant path, and CA3 pyramidal cells
sprouting greatly increases their connectivity. Even with through upregulation of presynaptic metabotropic glutamate
sprouting, it is difficult to recruit dentate granule cells into receptor activity (Doherty and Dingledine, 2001).
epileptiform activity perhaps because of the extensive and Loss of paired-pulse inhibition was proposed to be an
increased synaptic input from GABAergic interneurons. epileptogenic phenomenon, but subsequent studies have
Epileptiform activity can be induced in the dentate granule shown that paired-pulse inhibition becomes increased during
cells when mossy ber sprouting is present by increasing the latent period (i.e., with epileptogenesis) and that epilepto-
extracellular potassium or by reducing GABAA receptor- genesis is associated with increased recruitment of basket cells
mediated inhibition (Cronin et al., 1992; Wuarin and Dudek, (Milgram et al., 1991; Sloviter, 1992).
1996). This has led to the compelling hypothesis that epilep- Synaptic rearrangement also occurs in CA1; and the hyper-
togenesis results in a potentially hyperexcitable granule cell excitability in CA1 with epileptogenesis has been proposed to
layer that can be recruited into epileptic activity either by a be secondary to reduced recruitment of inhibitory interneu-
rise in extracellular potassium, which could occur secondary rons (Bekenstein and Lothman, 1993). The main evidence for
to a sustained discharge from the entorhinal cortex, or this is the loss of the IPSP from the EPSPIPSP sequence on
through breakdown of inhibition (see below). distant stimulation of the Schaffer collaterals. This evidence is
Is mossy ber sprouting necessary for epileptogenesis? perhaps awed because this study does not adequately differ-
Blocking mossy ber sprouting with the protein synthesis entiate loss of the IPSP from prolongation of the EPSP.
inhibitor cycloheximide does not prevent epileptogenesis Subsequent data have shown increased excitability of CA1
(Longo and Mello, 1998). The interpretation of these experi- interneurons following epileptogenesis (Sanabria et al., 2001).
ments is confounded, however, by the likelihood that in- There are, however, other mechanisms underlying possible
hibiting protein synthesis inhibits other antiepileptogenic disinhibition in the CA1 region including changes in the pat-
processes. In a recent study, the presence (or not) of dynor- tern of inhibition (see above) and postsynaptic changes (see
phin-positive mossy ber sprouting in hippocampal sclerosis below).
correlated with the postsurgical outcome; those with sprouting
more often having a seizure-free outcome (de Lanenerolle et Glial Cells and Extracellular Space. There are alterations in
al., 2003). It is also becoming increasingly apparent that there astrocytic function in the gliotic hippocampus. Astrocytes
is sprouting of axons from other neuronal populations, includ- show physiological changes characteristic of immature astro-
 Hippocampus in Human Disease 785

cytes, including prolonged depolarization, that may con- observed, highlighting the plasticity of this neurotransmitter
tribute to seizure generation (Hinterkeuser et al., 2000; system in hippocampal sclerosis (Loup et al., 2000). The
Schroder et al., 2000). Indeed, there is altered expression of changes that are observed differ with time and between
ionotropic glutamate receptors on astrocytes in hippocampal regions. Thus, specic GABAA receptor changes occur during
sclerosis which may facilitate seizure spread (Seifert et al., acute seizures. As acute seizures continue they can become less
2004). In another study of rat and human hippocampi in TLE, responsive to benzodiazepines, which is mirrored by de-
the proliferation of glial cells in areas of neuronal loss were creased potency of benzodiazepines on GABA-mediated
associated with alterations in extracellular potassium, which synaptic currents in dentate granule cells (Kapur and
also may affect conduction of seizure activity (Heinemann et Macdonald, 1997). In contrast, the potency of GABA itself and
al., 2000). Altered levels of glial glutamate transporters (e.g., pentobarbitone remained unaltered, suggesting that rapid
EAAT2), which has been shown in hippocampal sclerosis changes in GABA receptor properties occur during seizures.
(Proper et al., 2002), may also inuence the extracellular pool Although the pathophysiological consequences of these
of glutamate. Furthermore, calcium oscillations in astrocytes changes are difficult to predict, they have implications for the
can result in glutamate release, which may contribute to treatment of acute, prolonged seizures. During epileptogene-
epileptic activity (Tian et al., 2005). sis the GABAA receptor changes are more complex and are
During seizures there is considerable shrinkage of the region-specic. In the dentate granule cells, there is an
extracellular space due to intracellular accumulation of increase in the number of GABAA receptors per synapse, lead-
sodium chloride; indeed, a single seizure can result in a 10% ing to increased quantal size (Nusser et al., 1998). An espe-
to 30% decrease in extracellular space (Lux et al., 1986). This cially interesting finding is that the increased GABA
may result in increased nonsynaptic transmission through receptor-mediated signaling to dentate granule cells becomes
ephaptic and ionic mechanisms. Indeed, decreased extracellu- more sensitive to zinc (Buhl et al., 1996; Brooks et al., 1998).
lar space during seizures has been indirectly shown to have To understand the potential involvement of zinc in epilepsy, it
a role in seizure maintenance and spread; hypotonic extra- is necessary to consider the role of the sprouted mossy bers.
cellular solutions that decrease extracellular space are pro- Mossy ber terminals contain zinc and release it during
convulsant, whereas hypertonic solutions (increasing the synaptic activity (Assaf and Chung, 1984; Howell et al., 1984).
extracellular space) can terminate seizure discharges (Roper et Thus, it is conceivable that in the epileptic hippocampus zinc
al., 1992). Although overall there is an expansion of the extra- released from mossy bers results in disinhibition, unmasking
cellular space in hippocampal sclerosis (Hugg et al., 1999; the potentiated excitatory dentate granule cell circuits (Buhl et
Wieshmann et al., 1999), how this relates to local changes in al., 1996). This hypothesis is confounded by the observation
extracellular space and neurotransmission are unknown. that the zinc released from sprouted mossy bers failed to
affect GABAA receptor-mediated currents induced by local
Neurotransmitter Systems release of caged GABA (Molnar and Nadler, 2001). Further-
more, mice that lack the zinc transporter (ZnT3) and so lack
GABAergic Mechanisms. Alteration in the distribution of synaptically available zinc have an exaggerated response to a
neurotransmitter receptors has been extensively investigated convulsant; this observation does not, however, preclude a
as a pathogenic mechanism in the hyperexitability of the hip- role for zinc in epileptogenesis (Cole et al., 2000).
pocampus in TLE. The GABA hypothesis proposes that a Decreased sensitivity of GABAA receptor-mediated signals
decit in inhibitory GABAergic transmission is implicated in to zolpidem (a selective benzodiazepine agonist) has also been
seizures. GABAA and to a lesser extent GABAB receptor sub- noted (Brooks et al., 1998). Using a combination of patch-
type expression (Barnard et al., 1998) and reuptake mecha- clamp recording and single-cell mRNA amplication, it was
nisms have been studied in human hippocampal sclerosis found that the increased zinc sensitivity and decreased benzo-
tissues. Many alterations in GABA transmission may represent diazepine sensitivity of the GABAA receptor was associated
an adaptive mechanism in the brain in response to repetitive with (and possibly explained by) decreased expression of the
seizures, and increased expression of GABAA receptors has 1 subunit and increased expression of the 4 subunit
been documented in animal models of epilepsy as a compen- (Brooks et al., 1998). In addition, there were changes in sub-
satory mechanism (Fritschy et al., 1999). There is also upreg- unit expression that may affect benzodiazepine efficacy and
ulation of GAD, the main GABA-synthesizing enzyme, in the efficacy of barbiturates, steroids, zinc, and loreclezole, a
interneurons following acute seizures. GABA and GAD are new antiepileptic drug (Brooks et al., 1998). These changes
also upregulated in the mossy bers and dentate granule cells were seen during the latent period that predated the onset of
(Sloviter et al., 1996). The nding of a mossy ber-like epilepsy, suggesting a role in the epileptogenic process. GABAA
GABAergic signal has raised the possibility that mossy bers receptor-mediated transmission in CA1 undergoes different
co-release glutamate and GABA (Walker et al., 2001); seizures changes. In contrast to dentate granule cells, GABAA receptors
may upregulate the GABAergic component (Gutierrez and on CA1 pyramidal cells are less responsive to applied GABA
Heinemann, 2001). In human hippocampal sclerosis, selective following epileptogenesis (Gibbs et al., 1997). There are also
upregulation of GABAA2 subunit in granule cells has been changes that suggest there may be a decrease in the presynap-
786 The Hippocampus Book

tic GABA reserve, although the synaptic consequences of this culline and by glutamate receptor antagonists, suggesting that
are unknown (Hirsch et al., 1999). both excitatory and inhibitory signaling contributed to the
The effects of GABAA receptor activation on membrane spontaneous interictal-like events. Further studies in these
potential depend on the chloride reversal potential. High cells demonstrated that the GABAergic synaptic events
internal chloride such as occurs developmentally can result in reversed at depolarized potentials (Fig. 165). Thus, depolar-
depolarizing GABAA receptor-mediated potentials. Could izing GABAergic responses potentially contributed to human
such depolarizing GABAA receptor-mediated potentials occur ictal activity in the subiculum. The mechanism by which
in epileptic tissue and contribute to epileptogenesis? Evidence such depolarizing GABAA receptor-mediated potentials occur
from a study of hippocampal slices from patients who under- is unknown, but might be downregulation of the K/Cl
went temporal lobectomy suggest that this is so (Cohen et al., co-transporter KCC2 that maintains the low intracellular
2002). Synchronous rhythmic activity was found to be gener- chloride.
ated in the subiculum of slices of temporal lobe from patients GABAB receptor changes have also been demonstrated in
with temporal lobe epilepsy, and this synchronous rhythmic human hippocampal sclerosis tissue. GABAB receptors inhibit
activity was abolished by the GABAA receptor antagonist bicu- neurotransmitter release from presynaptic terminals and

Figure 165. Depolarizing -aminobutyric acid-A (GABAA) recep- by blocking GABAA receptors (lower trace). BIC, bicuculline. B, C.
tor-mediated potentials may contribute to seizure generation. Reversal potential for GABAergic transmission in cells that red
Pyramidal cells active during interictal-like events showed excitatory during interictal events was more depolarized than at rest (B, left; C,
GABAergic transmission. A. Discharges in pacemaker cells were squares). Reversal in inhibited cells was hyperpolarized from rest (B,
triggered by orthodromic stimulation (right, upper trace). right; C, circles). Records were obtained in NBQX and APV. Linear
Depolarizing synaptic potentials persisted when glutamatergic ts are indicated. PSP, postsynaptic potential. (Source: Adapted from
transmission was suppressed (middle trace). They were suppressed Cohen et al., 2002, with permission.)
 Hippocampus in Human Disease 787

cause late inhibitory synaptic potentials (Barnard et al., 1998). One of the major mechanisms underlying this is undoubtedly
Increased expression of GABAB1 receptor has been shown in an increase in NMDA receptor-mediated neurotransmission
the subiculum of hippocampal sclerosis cases and in surviving (Mody and Heinemann, 1987). Kindling results in fast, long-
CA1 neurons and granule cells with augmented receptor bind- lasting posttranslational modications in the function of den-
ing in CA3 (Billinton et al., 2001), although functional inter- tate granule cell NMDA receptor channels, leading to
pretation of these ndings is difficult. The upregulation could increases in the mean open time and burst and cluster dura-
represent a greater number of inhibitory synapses, increased tion and to decreases in the channel-blocking effect by mag-
postsynaptic GABAB receptors, or increased presynaptic nesium (Kohr et al., 1993). Similar changes in NMDA
GABAB receptor number, leading to decreased neurotransmit- channels have been reported in human epileptic tissue
ter release mainly from inhibitory axonal terminals. More (Lieberman and Mody, 1999). Modication of the NMDA
recently, downregulation of mossy ber presynaptic GABAB receptor channels probably results from a decrease in the
receptors has been found in tissue from epileptic animals. It activity of intracellular phosphatases, leading to increased
resulted in decreased mossy ber heterosynaptic depression, phosphorylation of the receptors (Kohr et al., 1993;
and may contribute to increased signal ow through the hip- Lieberman and Mody, 1994).
pocampus (Chandler et al., 2003). Decreased presynaptic There is more uncertainty concerning changes in AMPA
GABAB receptor activity on interneurons has also been pro- receptor neurotransmission. Certainly, changes in AMPA
posed to underlie the enhanced inhibition that occurs in the receptor subunit composition are seen in animal models prior
dentate gyrus during epileptogenesis (Haas et al., 1996). to neurodegeneration; there is a decrease in the expression of
Changes in GABA uptake have also been described in the the GluR2 subunit in vulnerable cells (Grooms et al., 2000).
dentate gyrus (During et al., 1995; Patrylo et al., 2001). Both This evidence supports a role for calcium ux through GluR2
animal and human data support a decrease in clearance of lacking AMPA receptors in mediating neuronal death
synaptically released extracellular GABA, perhaps owing to (Grooms et al., 2000). Conversely, there appears to be upregu-
decreased expression or impairment of GABA transporters lation of GluR2 in less vulnerable neurons such as the dentate
(During et al., 1995; Mathern et al., 1999; Patrylo et al., 2001). granule cells (de Lanerolle et al., 1998).
This has been speculated to lead to increased interictal Changes in glutamate uptake during epileptogenesis could
inhibitory efficacy. Because the rises in extracellular potas- also have an important role. There is burgeoning evidence
sium that occur during seizures may result in reversal of that glutamate may escape from the synaptic cleft to activate
GABA uptake, decreased GABA transporter function results extrasynaptic receptors, or even receptors at neighboring
in impairment of extracellular GABA rises during seizure synapses (Kullmann and Asztely, 1998). Glutamate spillover
activity, possibly resulting in impaired inhibition during can activate presynaptic glutamate receptors on GABAergic
seizure activity. The change in GABAA receptor subunits terminals, resulting in decreased inhibitory drive, and can
resulting in a greater response to endogenously applied GABA increase NMDA receptor-mediated signaling (Min et al., 1999;
(Brooks et al., 1998) could compensate for decreased extracel- Semyanov and Kullmann, 2000). Extrasynaptic accumulation
lular GABA rises. Nevertheless, these data raise the possibility of glutamate may play a role in epilepsy: Rodents lacking the
that decreased GABA transporter function could be pro- gene coding for the glial glutamate transporter GLT-1
epileptogenic during times of seizure activity but promote (EAAT2) show lethal spontaneous seizures (Tanaka et al.,
inhibitory transmission during the interictal period. 1997). Rather surprisingly, chronic administration of anti-
sense oligonucleotide to knock down the same transporter
Glutamatergic Mechanisms. Upregulation of excitatory produces a different phenotype, characterized by neurode-
metabotropic glutamate receptors (mGluR1 subunit) has generation rather than seizures (Rothstein et al., 1996).
been observed in the dentate gyrus in both human and animal Reduction of expression of the neuronal transporter EAAC1
models of hippocampal sclerosis, and it could contribute to (EAAT3), however, also causes seizures in rats. Subtle alter-
the development of chronic seizures through increased excita- ations in transporter levels have been reported in hippocam-
tory transmission (Blumcke et al., 2000a). In addition, upreg- pal tissue taken from patients with TLE (Mathern et al., 1999),
ulation of the presynaptic inhibitory metabotropic receptor although it is difficult to determine to what extent this reects
subunit mGluR4 in the dentate gyrus and granule cells in hip- selective neurodegeneration. In an animal model of hip-
pocampal specimens was also observed, which may contribute pocampal sclerosis there was downregulation of the glial glu-
to the dampening of seizure activity (Lie et al., 2000). In stud- tamate transporter that could contribute to glutamate
ies employing in situ hybridization techniques, increases in spillover but upregulation of the neuronal glutamate trans-
pyramidal and granule cell AMPA receptor mRNA (Mathern porter, which has been hypothesized to play a dominant role
et al., 1997) and in granule cell NMDAR1 and NMDAR2 sub- in reversed glutamate uptake (Ueda et al., 2001).
unit mRNA have been shown, results that are supported by Furthermore, marked extracellular glutamate rises have been
autoradiographic studies (Brines et al., 1997). Studies in kin- recorded in humans using in vivo microdialysis prior to
dled models have supported the hypothesis that during seizure onset, leading to the suggestion that these glutamate
epileptogenesis there is enhanced transmission from the rises are an initiating factor in spontaneous seizures (During
entorhinal cortex to dentate granule cells (Behr et al., 2001). and Spencer, 1993).
788 The Hippocampus Book

Other Neurotransmitters. Alterations of many other trans- stores (Mazarati et al., 1998). This would have a disinhibitory
mitter systems have been described in association with acute effect. Soon after the status epilepticus, however, galanin
limbic seizures and with hippocampal sclerosis. Perhaps one immunoreactive-positive neurons appeared in the hilus; they
of the most intriguing roles for many of these transmitters is increased in number after the rst day but gradually declined
in seizure termination, and acute alterations have been impli- a few days later. This increase in galanin-immunoreactive neu-
cated in the progression of seizures to status epilepticus. rons in the hippocampus is possibly a compensatory response
Adenosine, opioids, NPY, and galanin have all been proposed to prolonged ictal activity and depletion of galanin from sep-
to play an important role in seizure termination (Young tal afferents. Galanin injected into the hilus prevented the
and Dragunow, 1994; Mazarati et al., 1998, 1999; Vezzani et induction of status epilepticus and also stopped established
al., 1999), whereas accumulation of substance P has a pro- status epilepticus. Conversely, antagonists of galanin receptors
epileptogenic effect (Mazarati et al., 1999). facilitated the development of status epilepticus (Mazarati et
Adenosine is a potent inhibitor of neurotransmitter release al., 1998). Further conrmatory evidence of the importance of
and has been shown to be effective for terminating brief galanin comes from studies of transgenic mice in which over-
seizures. Indeed, accumulation of adenosine seems to be a expression of galanin had an antiepileptic effect, and galanin
credible contender for a prominent role in seizure termina- knockouts were more susceptible to the induction of status
tion, as seizures promote adenosine release (Berman et al., epilepticus (Mazarati et al., 2000).
2000). Adenosine antagonists shorten the stimulation proto-
col or lessen the chemoconvulsant dose necessary to induce Changes in Neuronal Properties
status epilepticus (Young and Dragunow, 1994). Also, adeno-
sine agonists are effective at stopping both the induction and Although most recent research into epileptogenesis and
maintenance of status epilepticus (Handforth and Treiman, seizure generation has concentrated on changes in the neu-
1994). To what degree changes in adenosine anticonvulsant ronal network, alterations in intrinsic neuronal properties
activity contribute to epileptogenesis or to the failure of could also contribute to this process. Importantly, ion channel
seizure termination (status epilepticus) is unknown. Although mutations in which there may be only subtle changes to the
regional changes in adenosine receptor density have been kinetics of ion channels can result in epilepsy.
described during epileptogenesis (Ekonomou et al., 2000), it As discussed in the section on the interictal spike, CA3
may reect cell loss and synaptic rearrrangement. pyramidal cells can generate burst ring, whereas few pyram-
Opioid release has also been suggested as a major mecha- idal cells in CA1 demonstrate such firing properties.
nism underlying the termination of seizures. The observed Alterations in intrinsic membrane properties can dramatically
loss of dynorphin-like immunoreactivity in the hippocampus affect the ring properties of such neurons and could promote
during sustained seizure activity is consistent with loss of burst ring. Such burst ring in a dense excitatory network
a potent endogenous antiepileptic (Mazarati et al., 1999). has the potential to generate synchronized bursts and may
Opioid antagonists facilitate the establishment of status thus promote epileptic activity. In the pilocarpine model of
epilepticus, and agonists inhibit both the induction and main- epileptogenesis, the proportion of bursting CA1 pyramidal
tenance of status epilepticus. cells increases dramatically, such that more than half demon-
In addition to opioids, a variety of other modulatory neu- strate bursting properties (Fig. 166) (Su et al., 2002). This
ropeptides exist. NPY is such a peptide that has potent effects may be due to upregulation of a T-type calcium channel that
on neurotransmission. Cloning has revealed ve NPY recep- can produce a signicant calcium tail current following an
tors, Y1Y5 (Vezzani et al., 1999). In human hippocampal action potential, resulting in signicant afterdepolarization
sclerosis, there is increased NPY, upregulated presynaptic Y2 (Su et al., 2002). Persistent sodium currents may also con-
receptors that inhibit neurotransmitter release, and downreg- tribute to this propensity for bursting.
ulated Y1 receptors that are expressed postsynaptically and are More recently, downregulation of dendritic A-type potas-
excitatory (Furtinger et al., 2001). Furthermore, Y5 receptor sium channels has been found in the pilocarpine epilepsy
knockout mice and NPY knockout mice have an exaggerated model (Bernard et al., 2004). This downregulation is partly
response to kainic acid with prolonged seizures (Baraban et due to increased channel phosphorylation by extracellular
al., 1997; Marsh et al., 1999). signal-regulated kinase but also to decreased transcription.
Galanin is another bioactive peptide that is widely distrib- These potassium channels limit the back-propagation of
uted throughout the CNS. Galanin in the hippocampus is pre- action potentials from the soma into the distal dendrites. The
dominantly inhibitory, decreasing the release of excitatory functional consequence of back-propagating action potentials
amino acids. In the hippocampus, galanin immunoreactivity is likely to be an amplicatioin of EPSPs and thus increased
is conned to axons, the bulk of which are the axons of medial excitation. The effect of downregulation of A-type potassium
septal neurons (Mazarati et al., 1998). Status epilepticus in two channels on dendritic calcium spikes and burst ring is
modelsperforant path stimulation and lithium pilo- unknown.
carpineresulted in the disappearance of galanin immunore- Epileptogenesis can thus lead to an acquired channelopa-
active bers in the hippocampus; this may have resulted from thy in neurons that may promote burst ring and hyperex-
loss of medial septal neurons or through exhaustion of galanin citability.
 Hippocampus in Human Disease 789

the hippocampus in epileptic activity. In association with neu-

ronal damage associated with hippocampal sclerosis, there is a
vast array of changes in organization, connectivity, receptors,
intrinsic neuronal properties, and astrocyte function that can
contribute to epileptogenicity. The great challenges are to dif-
ferentiate pro-epileptogenic changes from antiepileptogenic,
compensatory changes, and to determine which are the criti-
cal processes. It is likely that epileptogenesis is not a single
process but that many diverse processes can result in the
expression of epilepsy.

16.3 Alzheimers Disease

16.3.1 Introduction

Alzheimers disease (AD) is the most common cause of

dementia worldwide. AD is a neurodegenerative disorder in
which the accumulation of amyloid plaques and neurobril-
lary tangles represents the pathological hallmark of the dis-
ease. Five million people in the United States and 400,000 in
the United Kingdom are estimated to have AD, with a disease
onset typically occurring early in the eighth decade of life.
Disease prevalence increases with age: 1 in 2000 people aged
below 60 years are affected and 1 in 200 aged over 60 years. In
the elderly population the prevalence rises dramatically, with
AD occurring in 20% of those aged  80 years and in 50% of
those  90 years of age.
Memory loss is the predominant feature of AD. Impair-
ment of episodic memory occurs early during the course of
the disease and is usually the most prominent symptom
throughout the disease course. Progressive dysfunction of
other cognitive domains is also observed in AD, and the nal
stages of the disease are characterized by severe global impair-
\Figure 166. Upregulation of intrinsic burst-ring in CA1 pyrami- ment of cognitive function.
dal cells from SE-experienced versus control animals. A. Represen- The neuropathological changes in AD are thought to be
tative responses of CA1 pyramidal cells from sham-control (a) and
manifest initially in the entorhinal cortex (EC), progressing
SE-experienced animals (b) 30 days after pilocarpine treatment. In
from there to the hippocampus, with increasing involvement
each panel, the responses of CA1 pyramidal cells to long (leftmost
traces) and brief (rightmost traces) depolarizing current pulses of the neocortex as the disease progresses (Braak and Braak,
(injected through the recording microelectrode) are shown. Top and 1991). However, signicant neocortical pathology is already
bottom traces depict the neuronal response and the current stimu- present by the time dementia is clinically diagnosed, which
lus, respectively. B. Most of the CA1 pyramidal cells in the sham- has prompted efforts to identify the clinical manifestations of
control group (n  42) and in the naive-control group (n  15) AD in its earliest stages, when the pathological changes are
were regular ring cells, with only a small fraction displaying burst primarily restricted to the medial temporal lobe structures.
discharges to threshold stimulation (white bars). In contrast, CA1 This has resulted in the introduction of the concept of mild
pyramidal cells in the SE-experienced group showed a high inci- cognitive impairment (MCI), representing the predementia
dence of intrinsic bursting (54%, n  97), with bursting neurons stage of AD. Within the broad overview of AD provided in this
displaying all-or-none bursts of three to four clustered spikes in
chapter, particular attention is devoted to the involvement of
response to either brief or long current injection (Ab). (Source:
the hippocampal formation in AD and MCI and on the impli-
Adapted from Su et al., 2002. with permission.)
cations that improved identication of this involvement has
for earlier diagnosis and treatment of AD.
16.2.5 Conclusion
16.3.2 Clinical Features
The hippocampus through its physiological role can initiate
and maintain oscillatory behavior. It is this property along Memory impairment of insidious onset typies the initial
with the plasticity of the hippocampus and its vulnerability to stages of AD. Patients exhibit poor memory for autobiograph-
neuronal damage that contribute to the pathological role of ical events, current affairs, and the names and faces of acquain-
790 The Hippocampus Book

tances. Frequently they exhibit difficulty remembering familiar scribed and insufficiently severe to warrant a diagnosis of
routes, and getting lost is a commonly experienced symptom dementia. The introduction of drug treatments for AD, in the
early in the disease. These early symptoms may be attributed form of the acetylcholinesterase inhibitors, has helped stimu-
initially to cognitive decline as a consequence of normal aging; late efforts to identify this prodromal phase in AD, culminat-
but as the disease progresses the memory impairment becomes ing in the introduction of the concept of mild cognitive
severe enough to disrupt activities of daily living, and patients impairment (MCI) (Smith et al., 1996). Although there exists
begin to lose their functional independence. The central a degree of debate about the dening criteria for MCI, the
nature of the memory problem in AD is reected in the most widely accepted criteria for the diagnosis of MCI are
National Institute of Neurological and Communicative based on the presence of signicant objective memory decline
Diseases and Stroke/Alzheimers Disease and Related in the context of normal activities of daily living and intact
Disorders Association (NINCDS-ADRDA) criteria (McKhann function in other cognitive domains.
et al., 1984) for the diagnosis of probable AD, which state that Figures for the prevalence of MCI in the population vary
the presence of progressive memory impairment and impair- considerably, ranging from 3.0% to 16.8% depending on def-
ment of at least one other cognitive function are required for a initions (DeCarli 2003). Epidemiological data support the
clinical diagnosis of AD. However, atypical forms of AD are notion that MCI represents a precursor state of AD: Patients
well recognized, in which the prevailing symptomatology with MCI convert to AD at a rate of approximately 12% per
reects dysfunction of nonmemory cognitive domains, such as annum, compared with an annual rate of 1% to 2% in the
speech or visuoperceptual function, but these are much less normal elderly population.
common than the classic amnestic presentation. The current denition of MCI is predicated on impaired
Impairment of other cognitive functions becomes increas- memory. However, the heterogeneity in the clinical presenta-
ingly prominent during the course of the disease. Word- tion of AD suggests amnestic MCI that may represent a par-
nding difficulty is a relatively frequent early symptom, and ticular precursor state of AD. Future efforts to categorize the
speech output is often diminished. Executive functions, such initial clinical phases of AD will include the identication of
as problem solving and abstract reasoning, decline progres- equivalent MCI states affecting other cognitive domains and
sively. Later in the disease there is impairment of word com- will aim to differentiate MCI (due to AD) from cognitive
prehension, reecting a breakdown in semantic memory impairment as a consequence of other cognitive disorders, as
function. Other characteristic early symptoms include limb well as from the memory decline that accompanies the normal
apraxias (higher order motor disorders affecting the execution aging process. In this context, the fact that the pathological
of skilled or learned limb movements), impaired calculation damage in the earliest phases of AD is largely restricted to the
skills, and disorders of visuospatial and visuoperceptual func- entorhinal cortex and hippocampus may be used to direct
tion. The progressive involvement of multiple cognitive diagnostic investigations, including structural and functional
domains in AD reects the accumulation of pathological neuroimaging techniques and clinical neuropsychological
changes predominantly in the frontal, temporal, and parietal assessments of memory.
lobes; the occipital lobes are usually relatively spared in the
early stages of AD. Pattern of Cognitive Decits in AD
The disorders of cognitive function are accompanied by a
variety of neuropsychiatric symptoms, which include depres- The decit in memory that characterizes early AD is primarily
sion, changes in personality, delusions, and hallucinations, impaired episodic memory. Patients are unable to learn new
with depression often preceding diagnosis. Behavioral distur- material and have difficulty recalling recent events. A number
bances such as aggression, agitation, and nocturnal wandering of factors contribute to this disruption of episodic memory,
increase with disease severity. Loss of insight is common. In including deciencies in encoding and storing new informa-
the late stages of the disease, parkinsonian signs (e.g., limb tion and heightened sensitivity to the disruptive effects of
rigidity, motor slowing) and various involuntary limb move- proactive interference. AD patients also exhibit impaired
ments are observed, and seizures may also be observed in priming performance (facilitated performance by prior expo-
severe AD. Finally, patients become bedbound and entirely sure to stimuli); patients with AD, Huntingtons disease, and
dependent on caretakers, and in this state they are vulnerable Korsakoff syndrome are equally impaired on tests of verbal
to intercurrent medical complications such as sepsis. recognition and recall, but only AD patients show additional
From the time that AD is diagnosed initially, the disease impairment of verbal priming, indicating that AD is also asso-
runs approximately 5 to 10 years until the time of death, with ciated with a decit of implicit memory.
a mean survival of 8 years. Studies on the rate of long-term forgetting, or the rate of
loss of memory after successful learning, have yielded con-
Mild Cognitive Impairment icting results. Some studies have shown that AD patients
have faster forgetting rates than either controls or patients
There is increasing awareness that AD may be associated with with depression or Korsakoff syndrome (Hart et al., 1987), but
a prolonged prodromal or preclinical phase, during which others have failed to demonstrate accelerated forgetting in AD
the cognitive dysfunction is relatively subtle and circum- (Becker et al., 1987). Studies of retrograde amnesia in AD have
 Hippocampus in Human Disease 791

revealed that recent memories are more affected than remote with damage to the parietal association areas as well as dam-
memories. age to the premotor cortex and the supplementary motor
The diagnostic utility of memory tests in AD varies accord- areas.
ing to the severity of cognitive impairment. In the presympto- Visual agnosia (inability to recognize objects) is a common
matic phase of the disease, the tests currently considered to feature of AD in its more advanced stages and results from
predict with greatest accuracy the progression to AD are tests damage to the visual association areas. Subtypes of visual
of verbal learning and immediate visual recall. By contrast, for agnosia include apperceptive agnosia, in which the disorder of
established AD, tests of delayed recall are most sensitive at dif- object perception is exemplied by difficulty recognizing
ferentiating between early AD and normal controls but are of unusual views of common objects; it is typically associated
limited value in tracking disease severity in AD because per- with damage to the right parietal lobe. In associative agnosia,
formance on these tests often declines rapidly to a plateau. there is no perceptual decit but, instead, inability to assign
Recognition memory tests, involving verbal and visual sub- the correct semantic meaning to the perceived objects, result-
tests, are less sensitive than recall tasks for detecting early AD ing again in misidentication of objects. In this instance, the
but are more useful for staging disease severity. cortical damage most frequently involves the left occipitotem-
Disorders of speech and language also occur early in the poral region. More specific forms of agnosia include
course of AD. Word-nding difficulty is often the rst prob- prosopagnosia, in which there is impairment of familiar face
lem to become manifest, and it is associated with compensa- recognition, typically associated with damage to the right
tory circumlocution. Naming is initially preserved but temporal lobe.
becomes progressively more impaired during the course of the Together, amnesia, aphasia, apraxia, and agnosia form the
disease. Neologisms, verbal and literal paraphasias, also core disorders of cognitive function in AD. However, the
become more prominent and are accompanied by loss of global nature of the cortical involvement in established AD is
word comprehension, reflecting a breakdown of verbal reected in a multitude of additional cognitive decits,
semantic knowledge. Severe AD may be associated with prominent among which are disorders of attention and calcu-
palilalia (the repetition of words and phrases) and logoclonia lation. The involvement of the frontal lobes in the pathologi-
(repetition of the nal syllable of a word), and speech may cal process results in neurological symptoms such as
deteriorate into unintelligibility. Ultimately, some patients impaired executive function and reduced problem-solving
may become entirely mute. ability, as well as a variety of neuropsychiatric symptoms
Topographical disorientation is another characteristic early including changes in personality and disturbances of social
feature of AD. Patients complain of getting lost, initially only conduct.
in unfamiliar environments, but subsequently they experience
difficulty nding their way around familiar places, including Structural Imaging
their own homes. The presence of topographical disorienta-
tion may help differentiate early AD from other dementias; for Generalized cerebral atrophy is a characteristic gross patho-
instance, patients with frontotemporal lobe degeneration, logical feature of AD. The utility of cerebral atrophy as a bio-
typied by focal atrophy of the frontal and anterior temporal marker of disease is reected in the NINCDS-ADRDA
lobes, typically do not get lost. This topographical disorienta- diagnostic criteria (McKhann et al., 1984), which state that the
tion in AD has been variously ascribed to impaired visuospa- diagnosis of probable AD is supported by evidence of cere-
tial function as a result of damage to the parieto-occipital bral atrophy on CT or MRI and progression documented by
region; to topographical agnosia, representing an impairment serial observation. The presence of cerebral atrophy in AD
of the ability to recognize those cues or landmarks that are may be determined using a variety of techniques. Qualitative
required to permit successful navigation through an environ- assessment of brain atrophy (visual inspection of brain scans)
ment; and to an impaired memory for places as a consequence has the benet of general applicability but the disadvantage
of damage to the hippocampus or parahippocampal regions. of wide interobserver variability. Of the various quantitative
The last possibility is consistent with the observation that the techniques currently in use, volumetric analyses have
medial temporal lobe structures are preferentially affected in been shown to have greater diagnostic specicity and sensitiv-
the earliest stages of the AD disease process. ity than linear measurements of atrophy. Analyses may be
Apraxia (impairment of sensorimotor integration due to cross-sectional (i.e., based on a single scan) or longitudinal
disorders of higher cerebral function) is observed in most (repeated measurements performed over serial scans). The
patients with mild to moderate AD. Ideational apraxia, in diagnostic utility of data obtained from cross-sectional imag-
which there is an inability to construct the idea of a purpose- ing studies is limited by the variability in brain size among
ful movement, such that patients are unable to perform individuals, reecting differences in head size in the normal
these movements (e.g., using a manual tool), may occur as a population, and by the reduction in total brain volume that
result of damage to the parietal and frontal associations areas, occurs as a consequence of normal aging. Longitudinal stud-
particularly in the left hemisphere. Ideomotor apraxia, in ies have greater diagnostic specicity and sensitivity than
which the construct of a purposeful movement is intact but cross-sectional studies but are necessarily disadvantaged by
the execution of the movement is faulty, is associated more the need for at least two scans and as a consequence cannot
792 The Hippocampus Book

provide corroborative information at the time of the initial the degree of atrophy is classied according to a ve-point
diagnostic inquiry. grading scale following visual inspection of MRI scans
Most volumetric MRI studies have relied on manual seg- (Wahlund et al., 2000). Linear measures of atrophy include
mentation of brain regions of interest. Information on the measuring the height of the hippocampus and parahip-
structural brain changes in AD can also been obtained using pocampal gyrus, the interuncal distance, and the width of the
semiautomated, increasingly sophisticated techniques such as temporal horn of the lateral ventricles. Area measurements
voxel-based morphometry, in which the distribution of atro- include changes in the cross-sectional area through the hip-
phy in the AD brain is determined by comparison with a pocampus and in the surface area of the entorhinal cortex. Of
nonatrophied template brain (Good et al., 2002), and uid these various measurement techniques, perhaps the most use-
registration MRI, in which longitudinal patterns of atrophy ful in the clinical domain is that based on visual rating of
throughout the brain can be observed by tracking the brain medial temporal lobe atrophy, which has benets in terms of
changes over time on a voxel-by-voxel basis (Fox et al., 2001) ease of use and widespread applicability and compares favor-
(Fig. 167, see color insert). ably with quantitative volumtric analysis in the diagnostic dif-
Prior understanding of the pattern of pathological involve- ferentation of patients with AD.
ment of the cerebral cortex has prompted the assessment of
regional brain volume measurements in AD as an alternative Hippocampus. In recent years increasing emphasis has been
to measurements of whole-brain volume. Most studies have devoted to volumetric analyses of structural change derived
concentrated on the medial temporal lobe, but other using quantitative MRI techniques, with the hippocampus
researchers have investigated structures such as the frontal representing the primary region of interest in most instances.
lobes, the cingulate gyrus, the superior temporal gyrus, and The use of volumetric MRI measures of hippocampal atro-
the corpus callosum. As with the measurement of whole-brain phy as surrogate markers of disease in AD is validated by
atrophy, there are several methods for assessing atrophy of the the demonstration of a strong correlation between MRI-
medial temporal lobe structures. They include the use of a determined hippocampal volumes and neuronal numbers in
visual rating scale for medial temporal lobe atrophy, in which the hippocampus in AD (Bobinski et al., 1999). A number of

Figure 167. Coronal MRI at the mid-hippocampal level using voxel-compression mapping
overlay to show the change in brain volume over 12 months in a patient with Alzheimers dis-
ease (AD). Particular features to note are the marked involvement of the temporal lobes includ-
ing the hippocampi, the relative symmetry of the structural changes, and the diffuse
involvement of both gray and white matter.
 Hippocampus in Human Disease 793

studies have demonstrated that AD is associated with signi- cortex along the medial lip of the collateral sulcus), this tech-
cant hippocampal volume loss. Data from cross-sectional nique has been used to demonstrate signicant bilateral EC
studies are complemented by longitudinal data indicating that atrophy in established AD (Fig. 167, see color insert).
AD is associated with an increased rate of hippocampal atro- Longitudinal studies on AD patients, with volume measure-
phy (Jack et al., 1998). Although most of these studies have ments of the EC and the hippocampus performed on scans
concentrated on the structural changes in patients with estab- with an average scan interval of 21 months, reveal a signi-
lished dementia, hippocampal atrophy has also been found to cantly greater rate of atrophy affecting the EC (mean annual
be present in the presymptomatic stage of AD, as determined volume loss 7.1% per annum) than the hippocampus (mean
by scan data acquired on familial AD (FAD) patients prior to annual volume loss 5.9%), which would support the idea that
symptom onset (Convit et al., 1997). the pathological changes in AD are more severe in the EC than
Determination of hippocampal atrophy can distinguish in the hippocampus.
AD from normal aging with a high degree of specicity and
sensitivity. At a xed specicity of 80%, the sensitivity of hip- Structural Imaging in MCI. Neuroimaging studies reveal that
pocampal volumetric measurements for differentiating MCI is also associated with atrophy of the EC and the hip-
patients with mild-to-moderate AD is approximately 88% pocampus. Volumetric MRI analysis indicates that there is a
(Jack et al., 1997). progression of atrophy of these structures from normal aging
Several imaging techniques have been applied to the hip- through MCI to AD; the degree of atrophy in these structures
pocampus in recent years. MRI-based high-dimensional brain is sufficiently great to differentiate effectively between MCI
mapping (Csernansky et al., 2000) is a technique in which a patients and age-matched controls and between MCI and AD
control MRI template is transformed onto individual scans in patients, with signicantly greater atrophy observed in the lat-
such a way that changes in the shape of the hippocampus are ter group. The etiological relation between MCI and AD is
detectable. Using this technique, AD was found to be associ- underlined further by the observation that the presence of
ated with a symmetrical deformity of hippocampal shape that hippocampal atrophy in MCI patients is predictive of future
differentiated it from the morphological changes seen with conversion to AD (Jack et al., 1999).
normal aging. The particular deformities affecting the head of A comparison of hippocampal and EC volumes in normal
the hippocampus and along the lateral aspect of the body of control subjects, patients with MCI, and patients with early
the hippocampus are consistent with pathological changes AD revealed that assessment of EC volumes is most effective
affecting the CA1 eld in AD. A similar technique was subse- for discriminating between control subjects and MCI patients,
quently used by the same authors to demonstrate hippocam- whereas measurement of hippocampal volumes provided bet-
pal deformations in patients with mesial TLE (Hogan et al., ter discrimination between patients with MCI and those with
2004). AD, which suggests that atrophy of the EC precedes that of the
A technique has also been developed for visualizing neu- hippocampus and is more pronounced in the initial stages of
ritic plaques in autopsy-acquired human brain tissue using AD (Pennanen et al., 2004).
magnetic resonance (MR) microscopy, which provides greater
spatial resolution than MRI (Benveniste et al., 1999). Technical Medial Temporal Lobe Atrophy in Non-Alzheimer Dementias.
limitations currently militate against in vivo application, but Atrophy of medial temporal lobe structures discriminates
advances in scan technology may permit the future use of a effectively between AD and age-matched controls but is less
similar technique in the antemortem diagnosis of AD. effective at differentiating between AD and other diseases that
cause dementia. Hippocampal atrophy has been found in
Entorhinal Cortex. In view of the early, severe involvement of other neurodegenerative disorders such as dementia with
the EC in the AD pathological process, it has been argued that Lewy bodies and frontotemporal lobar degeneration (FTLD),
EC atrophy may be a more sensitive marker of AD than hip- as well as in vascular dementia. With regard to the latter,
pocampal atrophy. However, the theoretical benets of assess- the determination of hippocampal atrophy to differentiate
ing EC volume changes in AD are partially offset by the between different dementias is complicated further by the fre-
difficulty of determining with condence the boundaries of quent coexistence of AD and vascular pathology. With regard
the EC on MRI scans, which have led some to undertake to other temporal lobe structures, atrophy of the EC has also
measurements of the parahippocampal gyrus (which contains been observed in FTLD and, in particular the clinical subtype
the EC in its entirety) instead, with the presence of atrophy in of FTLD described as semantic dementia (SD). In this
this structure taken as a surrogate marker of EC atrophy. instance, the severity of EC atrophy exceeds that noted in AD
Despite these perceived difficulties, a technique for delineating patients of comparable disease severity, although the atrophy
the EC from volumetric MRI scans was developed (Insausti et is predominantly left-sided, in keeping with the language
al., 1998) using the cytoarchitectonic boundaries of the EC as dominance of the left hemisphere (Chan et al., 2001).
the guidelines for segmentation of the EC volume. In various The fact that the presence of hippocampal or EC atrophy
modied forms (modications in segmentation protocol alone is insufficiently specic to discriminate effectively be-
resulting primarily from the difficulty of establishing the tween these disorders suggests that greater diagnostic differ-
medial border of the EC at the junction with the perirhinal entiation may require evaluation of the particular distribution
794 The Hippocampus Book

of atrophy in these regions. One example of this is provided by puted tomography (CT) and MRI. To date, most of the func-
the left/right asymmetry of medial temporal lobe atrophy in tional imaging studies in AD have compared patients with AD
SD, which contrasts with the symmetrical atrophy that is typ- with age-matched normal control subjects rather than with
ical of AD. An alternative approach is to examine the distribu- patients with other neurodegenerative disorders.
tion of atrophy in regions of interest. For instance, in AD there
is an even distribution of atrophy along the rostrocaudal Positron Emission Tomography. Cerebral glucose metabolism
length of the hippocampus, whereas in SD there is asymmet- can be measured by positron emission tomography (PET)
rical atrophy affecting primarily the rostral portion of the left imaging of the radioactive tracer 18F-uorodeoxyglucose
hippocampus. (18F-FDG). PET can also be used to measure oxygen metabo-
lism or CBF using 15O2- or 15O2-labeled water. Statistical para-
Memory Impairment and Medial Temporal Lobe Atrophy in metric mapping (SPM) (Friston et al., 1995) is commonly
Alzheimers Disease. Decits of episodic memory correlate used to analyze scan data on a voxel-by-voxel basis.
with atrophy of the hippocampus but not with atrophy of Positron emission tomography (PET) scans in AD demon-
structures outside the medial temporal lobe, such as the cau- strate bilateral temporoparietal hypometabolism and hypop-
date nucleus and the lateral temporal cortex. Although differ- erfusion. The reductions in CBF and oxygen uptake have been
ing results have been observed across a number of studies, found to correlate with the severity of the dementia. A num-
partly as a consequence of differences in methodology and in ber of studies have demonstrated correlations between mem-
data interpretation, most studies have revealed that perform- ory scores and metabolism or blood ow in AD, and an
ance on tests of memory correlate with hippocampal volume association between hippocampal atrophy and regional glu-
but not with the volumes of the amygdala or of the whole cose metabolism has also been demonstrated. In advanced
temporal lobe. Hemispheric differences are also found; the AD, hypoperfusion changes are more widespread and addi-
volume of the left hippocampus has been found to correlate tional reductions in CBF are seen in the frontal lobes.
with verbal recall, whereas the volume of the right hippocam- The efforts to identify the earliest structural abnormality
pus correlates with performance on tests of visual or spatial in AD using MRI are mirrored by studies performed using
recall. Studies of the relation between hippocampal volume functional imaging paradigms. These have shown that hypop-
and immediate and delayed recall have yielded differing erfusion of the EC in cognitive normal elderly subjects is pre-
results: In one study, hippocampal volumes correlated with dictive of progression to MCI (de Leon et al., 2001), indicating
both recall tasks (de Toledo-Morrell et al., 2000), whereas in that abnormalities of brain function may be detected at the
another a positive correlation was observed between the vol- very earliest stages of the disease.
umes of the hippocampus and parahippocampal gyrus with The relation between the functional imaging data obtained
delayed, but not immediate, recall (Kohler et al., 1998). The from PET studies and data from structural imaging stud-
volume of the right parahippocampal gyrus was positively ies remains unclear. Although both imaging modalities have
associated with delayed visual recall. Hippocampal and identied abnormalities involving the EC in the earliest stages
parahippocampal gyrus volumes have also been found to cor- of AD, PET studies have also shown hypofunctioning of the
relate with a different verbal learning task (Libon et al., 1995). temporoparietal regions and the posterior cingulate cortex in
In summary, most studies attempting to correlate the early AD, whereas structural scans have revealed atrophy pre-
memory decits in AD with atrophy of specic brain regions dominantly affecting the medial temporal lobe regions at this
have implicated the hippocampus as the main region of inter- stage of the disease process. Given that these regions receive
est. The difficulty of establishing the nature of the memory signicant inputs from the hippocampal formation, it is pos-
task that provides the best correlation with hippocampal vol- sible that the PET data reect a disconnection syndrome, in
ume in AD are reminiscent of the problems experienced in the which decits are observed in regions with disturbed activity
clinical setting with identifying the memory tests that are due to reduced afferent input from damaged regions
most able to detect early AD. upstream in the projection. An alternative explanation for this
apparent discrepancy might rest with methodological consid-
Functional Imaging erations. With regard to the structural imaging data, repro-
ducible and easily validated protocols for the measurement of
The neuronal death and the dysfunction of surviving neurons brain volumes are primarily restricted to the temporal lobe
in affected brain regions in AD results in a reduction in neu- structures, within which atrophy is also readily detected on
ronal activity, which in turn produces an alteration in meta- visual inspection. By contrast, the anatomical landmarks of
bolic demands, with a lowering of glucose metabolism and regions such as the posterior cingulate gyrus are less easily
oxygen uptake. Cerebral blood ow (CBF) to affected brain identied on MRI, as a consequence of which volume loss in
regions is likewise reduced. Cerebral hypometabolism and these regions is more difficult to detect and may therefore be
hypoperfusion are detectable using various functional imag- underreported. In terms of the PET data, the absence of
ing techniques, and the information derived from functional observed hypometabolism in the hippocampus and parahip-
imaging complements the structural data obtained from com- pocampal regions in early AD may reect in part the low spa-
 Hippocampus in Human Disease 795

tial resolution of the current generation of PET scans. Finally, risk for AD exhibit different patterns of brain activation in the
the nature of the relation between regional brain atrophy and absence of any clear cognitive impairment.
detection of regional hypofunction has yet to be established Particular attention has been devoted to the regional acti-
fully for progressive degenerative disorders such as AD. vation patterns with the hippocampal formation in view of
In one study, Klunk et al. (2004) employed PET imaging the early pathological involvement of the hippocampus in AD.
using a novel tracer, named Pittsburgh compound-B (PIB), A comparison of the fMRI activation patterns in patients with
designed as a marker for brain amyloid. In patients with early AD and patients with isolated memory decline reveals that AD
AD, PET scans demonstrated increased retention of PIB (com- patients exhibit reduced activation in all hippocampal
pared with control subjects) in regions of association cortex regions. By contrast, patients with selective memory impair-
that are known to contain signicant numbers of amyloid ment either exhibited diminished activation in all hippocam-
deposits. By contrast, there was no signicant difference in PIB pal regions (similar to the pattern observed in AD) or reduced
retention between AD patients and controls in those brain activation affecting only the subiculum. In the latter cases, the
regions largely devoid of amyloid deposition, such as the cere- preferential involvement of the subiculum in these cases may
bellum. The evidence from this study that PET, in conjunction reect neuronal loss in the subiculum or loss of input to the
with targeted tracer compounds, may be used to provide quan- subiculum in patients at risk of developing AD.
tication of the pathological changes in AD raises the possibil- The observation that regional BOLD fMRI activation cor-
ity that PET may be used in the future to detect, and possibly relates with excitatory input, as manifested by the EPSP, rather
to track over time, the key pathological changes in AD. than the regional output (spiking activity), suggests that
reductions in regional fMRI activation patterns in AD may
Single Photon Emission Computed Tomography. Single pho- reect damage in upstream neuronal populations providing
ton emission computed tomography (SPECT) has an advan- excitatory inputs to the region in question (Logothetis et al.,
tage over PET in that it is less expensive and more widely 2001). However, several outstanding issues remain with regard
available, but it is less informative in that it provides only to the association of abnormal fMRI activation and structural
semiquantitative perfusion images and has poorer spatial pathology in AD. Specically, the relation between fMRI acti-
resolution. CBF is measured by detection of radioactive trac- vation and cerebral atrophy, as demonstrated on structural
ers such as the lipophilic technetium 99mTc-hexamethyl MRI, remains unclear, particularly in terms of the potential
propyleneamine oxamine (99mTc-HMPAO). There is a reduc- confounding effect of atrophy on fMRI activation patterns. In
tion in CBF in the temporoparietal regions of patients with addition, the presence of vascular pathology in AD may repre-
mild-to-moderate AD (Battistin et al., 1990), with a correla- sent another confounding factor in fMRI analysis in that
tion between the reduction in blood ow and hippocampal changes in activation may be attributable to alterations in
atrophy. However, the low diagnostic sensitivity and relatively hemodynamic response as well as to changes in neural activity.
poor spatial resolution of SPECT has limited its use as a diag-
nostic tool. Magnetic Resonance Spectroscopy. Detection of changes in
the concentration of brain metabolites using magnetic reso-
Functional MRI. The different magnetic properties of oxy- nance spectroscopy (MRS) represents an alternative imaging
genated and deoxygenated blood can be measured using the modality that can be applied to disease states affecting
technique of blood oxygen level-dependent (BOLD) func- the brain. Two metabolites that are of greatest interest are
tional MRI (fMRI), with increased levels of blood oxygenation N-acetyl aspartate (NAA) and myoinositol (MI), which are
resulting in greater signal intensity. As with PET, fMRI scan markers of neuronal and glial cell metabolism, respectively.
data can be analyzed using SPM. MRS has demonstrated levels of NAA that are reduced by 10%
During a verbal episodic memory task patients with mild to 15% in AD, with the magnitude of metabolite reduction
AD showed reduced activation of anterior prefrontal cortex correlating with disease severity. Other studies (Jessen et al.,
when compared with control subjects and, instead, exhibited 2000) have shown a 15% to 20% increase in MI, and a combi-
increased activation in a number of other brain regions, with nation of NAA and MI measurements increases the ability of
the latter believed to represent compensatory reallocation of MRS to differentiate AD from normal aging. At present, MRS
brain resources in response to the frontal lobe dysfunction. In is less useful for distinguishing AD from other neurodegener-
a cued recall task, AD patients failed to demonstrate any ative diseases, and the utility of this imaging technique beyond
increased activity in the left hippocampusa phenomenon the research domain has yet to be established.
observed in control subjectsbut, instead, were found to have
increased activity in other cortical regions. These observations 16.3.3 Genetics
of increased functional activation during cognitive tasks as
compensation for dysfunction of brain regions normally asso- Most AD cases occur in sporadic form, with familial AD
ciated with those tasks raises the possibility that fMRI may be (FAD) accounting for less than 5% of all cases. Apart from the
used in the diagnosis of AD. Functional brain changes have earlier age at onset (typically before the age of 65 years) no
also been detected prior to the clinical onset of AD; subjects at consistent differences in the clinical features of sporadic and
796 The Hippocampus Book

familial AD have been identied. This similarity in clinical observation that mice underexpressing PS1 exhibit a reduc-
presentation has underpinned the belief that greater under- tion in LTP following repeated tetanic stimulation of the CA1
standing of the defects occurring as a result of the genetic region (Morton et al., 2002) suggests that PS1 is also impli-
mutations associated with FAD will, in turn, yield key insights cated in the maintenance of LTP.
into the mechanisms of disease in AD.
Apolipoprotein E
Amyloid Precursor Protein
Linkage to chromosome 19 was observed in families with late-
Most cases in which the patients have early-onset AD (aged  onset FAD (Pericak-Vance et al., 1991). The responsible sus-
50 years) are attributable to familial forms of AD. Causative ceptibility gene was found to encode for apolipoprotein E
genetic mutations have been identied in these cases. The rst (ApoE), a 299-amino-acid lipid transport protein that medi-
reported FAD-associated mutations were those in the amyloid ates the intracellular uptake of lipids through binding to the
precursor protein (APP) on chromosome 21 (Chartier Harlin low density lipoprotein (LDL) receptor. Three alleles for the
et al., 1991). The exact function of APP remains undeter- ApoE gene exist: APOE- 2, APOE- 3 (the common form),
mined, with a role in growth promotion, signaling mecha- and APOE- 4. The likelihood of developing AD has been
nisms, and cell adhesion having been variously suggested found to correlate with the number of APOE- 4 genes
(Breen et al., 1991; Milward et al, 1992; De Strooper and (Saunders et al., 1993);  4 heterozygotes had a greater risk and
Annaert, 2000). APP is cleaved at its N- and C-termini by - earlier disease onset than non- 4 individuals, and  4 homozy-
and -secretases, respectively, to produce the peptide A , com- gotes had the greatest risk of all. Homozygous  4 patients were
prising 40 to 42 amino acids, which is the main constituent of also found to have a greater number of amyloid plaques than
the amyloid plaques in AD. Despite the uncertainty over the patients homozygous for the  3 allele. When compared with
role of APP, its role in the pathogenesis of AD appears age-matched normal controls (in whom the  4 allele is found
unequivocal; all currently identied mutations of the APP in 16% of cases), the presence of the  4 allele is more frequent
gene result in an increased amount of A or an increased pro- in both late-onset AD with a positive family history (52% of
portion of A containing 41 or 42 amino acids, which are all cases) and sporadic AD (40% of all cases).
more amyloidogenic and therefore predispose to the forma- One theory concerning the role of ApoE4 in the pathogen-
tion of amyloid plaques. esis of AD proposes that ApoE4 binds more easily to A than
ApoE3, resulting in increased deposition of amyloid in
Presenilins plaques. Another theory is based on the observation that
ApoE4 binds less well to the microtubule-associated protein
The APP mutations account only for a small proportion of tau than ApoE3 or ApoE2. This has the effect of destabilizing
early-onset FAD. Most of these cases are caused by mutations microtubules, which in turn results in the formation of neu-
of the PS1 gene on chromosome 14, coding for the protein robrillary tangles.
presenilin 1 (PS1). In all, PS1 mutations are responsible
for 30% to 50% of all cases of familial AD. Shortly after the 16.3.4 Pathophysiology
discovery of the PS1 gene, a second presenilin gene, PS2,
on chromosome 1 was identied (Li et al., 1995). Known PS2 Neuropathology
mutations account for less than 10% of all FAD cases. More
than 50 different mutations of PS1 have been described, On macroscopic examination, the AD brain can vary in
but to date only two causative PS2 mutations have been iden- appearance from normal to severely atrophic, with cases of
tied. early-onset AD often exhibiting the most marked atrophy.
The presenilins are transmembrane domain proteins. As Typically, the AD brain features widening of the sulci and ven-
with APP, the function of the presenilins remains unclear, tricular enlargement, which is most prominent in the lateral
although some clues can be derived from understanding the ventricles. There is generalized atrophy of the cerebral cortex,
function of the homologous proteins SEL-12 and SPE-4 in but closer inspection of the temporal lobes may reveal a
Caenorhabditis elegans. SEL-12 is involved in receptor traffick- greater degree of atrophy affecting the amygdala, hippocam-
ing and localization, mediated via the lin-12/Notch pathway, pus, and parahippocampal gyrus.
and SPE-4 plays a role in intracellular protein sorting during The denitive diagnosis of AD relies on the demonstration
spermatogenesis. The demonstration of a relation between of histological features that were rst described around the
PS1 and -secretase function (De Strooper et al., 1999) has turn of the twentieth century. Most prominent among them
fueled speculation that PS1 represents -secretase or that PS1 are the amyloid plaques and neurobrillary tangles, and their
and -secretase form part of a macromolecular complex that signicance is reected in the various published criteria for
mediates the cleavage of APP. Accordingly, presenilin muta- the pathological diagnosis of AD, which are based on an eval-
tions may alter the membrane conformation of APP, resulting uation of the frequency of neuritic plaques, on quantitative
in a different position of cleavage by -secretase; and thus assessment of both plaques and tangles, and most recently on
generation of the more amyloidogenic forms of A . The the number of tangles and neuropil threads in the cerebral
 Hippocampus in Human Disease 797

cortex. Other pathological changes in AD include granulovac- numbers of NFTs and NTs in the transentorhinal cortex, EC,
uolar degeneration and Hirano bodies, both of which prima- and CA1, with additional scant numbers of NFTs in CA4, the
rily affect the hippocampus, as well as amyloid angiopathy subiculum, and the parasubiculum. Small numbers of NFTs
and in severe AD mild spongiosis. Little is known about the and NTs are also found in association cortices. In stages V and
pathogenesis of granulovacuolar degeneration or Hirano VI (the isocortical stages) all hippocampal subelds and iso-
bodies or their relation to the natural history of AD, but the cortical association areas are severely affected. The progres-
predominance of hippocampal involvement in both instances sion of these pathological changes is depicted in Figure 168,
may provide another explanation for the prominence in AD of see color insert.
symptoms of hippocampal dysfunction.
Granulovacuolar Degeneration. In marked contrast to the
Amyloid Plaques. Extracellular amyloid plaques (APs) are widespread distribution of APs and NFTs, granulovacuolar
visualized best using silver stains or immunohistochemical degeneration is restricted primarily to one neuronal popula-
techniques using an antibody to A . APs are divided into two tion: the pyramidal cells of the hippocampus. This pathologi-
main types: diffuse plaques and neuritic plaques. Diffuse cal change is observed in up to 50% of AD cases. Vacuoles
plaques are composed of homogeneous deposits of brillary (35 m diameter) are found in the cytoplasm of the pyram-
material but contain only scant numbers of amyloid brils idal neurons, either singly or in combination. Within each
and do not stain with the Congo red stain for amyloid. vacuole is an electron-dense granular core. These features can
Neuritic plaques are more heterogeneous in composition, be seen on light microscopy using either silver staining or
with a central dense core of amyloid brils surrounded by hematoxylin and eosin (H&E) preparations.
glial and abnormally swollen neuritic processes, occasionally Granulovacuolar degeneration is not specic to AD. It is
containing paired helical laments. Neuritic plaques stain also a pathological feature of other neurodegenerative disor-
with Congo red. The two types of plaque differ crucially in ders such as amyotrophic lateral sclerosis (ALS) and the
that the A in neuritic plaques occurs in the form of insolu- Parkinson-dementiaALS complex of Guam and has been
ble, possibly neurotoxic -pleated sheets. APs are observed found in young adults with Downs syndrome.
mainly in the neocortex, with only small numbers seen in the
hippocampal formation during the early stages of AD; plaque Hirano Bodies. Hirano bodies are ovoid eosinophilic inclu-
density in the cortex increases with disease severity, although sions about 10 to 30 m in length. They can be visualized
the progression of AP deposition does not follow a clear hier- using the H&E stain. They are most commonly observed adja-
archical pattern (Arriagada et al., 1992). cent to the hippocampal pyramidal cells, when they indent the
neuronal perikaryon, although they can also be found in iso-
Neurobrillary Tangles. Neurobrillary tangles (NFTs) are lation in the stratum lacunosum. Electron microscopy reveals
found in the perikarya of neurons. They are stained with the that Hirano bodies are comprised of parallel laments 60 to
Bielschowsky silver stain (a stain for thioavine S) and by 100 nm in length.
immunohistochemistry using an antibody to the tau protein. Hirano bodies have been observed in many disorders as
They are commonly ame-shaped in appearance and occupy well as in the aged normal brain. Although they are most com-
the cell body and proximal portion of the apical dendrite of monly seen in the hippocampus, Hirano bodies have been
the affected neuron. NFTs are usually intraneuronal but occa- observed in most other structures of the CNS.
sionally extracellular and represent the insoluble remains of a
dead neuron (the ghost tangle). Ultrastructural examination Medial Temporal Lobe Pathology in Alzheimers Disease.
reveals that NFTs are primarily comprised of paired helical l- Entorhinal cortex: Within the EC, the stellate cells of layer II
aments, which are composed of a number of proteins includ- are the rst to exhibit NFTs and NTs. These cells are consis-
ing tau, -amyloid, ubiquitin, and neurolament proteins tently involved in AD. In one study of patients with patholog-
such as actin. ical diagnoses of denite AD, severe inltration of stellate cells
A hierarchical staging system for the neuropathological by NFTs was observed in 100% of cases (Hyman et al., 1990).
changes in AD has been elaborated based on the distribution These degenerative changes are accompanied by neuronal
of NFTs in the cerebral cortex (Braak and Braak, 1991). The loss, which is prominent even in the early stages of AD. By late
rst neurons to exhibit NFTs and neuropil threads (NTs) AD, severe loss of layer II cells in observed. Examination of the
straight and paired helical laments composed of abnormally perforant path reveals a number of associated changes. Myelin
phosphorylated tau proteinare the pre-alpha projection cuffing and argyrophilic degenerative changes are seen
neurons in the transentorhinal cortex, a transition zone throughout the course of the perforant path as well as in its
between the entorhinal cortex and the adjacent isocortex. termination zone in the outer two-thirds of the molecular
Other areas with early development of NFTs are the entorhi- layer of the dentate gyrus. Increased acetylcholinesterase
nal cortex (EC) and eld CA1 of the hippocampus. In stages I staining in the outer two-thirds of the dentate molecular layer
and II (the transentorhinal stages) these minor pathological suggests that there is partial cholinergic reinnervation of the
changes are restricted to the EC and hippocampus. Stages III hippocampus in response to the deafferentation of the dentate
and IV (the limbic stages) are characterized by moderate gyrus.
798 The Hippocampus Book

Figure 168. Distribution patterns of neurobrillary changes in in CA1 and the subiculum. The isocortex is minimally affected.
AD, including both neurobrillary tangles and neuropil threads. In stages V and VI the entire hippocampal formation is involved,
In stages I and II the pathological changes are largely restricted to with severe neuronal loss in CA1. The temporal lobes and cortical
the transentorhinal region (tangles may also be found in CA1). association areas are severely affected, but typically there is relative
Stages III and IV are characterized by severe involvement of the sparing of the primary sensory cortices. (Source: Courtesy of Dr.
transentorhinal and entorhinal regions, with additional changes Heiko Braak.)

Of the other layers of the EC, NFT formation is observed chical staging of AD pathology, the CA1 pyramidal cells are
in most of the pyramidal cells of layer IV. By contrast, signi- the rst hippocampal neurons to exhibit changes and, in fact,
cantly fewer NFTs are observed in layers III, V, and VI, represent the second neuronal population to be affected in
although in layer III the supercial layer of neurons is more AD, after the stellate cells of the EC. The nonpyramidal cells of
severely affected, compounding the disruption of perforant CA4 and the subicular neurons are next to be affected; the
path input to the hippocampus. Assessment of neuritic plaque granule cell layer of the dentate gyrus, CA3, and the pre-
density reveals a different pattern of laminar involvement, subiculum are involved only in the late stages of AD.
with most plaques observed in layer III. Layers IV, V, and VI Interestingly, NFTs are observed in the inner third of the den-
are associated with similar plaque density, but relatively few tate molecular layer in the most severely affected AD cases,
plaques are seen in layer II. Neither NFTs nor neuritic plaques which suggests that in the late disease stages there is additional
have been demonstrated in layer I. deafferentation of the input to the dentate gyrus from the hilar
Hippocampus: Of the hippocampal subelds, NFT and cells. Although the hippocampus is affected throughout its
AP density is greatest in CA1 and the subiculum. The CA1/ extent, morphometric studies have found that there is a pro-
subiculum interface zone is particularly affected, with large portional increase in the number of neurons exhibiting NFTs
number of plaques and tangles observed in all cases. In CA3 and granulovacuolar degeneration in the posterior hippocam-
and CA4, only small numbers of plaques and tangles are pus when compared with the distribution of changes observed
observed. The outer two-thirds of the molecular layer of the in the hippocampi of age-matched control subjects (Ball,
dentate gyrus is heavily inltrated by NFTs and neuritic 1987).
plaques, and NFTs are also seen in the dentate granule The degree of overall hippocampal pyramidal cell loss in
cells. Pathological changes in the mossy ber zone are negligi- AD has been estimated to be around 43% to 47%, with an
ble. In stark contrast to the subiculum, NFTs and plaques are increase in neuronal loss correlating with disease severity.
largely absent from the presubiculum. In terms of the hierar- Some disagreement exists with regard to the distribution of
 Hippocampus in Human Disease 799

neuronal loss affecting the hippocampus proper; some Comparison with the Pathological Changes of Normal Aging.
observers have reported that the greatest proportion of neu- The APs and NFTs are also observed in the brains of nonde-
ronal loss occurs in CA1, with additional neuronal loss mented aged individuals. Diffuse plaques are found through-
observed in CA4, the subiculum, and the prosubiculum and out the cerebral cortex, with additional plaques in the
relative sparing of the dentate granule cells and neurons in amygdala, EC, and CA1. Smaller numbers of neuritic plaques
CA3 and CA2 (Doebler et al., 1987; West et al., 1994). In a later are also found in these regions but with proportionally greater
study, no signicant difference was noted in the amount of quantities in the medial temporal lobe structures. NFTs are
cell loss in CA1 in AD and normal aging; instead, the greatest commonly present in the nondemented elderly brain and are
differences in neuronal numbers were observed in the granule most prominent in the hippocampus and parahippocampal
cell layer and the subiculum (Simic et al., 1997). An inverse regions (including the EC), with NFT numbers in CA1 corre-
correlation exists between hippocampal neuronal density and lating with age.
the number of neurons with NFT inltration or granulovac- These AD-like changes have led to the suggestion that some
uolar degeneration. In terms of the clinical signicance of the normal elderly subjects may in fact have covert, or preclini-
severity of hippocampal involvement, the degree of neuronal cal, AD. However, the observation that normal aging and AD
loss in CA1, CA4, and the subiculum is found to correlate with are associated with different patterns of neuronal loss in the
the duration and severity of AD. hippocampus suggest, instead, that the two do not share a
Amygdala: In the amygdala, NFT density is highest in common pathological substrate. The demonstration of
the accessory basal and cortical nuclei and lowest in the Alzheimer-type pathology in normal elderly individuals and
medial, lateral, and central nuclei. APs were most prominent the concomitant difficulty distinguishing these changes from
in the accessory basal and medial basal nuclei and least numer- those observed in the earliest stages of AD mirrors the prob-
ous in the medial, lateral, lateral basal, and central nuclei. lem in clinical practice with respect to the differentiation
Neuronal loss is greatest in the medial group of nuclei. The between individuals exhibiting minor cognitive decline in
projections between the amygdala and the hippocampal keeping with increasing age, and patients manifesting the ear-
formation are severely disrupted in AD: There is a prominent liest symptoms of AD.
afferent projection from the accessory basal nucleus both to
the hippocampus and to layer III of the EC. The return projec- Cholinergic Decit in AD
tion to the amygdala arises primarily from CA1, the subicu-
lum, and layer IV of the EC, all of which exhibit marked The cholinergic innervation to the hippocampal formation
pathological damage. arises from various components of the basal forebrain. The
medial septal nucleus and the nucleus of the diagonal band
Pathological Changes in MCI. Postmortem analysis reveals provide most of the inputs, with a smaller afferent projection
that the pathological changes associated with MCI primarily originating from the basalis of Meynert. By comparison, the
affect the EC and the hippocampus. Patients with very mild cerebral cortex receives its major cholinergic input from
dementia at the time of death exhibit severe neuronal loss pri- the basalis of Meynert, with additional lesser projections from
marily affecting layer II, with neuronal numbers being reduced the pedunculopontine and lateral dorsal nuclei. Cholinergic
by almost 60% (Gomez-Isla et al., 1996). In layer IV, there is a afferents are distributed to all cortical regions, but the limbic
40% reduction in neuronal numbers, whereas layers I, III, and and paralimbic cortices (including the parahippocampal
V are less affected. The degree of neuronal loss is greater in areas) are the recipients of a particularly strong projection. As
cases of severe dementia, with the drop in neuronal counts in a consequence, lesions of the basal forebrain in monkeys
layers II and IV rising to about 90% and 70%, respectively. result in widespread behavioral abnormalities, within which
These reductions in neuronal numbers are accompanied by disruption of memory functions are particularly prominent
increased deposition of NFTs and neuritic plaques. A compar- (Berger-Sweeney et al., 1994).
ison of the severity of EC pathology in MCI and AD reveals In AD, NPs and NFTs are observed in the basalis of
that AD is associated with greater volume loss affecting layer Meynert, the nucleus of the diagonal band, and the medial
II, but the absence of any corresponding decrease in the num- septal nuclei. There is marked neuronal loss in the nucleus
ber of layer II neurons indicates that the development of frank basalis and the nucleus of the diagonal band. The severity of
dementia may be associated with other structural changes, the neuropathological changes in the basalis of Meynert have
such as alterations in the extent of dendritic arborization. been found to correlate with clinical disease severity. The con-
The distribution of pathological damage in the EC in MCI comitant depletion of cortical cholinergic axons results in a
patients bears comparison with the initial stages of pathology reduction in the activity of choline acetyltransferase (ChAT)
described in AD (Braak stages I and II). Given that the onset in the cortex and diminished choline uptake in AD. ChAT
of frank dementia in AD is associated with more advanced activity is reduced by 60% in cortical biopsies obtained from
pathological involvement, equivalent to Braak stages III and patients with AD.
IV, it might reasonably be assumed that the syndrome of MCI The observation that in AD the basal forebrain nuclei are
represents the clinical correlate of the earliest pathological affected by plaques and tangles, in conjunction with the
(transentorhinal) stages of the AD disease process. reduction in cortical cholinergic innervation and the demon-
800 The Hippocampus Book

stration that disruption of the cholinergic system causes rations, but this was associated with an abnormally rapid
impaired learning and memory underpin the cholinergic decay function. In vivo studies using PDAPP mice have
hypothesis of AD. The hypothesis proposes that the cognitive demonstrated impaired induction and maintenance of LTP in
dysfunction associated with AD is at least partly attributable CA1 following high-frequency stimulation (Giacchino et al.,
to impairment of cholinergic neurotransmission. Support 2000). PDAPP mice also exhibited attenuation of paired-pulse
for the hypothesis comes from studies demonstrating that facilitation, indicating impaired presynaptic function in these
the reductions in ChAT activity and acetylcholine (ACh) animals. Although these changes were most prominent in
synthesis correlate with dementia severity in AD (Wilcock et aged transgenic mice, the demonstration of abnormalities in
al., 1982). However, the primary role of the cholinergic system young transgenic mice prior to the development of AD-
in the pathogenesis of AD is cast into question by the obser- related pathological changes indicates that defects of hip-
vation that there is relative preservation of the cholinergic pocampal synaptic transmission may precede the onset of
neurons of the nucleus basalis in MCI and early AD (Gilmor gross neurodegenerative changes.
et al., 1999). Furthermore, the cholinergic system is not selec-
tively affected in AD; pathological changes are also observed Mouse Models of AD
in a number of brain stem nuclei, including the locus
coeruleus, the ventral tegmental area, and the rostral raphe The discovery of the pathogenic APP mutation in 1991 stim-
nuclei. These nuclei, in turn, provide major components ulated efforts to create a mouse model in which the character-
of the noradrenergic, dopaminergic, and serotoninergic istic cognitive and pathological features were expressed.
innervation of the cerebral cortex, and it is likely that Gene-targeted knockout models provide information on the
their involvement contributes to the cognitive dysfunction possible actions of the proteins coded by the mutant genes,
observed in AD. whereas studies using transgenic mice explore the conse-
quences of the overexpression of mutant AD genes.
Impairment of Synaptic Function in AD
Knockout Models. APP knockout mice with functionally
Synaptic density microdensitometry performed on patholog- inactivated alleles of APP are observed to have mild impair-
ical specimens obtained from the frontal and temporal lobes ment of forelimb grip strength and decreased locomotor
has revealed signicant reductions in the density of presynap- activity associated with reactive gliosis. PS1 knockout mice
tic boutons in AD (approximately 60% of that observed were found to have disrupted development of the axial skele-
in control brains). The antemortem Mini Mental State ton as a result of impaired somitogenesis. Examination of the
Examination score, employed as a global measure of demen- brains of these mice reveals thinning of the ventricular zone
tia severity, was more closely correlated with synaptic density and severe regional neuronal loss. Cerebral hemorrhages were
than with amyloid plaque density or ChAT activity. Consistent seen in all mouse embryos. PS2 knockout mice were found to
with the known involvement of the entorhinal cortex in have mild pulmonary brosis and hemorrhage but no patho-
AD, there is a reduction in the markers for synaptic vesicle logical brain changes. Absence of PS2 did not affect APP pro-
proteins in the termination zone of the perforant path in cessing. However, the double homozygous PS1/PS2 knockout
the outer molecular layer of the dentate gyrus. This is accom- mice are more severely affected than the PS1 mice, suggesting
panied by a reduction in synaptic density in the inner molec- that PS1 and PS2 have partially overlapping functions.
ular layer, although this is associated with an expansion in the
size of the remaining synapses, resulting in the maintenance of Transgenic Models. Several lines of transgenic mice are cur-
the total synaptic contact area in this region (Scheff et al, rently in existence. The PDAPP mice overexpressing V717F
1996). -APP (an FAD-associated mutation) exhibit amyloid
Abnormalities of synaptic transmission in early AD have plaques with dystrophic neurites surrounding -amyloid
been demonstrated using in vitro and in vivo preparations. In cores. Transgenic mice with overexpression of human APP695
vitro studies using hippocampal slices prepared from PDAPP (a 695-amino-acid length APP isoform representing one of
transgenic mice overexpressing human mutant amyloid pre- the more abundant APP isoforms in the AD brain and a
cursor protein have demonstrated alterations in synaptic potential source of the A peptide) have abnormally high lev-
transmission and LTP (Larson et al., 1999). Enhanced paired- els of A and -amyloid deposits in the amygdala, hippocam-
pulse facilitation of synaptic transmission was observed in pus, and cortex between 6 and 12 months of age. Neither of
slice preparations taken from young PDAPP mice. In addi- these two lines of mice was found to have signicant neuronal
tion, there was a small (about 10%) reduction in the size of loss. None of the APP transgenic mice developed NFTs,
the CA1 dendritic EPSPs following stimulation of the Schaffer although abnormally phosphorylated tau immunoreactivity
collaterals and commissural bers. By contrast, slice experi- has been observed. Decits of spatial memory are noted by 9
ments performed on slices taken from aged mice revealed to 10 months of age, and APP transgenic mice also exhibit
diminution, rather than enhancement, of paired-pulse facili- impaired LTP (Chapman et al., 1999). PDAPP mice have been
tation and a marked (around 55%) reduction in CA1 eld found to exhibit an age-related decit in learning a succession
EPSPs. LTP could be induced in both young and aged prepa- of spatial locations in the watermaze (Chen et al., 2000). By
 Hippocampus in Human Disease 801

contrast, object recognition was normal. In these mice, trations in the synaptic cleft. Clinical trials have demonstrated
impaired performance on the watermaze task was found to that usage of these drugs results in improved cognitive func-
correlate with amyloid plaque density. tion. AChE inhibitors are currently licensed for the treatment
The PS1 transgenic mice have elevated levels of A 142. In of mild to moderate AD. There is accumulating evidence that
view of the fact that patients with familial AD due to PS1 they also attenuate the progression of noncognitive symptoms
mutations have amyloid plaques comprised primarily of this in AD, such as agitation and aggression, and may serve to
longer A peptide, these observations provide further evi- maintain activities of daily living. However, meta-analyses of
dence that PS1 is involved in this aspect of plaque deposition. randomized clinical trials relating to the use of AChE
Earlier studies involving APP and PS1 transgenic mice inhibitors indicate that these drugs provide only a modest
were able to provide reliable models of amyloid deposition, benet in AD (Doody et al, 2001; Kaduszkiewicz et al., 2005),
but the absence of any signicant tau pathology meant that and to date there is little evidence that these drugs have the
these studies were of limited value as realistic models of the capacity to alter the natural history of the disease.
human AD disease process. However, subsequent research has AChE inhibitors are at present the only drugs licensed for
now provided evidence of a causal association between amy- the treatment of mild to moderate AD. However, other drugs
loid and tau pathology. The observation that bigenic mice are being developed that are designed to enhance the cholin-
expressing both mutant APP and mutant tau are found to ergic system by alternative means. They include ACh agonists,
have signicantly greater quantities of intracytoplasmic tau m1 muscarinic receptor agonists, and stimulators of ACh
tangles in the limbic system and olfactory cortex than simi- release. There have been some reports that cigarette smoking
larly aged mice expressing the mutant tau gene alone can be may have a protective effect in AD, which may relate to the
taken as evidence that that the formation of NFTs is inu- effect on nicotinic receptors, but these ndings have yet to be
enced by amyloid protein (Lewis et al., 2001). Furthermore, fully substantiated. In addition, any potentially benecial
injection of A 142 brils into the hippocampi of transgenic effect on cholinergic transmission is at least partly offset by
mice expressing the mutant P301L tau gene results in a ve- the increased risk of cerebrovascular disease associated with
fold increase in the number of NFTs in the amygdala (Gotz et smoking, which increases not only the risk of AD per se but
al., 2001), which suggests that NFT deposition may be driven also the risk of developing concomitant vascular dementia.
by A 142. These data represent a signicant advance in our
understanding of the underlying nature of the neurodegener- NMDA Receptor Antagonists
ation of AD, particularly in terms of the link between the two
key aspects of AD pathology: -amyloid deposition and tan- Enhancement of glutamate-mediated synaptic function has
gle formation. been implicated in the pathophysiology of several neurode-
generative disorders including AD, Huntingtons disease,
16.3.5 Treatment Options and Parkinsons disease. The aminoadamantane compounds
amantadine and memantine have been found to confer neu-
The introduction of pharmacological treatments for AD dur- roprotection by noncompetitive inhibition of glutamate activ-
ing the last decade has resulted in a fundamental change in the ity at NMDA receptor. The low affinity of both compounds
approach to the clinical management of a condition previously for the NMDA receptor and their fast voltage-dependent
considered to be associated with an inexorable and unalterable channel unblocking kinetics have been cited as the explana-
decline. At present, cholinesterase inhibitors are licensed for tion for their low toxicity and good clinical tolerance. This
use in Europe and the United States, and the NMDA antago- contrasts with the neurotoxicity and psychotogenicity associ-
nist memantine is currently licensed for use in certain ated with compounds that are more potent NMDA receptor
European countries. Both treatment options represent symp- antagonists, such as MK-801 and phencyclidine. A clinical
tomatic therapy aimed at attenuating the rate of cognitive and trial involving the use of memantine in patients with severe
functional decline by enhancing synaptic function. However, AD has demonstrated a good safety prole and clinical
neither of these licensed treatments has been demonstrated to improvement as measured using cognitive and functional
exert any effect on the underlying pathological process and so assessment scores (Winblad and Poritis, 1999). As a conse-
neither is likely to affect the natural history of the disease. As a quence, memantine has been licensed for use in patients with
consequence, efforts have been directed toward the develop- moderately severe to severe AD.
ment of treatments that may interrupt the disease process, in
the anticipation that any disease-modifying drug may prove Antiamyloid Immunotherapy
successful in altering the natural history of AD.
PDAPP mice immunized with A 142 at 6 weeks of age, prior
Enhancement of Cholinergic Function to the development of pathological damage, developed anti-
A 142 antibodies and were observed to have signicant reduc-
Therapeutic strategies aimed at redressing the cholinergic tions in amyloid deposition and plaque formation (Fig. 169)
decit have led to the development of AChE inhibitors, which (Schenk et al., 1999). Injections of A 142 into older mice, in
block the breakdown of ACh, thereby increasing ACh concen- whom AD pathological changes were established, resulted in
802 The Hippocampus Book

Figure 169. Hippocampal A deposition, neuritic plaque forma- labeled with the APP-specic monoclonal antibody 8ES were found
tion, and astrocytosis in 13-month-old mice injected with phos- in the hippocampal sections of the PBS-injected (c) but not the
phate-buffered saline (PBS) and A 42. There is marked A A 42-injected mice (d). Plaque-associated astrocytosis is abundant
deposition in the outer molecular layer of the dentate gyrus of in the retrosplenial cortex of the PBS-injected (e) but not the A 42-
the PBS-injected mice (a), which contrasts with the absence of injected (f) mice. (Source: Schenk et al., 1999, with the permission
A observed in the A 42-injected mice (b). Dystrophic neurites of Dr. Dale Schenk and Nature Publishing Group.)

reduced neuritic plaque burden and associated neuritic dys- reduced amyloid plaque density when compared with nonim-
trophy and gliosis compared with nonimmunized mice. The munized mice.
pattern of amyloid pathology in the hippocampus was strik- Although initial safety trials of antiamyloid immunization
ingly different in the treated mice, with an absence of diffuse indicated that the treatment could be used in human subjects,
deposits and an altered pattern of A immunoreactivity. a subsequent Phase II safety trial was terminated following
Transgenic mice expressing the mutant human APP695 the development of subacute meningoencephalitis in 5% of
transgene that were vaccinated with the A 142 peptide exhib- the study population. Interestingly, a pathological study of the
ited a reduction of impairment when tested ozn a reference brain of the rst immunized patient to come to postmortem
memory version of the watermaze task over a number (Nicoll et al., 2003) revealed large areas of cerebral cortex con-
of weeks subsequent to immunization (Janus et al., 2000). taining very few amyloid plaques and plaque-associated dys-
However, immunized transgenic mice remained impaired trophic neurites, although quantities of NFTs and neuropil
compared with nontransgenic control mice. As with the threads were comparable to those observed in nonimmunized
PDAPP mice, A 142 immunization was associated with cases. Microglial cells were also found to be associated with
 Hippocampus in Human Disease 803

immunoreactivity to A , indicating that the immunization patients with mild AD. The rst published reports have indi-
had been successful in generating an appropriate immune cated that patients on statin therapy experience a small, but
response, which in turn resulted in clearance of amyloid not signicant, reduction in the rate of cognitive decline over
plaques. a 6-month period.
Certain metals have been implicated in the AD disease
Secretase Inhibitors process. Zinc and copper cations play a critical role in the
aggregation of amyloid; in addition, the combination of
Whereas the vaccination approach is designed to clear amy- these cations with amyloid results in oxidative damage as a
loid from the brain, an alternative approach is to prevent for- consequence of the generation of hydrogen peroxide. Clinical
mation of amyloid plaques by interfering with the production trials involving the use of metal chelators (agents that bind to
of A 142 from its precursor protein. With this aim in mind, metal ions) are currently under consideration (Scarpini et al,
attention has been focused on the discovery of potential 2003).
inhibitors of the - and -secretase enzymes that cleave APP Ultimately, any drug with true disease-modifying potential
to form A 142. Despite extensive animal studies, at present no must fulll a number of core criteria. First, a reduction in the
secretase inhibitor has been put forward openly for clinical rate of clinical decline must be demonstrated over several
trials. years, in view of the duration of the disease. Second, treatment
benets would have to be observed beyond the period of dos-
Other Therapeutic Options ing to exclude the possibility of a symptomatic effect only.
Ideally, benets would be noted not only in memory and other
Information derived from epidemiological observations and
cognitive functions but also in terms of more global measures
from research into factors inuencing the development of AD
such as activities of daily living. Finally, it would be desirable
pathological changes have stimulated trials with a number of
to supplement these clinical improvements with some evi-
treatments with potentially disease-modifying effects.
dence of attenuation of disease progression. In the absence of
Following the initial report in 1990 of a reduced incidence of
the ability to monitor directly the effect of treatments on the
AD in patients with arthritis taking nonsteroidal antiinam-
pathological brain changes, various surrogate markers of dis-
matory drugs (NSAIDs), a number of other reports have
ease progression have been proposed, including structural
drawn attention to the apparent protective effect of long-term
(regional and global cerebral atrophy) and functional (cere-
NSAID use (in tVeld et al, 2001). The underlying mechanism
bral hypometabolism) markers. Signicant, consistent alter-
for this neuroprotection has not been established, but the
ations in the longitudinal measurement of these surrogate
observation that amyloid plaques are surrounded by immune
markers of disease, observed over meaningful time spans,
cells (e.g., microglia) suggests that NSAIDs may serve to
would provide convincing evidence of disease modication.
reduce the degree of immune-mediated neuronal destruction.
The discovery of multiple potential treatment avenues has
An alternative explanation relates to the suppression by
markedly altered the clinical approach to AD. The anticipated
NSAIDs of oxygen free radical-mediated cellular damage.
development of genetic models of disease that more accu-
Finally, there may also be a direct effect on the key underlying
rately reect the core aspects of the AD disease process in con-
pathological processes in AD; some NSAIDs have been
junction with advances in our understanding of the
demonstrated to suppress A 142 formation. Although the epi-
pathophysiology of the disease are likely to give rise in the
demiological data are convincing, conclusions arising from
future to treatments that have the potential not only for effect-
clinical trials assessing the value of these drugs in AD must be
ing symptomatic alleviation but also for slowing downper-
tempered by the known prole of adverse effects associated
haps even arrestingthe pathological progression of
with these drugs, particularly the risk of peptic ulceration and
Alzheimers disease.
gastrointestinal bleeding.
Other long-term studies have documented the benecial
effects of compounds with antioxidant properties; most
prominent among them are gingko biloba and vitamins C and ACKNOWLEDGMENTS
E. Preliminary studies have indicated that all three prepara-
We thank the editors for their help in preparing this chapter. We also
tions may delay the progression of AD. Similar benets have
thank Eberhard Buhl and Dimitri Kullmann for their helpful com-
been shown for estrogen preparations, but progress in clinical
ments and criticisms on an earlier version.
trials has been delayed following the demonstration of an
increased risk of breast cancer, stroke, and myocardial infarc-
tion in a large study using hormone replacement therapy
REFERENCES
(Rapp et al, 2003).
The observation that individuals with high serum choles- Adams B, Sazgar M, Osehobo P, Van der Zee CE, Diamond J,
terol levels are more likely to develop AD, coupled with the Fahnestock M, Racine RJ (1997) Nerve growth factor accelerates
results from some studies suggesting that cholesterol pro- seizure development, enhances mossy ber sprouting, and
motes the formation of amyloid, has led to clinical trials attenuates seizure-induced decreases in neuronal density in the
involving the use of statins (cholesterol-lowering agents) in kindling model of epilepsy. J Neurosci 17:52885296.
804 The Hippocampus Book

Ajmone-Marsan C, Zivin LS (1970) Factors related to the occurrence Baulac M, De Grissac N, Hasboun D, Oppenheim C, Adam C,
of typical paroxysmal abnormalities in the EEG records of Arzimanoglou A, Semah F, Lehericy S, Clemenceau S, Berger B
epileptic patients. Epilepsia 11:361381. (1998) Hippocampal developmental changes in patients with
Ala TA, Beh GO, Frey WH (2000) Pure hippocampal sclerosis: a rare partial epilepsy: magnetic resonance imaging and clinical
cause of dementia mimicking Alzheimers disease. Neurology aspects. Ann Neurol 44:223233.
54:843848. Baulac S, Picard F, Herman A, Feingold J, Genin E, Hirsch E,
Alger BE, Nicoll RA (1980) Epileptiform burst afterhyperpolariza- Prudhomme JF, Baulac M, Brice A, LeGuern E (2001) Evidence
tion: calcium-dependent potassium potential in hippocampal for digenic inheritance in a family with both febrile convulsions
CA1 pyramidal cells. Science 210:11221124. and temporal lobe epilepsy implicating chromosomes 18qter
Amaral DG, Campbell MJ (1986) Transmitter systems in the primate and 1q25-q31. Ann Neurol 49:786792.
dentate gyrus. Hum Neurobiol 5:169180. Becker JT, Boller F, Saxton J, McGonigle-Gibson KL (1987) Normal
Annegers JF, Hauser WA, Shirts SB, Kurland LT (1987) Factors prog- rates of forgetting of verbal and non-verbal material in
nostic of unprovoked seizures after febrile convulsions. N Engl J Alzheimers disease. Cortex 23:5972.
Med 316:493498. Behr J, Heinemann U (1996) Low Mg2 induced epileptiform activ-
Arellano JI, Munoz A, Ballesteros-Yanez I, Sola RG, DeFelipe J (2004) ity in the subiculum before and after disconnection from rat
Histopathology and reorganisation of chandelier cells in the hippocampal and entorhinal cortex slices. Neurosci Lett
human epileptic sclerotic hippocampus. Brain 127:4564. 205:2528.
Arnold SE, Franz BR, Gur RC, Gur RE, Shapiro RM, Moberg PJ, Behr J, Heinemann U, Mody I (2001) Kindling induces transient
Trojanowski JQ (1995) Smaller neuron size in schizophrenia in NMDA receptor-mediated facilitation of high-frequency input
hippocampal subfields that mediate cortical-hippocampal in the rat dentate gyrus. J Neurophysiol 85:21952202.
interactions. Am J Psychiatry 152:738748. Bekenstein JW, Lothman EW (1993) Dormancy of inhibitory
Arriagada PV, Growdon JH, Hedley Whyte T, Hyman BT (1992) interneurons in a model of temporal lobe epilepsy. Science
Neurobrillary tangles but not senile plaques parallel duration 259:97100.
and severity of Alzheimers disease. Neurology 42:631639. Benes FM, Kwok EW, Vincent SL, Todtenkopf MS (1998) A reduction
Arruda F, Cendes F, Andermann F, Dubeau F, Villemure JG, Jones- of nonpyramidal cells in sector CA2 of schizophrenics and
Gotman M, Poulin N, Arnold DL, Olivier A (1996) Mesial atro- manic depressives. Biol Psychiatry 44:8897.
phy and outcome after amygdalohippocampectomy or Bengzon J, Kokaia Z, Elmer E, Nanobashvili A, Kokaia M, Lindvall O
temporal lobe removal. Ann Neurol 40:446450. (1997) Apoptosis and proliferation of dentate gyrus neurons
Assaf SY, Chung SH (1984) Release of endogenous Zn2 from brain after single and intermittent limbic seizures. Proc Natl Acad Sci
tissue during activity. Nature 308:734736. USA 94:1043210437.
Auer RN, Siesjo BK (1988) Biological differences between ischemia, Benveniste H, Einstein G, Kim KR, Hulette C, Johnson GA (1999)
hypoglycemia, and epilepsy. Ann Neurol 24:699707. Detection of neuritic plaques in Alzheimers disease by mag-
Babb TL, Brown WJ, Pretorius J, Davenport C, Lieb JP, Crandall PH netic resonance microscopy. Proc Natl Acad Sci USA
(1984) Temporal-lobe volumetric cell densities in temporal- 96:1407914084.
lobe epilepsy. Epilepsia 25:729740. Berger-Sweeney J, Heckers S, Mesulam MM, Wiley RG, Lappi
Babb TL, Kupfer WR, Pretorius JK, Crandall PH, Levesque MF (1991) DA, Sharma M (1994) Differential effects on spatial navigation
Synaptic reorganization by mossy bers in human epileptic fas- of immunotoxin-induced cholinergic lesions of the medial
cia dentata. Neuroscience 42:351363. septal area and nucleus basalis magnocellularis. J Neurosci
Ball MJ (1978) Topographical distribution of neurobrillary tangles 14:45074519.
and granulovacuolar degeneration in hippocampal cortex of Berman RF, Fredholm BB, Aden U, OConnor WT (2000) Evidence
aging and demented patients: a quantitative study. Acta for increased dorsal hippocampal adenosine release and metab-
Neuropathol (Berl) 42:7380. olism during pharmacologically induced seizures in rats. Brain
Baraban SC, Hollopeter G, Erickson JC, Schwartzkroin PA, Palmiter Res 872:4453.
RD (1997) Knock-out mice reveal a critical antiepileptic role for Bernard C, Anderson A, Becker A, Poolos NP, Beck H, Johnston D
neuropeptide Y. J Neurosci 17:89278936. (2004) Acquired dendritic channelopathy in temporal lobe
Baram TZ, Gerth A, Schultz L (1997) Febrile seizures: an appropri- epilepsy. Science 305:532535.
ate-aged model suitable for long-term studies. Brain Res Dev Bernasconi N, Bernasconi A, Caramanos Z, Dubeau F, Richardson J,
Brain Res 98:265270. Andermann F, Arnold DL (2001) Entorhinal cortex atrophy in
Barbarosie M, Avoli M (1997) CA3-driven hippocampal-entorhinal epilepsy patients exhibiting normal hippocampal volumes.
loop controls rather than sustains in vitro limbic seizures. J Neurology 56:13351339.
Neurosci 17:93089314. Billinton A, Baird VH, Thom M, Duncan JS, Upton N, Bowery NG
Barnard EA, Skolnick P, Olsen RW, Mohler H, Sieghart W, Biggio G, (2001) GABA(B) receptor autoradiography in hippocampal
Braestrup C, Bateson AN, Langer SZ (1998) International sclerosis associated with human temporal lobe epilepsy. Br J
Union of Pharmacology. XV. Subtypes of gamma-aminobutyric Pharmacol 132:475480.
acid(A) receptors: classication on the basis of subunit struc- Blumcke I, Beck H, Nitsch R, Eickhoff C, Scheffler B, Celio MR,
ture and receptor function. Pharmacol Rev 50:291313. Schramm J, Elger CE, Wolf HK, Wiestler OD (1996)
Battistin L, Pizzolato G, Dam M, Ponza I, Borsato N, Zanco PL, Preservation of calretinin-immunoreactive neurons in the hip-
Ferlin G (1990) Regional cerebral blood ow study with 99mTc- pocampus of epilepsy patients with Ammons horn sclerosis. J
hexamethyl-propyleneamine oxime single photon emission Neuropathol Exp Neurol 55:329341.
computed tomography in Alzheimers and multi-infarct Blumcke I, Zuschratter W, Schewe JC, Suter B, Lie AA, Riederer BM,
dementia. Eur Neurol 30:296301. Meyer B, Schramm J, Elger CE, Wiestler OD (1999a) Cellular
 Hippocampus in Human Disease 805

pathology of hilar neurons in Ammons horn sclerosis. J Comp Cain DP, Boon F, Hargreaves EL (1992) Evidence for different neuro-
Neurol 414:437453. chemical contributions to long-term potentiation and to kin-
Blumcke I, Beck H, Suter B, Hoffmann D, Fodisch HJ, Wolf HK, dling and kindling-induced potentiation: role of NMDA and
Schramm J, Elger CE, Wiestler OD (1999b) An increase of hip- urethane-sensitive mechanisms. Exp Neurol 116:330338.
pocampal calretinin-immunoreactive neurons correlates with Cavazos JE, Das I, Sutula TP (1994) Neuronal loss induced in limbic
early febrile seizures in temporal lobe epilepsy. Acta Neuropathol pathways by kindling: evidence for induction of hippocampal
(Berl) 97:3139. sclerosis by repeated brief seizures. J Neurosci 14:31063121.
Blumcke I, Becker AJ, Klein C, Scheiwe C, Lie AA, Beck H, Waha Cendes F, Cook MJ, Watson C, Andermann F, Fish DR, Shorvon SD,
A, Friedl MG, Kuhn R, Emson P, Elger C, Wiestler OD (2000a) Bergin P, Free S, Dubeau F, Arnold DL (1995) Frequency and
Temporal lobe epilepsy associated up-regulation of metabo- characteristics of dual pathology in patients with lesional
tropic glutamate receptors: correlated changes in mGluR1 epilepsy. Neurology 45:20582064.
mRNA and protein expression in experimental animals and Chan D, Fox NC, Scahill RI, Crum WR, Whitwell JL, Leschziner G,
human patients. J Neuropathol Exp Neurol 59:110. Rossor AM, Stevens JM, Cipolotti L, Rossor MN (2001) Patterns
Blumcke I, Suter B, Behle K, Kuhn R, Schramm J, Elger CE, Wiestler of temporal lobe atrophy in semantic dementia and Alzheimers
OD (2000b) Loss of hilar mossy cells in Ammons horn sclero- disease. Ann Neurol 49:433442.
sis. Epilepsia 41:S174S180. Chandler KE, Princivalle AP, Fabian-Fine R, Bowery NG, Kullmann
Blumcke I, Schewe JC, Normann S, Brustle O, Schramm J, Elger CE, DM, Walker MC (2003) Plasticity of GABA(B) receptor-
Wiestler OD (2001) Increase of nestin-immunoreactive mediated heterosynaptic interactions at mossy bers after status
neural precursor cells in the dentate gyrus of pediatric patients epilepticus. J Neurosci 23:1138211391.
with early-onset temporal lobe epilepsy. Hippocampus 11: Chapman PF, White GL, Jones MW, Cooper-Blacketer D, Marshall VJ,
311321. Irizarry M, Younkin L, Good MA, Bliss TV, Hyman BT, Younkin
Bobinski M, de Leon MJ, Wegiel J, De Santi S, Convit A, Wisniewski SG, Hsiao KK (1999) Impaired synaptic plasticity and learning
HM (1999) The histologic validation of MRI hippocampal vol- in aged amyloid precursor protein transgenic mice. Nat Neurosci
ume measurements in Alzheimers disease. Neuroscience 2:271276.
95:721725. Chartier Harlin M-C, Crawford F, Houlden H, Warren A, Hughes D,
Borck C, Jefferys JG (1999) Seizure-like events in disinhibited ven- Fidani L, Goate A, Rossor M, Roques P, Hardy J, Mullan M
tral slices of adult rat hippocampus. J Neurophysiol 82: (1991) Early onset Alzheimers disease caused by mutations at
21302142. codon 717 of the beta amyloid precursor gene. Nature
Bouchet, Cazauvieilh (1825) De lpilepsie considr dans ses rap- 353:844846.
ports avec lalination mentale: recherches sur la nature et le Chen G, Chen KS, Knox J, Inglis J, Bernard A, Martin SJ, Justice A,
sige de ces deux maladies. Arch Gen Med 9:510542. McConlogue L, Games D, Freedman SB, Morris RGM (2000) A
Bouilleret V, Schwaller B, Schurmans S, Celio MR, Fritschy JM (2000) learning decit related to age and -amyloid plaques in a mouse
Neurodegenerative and morphogenic changes in a mouse model of Alzheimers disease. Nature 408:975979.
model of temporal lobe epilepsy do not depend on the expres- Cobb SR, Buhl EH, Halasy K, Paulsen O, Somogyi P (1995) Synchro-
sion of the calcium-binding proteins parvalbumin, calbindin, or nization of neuronal activity in hippocampus by individual
calretinin. Neuroscience 97:4758. GABAergic interneurons. Nature 378:7578.
Braak H, Braak E (1991) Neuropathological staging of Alzheimer- Cohen I, Navarro V, Clemenceau S, Baulac M, Miles R (2002) On the
related changes. Acta Neuropathol (Berl) 82:239259. origin of interictal activity in human temporal lobe epilepsy in
Bragin A, Penttonen M, Buzsaki G (1997) Termination of epileptic vitro. Science 298:14181421.
afterdischarge in the hippocampus. J Neurosci 17:25672579. Cole TB, Robbins CA, Wenzel HJ, Schwartzkroin PA, Palmiter RD
Breen KC, Bruce M, Anderton BH (1991) Beta amyloid precursor (2000) Seizures and neuronal damage in mice lacking vesicular
protein mediates neuronal cell-cell and cell-surface adhesion. J zinc. Epilepsy Res 39:153169.
Neurosci Res 28:90100 Conrad AJ, Abebe T, Austin R, Forsythe S, Scheibel AB (1991)
Briellmann RS, Newton MR, Wellard RM, Jackson GD (2001) Hippocampal pyramidal cell disarray in schizophrenia as a
Hippocampal sclerosis following brief generalized seizures in bilateral phenomenon. Arch Gen Psychiatry 48:413417.
adulthood. Neurology 57:315317. Convit A, De LM, Tarshish C, De SS, Tsui W, Rusinek H, George
Brines ML, Sundaresan S, Spencer DD, deLanerolle NC (1997) A (1997) Specic hippocampal volume reductions in indi-
Quantitative autoradiographic analysis of ionotropic glutamate viduals at risk for Alzheimers disease. Neurobiol Aging
receptor subtypes in human temporal lobe epilepsy: up-regula- 18:131138.
tion in reorganized epileptogenic hippocampus. Eur J Neurosci Corsellis JA, Goldberg GJ, Norton AR (1968) Limbic encephalitis
9:20352044. and its association with carcinoma. Brain 91:481496.
Brooks KA, Shumate MD, Jin H, Rikhter TY, Coulter DA (1998) Selec- Cossart R, Dinocourt C, Hirsch JC, Merchan-Perez A, De Felipe J,
tive changes in single cell GABA(A) receptor subunit expression Ben-Ari Y, Esclapez M, Bernard C (2001) Dendritic but not
and function in temporal lobe epilepsy. Nat Med 4:11661172. somatic GABAergic inhibition is decreased in experimental
Bruton CJ (1988) The neuropathology of temporal lobe epilepsy. epilepsy. Nat Neurosci 4:5262.
Oxford, UK: Oxford University Press (Maudsley Monographs). Cotter D, Wilson S, Roberts E, Kerwin R, Everall IP (2000) Increased
Buhl EH, Otis TS, Mody I (1996) Zinc-induced collapse of aug- dendritic MAP2 expression in the hippocampus in schizophre-
mented inhibition by GABA in a temporal lobe epilepsy model. nia. Schizophr Res 41:313323.
Science 271:369373. Cronin J, Obenaus A, Houser CR, Dudek FE (1992) Electrophysiology
Cain DP (1989) Long-term potentiation and kindling: how similar of dentate granule cells after kainate-induced synaptic reorgani-
are the mechanisms? Trends Neurosci 12:610. zation of the mossy bers. Brain Res 573:305310.
806 The Hippocampus Book

Csernansky JG, Wang L, Joshi S, Miller JP, Gado M, Kido D, McKeel common pathological feature of dementia in very old ( or 
D, Morris JC, Miller MI (2000) Early DAT is distinguished from 80 years of age) humans. Acta Neuropathol (Berl) 88:21