Crystallization of Nucleic Acids and Proteins

The Practical Approach Series

SERIES EDITOR

B. D. HAMES
Department of Biochemistry and Molecular Biology
University of Leeds, Leeds LS2 9JT, UK

See also the Practical Approach web site at http://www.oup.co.uk/PAS

* indicates new and forthcoming titles

Affinity Chromatography if Cell Growth, Differentiation

Affinity Separations and Senescence
Anaerobic Microbiology if Cell Separation
Animal Cell Culture Cellular Calcium
(2nd edition) Cellular Interactions in
Animal Virus Pathogenesis Development
Antibodies I and II Cellular Neurobiology
Antibody Engineering Chromatin
Antisense Technology if Chromosome Structural
Applied Microbial Physiology Analysis
Basic Cell Culture Clinical Immunology
Complement
Behavioural Neuroscience
* Crystallization of Nucleic
Bioenergetics
Acids and Proteins
Biological Data Analysis (2nd edition)
Biomechanics—Materials Cytokines (2nd edition)
Biomechanics—Structures and The Cytoskeleton
Systems
Diagnostic Molecular
Biosensors Pathology I and II
Caenorhabditis Elegans DNA and Protein Sequence
Carbohydrate Analysis Analysis
(2nd edition) DNA Cloning 1: Core
Cell-Cell Interactions Techniques (2nd edition)
The Cell Cycle DNA Cloning 2: Expression
Cell Growth and Apoptosis Systems (2nd edition)
 DNA Cloning 3: Complex * High Resolution
Genomes (2nd edition) Chromatography
DNA Cloning 4: Mammalian Histocompatibility Testing
Systems (2nd edition] HIV 1 and 2
* DNA Microarrays * HPLC of Macromolecules (2nd
* DNA Viruses edition)
Drosophila (2nd edition) Human Cytogenetics I and II
Electron Microscopy in (2nd edition)
Biology Human Genetic Disease
Electron Microscopy in Analysis
Molecular Biology * Immobilized Biomolecules in
Electrophysiology Analysis
Enzyme Assays Immunochemistry 1 and 2
Epithelial Cell Culture Immunocytochemistry
Essential Developmental * Immundiagnostics
Biology if In Situ Hybridization (2nd
Essential Molecular Biology I edition)
Essential Molecular Biology II lodinated Density Gradient
Media
* Eukaryotic DNA Replication
Ion Channels
Experimental Neuroanatomy
* Light Microscopy (2nd edition)
Extracellular Matrix
Lipid Modification of Proteins
Flow Cytometry (2nd edition)
Lipoprotein Analysis
Free Radicals
Liposomes
Gas Chromatography
Mammalian Cell
Gel Electrophoresis of Nucleic Biotechnology
Acids (2nd edition)
Medical Parasitology
if Gel Electrophoresis of Proteins
(3rd edition) Medical Virology
Gene Probes 1 and 2 MHC 1 and 2
Gene Targeting if Molecular Genetic Analysis of
Populations (2nd edition)
* Gene Targeting (2nd edition)
Molecular Genetics of Yeast
Gene Transcription
Molecular Imaging in
Genome Mapping Neuroscience
Glycobiology Molecular Neurobiology
if Growth Factors and Receptors Molecular Plant Pathology I
Haemopoiesis and II
 Molecular Virology * Post-translational Processing
Monitoring Neuronal Activity Preparative Centrifugation
* Mouse Genetics and Protein Blotting
Transfenics * Protein Expression
Mutagenicity Testing Protein Engineering
Mutation Detection Protein Function
Neural Cell Culture (2nd edition)
Neural Transplantation * Protein Phosphorylation (2nd
Neurochemistry (2nd edition) edition)
Neuronal Cell Lines Protein Purification
NMR of Biological Applications
Macromolecules Protein Purification Methods
Non-isotopic Methods in Protein Sequencing
Molecular Biology Protein Structure
Nucleic Acid Hybridization (2nd edition)
* Nuclear Receptors Protein Structure Prediction
Oligonucleotides and Protein Targeting
Analogues
Proteolytic Enzymes
Oligonucleotide Synthesis
Pulsed Field Gel
PCR 1 and 2 Electrophoresis
*PCR3:PCR In Situ
RNA Processing I and II
Hybridization
Peptide Antigens RNA-Protein Interactions
Photosynthesis: Energy Signalling by Inositides
Transduction * Signal Transduction
Plant Cell Biology (2nd edition)
Plant Cell Culture (2nd edition) Subcellular Fractionation
Plant Molecular Biology Signal Transduction
Plasmids (2nd edition) if Transcription Factors
Platelets (2nd edition)
Postimplantation Mammalian Tumour Immunobiology
Embryos * Virus Culture
Crystallization of
Nucleic Acids and
Proteins
A Practical Approach
Second Edition

Edited by
ARNAUD DUCRUIX
Laboratoire de Cristallographie et
RMN Biologiques, Faculte de Pharmacie
Universite de Paris V, Paris

and
RICHARD GIEGE
Institut de Biologie Moleculaire
et Cellulaire du CNRS, Strasbourg

OXFORD
UNIVERSITY PRESS
 OXFORD
UNIVERSITY PRESS
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford
and furthers the University's aim of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Athens Auckland Bangkok Bogota Buenos Aires Calcutta
Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul
Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai
Nairobi Paris Sao Paulo Singapore Taipei Tokyo Toronto Warsaw
and associated companies in Berlin Ibadan
Oxford is a registered trade mark of Oxford University Press
Published in the United States
by Oxford University Press Inc., New York
© Oxford University Press, 1999
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,
without the prior permission in writing of Oxford University Press.
Within the UK, exceptions are allowed in respect of any fair dealing for the
purpose of research or private study, or criticism or review, as permitted
under the Copyright, Designs and Patents Act, 1988, or in the case
of reprographic reproduction in accordance with the terms of licenses
issued by the Copyright Licensing Agency. Enquiries concerning
reproduction outside those terms and in other countries should be
sent to the Rights Department, Oxford University Press,
at the address above.
This book is sold subject to the condition that it shall not, by way
of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated
without the publisher's prior consent in any form of binding or cover
other than that in which it is published and without a similar condition
including this condition being imposed on the subsequent purchaser
Users of books in the Practical Approach Series are advised that prudent
laboratory safety procedures should be followed at all times. Oxford
University Press makes no representation, express or implied, in respect of
the accuracy of the material set forth in books in this series and cannot
accept any legal responsibility or liability for any errors or omissions
that may be made.
A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
Crystallization of nucleic acids and proteins : a practical approach /
edited by Arnaud Ducruix, Richard Giege — [2nd ed.]
(The practical approach series : 210)
Includes bibliographical references and index.
1. Proteins—Analysis. 2. Nucleic Acids—analysis. I. Ducruix.
A. (Arnaud) II. Giege, R. (Richard) III. Series.
QD431.25.A53C79 1999 547.7'5046—dc21 99-15222
ISBN 0-19-963679-6 (Hbk)
0-19-963678-8 (Pbk)
Typeset by Footnote Graphics,
Warminster, Wilts
Printed in Great Britain by Information Press, Ltd,
Eynsham, Oxon.
 Preface
With the development of genomics and proteomics, and their applications in
basic biological research and the biotechnologies, there is an increasing need
of three-dimensional structural knowledge of proteins, nucleic acids, and
multi-macromolecular assemblies by X-ray methods. To achieve this aim,
crystals diffracting at high resolution are needed. The major aim of the second
edition of this book in the Practical Approach series is to present an update of
the methods employed to produce crystals of biological macromolecules and
to outline the newest trends that have entered the field. Since the first edition,
which appeared in 1992, the science of crystallogenesis, which was then in its
infancy, has grown rapidly balancing between the physics of crystal growth and
blind-screen crystallizations. The advances can be appreciated from the Pro-
ceedings of the different International Conferences on the Crystallization of
Biological Macromolecules (ICCBM1 to 7) which appear every two years, the
latest covering ICCBM-7, held in June 1998 in Granada, and published in
J. Crystal Growth, Vol. 196, January 1999.
As usual in the series, the emphasis of the present book is to give detailed
laboratory protocols throughout the chapters. However, we have not given
protocols just as 'recipes', but instead we have intended to always present the
methods with reference to the theoretical concepts and principles underlying
them. In fact one of the aims of this book was to fight against the fallacious
idea according to which crystal growth of biological macromolecules is more
an 'art' than a science. Although this is probably sometimes true from a prac-
tical point of view, it is certainly incorrect in its principle. Therefore emphasis
has been given to the physical parameters involved in crystallization and on
the large knowledge on the crystal growth of small molecules, as well as to the
particular biochemical and physico-chemical properties of biological macro-
molecules.
This book is intended to be read by a wide range of scientists. First, by the
crystallographers who have to solve three-dimensional structures of macro-
molecules. Secondly, by all molecular biologists who have access to macro-
molecules but often do not know how to handle them for crystallization, and
who may consider crystallization as an esoteric undertaking, because of lack of
basic knowledge about the crystallization process. Thirdly, by the physico-
chemists and physicists who for other reasons consider biology as an esoteric
science. It is our wish that this book will contribute to a better understanding
of crystallogenesis by these scientists and to the improved perception of the
biological requirements that have to be taken into account for physical studies.
Finally, the book should be a laboratory guide for all students and beginners,
helping them to avoid making mistakes when entering the field of crystal
preparation.
 Preface
Chapter 1 is an introduction to crystallogenesis of biological macro-
molecules. It includes a brief historical survey of the subject and introduces the
general principles and major achievements of this new discipline. The prepara-
tion of biological macromolecules and the concept of 'crystallography-grade
purity' are developed in Chapter 2. Chapter 3 is new and introduces the use of
molecular biology methods to 'customize' domains for structural biology. It
also includes the preparation of protein crystals made of protein molecules
containing selenomethionine residues and outlines why they can be used for
the multiple anomalous dispersion (MAD) method. Screen-like methods are
now widespread but do not provide suggestions when they fail. The answer
may then come from statistical methods presented in Chapter 4 which explains
their theory and gives practical advice (and a computer program) for protocol
design. One of the goals of this book is to give to crystal growers of biomacro-
molecules the conceptual and methodological tools needed to control crystal-
lization. This is examined in Chapter 5 which includes a description of the
classical crystallization methods together with workshop examples. Crystal-
lization in gels is described with theoretical and practical considerations in
Chapter 6. This chapter includes a novel section describing the gel acupunc-
ture method; it also contains information on crystal growth under microgravity
and hypergravity conditions. Because it is sometimes difficult to reproduce
appropriate nucleation conditions, Chapter 7 is devoted to seeding procedures
with preformed crystalline material, including micro-, macro-, and cross-
seeding with numerous examples.
The special cases of nucleic acids (and their complexes with proteins and
nucleoprotein assemblies) and membrane proteins are covered in two indi-
vidual chapters (8 and 9). Their crystallization is still challenging, but the
novel developments in the field have led to a number of recent breakthroughs,
that are encouraging for experimenters entering the field. For nucleic acids,
Chapter 8 gives emphasis to the strategies for the design and preparation of
appropriate DNA or RNA fragments and to the specific features character-
izing their crystallization, either free or in complexes with proteins. For
membrane proteins, the already published genomes show that a good third of
the expressed proteins belong to this category and it is expected that the inter-
est for membrane proteins will expand quickly, especially among structural
biologists. It is our wish that the methods described in Chapter 9 will help them
to reach this goal.
The link between protein solubility and the physico-chemical parameters
governing crystal growth is presented in Chapter 10 with a strong emphasis on
the practical issues. Chapter 11 deals with physical methods and gives an intro-
duction to the physics of crystal growth. In particular, the use of light scattering
methods to monitor early nucleation events is advocated with examples and a
description of the material used.
Chapter 12 is new and covers the expanding field of the two-dimensional
crystallization of soluble proteins on planar lipid films. It presents many proto-
 Preface
cols which may be readily used. Soaking of crystals of biological macro-
molecules is of great interest for crystallographers, either for resolving a
structure (heavy-atom derivatives), or for diffusing inhibitors, activators, or
cofactors (eventually photoactivable). As in all previous chapters, practical
aspects were the driving force and Chapter 13 is illustrated by a variety of pro-
tocols. The editors thought that an introduction to X-ray crystallography
should be included in this book. This is done in Chapter 14, that is geared
toward biochemists wanting to characterize crystals by themselves rather than
explaining how to solve a structure.
It is a great pleasure to acknowledge our gratitude to a number of friends
and colleagues. First, our colleagues from Gif-sur-Yvette/Paris and Strasbourg
deserve particular thanks for having participated over the years in the develop-
ment of our studies on crystallogenesis; their enthusiasm was essential and
gave us the impetus for the preparation of a book covering this field and for
updating it in its second edition. However, without the invaluable help of
many friends from both sides of the Atlantic who agreed to cover specialized
topics, this venture would not have been possible. We would like to warmly
thank all of them. The French Centre National de la Recherche Scientifique
(CNRS) and Centre National d'Etudes Spatiales (CNES) are acknowledged
for their permanent support in developing biological crystallogenesis and their
interest for the physico-chemical aspects of the field.
While preparing this second edition, our dear colleague Roland Boistelle
closed his eyes. He was a source of inspiration for all of us and one of the first
who geared the biology oriented scientists to the physics of crystal growth. His
contribution to the field of macromolecules crystallogenesis was essential and
we would like to dedicate this book to his memory.
Paris and Strasbourg A. D.
June 1999 R. G.

IX
This page intentionally left blank
 Contents
List of Contributors
Abbreviations

1. An introduction to the crystallogenesis of

biological macromolecules 1
R. Giege and A. Ducruix
1. Introduction 1
2. Crystallization and biology: a historical background 4
Before the X-rays 4
Crystallogenesis and structural biology 5
3. General principles 7
A multiparametric process 7
Purity 7
Solubility and supersaturation 9
Nucleation, growth, and cessation of growth 10
Packing 10
4. From empiricism to rationality 10
Towards a better understanding of parameters 10
Towards an active control of crystal growth 11
The future of biological crystallogenesis 11
5. Search for crystallization strategies 12
References 13

2. Biochemical aspects and handling of

macromolecular solutions and crystals 17
B. Lorber and R. Giege
1. Introduction 17
2. The biological material 18
Sources of biological macromolecules 18
Macromolecules produced in host cells or in vitro 18
3. Isolation and storage of pure macromolecules 19
Preparative isolation methods 20
Specific preparation methods 22
 Contents
Stabilization and storage 22
Ageing 23
4. Characterization and handling of macromolecules 23
Analytical biochemical methods 24
Prevention of macromolecular damage 26
5. The problem of purity and homogeneity 27
The concept of 'crystallography-grade' quality 27
Purity of samples 27
Microheterogeneity of samples 28
Probing purity and homogeneity 31
Improving purity and homogeneity in practice 32
6. Characterization and handling of crystals 34
Analysis of crystal content 34
Crystal density 36
References 39

3. Molecular biology for structural biology 45

P. F. Berne, S. Doublie, and C. W. Carter, Jr
1. Introduction 45
2. Molecular biology to optimize protein expression systems 46
Expression systems 46
Factors influencing the level of expression 47
Inclusion bodies in E. coli 48
Tagged proteins to facilitate purification 50
Use of molecular biology to design an appropriate expression vector 51
3. Engineering physical properties of macromolecules 55
Problems encountered and possible solutions 55
Defining the optimal size of the molecule to be crystallized 57
Site-directed mutagenesis 59
Random mutagenesis 61
Selection of variants following mutagenesis 62
4. Preparation of selenomethionyl protein crystals 64
Background 64
Expression and cell growth in a prokaryotic system 65
Eukaryotes 68
Purification 68
Crystallization 69
Warning 70
5. Conclusion 70
References 71
xii
 Contents

4. Experimental design, quantitative analysis,

and the cartography of crystal growth 75
C. W. Carter, Jr
1. Introduction 75
2. Response surfaces and factorial design 77
Mathematical models and inference 77
What is a response surface? 77
Factorial experimental design 77
3. Sampling appropriate subsets from a full-factorial design 82
Screening versus optimization 82
Subsets for screening 82
Minimum variance sampling: Hardin-Sloane designs 85
Computer programs to generate designs for special purposes 85
4. Screening with factorial designs 87
Selecting experimental variables 87
Preparing the experimental matrix 89
'Floating' variables, initial values, and sampling intervals 91
Design for initial screening of variables for crystallizing a new protein 92
Experimental set-ups 95
5. Quantitative scoring 96
Hierarchical evaluation: interrogating nature by experimental design 96
Rating different solid phases 99
Size and shape 100
Scoring the best result or the 'average' from a given test? 100
6. Regression, the analysis of variance, and analysis of models 100
Analysis of contrasts 101
Analysis of models by multiple regression and the analysis of variance 102
7. Optimization 112
Steepest ascent and simplex optimization 112
Optimization using quadratic polynomial models 113
8. Resolution of polymorphs 116
References 118

5. Methods of crystallization 121

A. Ducruix and R. Giege
1. Introduction 121
2. Sample preparation 121
Solutions of chemicals 121
Preparing samples of biological macromolecules 123
xiii
 Contents
3. Crystallization by dialysis methods 126
Principle 126
Examples of dialysis cells 126
4. Crystallization by vapour diffusion methods 130
Principle 130
Experimental set-ups 132
Varying parameters 137
Kinetics of evaporation 137
5. Crystallization by batch methods 138
Classical methods 138
Advanced methods 140
6. Crystallization by interface diffusion 141
7. Correlations with solubility diagrams 141
Dialysis 141
Vapour diffusion 142
Batch crystallization 142
8. Practising crystallization 143
9. Concluding remarks 145
References 146

6. Crystallization in gels and related methods 149

M.-C. Robert, O. Vidal, J.-M. Garda-Ruiz, and F. Otalora
1. Introduction 149
2. General considerations 150
Formation and structure of gels 150
Gel properties related to crystal growth 152
3. Practical consideration 157
Gel preparation 158
Gel methods 161
4. Crystal preparation and characterization 170
5. Related methods 170
Microgravity 170
Hypergravity 172
References 173

7. Seeding techniques 177

E. A. Stura
1. Introduction 177
xiv
 Contents

2. Seeding 178
Supersaturation and nucleation 178
Crystal growth 179
Seeding techniques 179
3. Crystallization procedures 180
Pre-seeding: sitting drop vapour diffusion 180
Analytical seeding 185
4. Production seeding methods 188
Microseeding 188
Macroseeding 191
5. Heterogeneous seeding 196
Cross-seeding 197
Epitaxial nucleation 200
6. Crystallization of complexes 202
Considerations in the crystallization of complexes 202
Use of streak seeding in protein complex crystallization 204
Analytical techniques for determination of crystal content 204
7. Concluding remarks 206
Acknowledgements 207
References 207

8. Nucleic acids and their complexes 209

A.-C. Dock-Bregeon, D. Moras, and R. Giege
1. Introduction 209
2. Preparation of nucleic acids 210
Synthetic nucleic acid fragments for crystallogenesis 210
Preparation of natural small RNAs 214
3. Crystallization of nucleic acids 218
General features 220
Specific features: additives 221
Crystallization strategies 223
The special case of DNA:drug co-crystallization 224
4. Co-crystallization of nucleic acids and proteins 224
General features of nucleic acid:protein co-crystallization 225
Complexes of synthetic oligodeoxynucleotides and proteins 227
Complexes of RNAs and proteins 231
Ribosomes and their subunits 234
Viruses 235
References 238
xv
 Contents

9. Crystallization of membrane proteins 245

F. Reiss-Husson and D. Picot
1. Introduction 245
2. Crystallization principles 246
3. Detergents for crystallization 247
n-Alkyl-p-glucosides (CnG) 250
n-Alkyl-thioglucosides 251
n-Alkyl-maltosides (CnM) 254
n-AIkyl-dimethylamineoxides (CnDAO) 254
n-Alkyl-oligoethylene glycol-monoethers (CnEm) 254
4. Purification of membrane proteins before crystallization 255
Purity requirements 255
Detergent exchange 257
Sample concentration 258
5. Crystallization protocols 259
Detergent 260
Additives 261
Crystallizing agent 261
Optimization 263
6. Experimental techniques 263
7. Conclusion 264
References 265

10. From solution to crystals with a

physico-chemical aspect 269
M. Ries-Kautt and A. Ducruix
1. Introduction 269
2. The concept of solubility and methods for solubility diagram
determination 269
Solubility 270
Measurements of the solubility 271
Phase diagram 278
Kinetic aspects 281
3. Proteins as polyions 283
Estimation of the net charge 284
Desalting of proteins 286
Net charge and crystallization conditions 288
xvi
 Contents
4. Influence of physico-chemical parameter changes 291
Interactions in a protein solution 291
pH 293
Ionic strength 295
Nature of salts 298
Temperature 302
H2O versus D2O 303
Combined effects of crystallization variables 304
Stepwise replacement technique 304
5. Crystallization 305
Crystallization strategies 306
Polymorphism 308
References 310

11. Diagnostic of pre-nucleation and nucleation

by spectroscopic methods and background
on the physics of crystal growth 313
5. Veesler and R. Boistelle
1. Introduction 313
2. Concentration and supersaturation 313
3. Nucleation 315
Nucleation rate 315
Activation free energy for homogeneous nucleation 316
Activation free energy for heterogeneous nucleation 318
Examples 319
4. Pre-nucleation — investigation of the solution 321
Methods 321
Practical recommendations 324
Examples 326
Practical considerations 328
5. Crystal growth 329
Growth controlled by surface processes 330
Kinetic measurements 334
6. Crystallization in the presence of impurities and additives 335
General trends 335
Additives, phases, and polymorphs 336
Crystallization kinetics, impurities, and additives 337
Impurity incorporation 338
References 338
xvii
 Contents

12. Two-dimensional crystallization of soluble

proteins on planar lipid films 341
A. Brisson, O. Lambert, and W. Bergsma-Schutter
1. Introduction 341
2. Two-dimensional crystallization of soluble proteins on
planar lipid films 341
3. Setting up of a crystallization experiment 344
Lipid solutions 344
Protein solutions 345
Preparation and cleaning of Teflon supports 346
4. Transfer of protein-lipid films to an EM grid 349
General considerations 349
Preparation of EM grids 349
5. Characterization of the protein 2D crystals by EM 355
6. Characterization of the protein-lipid crystals by optical
diffraction 360
7. Conclusion 360
References 362

13. Soaking techniques 365

E. A. Stum and T. Gleichmann
1. Introduction 365
The crystal lattice 365
Reasons for soaking 366
Soaking of crystals versus co-crystallization 367
Soaking techniques 368
2. Soaking of substrates, activators, and inhibitors 371
Soaking techniques for crystals in drops 372
Soaking of crystals in capillaries 374
Soaking of crystals in dilute ligand solutions 376
Cross-linking of crystals 377
3. Soaking application 377
Heavy-atom soaking and isomorphous replacement 377
Selecting a heavy-atom compound 379
Soaking for cryo-crystallography 379
4. Conclusions 388
Acknowledgements 389
References 389
xviii
 Contents

14. X-ray analysis 391

L. Sawyer and M. A. Turner
1. Introduction 391
2. Background to X-ray crystallography 391
X-rays 392
What is a crystal? 392
How do X-rays interact with crystals 395
How is a protein crystal structure solved? 399
Importance of preliminary characterization 402
3. Mounting crystals 404
Initial examination with a microscope 404
The basic techniques 405
4. X-ray data 411
Oscillation methods for data collection 412
Optical alignment 413
Crystal characterization with an area detector/image plate 413
Determination of space group 416
Other techniques for diffraction data collection 417
5. Concluding remarks 418
References 419
Appendix 421
Index 429

XIX
This page intentionally left blank
 Contributors
W. BERGSMA-SCHUTTER
Department of Chemistry / Biophysical Chemistry, University of Groningen,
Nijenborgh 4, NL-9747 AG Groningen, The Netherlands,
P. F. BERNE
Rhone Poulenc Rorer, 13, quai Jules Guesde, 94403 Vitry sur Seine Cedex,
France.
R. BOISTELLE (+ DECEASED 1998)
Centre de Recherche sur les Me'canismes de la Croissance Cristalline du
CNRS, Campus de Luminy, Case 913, 13288 Marseille Cedex, France.
A. BRISSON
Department of Chemistry / Biophysical Chemistry, University of Groningen,
Nijenborgh 4, NL-9747 AG Groningen, The Netherlands.
C.W. CARTER JR
Department of Biochemistry, CB 7260, University of North Carolina at
Chapel Hill, Chapel Hill, NC 27599-7260, USA.
A.-C. DOCK-BREGEON
Institut de Genetique et de Biologie Moleculaire et Cellulaire, 1, rue Leon
Fries, Pare d'Innovation, BP163, F-67404 Illkirch Cedex, France.
S. DOUBLIE
Department of Microbiology and Molecular Genetics, The Markey Center for
Molecular Genetics, University of Vermont, Burlington, VT 05045, USA.
A. DUCRUIX
Laboratoire de Cristallographie et RMN Biologiques, Faculte de Pharmacie,
Universite de Paris V, 4, Avenue de l'Observatoire, 75270 Paris Cedex 06,
France.
J. M. GARCIA-RUIZ
Institute Andaluz de Ciencias de la Tierra, CSIC-Universidad de Granada,
Facultad de Ciencias, 18002-Granada, Spain.
R. GIEGE
Institut de Biologie Moleculaire et Cellulaire du CNRS, 15 rue Rene
Descartes, F-67084 Strasbourg Cedex, France.
T. GLEICHMANN
Anorganisch-Chemisches Institut, Westfalische Wilhelms-Universitat,
Wilhelm-Klemm Str. 8, D-48149 Munster, Germany.
 Contributors
O. LAMBERT
Department of Chemistry/Biophysical Chemistry, University of Groningen,
Nijenborgh 4, NL-9747 AG Groningen, The Netherlands.
B. LORBER
Institut de Biologic Moleculaire et Cellulaire du CNRS, 15 rue Rene
Descartes, F-67084 Strasbourg Cedex, France.
D. MORAS
Institut de Genetique et de Biologie Moleculaire et Cellulaire, 1 rue Leon
Fries, Pare d'Innovation, BP 163. F-67404 Illkirch Cedex, France.
F. OTALORA
Institute Andaluz de Ciencias de la Tierra, CSIC-Universidad de Granada,
Facultad de Ciencias, 18002-Granada, Spain.
D. PICOT
Institut de Biologie Physico-Chimique, 13 rue P. et M. Curie, F-75005 Paris.
F. REISS-HUSSON
Centre de Genetique Moleculaire du CNRS, Avenue de la Terrasse, F-91190
Gif-sur-Yvette Cedex, France.
M. RIES-KAUTT
Laboratoire de Cristallographie et RMN Biologiques, Faculte de Pharmacie,
Universite de Paris V, 4, Avenue de 1'Observatoire, 75270 Paris Cedex 06,
France.
M.-C. ROBERT
Laboratoire de Mineralogie-Cristallographie, Universites Pierre et Marie
Curie, 4 place Jussieu, F-75252 Paris Cedex 05, France.
L. SAWYER
Structural Biochemistry Group, The University of Edinburgh, Swann Build-
ing, King's Buildings, Mayfield Road, Edinburgh EH9 3JR, UK.
E. A. STURA
Dept. d'Ingenierie et d'Etudes des Proteines, Bat. 152, CEA/Saclay, 91191
Gif-sur-Yvette Cedex, France.
M. A. TURNER
X-ray Structure Laboratory, Department of Biochemistry, Hospital for Sick
Children, 555 University Avenue, Toronto, Ontario, Canada.
S. VEESLER
Centre de Recherche sur les Mecanismes de la Croissance Cristalline du
CNRS, Campus de Luminy, Case 913, F-13288 Marseille Cedex, France.
O. VIDAL
Laboratoire de Mineralogie-Cristallographie, Universites Pierre et Marie
Curie, 4 place Jussieu, F-75252 Paris Cedex 05, France.
xxii
 Abbreviations
BPTI bovine pancreatic trypsin inhibitor
BTP bis Tris propane
CMC critical micellar concentration
DEPC diethylpyrocarbonate
DiFP diisopropylfluorophosphate
DLS dynamic light scattering
DMSO dimethyl sulfoxide
DOPC dioleoylphosphatidylcholine
DTT dithiothreitol
EDTA ethylenediaminetetraacetic acid
ESI electronspray ionization mass spectra
GST glutathione-S-transferase
HEW hen egg white
HEWL hen egg white lysozyme
HIC hydrophobic interaction chromatography
IEF isoelectric focusing
LS light scattering
MAD multiple wavelength anomalous dispersion
MALDI matrix-assisted desorption/ionization
MeTEOS methyltriethoxysilane
MIR multiple isomorphous replacement
MPD 2-methyl-2,4-pentane diol
MR molecular replacement
OD optical density
OP osmotic pressure
PBC periodic bond chain
PC phosphatidylcholine
PCR polymerase chain reaction
PEG polyethylene glycol
pTS para-toluenesulfonate
RPC reverse-phase chromatography
SANS small angle neutron scattering
SAXS small angle X-ray scattering
SIR single isomorphous replacement
SLS static light scattering
TEOS tetraethoxysilane
TLC thin-layer chromatography
TMOS tetramethoxysilane
This page intentionally left blank
 1

An introduction to the
crystallogenesis of biological
macromolecules
R. GIEGE and A. DUCRUIX

'II y a la des mysteres, qui preparent a 1'avenir d'immenses travaux et appellent des
aujourd'hui les plus serieuses meditations de la science'
Pasteur, 1860, in Leqons de Chimie.

1. Introduction
The word 'crystal' is derived from the Greek root 'krustallos' meaning 'clear
ice'. Like ice, crystals are chemically well defined, and many among of them
are of transparent and glittering appearance, like quartz, which was for a long
time the archetype. Often they are beautiful geometrical solids with regular
faces and sharp edges, which probably explains why crystallinity, even in the
figurative meaning, is taken as a symbol of perfection and purity. From the
physical point of view, crystals are regular three-dimensional arrays of atoms,
ions, molecules, or molecular assemblies. Ideal crystals can be imagined as
infinite and perfect arrays in which the building blocks (the asymmetric units)
are arranged according to well-defined symmetries (forming the 230 space
groups) into unit cells that are repeated in the three-dimensions by trans-
lations. Experimental crystals, however, have finite dimensions. An implicit
consequence is that a macroscopic fragment from a crystal is still a crystal,
because the orderly arrangement of molecules within such a fragment still
extends at long distances. The practical consequence is that crystal fragments
can be used as seeds (Chapter 7). In laboratory-grown crystals the periodicity
is never perfect, due to different kinds of local disorders or long-range
imperfections like dislocations. Also, these crystals are often of polycrystal-
line nature. The external forms of crystals are always manifestations of their
internal structures and symmetries, even if in some cases these symmetries
may be hidden at the macroscopic level, due to differential growth kinetics of
the crystal faces. Periodicity in crystal architecture is also reflected in their
macroscopic physical properties. The most straightforward example is given
 R. Giege and A. Ducruix
by the ability of crystals to diffract X-rays, neutrons, or electrons, the
phenomenon underlying structural chemistry and biology (for introductory
texts see refs 1 and 2), and the major aim of this book is to present the
methods employed to produce three-dimensional crystals of biological macro-
molecules, but also two-dimensional crystals (Chapter 12), needed for
diffraction studies. Other properties of invaluable practical applications
should not be overlooked either, as is the case of optical and electronic
properties which are at the basis of non-linear optics and modern electronics
(for an introduction to physical properties of molecular crystals see ref. 3).
Crystals furnish one of the most beautiful examples of order and symmetry in
nature and it is not surprising that their study fascinates scientists (4).
What characterizes biological macromolecular crystals from small molecule
crystals? In terms of morphology, one finds with macromolecular crystals the
same diversity as for small molecule crystals (Figure 1). In terms of crystal
size, however, macromolecular crystals are rather small, with volumes rarely
exceeding 10 mm3, and thus they have to be examined under a binocular
microscope. Except for special usages, such as neutron diffraction, this is not
too severe a limitation. Among the most striking differences between the two
families of crystals are the poor mechanical properties and the high content of
solvent of macromolecular crystals. These crystals are always extremely
fragile and are sensitive to external conditions. This property can be used as a
preliminary identification test: protein crystals are brittle or will crush when
touched with the tip of a needle, while salt crystals that can sometimes
develop in macromolecule crystallization experiments will resist this treat-
ment. This fragility is a consequence both of the weak interactions between
macromolecules within crystal lattices and of the high solvent content (from
20% to more than 80%) in these crystals (Chapter 14). For that reason,
macromolecular crystals have to be kept in a solvent-saturated environment,
otherwise dehydration will lead to crystal cracking and destruction. The high
solvent content, however, has useful consequences because solvent channels
permit diffusion of small molecules, a property used for the preparation of
isomorphous heavy-atom derivatives needed to solve the structures (Chapters
13 and 14). Further, crystal structures can be considered as native structures,
as is indeed directly verified in some cases by the occurrence of enzymatic re-
actions within crystal lattices upon diffusion of the appropriate ligands (5, 6).
Other characteristic properties of macromolecular crystals are their rather
weak optical birefringence under polarized light: colours may be intense for
large crystals but less bright than for salt crystals (isotropic cubic crystals or
amorphous material will not be birefringent). Also, because the building blocks
composing macromolecules are enantiomers (L-amino acids in proteins—
except in the case of some natural peptides—and D-sugars in nucleic acids)
macromolecules will not crystallize in space groups with inversion sym-
metries. Accordingly, out of the 230 possible space groups, macromolecules
do only crystallize in the 65 space groups without such inversions (7). While
 1: An introduction to the crystnllogaiwsis

Figure 1. From precipitates to perfect crystals of biological maeromolecules. (a) Precipi-

tate of HEW lysozyme; (b) yeast aspartyl-tRNA synthetase microcrystals; (3) spherutites
of the complex between yeast aspartyl-tRNA synthetase and tRNAASO; (d) short needles of
tissue inhibitor protein of metalloprotease; (e) aspartyl-tRNA synthetase long needle-like
crystals; (f) yeast initiator tRNAMot thin plates with growth defects; (g) plate-shaped
crystals of Hypoderma lineatum collagenase; (h) tetragonal crystals of aspartyl-tRNA
synthetase showing growth defects together with brush-like needle bunches; (i) exampie
of 'skin' of denatured protein around a HEW lysozyme crystal; (j) crystal of mellitin with
a hollow extremity; (k, I) twinned and twinned-embedded crystals of collagenase and
HEW lysozyme; [m, n, o, p) perfect three-dimensional crystals; (m) polymorphism in a
same crystallization drop showing cubic and orthorhombic crystal-habits of aspartyl-
tRNA synthetase/tRNAA5p complex; (n) crystals of tRMAAsp with cracks due to ageing;
(o) crystals of HEW lysozyme with fourfold symmetry; (p) crystals of H. linealum
collagenase.
 R. Giege and A. Ducruix
small organic molecules prefer to crystallize in space groups in which it is
easiest to fill space, proteins crystallize primarily in space groups in which it is
easiest to achieve connectivity (8). Macromolecular crystals are also charac-
terized by large unit cells with dimensions that can reach up to 1000 A for
virus crystals (9). From a practical point of view, it is important to remember
that crystal morphology is not synonymous with crystal quality. Therefore, the
final diagnostic of the suitability of a crystal for structural studies will always
be the quality of the diffraction pattern which reveals its internal order, as is
reflected at first glance by the so-called 'resolution' parameter (Chapter 14).
Crystal growth, which is a very old activity that has always intrigued
mankind, and many philosophers and scientists have compared it with the
biological process of reproduction, and it has even been speculated that the
duplication of genetic material would occur through crystallization-like
mechanisms (10). Nowadays, the theoretical and practical frames of crystallo-
genesis are well established for small molecules, but less advanced for
macromolecules, although it can be anticipated that many principles under-
lying the growth of small molecule crystals will apply for that of macro-
molecules (11, 12). Until recently, crystallization of macromolecules was
rather empirical, and because of its unpredictability and frequent irrepro-
ducibility, it has long been considered as an 'art' rather than a science. It is
only in the last 15 years that a real need has emerged to better understand and
to rationalize the crystallization of biological macromolecules. It can be stated
at present that the small molecule and macromolecular fields are converging,
with an increasing number of behaviours or features known for small
molecules that are now found for macromolecules (13).

2. Crystallization and biology: a historical background

2.1 Before the X-rays
It is often forgotten that some advances in biochemistry and molecular
biology have their origin in crystallization data and that empirical crystal
growth of biological materials is as old as biochemistry. The first reports on
protein crystals were published more than a century ago when Funke,
Hiinefeld, Lehman, Teichman, and others crystallized haemoglobin from the
blood of various invertebrates and vertebrates, and it was prophesied that the
study of crystals would shed light on the exact nature of proteinic substances
(14-17). This was followed by the crystallization of hen egg white albumin and
a series of plant proteins (reviewed in refs 16 and 17). The beauty of crystals
certainly fascinated the physiological chemists in these early days, since an
atlas with extensive descriptions of the morphologies of haemoglobin crystals
was published in 1909 (15). In 1926 Sumner reported the crystallization of
urease from jack beans (18), soon followed by Northrop who crystallized
pepsin and a series of other proteolytic enzymes (19). It is interesting to note
 1: An introduction to the crystallogenesis
that for technical difficulties, the X-ray structure of urease has only be deter-
mined very recently (20). Besides being a method of purification, crystalliza-
tion experiments established the view that pure enzymes are proteins, a fact
not obvious to all at that time (21). Another scientific achievement arose in
1935 when Stanley crystallized tobacco mosaic virus. Influenced by Northrop's
conception of enzymes, and using methods developed for proteins, he prepared
the virus in pure crystalline state (he believed the virus was an autocatalytic
protein) and showed that it retains its infectivity after several recrystalliza-
tions (reviewed in ref. 10). The importance and the implications for biology of
these discoveries was recognized rapidly and in 1946 the Nobel Prize for
Chemistry was awarded to Sumner, Northrop, and Stanley.

2.2 Crystallogenesis and structural biology

The use of crystals in structural biology goes back to 1934, when Bernal and
Crowfoot (D. Hodgkin) produced the first X-ray diffraction pattern of a
protein, that of crystalline pepsin (22). Since then representatives of most
families of macromolecules have been crystallized, but using mainly empirical
methods and without rational control of the growth mechanisms. This can be
well understood because in the early days of structural biology the interest of
scientists was mainly directed at the development of the X-ray methods
needed to resolve the structures rather than in that of a rationalization of the
crystallization procedures. Therefore, only limited efforts were expended in
understanding or improving macromolecular crystallization procedures. At
present X-ray methods are well established (2, 7, 23). The overall scheme of
the various steps of the resolution of a three-dimensional structure is sum-
marized in Figure 2. But production of suitable crystals diffracting at high
resolution often remains the bottleneck in structure determination projects.
With the rapid development of biotechnologies and the unlimited potential of
macromolecular engineering (which requires structural knowledge for site-
directed mutagenesis experiments or for design of factitious macromolecules)
there is now an increasing need for macromolecular crystals diffracting at high
resolution, not only proteins but also nucleic acids and multi-macromolecular
assemblies. Thanks to the biotechnological tools it is now possible to obtain
rather easily the amounts of pure macromolecules (in most cases several
milligrams) needed to start a crystallization project (Chapter 3). In this
introductory chapter we present the general trends of the science of crystal
growth in biology. More extensive discussions and the practical details will be
presented in following chapters.
The first major breakthrough towards better and easier crystallizations was
the development, in the 1960s, of crystallization micromethods (e.g. dialysis
and vapour phase diffusion) (Chapter 5). It was promoted by structural pro-
jects on macromolecules reluctant to crystallize easily and available in limited
amounts (24). Further significant improvements came from the discovery of
 R. Giege and A. Ducruix

1- PURIFICATION OF MACROMOLECULES
from wild-type, engineered or overproducing organisms
(possibility of in vitro synthesis for nucleic acids and small peptides)

2- CRYSTALLIZATION
by de novo crystallization or seeding techniques

3- DATA MEASUREMENTS
characterization of space group and diffraction resolution;
measurements of diffraction intensities on an electronic area detector
(possible use of neutron and frequently of tunable X-ray synchrotron radiation);
frequent data acquisition by cryo-crystallographic methods.

4- PHASE DETERMINATION
using methods based on isomorphous replacement (preparation of heavy atom derivatives),
anomalous scattering, molecular replacement and non-crystallographic symmetry,
or direct calculations (e.g. from maximum entropy)

5- ELECTRON DENSITY MAP COMPUTATION AND INTERPRETATION

interpretation of mini-maps; model building on computer graphic displays

6- MODEL REFINEMENT
least-square refinements; restrained refinements;...
Figure 2. Steps involved in the resolution of the 3D structure of a biological macro-
molecule (for more details see Chapter 14).

specific properties of additives to be included in crystallization solvents, such

as the polyamines (25, 26) and the non-ionic detergents (27-30) which gave
the clue for crystallizing nucleic acids (Chapter 8) or membrane proteins
(Chapter 9). Also, a more systematic use of organic cosmotropes (compounds
promoting 'order'), like polyhydric alcohols that are stabilizers of protein
structure when at high concentration, may facilitate crystallization of flexible
proteins (31).
The perception of the importance of purity for growing better crystals (32)
was an important achievement in the field (Chapter 2), and the adequate
choice of the biological material was an important determinant for the success
of many crystallizations. With the ribosome for instance, the preparation of
homogeneous particles from halophilic or thermophilic bacteria instead from
mesophilic bacteria, considerably improved crystal quality (33, 34). Also, the
ease of synthesizing oligonucleotides with automated methods (Chapter 8), or
 1: An introduction to the crystallogenesis
the development of genetic engineering technologies for overexpression of
proteins (Chapter 3), explains the increasing number of crystallized nucleic
acids or rare proteins.
Today, crystal growth research is stimulated by macromolecular engineer-
ing requirements, but also to some extent by space-science projects (crystal-
lization under microgravity conditions, Chapter 6) and its descriptive stage is
moving towards a new, more quantitative discipline, biocrystallogenesis,
which includes biology, biochemistry, physics, and engineering related aspects.
For specialized literature see refs 35^42, and for general reviews 43—46.

3. General principles
3.1 A multiparametric process
Biocrystallization, like any crystallization, is a multiparametric process involv-
ing the three classical steps of nucleation, growth, and cessation of growth.
What makes crystal growth of biological macromolecules different is, first, the
much larger number of parameters than those involved in small molecule
crystal growth (Table 1) and, secondly, the peculiar physico-chemical pro-
perties of the compounds. For instance, their optimal stability in aqueous
media is restricted to a rather narrow temperature and pH range. But the
main difference from small molecule crystal growth is the conformational
flexibility and chemical versatility of macromolecules, and their consequent
greater sensitivity to external conditions. This complexity is the main reason
why systematic investigations were not undertaken earlier. Furthermore, the
importance of some parameters, such as the geometry of crystallization
vessels or the biological origin of macromolecules, had not been recognized. It
is only recently that the hierarchy of parameters has been perceived. A prac-
tical consequence of this new perception was the development of statistical
methods to screen crystallization conditions (Chapter 3). For a rational design
of growth conditions, however, physical and biological parameters have to be
controlled. One of the aims of this book is to give to crystal growers of
biological macromolecules the conceptual and methodological tools needed
to achieve such control.

3.2 Purity
Because macromolecules are extracted from complex biological mixtures,
purification plays an extremely important role in crystallogenesis (Chapter 2).
Purity, however, is not an absolute requirement since crystals of macro-
molecules can sometimes be obtained from mixtures. But such crystals are
mostly small or grow as polycrystalline masses, are not well shaped, and are of
bad diffraction quality, and thus cannot be used for diffraction studies. How-
ever, crystallization of macromolecules from mixtures may be used as a tool
for purification (47), especially in industry (48). For the purpose of X-ray
 R. Giege and A, Ducruix

Table 1. Parameters affecting the crystallization (and/or the solubility) of macro-

moleculesa

Intrinsic physico-chemical parameters

Supersaturation (concentration of macromolecules and precipitants)
Temperature, pH (fluctuations of these parameters)
Time (rates of equilibration and of growth)
Ionic strength and purity of chemicals (nature of precipitant, buffer, additives)
Diffusion and convection (gels, microgravity)
Volume and geometry of samples and set-ups (surface of crystallization chambers)
Solid particles, wall and interface effects (e.g. homogeneous versus heterogeneous
nucleation, epitaxy)
Density and viscosity effects (differences between crystal and mother liquor)
Pressure, electric and magnetic fields
Vibrations and sound (acoustic waves)
Sequence of events (experimentalist versus robot)

Biochemical and biophysical parameters

• Sensitivity of conformations to physical parameters (e.g. temperature, pH, ionic
strength, solvents)
• Binding of ligands (e.g. substrates, cofactors, metal ions, other ions)
• Specific additives (e.g. reducing agents, non-ionic detergents, polyamines) related with
properties of macromolecules (e.g. oxidation, hydrophilicity versus hydrophobicity,
polyelectrolyte nature of nucleic acids)
• Ageing of samples (redox effects, denaturation, or degradation)

Biological parameters
• Rarity of most biological macromolecules
• Biological sources and physiological state of organisms or cells (e.g. thermophiles
versus halophiles or mesophiles, growing versus stationary phase)
• Bacterial contaminants

Purity of macromolecules
• Macromolecular contaminants (odd macromolecules or small molecules)
• Sequence (micro) heterogeneities (e.g. fragmentation by proteases or nucleases—
fragmented macromolecules may better crystallize —, partial or heterogeneous post-
translational modifications)
• Conformational (micro) heterogeneities (e.g. flexible domains, oligomer and conformer
equilibria, aggregation, denaturation)
• Batch effects (two batches are not identical)

aAlthough all these parameters have not been screened systematically, especially for the crystal-
lization of a given macromolecule, all of them have been evaluated individually in isolated cases.

crystallography, high-quality monocrystals of appreciable size (0.1 mm at least

for the dimension of a face) are needed. It is our belief that poor purity is the
most common cause of unsuccessful crystallization, and for crystallogenesis
the purity requirements of macromolecules have to be higher than in other
fields of molecular biology. Purity has to be of 'crystallography grade': the
 1: An introduction to the crystallogenesis
macromolecules not only have to be pure in terms of lack of contaminants,
they have also to be conformationally 'pure' (32). Denatured macro-
molecules, or macromolecules with structural microheterogeneities, adversely
affect crystal growth more than do unrelated molecules, especially when
structural heterogeneities concern domains involved in crystal packing. On
the other hand, the presence of microquantities of proteases (or nucleases)
can alter the structure of the macromolecules during storage or the rather
long time needed for crystallization. As a consequence, when starting a
crystallization project one has to be primarily concerned with purification
methodologies and to take all precautions against protease and nuclease
action. To have reproducible results, the physiological state of cells should be
controlled, because protease (or nuclease) levels may vary as well as the
balance between cellular components. As a general rule, batches of macro-
molecules should not be mixed and crystallization experiments should be
conducted on fresh material so that ageing phenomena are limited. For more
details see Chapter 2.
In summary, we emphasize the importance of macromolecular purity in
biological crystallogenesis, and in cases of unsuccessful experiments we rec-
ommend first improvement or reconsideration of the purification procedure
of the molecules of interest.
3.3 Solubility and supersaturation
To grow crystals of any compound, molecules have to be brought in a super-
saturated, thermodynamically unstable state, which may develop in a crystal-
line or amorphous phase when it returns to equilibrium. Supersaturation can
be achieved by slow evaporation of the solvent or by varying parameters
(listed in Table 1). The recent use of pressure as a parameter is of note (49).
From this it follows that knowledge of macromolecular solubility is a pre-
requisite for controlling crystallization conditions. However, the theoretical
background underlying solubility is still controversial, especially regarding
salt effects (50), so that solubility data almost always originate from experi-
mental determinations. Specific quantitative methods permitting such deter-
minations on small protein samples are available (51-53). The main output
was the experimental demonstration of the complexity of solubility behaviours,
emphasizing the importance of phase diagram determinations for a rational
design of crystal growth (Chapter 10).
As to the nature of the salt used to reach supersaturation, one can wonder
why ammonium sulfate is so frequently chosen by crystal growers (54). This
usage is in fact incidental and results from the practices of biochemists for
salting-out proteins. Indeed many other salts can be employed, but their
effectiveness for inducing crystallization is variable (52). The practical
consequence is that protein supersaturation can be reached (or changed) in a
large concentration range of protein and salt, provided that adequate salts are
used.
9
 R. Giege and A. Ducruix

3.4 Nucleation, growth, and cessation of growth

Because proteins and nucleic acids require denned pH and ionic strength for
stability and function, biomacromolecule crystals have to be grown from
chemically rather complex aqueous solutions. Crystallization starts by a
nucleation phase (i.e. the formation of the first ordered aggregates) which is
followed by a growth phase. Nucleation conditions are sometimes difficult to
reproduce, and thus seeding procedures with preformed crystalline material
should not be overlooked as in many cases they represent the only method to
obtain reproducible results (55) (Chapter 7). It should be noticed that nucleation
requires a greater supersaturation than growth, and that crystallization rates
increase when supersaturation increases. Thus nucleation and growth should
be uncoupled, which is almost never done consciously but occurs sometimes
under uncontrolled laboratory conditions. From a practical point of view, inter-
face or wall effects as well as shape and volume of drops can affect nucleation
or growth, and consequently the geometry of crystallization chambers or drops
has to be defined. For additional information see Chapter 11.
Cessation of growth can have several causes. Apart from trivial ones, like
depletion of the macromolecules from the crystallizing media, it can result
from growth defects, poisoning of the faces, or ageing of the molecules. Better
control of growth conditions, in particular of the flow of molecules around
the crystals, may in some cases overcome the drawbacks as was shown in
microgravity experiments (refs 56, 57, and Chapter 6).

3.5 Packing
With biological macromolecules, crystal quality may be correlated with the
packing of the molecules within the crystalline lattices, and external crystal
morphology with internal structure. As shown by the periodic bond chain
(PBC) method, direct protein-protein contacts are essential in determining
packing and morphology (Chapter 11). Forces involved in packing of macro-
molecules may be considered as weak as compared to those maintaining the
cohesion of small molecule crystals. They involve salt bridges, hydrogen bonds,
Van der Waals, dipole-dipole, and stacking interactions (58-60). It must also
be borne in mind that the weak cohesion of macromolecular crystals results
from the fact that only a small part of macromolecular surfaces participate in
intermolecular contacts (61), the remaining being in contact with the solvent
(exceptions may be found for small proteins). This explains the commonly
observed polymorphism of biological macromolecular crystals.

4. From empiricism to rationality

4.1 Towards a better understanding of parameters
To date, the major parameters underlying crystallization of macromolecules
have been recognized (Table 1), and if their respective contributions in the
10
 1: An introduction to the crystallogenesis
crystallization process are not known with certainty, the theoretical frame
needed for explaining their role is well established (11-13, 46, 56). Also, the
experimental tools exist that are needed for measuring the contributions of
these parameters. This is the case for solubility and aggregation state
measurements as well as for monitoring pH and temperature (Chapters 10
and 11), or even the effect of microgravity (Chapter 6). Correlations between
the variation of a parameter and the ability of a given macromolecule to
crystallize are expected to be found. Finally, and perhaps the most important,
was the recognition of the importance of purity. Thus again we emphasize that
experimenters should primarily devote their efforts to starting crystallization
attempts with molecules of the highest biochemical quality.

4.2 Towards an active control of crystal growth

Only few attempts have been published describing the active control of
crystallization experiments. In general the history of experiments is not well
known, because crystal growers do not monitor parameters. This is especially
the case for temperature, which is almost never known with accuracy, even if
experiments are conducted in thermostated cabinets (this may be advan-
tageous because many microconditions may be screened, although at the cost
of reproducibility). Also the kinetics of events are practically never moni-
tored. In the following chapters methods to control macromolecular crystal
growth will be described (e.g. for the kinetics of evaporation in vapour diffusion
crystallization, see Chapter 5; for active video and temperature control, see
Chapter 11). Although these different aspects are all in their infancy, and
required instrumentation often does not exist, or exists only as prototypes, we
believe that user-friendly methodologies will be developed soon, and that
laboratories may be equipped with the adequate instrumentation.

4.3 The future of biological crystallogenesis

As mentioned before, modern genetic methods give access to molecules
present at very low amounts in cells and help to solve problems linked to
structural or conformational heterogeneities of proteins reluctant to crystal-
lize. In particular, engineering of active variants containing compact cores will
permit easier crystallizations of proteins with flexible domains. Macro-
molecular engineering will certainly find many applications in the RNA
world, where many ribozymes, pseudo-knots, and other RNA constructs
deserve structural investigations (Chapter 8). On the other hand, search of
crystal perfection, preparation of large size crystals, together with improve-
ment of data collection methods at synchrotron or neutron sources, are other
challenges, notably for the resolution of structures at highest resolution and
time-resolved crystallography (62).
For the biologist, studying crystal growth should be correlated with bio-
logical problems, and crystallization projects on macromolecular complexes,
11
 R. Giege and A. Ducruis
on membrane proteins (Chapter 9), and especially on engineered proteins,
are being developed. For the physicist, growing large monocrystals can be a
goal in itself, and one might speculate that exploration of optical, electrical,
mechanical, and other physical properties of crystalline arrays made from
biological macromolecules or assemblies can lead to novel frontiers in the
material science of tomorrow. Finally, for the chemist, usage of chemical and
molecular biology tools could lead in future to the design of molecular devices
and other nanostructures mimicking macromolecular crystals, as was discussed
for artificial self-assembling nucleic acids (63).
In conclusion, the rational approach for prompt crystallizations will de-
mand a synergy between biochemically and physically directed research and
usage of automated methods for the control of nuclealion and growth, as well
as the rapid preparation of high-quality monocrystals.

5. Search for crystallization strategies

No universal answer(s) can be given to the obvious questions about how to
start a crystallization project and what kind of strategy would be the most
appropriate to be successful. Because of the multiparametric nature of the
crystallization process and the diversity of the individual properties of pro-
teins, the advice would be to collect all possible information on the protein
one intends to crystallize, so that to hierarchize the variables of the process
and if possible to restrict their number. To this aim the questionnaire in
Protocol 1 addresses a number of basic questions.

Protocol 1. Before starting crystallizations, what are the

biological and biochemical characteristics of your
protein?

If your project concerns nucleic acids or their complexes with proteins

see Chapter 8; for membrane proteins see Chapter 9.

A. Biology and production

1. What is the biological origin of your protein? From a micro-organism
(mesophile orthermophile), a plant, an animal (which tissue)? etc.
2. Is the gene of the protein sequenced?
3. Is the protein cloned? In what expression vector and in what cells was
it overexpressed (E. coli, yeast, baculovirus, others)?
4. Is it a his-tagged protein? If yes do you plan to crystallize it with or
without the tag?

12
 1: An introduction to the crystallogenesis
5. Is it a fusion protein? If yes, which protease is used to cleave the
protein?
6. How many mg/litre of culture can you produce?
7. How many mg of protein can you obtain per standard purification?

B. Biochemistry
1. How long does it take to purify one batch of protein?
2. How do you assess the purity of the protein?
(a) Electrophoresis (native/denaturating conditions).
(b) HPLC (which phase).
(c) Ion spray (mass spectrometry).
(d) Checking of N-terminus.
(e) Activity assay.
2. What are the principal characteristics of the protein?
(a) Molecular weight.
(b) Isoelectric point (calculated or measured).
(c) Glycosylation (yes/no).
(d) Number of free cysteine(s).
(e) Number of disulfide bridges.
(f) Hydrophobicity or hydrophilicity.
(g) What are the ligands?
3. What are the friendly (or unfriendly) solvents?
4. Is the protein monomeric of oligomeric? How did you check it
(chromatography, light scattering, others)? Does your protein has a
tendency to aggregate?
5. What is the stability of the protein versus time, temperature, or pH?

The better these questions can be answered, the easier will be the design of
a crystallization strategy. Good knowledge of the characteristics of the pro-
tein, of its availability, will guide the experimenter. Quite often, it helps
people to be aware that the premise of crystallization may be as important as
the crystallization itself.

References
1. Pickworth Glusker, J. and Trueblood, K. N. (1985). Crystal structure analysis, a
primer. Oxford University Press, New York.

13
 R. Giege and A. Ducruix
2. Drenth, J. (1995). Principles of protein X-ray crystallography. Springer-Verlag,
Berlin.
3. Wright, J. D. (1987). Molecular crystals. Cambridge University Press, Cambridge.
4. Lima-de-Faria, J. (ed.) (1990). Historical atlas of crystallography. Kluwer,
Dordrecht.
5. Hajdu, J., Acharaya, K. R., Stuart, D. I., Barford, D., and Johnson, L. N. (1988).
Trends Biochem. Sci., 13, 104.
6. Mozzarelli, A. and Rossi, G. L. (1996). Annu. Rev. Biophys. Biomol. Struct., 25,
3430.
7. Blundell, T. L. and Johnson, L. M. (1976). Protein crystallography. Academic
Press, New York.
8. Wukovitz, W. and Yeates, T. O. (1995). Nature Struct. Biol., 2, 1060.
9. Usha, R., Johnson, J. E., Moras, D., Thierry, J.-C, Fourme, R., and Kahn, R.
(1984). J. Appl. Crystallogr., 17, 147.
10. Kay, L. L. (1986). ISIS/J. Hist. Sci. Soc., 77, 450.
11. Feigelson, R. S. (1988). J. Cryst. Growth, 90, 1.
12. Boistelle, R. and Astier, J.-P. (1988). J. Cryst. Growth, 90, 14.
13. Rosenberger, F., Vekilov, P. G., Muschol, M., and Thomas, B. R. (1996). J. Cryst.
Growth, 168, 1.
14. Lehman, C. G. (1853). Lehrbuch der physiologische Chemie. Leipzig.
15. Reichert, E. T. and Brown, A. P. (1909). The differentiation and specificity of
corresponding proteins and other vital substances in relation to biological classifi-
cation and evolution: the crystallography of hemoglobins. Carnegie Institution,
Washington DC.
16. Debru, C. (1983). L'esprit des proteines: histoire et philosophic biochimiques.
Hermann, Paris.
17. McPherson, A. (1990). J. Cryst. Growth, 110, 1.
18. Sumner, J. B. (1926). J. Biol. Chem., 69, 435.
19. Northrop, J. H., Kunitz, M., and Herriot, R. M. (1948). Crystalline enzymes.
Columbia University Press, New York.
20. Jabri, E., Carr, M. B., Hausinger, R. P., and Karplus, P. A. (1995). Science, 268,
998.
21. Dounce, A. L. and Allen, P. Z. (1988). Trends Biochem. Sci., 13, 317.
22. Bernal, J. D. and Crowfoot, D. (1934). Nature, 133, 794.
23. Jones, C., Mulloy, B., and Sanderson, M. R. (ed.) (1996). In Methods in molecular
biology. Crystallographic methods and protocols, Vol. 114, pp. 1-394. Humana
Press, Totowa, NJ, USA.
24. McPherson, A. (1982). Preparation and analysis of protein crystals. Wiley, New
York.
25. Kim, S. H. and Rich, A. (1968). Science, 162, 1381.
26. Dock, A.-C, Lorber, B., Moras, D., Pixa, G., Thierry, J.-C., and Giege, R. (1984).
Biochimie, 66, 179.
27. Michel, H. (1982). J. MoL Biol, 158, 567.
28. Kuhlbrandt, W. (1988). Q. Rev. Biophys., 21, 429.
29. Arnoux, B., Ducruix, A., Reiss-Husson, F., Lutz, M., Norris, J., Schiffer, M., et al.
(1989). FEES Lett, 258, 47.
30. Michel, H. (ed.) (1991). Crystallization of membrane proteins. CRC Press, Boca
Raton, FL, USA.
14
 1: An introduction to the crystallogenesis
31. Jeruzalmi, D. and Steitz, T. A. (1997). J. Mol. Biol, 274, 748.
32. Giege, R., Dock, A.-C., Kern, D., Lorber, B., Thierry, J.-C, and Moras, D. (1986).
J. Cryst. Growth, 76, 554.
33. Yonath, A., Frolow, F., Shoham, M., Mtissig, J., Makowski, I., Glotz, C, et al.
(1988). J. Cryst. Growth, 90, 231.
34. Trakhanov, S., Yusupov, M., Shirikov, V., Garber, M., Mitschler, A., Ruff, M., et
al. (1989). J. Mol. Biol, 209, 327.
35. Feigelson, R. S. (ed.) (1986). Proc. 1st Int. Conf. Protein Crystal Growth,
Stanford, CA, USA, 1985. J. Cryst. Growth, 76, 529.
36. Giege, R., Ducruix, A., Fontecilla-Camps, J., Feigelson, R. S., Kern, R., and
McPherson, A. (ed.) (1988). Proc. 2nd Int. Conf. Crystal Growth of Biological
Macromolecules, Bischenberg, France, 1987. J. Cryst. Growth, 90, 1.
37. Carter, C. W., Jr. (ed.) (1990). Methods: a companion to methods in enzymology,
Vol. 1, pp. 1-127.
38. Ward, K. and Gilliland, G. (ed.) (1990). Proc. 3rd Int. Conf. Crystal Growth of
Biological Macromolecules, Washington, DC, USA, 1989. J. Cryst. Growth, 110, 1.
39. Stezowsky, J. J. and Littke, W. (ed.) (1992). Proc. 4th Int. Conf. Crystal Growth of
Biological Macromolecules, Freiburg, Germany, 1991. J. Cryst. Growth, 122, 1.
40. Glusker, J. P. (ed.) (1994). Proc. 5th Int. Conf. Crystal Growth of Biological
Macromolecules, San Diego, CA, USA, 1993. Acta Cryst., D50, 337.
41. Miki, K., Ataka, M., Fukuyama, K., Higuchi, Y., and Miyashita, T. (ed.) (1996).
Proc. 6th Int. Conf. Crystal Growth of Biological Macromolecules, Hiroshima,
Japan, 1995. J. Cryst. Growth, 168, 1.
42. Drenth, J., and Garcia-Ruiz, J. M. (ed.) (1999). Proc. 7th Int. Conf. Crystal
Growth of Biological Macromolecules, Granada, Spain, 1998. J. Cryst. Growth, 196,
pp. 185-720.
43. Wood, S. P. (1990). In Protein purification applications: a practical approach (ed.
E. L. V. Harris and S. Angal), pp. 45-58. IRL Press, Oxford.
44. Weber, P. C. (1991). Adv. Protein Chem., 41, 1.
45. McPherson, A., Malkin, A. J., and Kuznetsov, Y. G. (1995). Structure, 3, 759.
46. Durbin, S. D. and Feher, G. (1996). Annu. Rev. Phys. Chem., 47,171.
47. Jakoby, W. B. (1971). In Methods in enzymology (ed. W. B. Jakoby), Vol. 22,
pp. 248-52. Academic Press, London.
48. Judge, R. A., Johns, M. R., and White, E. T. (1995). Biotechnol. Bioeng., 48, 316.
49. Lorber, B., Jenner, G., and Giege, R. (1996). J. Cryst. Growth, 103, 117.
50. Von Hippel, P. H. and Schleich, T. (1969). In Structure and stability of biological
macromolecules (ed. S. N. Timasheff and G. D. Fasman), Vol. 2, pp. 417-574.
Dekker.
51. Mikol, V. and Giege, R. (1989). J. Cryst. Growth, 97, 324.
52. Ries-Kautt, M. and Ducruix, A. (1989). J. Biol. Chem., 264, 745.
53. Cacioppo, E., Munson, S., and Pusey, M. L. (1991). J. Cryst. Growth, 110, 66.
54. Gilliland, G. L., Tung, M., Blakeslee, D. M., and Ladner, J. E. (1994). Acta Cryst.,
D50, 408.
55. Thaller, C., Weaver, L. H., Eichele, G., Wilson, E., Karlson, R., and Jansonius, J.
N. (1981). J. Mol. Biol., 147, 465.
56. Giege, R., Drenth, J., Ducruix, A., McPherson, A., and Saenger, W. (1995). Prog.
Cryst. Growth Charact., 30, 237.
57. McPherson, A. (1997). Trends Biotechnol. 15, 197.
15
 R. Giege and A. Ducruix
58. Bergdoll, M. and Moras, D. (1988). J. Cryst. Growth, 90, 283.
59. Salemme, F. R., Genieser, L., Finzel, B. C., Hilmer, R. M., and Wendolosky, J. J.
(1988). J. Cryst. Growth, 90, 273.
60. Wang, A. H. J. and Teng, M. K. (1988). J. Cryst. Growth, 90, 295.
61. Carugo, O. and Argos, P. (1997). Protein Set., 6, 2261.
62. Chayen, N. E., Boggon, T. J., Cassetta, A., Deacon, A., Gleichmann, T., Habash,
J., etal. (1996). Q. Rev. Biophys., 29, 227.
63. Seeman, N. C. (1991). Curr. Opin. Struct. BioL, 1, 653.

16
 2

Biochemical aspects and handling of

macromolecular solutions and
crystals
B. LORBER and R. GIEGE

1. Introduction
The quality and quantity of the macromolecular samples are important pre-
requisites for successful crystallizations. Proteins and nucleic acids extracted
from living cells or synthesized in vitro differ from small molecules by
additional properties intrinsic to their chemical nature and their larger size.
They are frequently difficult to prepare at a high degree of purity and
homogeneity. Besides traces of impurities, harsh treatments may decrease their
stability and activity through different kinds of alterations. Consequently, the
quality of biomacromolecules depends on the way they are prepared and
handled. As a general rule purity and homogeneity are regarded as conditions
sine qua non. Accordingly, purification, stabilization, storage, and handling of
macromolecules are essential steps prior to crystallization attempts. Other
difficulties in crystal growth may come from the source of the biological
material. It is advisable to have at disposal a few milligrams of material when
starting first crystallization trials although structures were solved with sub-
milligram quantities of protein (1). Once crystals suitable for X-ray analysis
can be produced, additional material is often needed to improve their quality
and size and to prepare heavy-atom derivatives. It is thus essential that
isolation procedures are able to supply enough fresh material of reproducible
quality. Similar situations are encountered with multi-macromolecular
assemblies (e.g. viruses, nucleosomes, ribosomal particles, or their subunits).
This chapter discusses biochemical methods used to prepare and character-
ize macromolecules intended for crystallization assays. Practical aspects
concerning manipulation and qualitative analyses of soluble proteins will be
emphasized. The cases of nucleic acids and membrane proteins are described
in more detail in Chapters 8 and 9. Peculiar aspects of molecular biology that
are important for crystallogenesis are presented in Chapter 3. They include
the design of engineered macromolecules with new physical properties or
 B. Lorber and R. Giege
modified to simplify purification or crystallographic analysis. Finally, methods
for identification of macromolecular content of crystals and measurements of
their density are presented as well.

2. The biological material

2.1 Sources of biological macromolecules
Many biological functions are sustained by classes of proteins and nucleic
acids universally present in living organisms so that the source of macro-
molecules may seem unimportant. In fact, better crystallization conditions or
better diffracting crystals are frequently found by switching from one
organism to another. Variability in sequences between heterologous macro-
molecular species may lead to different conformations and consequently to
different crystallization behaviours. Also, differences in crystal quality may
result from addition or suppression of intermolecular contacts because of the
high solvent content (50-80%) (2) and the existence of relatively few contacts
in macromolecular crystal lattices. In practice, proteins isolated from
eukaryotes are frequently more difficult to crystallize than their prokaryotic
counterparts due to the presence of additional flexible domains. In
eukaryotes, post-translational modifications are often responsible for struc-
tural and conformational microheterogeneity. Proteins isolated from thermo-
philic micro-organisms are more stable at higher temperatures than those
from other organisms and may be more amenable to crystallization. Com-
ponents of the protein biosynthesis machinery are examples (3, 4). Proteins
from halophilic micro-organisms are alternative candidates whose stability is
optimal in the presence of high salt concentrations close to those needed to
reach supersaturation but their purification presents methodological diffi-
culties (5). Finally, the physiological state of the cells and the 'freshness' of the
starting material may be important. Certain proteins from unicellular organ-
isms are isolated in their native state only when cells are in exponential or pre-
stationary growth phase (6-8). Since catabolic processes are predominant in
tissues of dead organisms, they should not be stored before use unless they
are frozen immediately post mortem. Plant material should be processed
immediately after harvest or quick-frozen for storage.

2.2 Macromolecules produced in host cells or in vitro

In the past, macromolecules that were most abundant, easy to isolate, and
most stable were the first crystallized. Today, many researchers deal with
biological molecules that are only present in trace amounts and the pre-
paration of the quantities needed for crystallization assays can become a limit-
ing step. In a number of cases this problem can be circumvented owing to the
advancement of genetic engineering methods which make it possible to clone
and overexpress genes in bacterial or eukaryotic cells (9, 10). In recombinant
18
 2: Biochemical aspects and handling of macro-molecular solutions
bacteria with multicopy plasmids non-toxic proteins can accumulate and in
exceptional cases reach a quarter of total proteins (more typically a 100 mg of
protein are isolated with a good yield from less than a 100 g of bacterial host
cells). High intracellular concentration of certain proteins may lead to growth
inhibition (11) or formation of inclusion bodies of denatured, aggregated, or
even pseudo-crystallized material that requires adapted isolation procedures
(12). Therefore, proper vectors and host cells must be chosen to optimize
over-production levels and to retain native conformation and functional
activity of the proteins (13). Rare codons may reduce translation rate and
efficiency due to limiting concentrations of minor tRNA isoacceptors or amino
acid misincorporation (14, 15). Separation of foreign macromolecules from
endogenous ones can be difficult in the absence of specific biological assays
(16). Finally, the presence of ligands or chaperones may be indispensable to
maintain the native conformation of certain proteins (17). All these factors
have to be kept in mind when planning a purification strategy.
Overexpressed proteins can be hypomodified because maturation enzymes
responsible for the co- or post-translational modifications may not work with
the same efficiency on recombinant macromolecules. Protein glycosylation,
unknown in prokaryotes, occurs to various extents in eukaryotes. Mammalian
proteins are glycosylated in yeast and in insect cells infected by baculovirus
(18) but different modification patterns lead to structural variants (19)
especially when modification enzymes are limiting. Continuous cell-free
translation systems are potential alternatives to produce natural or factitious
proteins for crystallization (20). Similarly, modification of nucleotides may be
partial when RNA genes are overexpressed in host cells (21) and are absent
when transcribed in vitro. RNAs produced in vitro by polymerases (from
bacteriophages SP6 or T7) are frequently heterogeneous in length because
transcription does not always terminate at a unique position (Chapter 8).
Changes introduced in genes by site-directed mutagenesis (22) may result in
proteins with altered conformations, decreased stability and activity, per-
turbed folding, changed modified post-synthetic modifications and solubility,
or unforeseen degradations. Amino acid substitutions to increase stability or
solubility, deletion of flexible domains, fusion with sequences binding to
immobilized ligands in affinity chromatography or with polypeptides may
facilitate crystallization. Mutants that are more suitable for the preparation of
heavy-atom derivatives and selenomethionine containing proteins for
anomalous scattering studies are now of general use (Chapter 3). Crystal-
lization of protein-RNA complexes also profits from advantages of protein
engineering methods (23).

3. Isolation and storage of pure macromolecules

Methods for the purification of proteins (24) and nucleic acids (Chapter 8)
have been reviewed extensively. A few practical points are discussed here.
19
 B. Lorber and R. Giege
Readers are encouraged to consult manufacturer's or supplier's catalogues
and application notes for updated technical information.
3.1 Preparative isolation methods
3.1.1 Generals
The manifold properties of macromolecules makes it difficult to give a general
scheme to facilitate purification of an unknown biomolecule. Optimized pro-
tocols are usually achieved by trial-and-error approaches but more systematic
procedures have been described (25). In such protocols, the sequence of
events is important because macromolecules in crude extracts may be pro-
tected by interaction with Hgands or other macromolecules which will prob-
ably be eliminated during isolation. Harmful compounds, like hydrolases,
must be separated as early as possible. In no case they should be enriched or
co-fractionated. Table 1 lists major purification methods together with the
appropriate equipment.

Table 1. Methods and equipment for the purification of biological macromolecules

• Cell culture
Fermentors, culture flasks and plates, thermostated cabinets
High capacity centrifuges or filtration devices for cell recovery
• Cell disruption
Mechanical disruption devices (grinders, glass bead mills, French press)
Chemical treatments (e.g. phenolic extraction of small RNAs)
Biochemical treatments (e.g. cell lysis by enzymes)
Others (e.g. sonication, freezing/thawing)
• Centrifugation
Low speed centrifuge (to remove cell debris or recover precipitates)
High speed centrifuge (to fractionate subcellular components)
• Dialysis and ultrafiltration
Dialysis tubing (hollow fibres or membranes of various porosities and sizes)
Concentrators (from 0.5 ml to a few litres with high flow rate low macromolecule-
binding membranes of various cut-offs)
• Chromatography (prefer metal-free systems)
Low pressure equipment for fast separation (FPLC, hyperdiffusion, perfusion)
High pressure equipment (HPLC)
Columns of various capacities filled with various matrices (particle size 10-30 mm)
Pumps, programmer, on-line absorbance detector, fraction collector, recorder
• Preparative electrophoresis and isoelectric focusing
Electrophoresis apparatus for large rod or slab gels
Preparative liquid IEF apparatus (column or horizontal cells)
Power supplies
• Detection, characterization, and quantitation
Spectrophotometer, fluorimeter
pH meter, conductimeter, refractometer (to monitor chromatographic elution)
Liquid scintillation counter (for radioactivity detection)
Analytical gel electrophoresis, capillary electrophoresis, and IEF equipment

20
 2: Biochemical aspects and handling of macromolecular solutions
The preparation of a cellular extract and fractionation of its components
are the two stages common to most purification protocols (except for macro-
molecules secreted in culture media). Intracellular macromolecules are
released using physical, chemical, or biological disruption methods and
extracts are clarified by centrifugation or ultrafiltration. Membrane proteins
and proteins with hydrophobic surfaces are solubilized with detergents or
sulfobetaines (26) (Chapter 9). Extracellular compounds and macromolecules
synthesized in vitro may be recovered either by ultrafiltration, centrifugation,
flocculation, or liquid-liquid partitioning.
3.1.2 Proteins
Gross fractionation includes one or several precipitations induced either by
addition of salts (e.g. ammonium sulfate), organic solvents (e.g. acetone), or
organic polymers (e.g. PEG). Temperature or pH variation are applied to
decrease solubility or stability of unwanted macromolecules. Fractionation
between two liquid phases and selective precipitation (e.g. of nucleic acids by
protamine) are additional methods. The next steps involve more resolutive
methods, generally a combination of column chromatographies. These are
based on separation by charge (adsorption, anion or cation exchange,
chromatofocusing), hydrophobicity (hydrophobic interaction (HIC) or
reverse-phase (RFC) chromatographies), size (exclusion chromatography),
peculiar structural features (e.g. affinity for heparin, antibodies, metal ions, or
thiol groups), or activity (affinity for catalytic sites, receptors, or biomimetic
compounds). HPLC yields higher resolution than standard techniques
because of the monodispersity and small size of the spherical matrix particles
(27) and new matrices take advantage of hyperdiffusion (Beckman) or per-
fusion (PerSeptive Biosystems) to accelerate elution. Preparative IEF is
carried on in gels with free or immobilized ampholytes (Immobiline®,
Pharmacia) (28) in rotating cells divided in compartments by permeable nylon
grids (Rotofor®, Bio-Rad) or in multichamber units holding fixed pH mem-
branes (Isoprime™, Pharmacia). Differential centrifugation and free flow
electrophoresis (29) are other methods. Monitoring of specific activities
during the purification procedure helps to identify unsatisfactory steps in
which macromolecules are lost or inactivated. Guidelines for effective protein
purification may be summarized as follows:
• work in the cold room (i.e. at 4°C) with chilled equipment and solutions if
the protein is unstable at higher temperature
• use precipitation steps to speed up fractionation
• limit the number of chromatographies (to three or four)
• prefer quick assays to characterize macromolecules
• use short and efficient non-denaturing intermediary treatments (e.g.
repeated dialysis)
• add stabilizing agents and protease inhibitors (see Section 5.5).
21
 B. Lorber and R. Giege
In summary, success in crystallization is often dependent on rapid purifi-
cation. To reach this aim advanced equipment and chromatographic systems
enabling high flow rate (e.g. advanced HPLC and perfusion chromatography)
are recommended.
3.1.3 Nucleic acids
Purification of nucleic acids requires specific methods. For tRNAs, phenol
extraction precedes counter-current distribution or chromatography on
benzoylated DEAE-cellulose (also on Sepharose® or other matrices).
Further purification is based on anion exchange, adsorption, reverse-phase,
mixed-mode, hydrophobic interaction, perfusion, or affinity chromato-
graphies. Intermediary treatments include precipitation by ethanol, dialysis,
and concentration by evaporation under vacuum. HPLC on coated silica
substituted by short aliphatic chains (C4) gives separations with good
resolution. Oligo-DNA or RNA are synthesized chemically on solid phase
supports or enzymatically in vitro. Abortive sequences are eliminated by
denaturing gel electrophoresis (Prepcell™, Bio-Rad) or HPLC (30-32). For
further details see Chapter 8.

3.2 Specific preparation methods

Re-chromatography of samples prior to crystallization removes minor con-
taminants (e.g. degradation or ageing products), aggregates appearing during
storage or small molecules (e.g. additives like glycerol) (33), and favours
crystallization. HIC (at a 1-100 mg scale) can bring proteins and nucleic acids
directly into solutions containing crystallizing agents (e.g. salt, MPD, or PEG)
(34, 35). Purification on hydroxyapatite was cardinal for crystallizing some
proteins. Microscale concentration devices are useful to remove small size
contaminants and exchange buffers prior to crystallization. Additional advice
is given in Section 5.5.

3.3 Stabilization and storage

Macromolecules extracted from cells must be kept in solutions having pro-
perties close to those of the cellular medium in order to maintain native con-
formations. Storage under improper conditions spoils the precious material
obtained after long and hard work. Buffers whose pK is only weakly affected
by temperature (36) are recommended to avoid pH variations in samples
stored frozen and assayed at room temperature. Macromolecules prone to
aggregation require a minimal ionic strength to stay soluble but all ions may
not be compatible with their native structure or activity. Denaturation is
minimized by avoiding pH or temperature extremes as well as contact with
organic solvents, chaotropic agents, or oxidants (37, 38). Thiol groups in
proteins require a reducing agent (e.g. DTE, DTT, 2-mercaptoethanol, or
glutathione). Finally, diluted proteins may adsorb onto the walls of glass or
plastic containers (39).
22
 2: Biochemical aspects and handling of macromolecular solutions
Structure and stability of globular proteins depend upon intramolecular
hydrogen bonds. Interaction with ions of the Hofmeister series affects stability
(40) (Chapter 10). Ice formation during freezing or freeze-drying induce
partial unfolding (41). Therefore, do not freeze or lyophilize unprotected
macromolecules (the latter process removes bound water belonging to the
macromolecular solvation shell). Glycerol at high concentration (e.g. 50-60%,
v/v) stabilizes proteins (42) and stays liquid at -20 °C (its high viscosity reduces
diffusion by about two orders of magnitude). Trehalose, sucrose (43), and
other cosmotropic agents (44) preserve biomaterials by favouring more com-
pact structures and by reducing backbone thermal fluctuation (45). Storage as
a suspension in an ammonium sulfate solution is efficient but sometimes less
convenient since after some time it generates heterogeneities due to amidation
of glutamate and aspartate side chains. Ligands enhance protein stability.
Bactericidal or fungicidal agents (like highly toxic sodium azide, ethyl-
mercurithiosalicylate, or volatile thymol) should be added for storage or
crystallization. Nucleic acids stored dry or as alcoholic precipitates should be
free of phenol which leads to alkaline-type hydrolysis. RNA molecules are
chemically and structurally stable at slightly acidic pH (4.5-6.0) and in the
presence of Mg2+ (Chapter 8).

3.4 Ageing
Properties of macromolecular samples change with time as explicitly illus-
trated for lysozyme crystallization (46). Ageing results from the action of
contaminants present or introduced in samples or of modifications generated
by oxidants. In the example of lysozyme, changes in crystallizability are due to
the presence of fungi that multiply in the stored protein solution (46). Water
molecules or metal ions induce slow hydrolysis in RNAs (Chapter 8).
Self-cleaving macromolecules, like certain proteases and ribozymes, pose
specific problems. For protease crystallization the problem can be solve by
storing and co-crystallizing the protease with an active site inhibitor. For
instance, the three-dimensional structure of a human protease of the ICE type
participating in apoptosis could be solved because it was crystallized as a
covalent complex with a tetrapeptide inhibitor (47). For hammerhead
ribozyme, self-cleavage was prevented by introducing modified bases into the
molecule (48).

4. Characterization and handling of macromolecules

Numerous analytical tools are available to detect, characterize, and quantitate
macromolecules. This section deals with general methods of particular
interest for crystal growers and gives practical advice for handling pure
macromolecules.
23
 B. Lorber and R. Giege

4.1 Analytical biochemical methods

Gel electrophoresis quickly visualizes the macromolecular content of a
sample. Procedures for proteins are well known (49) and Protocol 1 is
adapted for small nucleic acids. Electrophoresis gives an estimate of the
apparent size of proteins (49) and nucleic acids (50). IEF in polyacrylamide
gels containing free or immobilized pH gradients separates proteins with
differences in isoelectric points smaller than 0.01 pH unit (51). Capillary
electrophoresis and mass spectrometry give rapid purity and homogeneity
diagnostics (52,53). Electronspray ionization mass spectra (ESI) inform about
the counterion distribution around a protein (54). ESI and MALDI (matrix-
assisted desorption/ionization) are the most sensitive and rapid methods to
detect microheterogeneities in macromolecular samples. HPLC is useful to
separate small amounts of macromolecules for further analysis (28). Sequence
analysis is complementary to amino acid composition determination (55).

Protocol 1. Gel electrophoresis of nucleic acids (up to 150-mers)

Caution! Acrylamide, A/,/Vmethylene bisacrylamide, 'Stains all', and

ethidium bromide are highly toxic. Wear gloves when working with their
solutions. Also protect your eyes by wearing glasses or a facial screen
when working with UV light.

Equipment and reagents

• PAGE equipment (e.g. Mini-Protean, Bio- » Urea (ultrapure or electrophoresis grade)
Rad) sold for protein electrophoresis . Saccharose or glycerol
• Acrylamide-bisacrylamide solution: 38% . Bromophenol blue
(w/v) acrylamide, 2% (w/v) N./V'methylene . Ultrafiltration membrane (0.45 um pore
bisacrylamide size) (Millipore)
. 20 x Tris-borate buffer: 243 g Tris base, 110 . 'Stains all, dye powder (Eastman Kodak,
g boric acid, 18.6 gEDTA for 1 litre, pH 8.3 Aldrich)
. TEMED . Ethidium bromide
• Fresh 5% (w/v) solution of ammonium
peroxodisulfate

Method
1. Prepare a stock solution to make gels having a total acrylamide con-
centration T = 8% (w/v), a cross-linker concentration C = 5% (w/w),
and containing 8 M urea by mixing:
• 20 ml acrylamide-bisacrylamide solution
• 5 ml of 20 x Tris-borate buffer
• 50 g urea
• distilled water up to 100 ml
2. Filter the above stock solution on a 0.45 um pore size membrane.
24
 2: Biochemical aspects and handling of macro-molecular solutions
3. To 10 ml solution add 10 ul TEMED and 100 ul fresh 5% (w/v)
ammonium peroxodisulfate solution. Mix and pour in the mould.
Polymerization occurs in about 30 min.
4. Denature nucleic acids in a solution containing 8 M urea, 20% (w/v)
saccharose (or 20% (v/v) glycerol), and 0.025% (w/v) bromophenol
blue. Load samples onto the gel and run electrophoresis under
appropriate voltage.
5. Stain nucleic acids by soaking gels in the dark in a solution containing
30 mg 'Stains all', 100 ml dimethylformamide, and distilled water up to
1 litre. Destain in the light. Another technique employs electrophoresis
buffer containing 0.5 ug/ml ethidium bromide. Wait 10 min and view
the gel in UV light (254 nm). Silver stain techniques for proteins are
also suitable to visualize nucleic acids (49).

Methods to quantitate proteins are dictated by sample size (volume and

concentration) and required degree of specificity (56). Spectrophotometry
is non-invasive and accurate when extinction coefficients are known.
Theoretical molar absorption coefficients e of polypeptides can be calculated
from tryptophan and tyrosine content using:

where 5690 and 1280 are the rnolar absorption coefficients at 280 nm of
tryptophan and tyrosine, and nx and ny the numbers of tryptophan and
tyrosine residues, respectively (57). Hence, protein concentrations are
obtained from:

If the amino acid composition of a protein is unknown, the e coefficient

can be determined from ponderal, spectrophotometric, or refractive index
measurements, as well as from colorimetric dye binding assays, but results are
skewed when its composition deviates from that of the reference protein or
when contaminants interfere (57). In mixtures, protein concentrations are
estimated by empirical formulas eliminating the contribution of nucleic acids.
For quartz cuvettes of / cm optical pathway (58, 59):

or for a better estimate

Active site titration monitors the functionality of individual enzyme

molecules (60) but it cannot detect altered molecules whose activity is
unaffected. Immunological properties may be used to assess the ability of
antibodies to recognize conformational states or parts of molecules.
25
 B. Lorber and R. Giege
Approximate concentrations of RNA are obtained assuming that 1 A2eo nm
unit (for 1 cm path length) corresponds to 0.040 mg/ml. Values are more
accurate when extinction coefficients are known (61) from absorbance and
phosphorus content measurements (62).

4.2 Prevention of macromolecular damage

Pure macromolecules require special care to prevent damage or loss.
Following practical advice can contribute to the success of a crystallization
project:
(a) Concentrate samples by ultrafiltration in devices using pressure or
centrifugal force (cylindrical cells or parallel flow plates) or by dialysis
against hygroscopic compounds (e.g. dry, high MT PEG or gel filtration
matrices). Choose membranes with low binding capacity for stirred
pressure cells. Optimize stir-rate to impede denaturation through shear-
ing or adsorption onto membranes. Do not create oxidizing environments
(foam or air bubbles) to avoid formation of disulfide bonds in proteins.
(b) Centrifuge aggregates forming as a consequence of pH decrease,
oxidation, or increase of salt or protein concentration.
(c) Concentrate biomolecules by precipitation (e.g. add ammonium sulfate)
and dissolve the precipitate in a small volume, or adsorb them on a
chromatographic matrix and elute them at higher ionic strength.
(d) Beware of techniques concentrating contaminants (e.g. proteases and
nucleases).
(e) Exchange buffers and concentrate macromolecules over membranes with
appropriate cut-offs to eliminate small Mr compounds.
(f) Solubilize proteins with mild non-ionic detergents (e.g. octylglucoside) or
non-detergent sulfobetaines (26) (Chapter 9).
(g) Avoid high concentrations of denaturing agents (e.g. guanidinium
chloride, urea, or chaotropic detergents) which inactivate or unfold
macromolecules.
(h) Prepare buffers freshly with ultra-pure water and high-grade chemicals
and adjust their pH after mixing all ingredients (pH may change after
dilution or in the presence of other compounds). Purify suspicious
chemicals by crystallization, distillation, or chromatography.
(i) Add bactericidal or fungicidal agents (e.g. sodium azide, sodium
ethylmercurithiosalicylate, or thymol at 0.02%, w/v) in solutions (some
buffers, like toxic cacodylate ions, have bactericidal properties).
(j) Experiment on aliquots to limit handling of stock solutions and avoid
repeated freezing/thawing of macromolecules.
(k) Remove undesired molecules by dialysis, ultrafiltration, or size exclusion
chromatography.
26
 2: Biochemical aspects and handling of macromolecular solutions
(1) Prepare macromolecules with or without their ligands (e.g. coenzyme,
metal ions) or try additives (e.g. ions, reducing agents, chelators) to
search for conformers crystallizing more readily.
(m) Sterilize glass- or plasticware in contact with nucleic acids. Wear gloves
during manipulations; fingers are always contaminated by nucleases (63),
proteases, and bacteria.
(n) Store solutions in air-tight bottles to prevent contamination by airborne
micro-organisms.

5. The problem of purity and homogeneity

5.1 The concept of crystallography-grade' quality
The concept of purity takes a peculiar meaning in biological crystallogenesis
(36). Not only must molecules be pure, i.e. deprived of unrelated macro-
molecules or undesired small molecules, but they must be 'pure' in terms of
structure and conformation. In other words crystallization trials should be
done with homogeneous populations of conformers. This concept is based on
the fact that the best crystals can only be grown from solutions containing
well-defined entities with identical conformations and physico-chemical pro-
perties. Over the years it has been refined with the improvements of analytical
biochemical and biophysical micromethods.
The importance of purity may appear exaggerated and contradictory with
earlier views because crystallization can be used as a purification method in
chemistry and biochemistry (64). For structural studies, however, the aim is to
prepare monocrystals diffracting at high resolution with a good mosaicity and
a prolonged stability in the X-ray beam. It is thus understandable that con-
taminants may compete for sites on the growing crystals and generate lattice
errors leading to internal disorder, dislocations, irregular faces and secondary
nucleation, twinning, poor diffraction, or early cessation of growth (65).
Dynamic light scattering of contaminated protein solutions and in situ atomic
force microscopy (AFM) on growing crystals have provided convincing
evidence that these phenomena occur for biomolecules (66, 67). Because of
the high number of molecules in a single crystal (~ 1020 per mm3), p.p.m.
amounts of contaminant may induce formation of non-specific aggregates,
alter macromolecular solubility, or interfere with nucleation and crystal
growth (66, 67). The effects of impurities are reduced in gelified media (68, 69,
and Chapter 6). Successful crystallization of rare proteins and nucleic acids
support the importance of purity and homogeneity (1, 4, 33, 70).

5.2 Purity of samples

The level of confidence for the purity of a macromolecule depends upon the
resolution, specificity, and sensitivity of the methods used to identify con-
taminants. Protein or nucleic acid samples are seldom analysed for contamin-
27
 B. Lorber and R. Giege
ation by other classes of molecules. Although most of the contaminants are
eliminated along the purification steps through which the proteins or nucleic
acids have gone, traces of polysaccharides, lipids, proteases, or nucleases may
be sufficient to hinder crystallization. Small molecules, like peptides, oligo-
nucleotides, ammo acids, carbohydrates, or nucleotides as well as uncontrolled
ions should also be considered as contaminants. Buffer molecules remaining
from a former purification step can be responsible for irreproducible crystal-
lization (e.g. phosphate ions are relatively difficult to remove and may
crystallize in the presence of other salts). Counterions play a critical role in
the packing of biomolecules. Often macromolecules do not crystallize or yield
different habits in the presence of various buffers adjusted at the same pH.
Consequently, 'purity' means also that reagents used with pure macro-
molecules (e.g. precipitants, buffers, detergents, or additives) should be of the
highest grade. This is especially true for precipitants present at molar
concentrations. Contamination of a 2 M ammonium sulfate solution by only
0.001% (w/w) (1 mM) Pb2+ equals a stoichiometry of one molecule of
impurity per molecule of a protein of Mr 50 000 present at 5 mg/ml (note that
such contamination level may promote crystallization; for instance Cd2+ is
needed for ferritin crystallization because it is involved in packing contacts)
(71). Purification techniques for common precipitants are listed in Table 2.
Commercial detergents should also be repurified (75, 76). Chemicals in which
contaminants do not exceed a few p.p.m. are commercially available but the
label 'ultra pure' is sometimes exaggerated. Molecules released from non-
inert chromatography matrices (e.g. Sephadex, celluloses) by enzymatic
digestion or by desorption of organic compounds (e.g. organic phases bound
to silica matrices) fall in the category of impurities.

5.3 Microheterogeneity of samples

Microheterogeneity in pure macromolecules is only revealed by very
resolutive methods (Section 5.4). Although its causes are multiple, the most

Table 2. Techniques for the purification of major crystallization agents

Chemicals Major contaminants in Purification techniques

commercial batches and references
Ammonium sulfate Ca2+, Fe ions, Mg2+, Recrystallization
PbS04,a CN-, NO3-
PEG6 Cr-, F-, NO3-, PO42-, SO42-, Column chromatographyb (72, 73),
peroxides, aldehydes recrystallizationb
MPD cr-, k-, Na+, SO42- Distillation under vacuum,
column chromatography (74)

aNon-soluble species.
bSee Chapter 5, Protocol 1.

28
 2: Biochemical aspects and handling of macromolecular solutions

Table 3. Frequent sources of microheterogeneity in pure proteins and ribonucleic acids

Variation in primary structure (genetic variations, synthesis errors, hydrolysis)

Variation in secondary structure (misfolding or partial unfolding)
Variation in tertiary structure (conformers)
Variation in quaternary structure (oligomerization)
Molecular dynamics (flexible domains)
Incomplete post-transcriptional or post-translational modifications
Partial binding of ligands or foreign molecules
Aggregation (specific or non-specific)
Fragmentation (i.e. chemical or enzymatic hydrolysis)
Chemical alterations (e.g. partial oxidation of sulfhydryl groups, deamidation)
Others

common ones are uncontrolled fragmentation and post-synthetic modifica-

tions (Table 3). It must be emphasized that the strict identity in sequence of in
vivo or in vitro produced polypeptides or nucleic acids with that of natural
molecules should be confirmed.
5.3.1 Structural microheterogeneity
Proteolysis normally takes part in many physiological processes (e.g. mat-
uration, regulation of enzymatic activity, and catabolism) (77-82), and
represents a major difficulty to overcome during protein isolation because
proteases (Mr 20000-800000) are localized in various cellular compartments
or secreted in the extracellular medium. Proteases are distinguished by the
structure of their catalytic site containing either a serine, aspartic acid, or
cysteine residue, or a metal ion. Not all can be inhibited by commercial
compounds (Table 4). Upon cell disruption, cellular compartments are mixed
with extracellular proteases and control over proteolysis is lost. Decrease of
protein size or stability, modification of their charge or hydrophobicity, partial
or total loss of activity or of immunological properties are signs of proteolysis.
Hydrolysis of nucleic acids by nucleases is frequently detected on sequencing
gels (83). Traces of proteases or nucleases may not be detectable even when
overloading electrophoresis gels but they can cause damage during concentra-
tion or storage of samples. Fragmentation of RNA by chemical hydrolysis is
catalysed by metal ions and enhanced at alkaline pH (84) (Chapter 8).
Co- or post-translational enzymatic modifications generate microhetero-
geneities in proteins when different groups (e.g. oligosaccharide chains)
occupy all sites or when correct modifications are unevenly distributed over
the polypeptide chains (e.g. when all sites are not substituted). Over a
hundred modifications are known of which some are listed in Table 5. Most of
them require special methods for their analysis (85, 86). Only some modifica-
tions are reversible (e.g. phosphorylation) but not glycosylation or methyl-
ation. Heterogeneity in carbohydrate chains, either N-linked at asparagine or
O-linked at serine or threonine residues, is frequent in eukaryotic proteins
29
 B. Lorber and R. Giege

Table 4. Some commercially available protease or nuclease inhibitors

Proteases or nucleases Inhibitorsa

All protease classes Possibly a2-macroglobulin or DEPC
Serine proteases DIFP, PMSF, PefablocR SC,baminobenzamidine, 3,4-dichloro
isocoumarin, antipain, chymostatin, elastinal, leupeptin,
boronic acids, cyclic peptides, trypsin inhibitors (e.g. aprotinin,
peptidyl chloromethyl ketone)
Aspartic acid proteases Pepstatins and statin-derived inhibitors
Cysteine proteases All thiol binding reagents, peptidyldiazomethanes,
epoxysuccinyl peptides (e.g. E-64), cystatins, peptidyl
chloromethanes
Metalloproteases Chelators (e.g. EDTA, EGTA), phosphoramidon and phosphorus
containing inhibitors, bestatin, amastatin and structurally
related inhibitors, thiol
derivatives, hydroxamic acid
Ribonucleases RNasin® (Promega), ribonucleoside-vanadyl complexes, DEPC
Deoxyribonucleases DEPC, chelators (e.g. EDTA, EGTA)
a
These compounds are toxic and must be manipulated with caution. For more inhibitors, see ref. 96.
b
According to the manufacturer (Pentapharm Ltd.), Pefabloc® SC or AEBSF (4-(2-aminoethyl)-
benzensulfonyl fluoride) is a non-toxic alternative to PMSF and DIFP.

Table 5. Some co- or post-translational modifications of proteins

Amino acid residues Chemical group added or modification

or chemical groups
Arnino terminal -NH3+ Formyl-, acetyl-, glycosyl-, aminoacyl-, cyclization of Gln
Carboxy terminal -COOH Amide, amino acyl-
Arg ADP ribosyl-, methyl-, ornithine
Asp, Glu Carboxyl-, methyl-
Asn, Gln Glycosyl-, deamination
Cys Seleno-, heme, flavin-
His Flavin-, phospho-, methyl-
Lys Glycosyl-, pyridoxyl-, biotinyl-, phospho-, lipoyl-, acetyl-, methyl-
Met Seleno-
Phe Hydroxyl-
Pro Hydroxy-
Ser Phospho-, glycosyl-, ADP ribosyl-
Thr Phospho-, glycosyl-, methyl-
Tyr lodo-, hydroxy-, bromo-, chloro-

(87). Microheterogeneities may appear during storage, e.g. by deamidation of

asparagine or glutamine residues.
Similarly, nucleotides in RNA and especially in tRNAs are modified
or hypermodified co- or post-transcriptionally (88). Purification from bulk
and subsequent crystallization of individual tRNA species is affected by
the physiological state of cells (21) as well as by the amplitude of the
modifications which modulates their overall charge and hydrophobicity.
30
 2: Biochemical aspects and handling of macromolecular solutions

5.3.2 Functional versus conformational heterogeneity

Pure macromolecules can be fully functional in a biochemical activity assay
even though they are microheterogeneous. Conformational heterogeneity
may have several origins: binding of ligands, intrinsic flexibility of molecular
backbones, oxidation of cysteine residues, or partial denaturation. In the first
case, macromolecules should be prepared in both forms, the one deprived of
and the other saturated with ligands (89). In the second case, controlled
fragmentation may be helpful. In the last one, oxidation of a single cysteine
residue leads to a complex mixture of molecular species for which the chances
of growing good crystals are low (90); reducing agents reverse such redox
effects (Section 5.5).
5.4 Probing purity and homogeneity
Although macromolecules may crystallize readily in an impure state (63), it is
always preferable to achieve a high level of purity before starting crystal-
lization trials. Biochemical quality controls can be performed at relatively
little expense in comparison to time-consuming crystallizations. It is recom-
mended to combine several independent analytical methods (e.g. those
described in Section 3.1) to assert the absence of contaminants or of
microheterogeneities in samples (91).
Spectrophotometry and fluorimetry give information about the quality of
samples if macromolecules or their contaminants have special absorbance or
emission properties. As illustrated in Figure 1, a sharp peak in HPLC is
insufficient evidence for the quality of a product. SDS-PAGE indicates the
size of protein contaminants (visualized by staining, autoradiography, or
immunodetection) but not that of non-protein contaminants. Gel IEF gives
an estimate of the pI of protein components in a mixture and electrophoretic

Table 6. Selected techniques to detect structural and conformational heterogeneity

Techniques Amount required Information on

d-g)
Activity assay <1 Biological activity
Active site titration <1 Ligand binding, affinity
Gel electrophoresis < 1-10 Mobility, size, charge
Gel filtration <10-100 Size, shape
lEF/titration curve < 1-10 Charge, pl, mobility
Capillary electrophoresis/IEF <1 Size, charge
Immunological titration < 1-10 Antigenic determinants
Scattering methods (light, X-rays, neutrons) 100-1000 Size, shape
Spectrophotometry, fluorimetry <10-100 Absorption, emission
Ultracentrifugation <100-1000 Size, shape
Mass spectrometry <1 Mass

31
 B. Lorbor and R. Giege
titration shows the mobility of individual proteins as a function of pl I (92).
The latter method can also suggest the type of chromatography (i.e. anion or
cation exchange, chromatofocusing) suitable for further purification or guide
toward other chromalographies (adsorption. size exclusion, hydrophobic
interaction, or affinity). Capillary electrophoresis is well adapted for purity
analysis (52). Ammo acid composition and sequencing of N- and C-termim
verify in part the integrity of primary structure (54), ESI and MA1.DI mass
spectrometries are powerful tools in recombinant protein chemistry (93).
Homogeneity of nucleic acids is probed by electrophoresis in gels containing
urea (50, 89). Radioactive end-labelling enables detection of low levels of
cleavage in ribose-phosphate chains (84). NMR detects small size con-
taminants and gives structural information on biomolecules (94). Useful
methods for detecting conformational heterogeneity are given in Table 6.

5.5 Improving purity and homogeneity in practice

Some difficulties in crystallization (e.g. absence of crystals or poor diffraction)
may be overcome by reconsidering the purification protocol. Start with fresh

Figure 1. (Left) Comparison of the resolution of ion exchange HPLC and IEF. Aspartyl-
tRNA synthetase from yeast (250 ug, pure according to standard ion exchange
chromatography) was fractionated by anion exchange HPLC on a Mono Q column (i.d.
5 mm x length 5 cm, v 1 ml, Pharmacia) in 50 mM Tris-HCI buffer pH 7.5, and was
eluted at 0.5 ml/min with increasing NaCI concentration. IEF was performed on aliquots
(3 (j.g protein) of the fractions. The polyacrylamide gel was 10 x 10 cm2 (thickness
0.5 mm) and contained 2% (w/v) ampholytes (pH range 4-7). Staining with Coomassie
Blue R-250 reveals several protein populations differing by charge. (Right) Batch-
dependent variation in the microheterogeneity of pure aspartyl-tRNA synthetase. Six
batches of protein purified according to a standard procedure and having the same
specific activity, were compared by IEF under native conditions (samples of 5 ug protein,
a dimer with a subunit Mr of 60000, were analysed. Differences in charge result from
uncontrolled proteolysis between positions 14 and 33 in the polypeptide chain.

32
 2: Biochemical aspects and handling of macromolecular solutions
material, change the sequence of events (by inverting chromatographic steps)
or the steps themselves (by using other chromatographic matrices). To avoid
cross-contamination never mix batches of pure macromolecules even when
they look apparently identical. A small shift in the elution from a chroma-
tography column or a preparation done on the same columns but at another
scale or temperature can introduce other contaminants in active fractions.
Such variability can sometimes be detected by IEF (Figure 1). Clean and
sterilize by filtration (e.g. over 0.22 um porosity membranes) all solutions in
contact with pure macromolecules. Use chemically inert and autoclavable
chromatography matrices which do not release molecules (e.g. Trisacryl, IBF
Biotechnics, or TSK gels, Merck).
Macromolecules can be rendered more homogeneous in various ways.
Addition of protease inhibitors is generally effective (Table 4) (95, 96). Assays
to detect proteases by solubilization of clotted protein or by degradation of
labelled peptides are commercially available (e.g. Peptag™, Boehringer). A
cocktail of inhibitors should contain at least one specific for each protease
class; an example is given in Protocol 2, On a small scale, chromatography
over a column of immobilized inhibitors (e.g. a2-macroglobulin) or substrate
analogues (like arginine or benzamidine) may trap proteases. The major
drawback of inhibitors lies in their possible binding to or inactivating of the
proteins they should protect. Over-production in strains deprived of harmful
proteases is a common solution to proteolysis (97).

Protocol 2. Preparation of buffered solutions of protease

inhibitors and stabilizing agents

Caution! DIFP is a powerful inhibitor of human acetylcholine esterase. It

must be handled with extreme caution. See manufacturer's safety data
sheet. Other protease inhibitors are also very toxic. Wear gloves when
handling solutions.

Reagents
• Buffer solutions containing 10% (v/v) • 10-3 M stock solutions of peptidic inhib-
glycerol and 10-3 M EDTA itorsa (pepstatin, bestatin, and E-64 from
. 0.1 M stock solution of DIPF (Sigma) pre- Sigma) in ethanol:water (50:50)
pared by diluting a 1 g commercial sample • Reducing agents (2-mercaptoethanol and
(about 1 ml) in 50 ml cold anhydrous isopro- DTE or DTT) as stock solutions at 10-1 M
panol; always keep this solution at -20°C

Method
1. Add DIPF, peptidic inhibitors, and 2-mercaptoethanol (DTE or DTT)b in
buffer solutions just prior to use (final concentrations 5 x 10-4 M, 5 x
10-6 M, and 5 x 10-3 M, respectively).

33
 B. Lorber and R. Giege
Protocol 2. Continued
2. Add inhibitors afresh before cell disruption and at each step of the
isolation procedure.
a
Experimenters should be aware of low solubility, limited stability, affinity, and reversibility of
inhibitors.
b
Final concentration 10-3 to 10-4 M.

The action of nucleases is minimized by the addition of non-specific in-

hibitors, e.g. ribonucleoside-vanadyl complexes (98) (Table 4). Diethyl-
pyrocarbonate (DEPC) reacts with histidyl residues at the surface of proteins
and blocks the catalytic site of nucleases. RNasin® (Promega), a protein of Mr
51000 isolated from human placenta, inactivates RNases by stoichiometric
non-covalent and non-competitive binding (99). When added in transcription
media to protect in vitro synthesized RNAs, it must be removed prior to
crystallization. Nucleases show affinity for blue dextran (100) and 5'-(4-
aminophenyl)-uridine-(2',3') phosphate (101) bound to agarose and chroma-
tography on these media can remove them. Metal ions responsible for
chemical hydrolysis of RNAs are eliminated by ion exchange or chelation
(e.g. with EDTA, EGTA, or chelators immobilized on agarose beads). Mag-
nesium is an exception and stabilizes RNAs at physiological pH (Chapter 8).
To enhance compactness and homogeneity undesirable parts of multi-
domain macromolecules may be removed by controlled fragmentation.
Indeed, some proteins yield crystals of better quality after limited proteolysis
(102). Unsuspected contamination by proteases may lead to a similar result.
Mammalian glycoproteins must be deprived of their heterogeneous flexible
carbohydrate moieties unless they are over-produced in prokaryotes.
Treatment with specific endoglycosidases leaves only a single or a few sugar
groups at each site (103-105). Crystals of deglycosylated proteins can diffract
X-rays to high resolution (106). After controlled cleavage, the macro-
molecular core must be purified in order to remove proteases, glycosidases,
peptides, or oligosaccharides. To be reproducible, enzymatic tools must be
free of contaminant proteases that could introduce undesirable cleavages.
The isolation of a single molecular species is essential to grow large mono-
crystals. In many instances additional purification by HPLC or IEF yielded
better crystals (107-111).

6. Characterization and handling of crystals

6.1 Analysis of crystal content
Once crystals have been obtained, their macromolecular content must be
verified. An X-ray diffraction pattern is convincing proof (Chapter 14) but
simple analytical biochemical and biophysical methods give complementary
information. Table 7 lists two classes of biochemical and biophysical methods
34
 2: Biochemical aspects and handling of macromolecular solutions

Table 7. Selected methods for biochemical and biophysical characterization of crystals

Analytical methodsa Expected information and references

Crystalline material
Mechanical stability test (with glass needle) Softness of macromolecular crystals
Soaking with selective dyesb Chemical nature of components
Enzymic digestion Biochemical nature of components
X-ray diffractionc Space group, resolution, unit cell
parameters
Density measurementsd Solvent content, molecular mass of
protomer
Microscopy !EM, AFM) Crystallinity, surface topology (65)
Microspectrophotometry Chemical composition (112)
Laser Raman spectroscopy Conformational characteristics (113)
Soaking with ligands Binding affinity (crystals may crack!)
In situ catalysis (for enzymes) Catalytic activity (114)

Dissolved crystals
Spectrophotometry/fluorimetry Characterization and quantitation of
molecules
Gel or capillary electrophoresis Characterization of macromolecules, size
Gel or capillary IEF (for proteins only) Characterization and charge
Column chromatography (microscale) Characterization and quantitation (115)
Activity assays Biological activity
Mass spectrometry Molecular mass of components, detection
of microheterogeneities, of counterions,
sequencing (52-54, 93)
Sequencing (protein, nucleic acid) Integrity of primary structure

'Other methods may be employed in particular cases, e.g. dichroism (116), analytical ultracentrifuga-
tion(117).
b
c
For proteins, most stains used in light microscopy. For nucleic acids, see Protocol 1.
d
See Chapter 14.
See Section 6.

for the analysis of crystals. Some are applicable to the crystalline material
itself whereas others require solubilized molecules. Since crystals contain only
micrograms of macromolecules most methods must be scaled down. In all
cases, the aims are:
(a) To verify that crystals contain the desired macromolecules and in the
right stoichiometry in the case of co-crystals.
(b) To ensure that macromolecules are in an active conformation within
crystals (for enzymes, this can be asserted by in situ catalytic assays pro-
vided active sites are accessible and ligands can diffuse harmlessly within
the crystalline lattice).
(c) To compare the macromolecules in the crystalline state with those in
solution (e.g. by laser Raman spectroscopy).
35
 B. Lorber and R. Giege
Prior to any analysis, uncrystallized macromolecules or amorphous material
(present within the mother liquor or deposited onto the crystal faces) must be
washed away. This is done by transferring crystals several times in large
volumes of mother liquor.

6.2 Crystal density

Unlike crystals of small molecules made of densely packed matter, macro-
molecular crystals contain non-negligible amounts of solvent (2). The density
of a crystal not only reflects the way molecules and solvent occupy the three-
dimensional lattice but also provides structural information about the
molecules contained in the unit cells. In the past, crystal density was used to
estimate molecular weights (118) but sometimes wrong estimations have led
to erroneous structure models.
In macromolecular crystals, the solvent represents on average 30-80% of
the unit cell volume (2). It is made of free solvent (mother liquor) and bound
solvent (water, ions, but also precipitant or buffer molecules, additives)
forming the solvation (hydration) shell of the crystallized macromolecule
(119). The knowledge of crystal density may be of importance for crystals of
complexes (either of protein-ligand, protein-protein, or protein-nucleic acid
nature) or co-crystals. The number « of macromolecules (or protomers)
contained in one unit cell of volume V (cm3) can be derived from crystal
density rc (g/cm3) using the relation:

where N is Avogadro's number (6.02 X 1023), rs the density of free and bound
solvent (g/cm3), M the protomer molar weight (g/mol), and vp the partial
specific volume of the dry protein (cm3/g). Hence, the solvent fraction:

In the case of proteins, the mass per asymmetric unit Mp is obtained with good
approximation with the simplified formula (120):

assuming the density of the solvent equals that of water (rw = 1.0) and vp =
0.74 cm3/g. In practice, the volume of the unit cell is calculated from cell
parameters measured on X-ray diffraction patterns (Chapter 14). The molar
weight of the macromolecule is determined by biochemical or biophysical
methods (Table 6). Partial specific volumes of proteins (vp) are either assumed
to be equal to the inverse of their density, computed from partial specific
volumes of individual amino acids (121) and the amino acid composition, or
approximated as above. Partial specific volumes of nucleic acids can be
approximated to 0.54 cm3/g (122). The densities of solvent and crystals are
determined experimentally. The resulting number of protomers is rounded to
the nearest integer.
36
 2: Biochemical aspects and handling of macromolecular solutions
Two common methods for the measurement of crystal densities use either
organic solvents or solutions of large polymers in which the sedimentation of
the crystals is compared with internal references. Both are applicable to
proteins, nucleic acids, and viral or ribosomal particles. They have their
advantages and limitations. Other methods are described separately below.
The densities of most protein crystals range from 1.10-1.60 g/cm3 (123).
6.2.1 Organic solvent and Ficoll™ methods
The first method, adapted from the small molecule field, uses mixtures of
water saturated carbon tetrachloride or chloroform and xylene. Experimental
details are given in Protocol 3, part A. Measurements are sensitive to the
presence of mother liquor around the crystals and difficulties may arise when
trying to remove the excess solution. Readers are referred to ref. 123 for
further details.

Protocol 3. Experimental determination of crystal density

(adapted from ref. 126)

Caution! Organic solvents are toxic, especially carbon tetrachloride.

Manipulate them under a fume-hood. Wear gloves and glasses.

Equipment and reagents

• Transparent glass tubes of 0.7-1 mm inner • X-ray diffraction equipment
diameter . Water saturated xylene, carbon tetrachloride,
« Pycnometer or electronic densimeter chloroform, toluene
equipped with a vibrating tube (DMA, . Salt (e.g. sodium phosphate) solutions
Anton Paar) . FicollTM (Pharmacia)
• Glass micropipettes
• Centrifuge (reaching 3000 g)

A. Organic solvent method

1. Prepare gradients, in transparent glass tubes of 0.7-1 mm inner
diameter, by combining various proportions of water saturated xylene
and carbon tetrachloride (or chloroform).
2. Calibrate the gradients with salt (e.g. sodium phosphate) solutions
whose densities are measured with a pycnometer or an electronic
densimeter equipped with a vibrating tube.
3. Deposit one or two crystals with a minimal volume of mother liquor
onto the gradient with a glass micropipette.
4. Estimate the final position of the crystals and compare with
references.

B. Ficoll™ method
1. Prepare a series of solutions (from 30-60%, w/w) by mixing appropriate
37
 B. Lorber and R. Giege
Protocol 3. Continued
amounts of Ficoll™ powder and water. Heat at 55°C to dissolve.
Cooled solutions are yellowish and viscous.
2. Work at constant temperature. Prepare gradients in transparent glass
tubes of 0.7-1 mm inner diameter. Use a gradient maker or proceed by
layers of decreasing density. Centrifuge after each layer (5 min, 1000 g)
and end with a longer centrifugation (e.g. 1 h at 3000 g) to smooth the
gradient.
3. Calibrate gradients with drops of carbon tetrachloride (or chloroform)
and toluene mixtures (all saturated with water). The densities of these
organic solutions are measured as in part A.
4. Deposit one or two crystals and a droplet of mother liquor onto the
gradient with a glass micropipette. Centrifuge at 1000 g for 1-60 min
and compare the displacement of the crystals with that of calibrated
drops.
5. Take X-ray diffraction pictures to verify that unit cell parameters are
unchanged.

Designed for macromolecular crystals, the second method uses a large

sucrose polymer (Ficoll™, MT 400000, Pharmacia) as the only solute (124).
The density of Ficoll solutions varies linearly with concentration. Density
gradients are prepared by centrifugation and calibrated with mixtures of
organic solvents obtained as in the previous method. Technical details are
summarized in Protocol 3, part B. The method may lead to overestimation of
crystal densities presumably due to the slow diffusion of polymer in the crystal
lattice (120). Extrapolation of successive measurements to time zero should
give a good estimate of density.

6.2.2 Other methods

Other methods can be used to estimate the solvent content of a crystal. Cross-
linking may be helpful when crystals are too fragile. It is carried out by
soaking them for a few hours in mother liquor containing either 1% or 2%
(v/v) glutaraldehyde or formaldehyde to bind proteins or nucleic acids and
proteins, respectively (125). For proteins or nucleic acids whose Mr and
sequence are known, the molar absorption coefficient e can be calculated and
used to measure the amount contained in single crystals. To be applicable, the
volume of the crystal must be known precisely (the task is easier when crystals
have a regular geometrical shape). Then the crystal is dissolved in a small
volume of buffer solution (e.g. 50 ml) and its absorption spectrum recorded.
(A rough estimate of protein or RNA amounts in a crystal can also be
obtained by comparison with known quantities analysed in parallel by gel
electrophoresis.) From the weight amount of macromolecules in the crystal
38
 2: Biochemical aspects and handling of macromolecular solutions

and the unit cell parameters, the fraction volume occupied by the solvent can
be estimated and an approximate density can be extrapolated. More accurate
results are expected when molar absorption coefficients are high. For crystals
containing more than one type of macromolecule, protein and nucleic acid or
a macromolecule and a smaller molecule (e.g. a protein and a ligand), their
stoichiometry can be calculated when molecules differ by their spectroscopic
properties.

References
1. Wierenga, R. K.? Lalk, K. H., and Hol, W. G. J. (1987). J. Mol. Biol, 198, 109.
2. Matthews, B. V. (1968). J. Mol. Biol., 33, 491.
3. Garber, M., Davydova, N., Eliseikina, I., Fomenkova, N., Grysaznova, O.,
Gryshkovskaya, I., et al. (1996). J. Cryst. Growth, 168, 301.
4. Thygesen, J., Krumholz, S., Levin, I., Zaytzev-Bashan, A., Harms, J., Bartels, H.,
et al. (1996). J. Cryst. Growth, 168, 308.
5. Zaccai, G. and Eisenberg, H. (1990). Trends Biochem. Sci., 15, 333.
6. Kern, D., Giege, R., Robbe-Saul, S., Boulanger, Y., and Ebel, J.-P. (1975).
Biochimie, 57, 1167.
7. Adelman, R. C. and Dekker, E. E. (ed.) (1985). Modification of proteins during
aging. Alan R. Liss, New York.
8. Lorber, B. and DeLucas L, J. (1990). FEBS Lett., 241, 14.
9. Sambrook, J., Fritsch, E. F., and Maniatis, T. (ed.) (1989). Molecular cloning: a
laboratory manual, 3 Vols. Cold Spring Harbor Laboratory Press.
10. Goeddel, D. V. (ed.) (1991). Methods in enzymology, Vol. 185, pp. 1-599.
Academic Press, London.
11. Kurland, C. G. and Dong, H. (1996). Mol. MicrobioL, 21, 1.
12. Schein, C. H. (1990). Biotechnology, 8, 308.
13. Miroux, B. and Walker, J. E. (1996). J. Mol. Biol., 260, 289.
14. Zahn, K. and Landy, A. (1996). Mol. Microbiol, 21, 69.
15. Calderone, T. L., Stevens, R. D., and Oas, T. G. (1996). J. Mol. Biol, 262, 407.
16. Carter, C. W. (1988). J. Cryst. Growth, 90, 168.
17. Horwich, A. L., Neupert, W., and Hartl, F. U. (1990). Trends Biotech., 8, 126.
18. Richardson, C. D. (ed.) (1995). Methods in molecular biology, Vol. 39, pp. 1-420.
Humana Press, Totowa, NJ.
19. Marston, F. A. O. (1986). Biochem. J., 240, 1.
20. Spirin, A. S., Baranov, V. L, Ryabova, L. A., Ovodov, S. Y., and Alakhov, Y. B.
(1988). Science, 242, 1162.
21. Perona, J. J., Swanson, R., Steitz, T. A., and Soil, D. (1988). J. Mol. Biol., 202, 121.
22. McPherson, M. J. (ed.) (1991). Directed mutagenesis: a practical approach,
pp. 1-288. IRL Press, Oxford.
23. Oubridge, C., Io, N., Teo, C. H., Fearnley, I., and Nagai, K. (1995). J. Mol. Biol.,
249, 409.
24. Harris, E. L. V. and Angal, S. (ed.) (1990). Protein purification applications: a
practical approach, pp. 1-192. IRL Press, Oxford.
25. Yin, Y. and Carter, C. W., Jr. (1996). Nucleic Acids Res., 24, 1279.
39
 B. Lorber and R. Giege
26. Vuillard, L., Baalbaki, B., Lehmann, M., Norager, S., and Legrand, P. (1996). /.
Cryst. Growth, 168, 150.
27. Oliver, R. W. A. (ed.) (1989). HPLC of macromolecules: a practical approach,
pp. 1-252. IRL Press, Oxford.
28. Righetti, P. G. (1983). Isoelectric focusing: theory, methodology and application.
Elsevier, Amsterdam.
29. Wagner, H. (1989). Nature, 341, 669.
30. Ott, G., Dorfler, S., Sprinzl, M., Muller, U., and Heinemann, U. (1996). Acta
Cryst., D52, 871.
31. Wahl, M. C., Ramakrishnan, B., Ban, C, Chen, X., and Sundaralingam, M. (1996).
Acta Cryst., D52, 668.
32. Gate, J. H., Gooding, A. R., Podell, E., Zhou, K., Golden, B. L., Kundrot, C. E., et
al (1996). Science, 273, 1678.
33. Aoyama, K. (1996). J. Cryst. Growth, 168, 198.
34. Gulewicz, K., Adamiak, D., and Sprinzl, M. (1985). FEBS Lett., 189, 179.
35. Giege, R., Dock, A.-C., Kern, D., Lorber, B., Thierry, J.-C., and Moras, D. (1986).
J. Cryst. Growth, 76, 554.
36. Stoll.V. S. and Blanchard, J. S. (1990). In Methods in enzymology (ed. M. P.
Deutscher), Vol. 182, pp. 24-38. Academic Press, London.
37. Yang, A. S. and Honig, B. (1993). J. Mol. Biol., 231, 459.
38. Darby, N. J. and Creighton, T. E. (ed.) (1989). Protein structure: a practical
approach, pp. 1-120. IRL Press, Oxford.
39. Norde, W. (1995). Cells Mater., 5, 97.
40. Baldwin, R. L. (1996). Biophys. J., 71, 2056.
41. Strambini, G. B. and Gabellieri, E. (1996). Biophys. J., 70, 971.
42. Sousa, R. (1995). Acta Cryst., D51, 271.
43. Crowe, L. M., Reid, D. S., and Crowe, J. H. (1996). Biophys. J., 71, 2087.
44. Jeruzalmi, D. and Steitz, T. A. (1997). J Mol. Biol., 274, 748.
45. Buttler, S. L. and Falke, J. J. (1996). Biochemsitry, 35, 10595.
46. Chayen, N. E., Radcliffe, J. W., and Blow, D. M. (1993). Protein Sci., 2, 113.
47. Mittl, P. R. E., Di Marco, S., Krebs, J. F., Bai, X., Karanewsky, D. S., Priestle, J. P.,
etal. (1997). J. Biol. Chem., 272, 6539.
48. Scott, W. G., Finch, J. T., Grenfell, R., Fogg, J., Smith, T., Gait, M. J., et al. (1995).
J. Mol. Biol., 250, 327.
49. Hames, B. D. and Rickwood, D. (ed.) (1990). Gel electrophoresis of proteins: a
practical approach, pp. 1-250. IRL Press, Oxford.
50. Rickwood, D. and Hames, B. D. (ed.) (1990). Gel electrophoresis of nucleic acids:
a practical approach, pp. 1-336. IRL Press, Oxford.
51. Righetti, P. G., Gianazza, E., and Gelfi, C. (1988). Trends Biochem. Sci., 13, 333.
52. Karger, B. L. and Hancock, W. S. (ed.) (1996). Methods in enzymology, Vol. 270,
pp. 1-611 and Vol. 271, pp. 1-543. Academic Press, London.
53. Muddiman, D. C., Bakhtiar, R., Hofstaler, S. A., and Smith, R. C. (1997). J. Chem.
Edu.,74, 1288.
54. Ries-Kautt, M., Ducruix, A., and Van Dorsselaer, A. (1994). Acta Cryst., D50, 366.
55. Matsudaira, P. (ed.) (1993). A practical guide to protein and peptide purification
and microsequencing. Academic Press, Inc., San Francisco.
56. Stoscheck, C. M. (1990). In Methods in enzymology (ed. M. P. Deutscher), Vol.
182, pp. 50-68. Academic Press, London.
40
 2: Biochemical aspects and handling of macro molecular solutions
57. Pace, C. N., Vajdos, F., Fee, L., Grimsley, G., and Gray, T. (1995). Protein Sci., 4,
2411.
58. Warburg, O. and Christian, W. (1941). Biochem. Z., 310, 384.
59. Ehresmann, B., Imbault, P., and Weil, J.-H. (1973). Anal. Biochem., 54, 454.
60. Fersht, A. (1977). Enzyme structure and mechanism. Freeman & Company,
Reading.
61. Gueron, M. and Leroy, J.-L. (1978). Anal. Biochem., 91, 691.
62. Ames, B. N. and Dubin, D. T. (1960). /. Biol. Chem., 235, 769.
63. Holley, R. W., Apgar, J., and Merrill, S. H. (1961). /. Biol. Chem., 236, PC42.
64. Judge, R. A., Johns, M. R., and White, E. T. (1995). Biotech. Bioeng., 48, 316.
65. Vekilov, P. G. and Rosenberger, F. (1996). J. Cryst. Growth, 158,540.
66. Skouri, M., Lorber, B., Giege, R., Munch, J.-P., and Candau, S. J. (1996). J. Cryst.
Growth, 152, 209.
67. McPherson, A., Malkin, A. J., Kuznetsov, Y. G., and Koszelak, S. (1996). J Cryst.
Growth, 168, 74.
68. Provost, K. and Robert, M.-C. (1995). J. Cryst. Growth, 156, 112.
69. Hirschler, J., Charon, M.-H., and Fontecilla-Camps, J. C. (1995). Protein Sci., 4, 2573.
70. Doudna, J. A., Grosshans, C., Gooding, A., and Kundrot, C. E. (1993). Proc. Natl.
Acad. Sci. USA, 90, 7829.
71. Lawson, D. M., Artymiuk, P. J., Yewdall, S. J., Smith, J. M. A., Livingstone, J. C.,
Treffry, A., et al. (1991). Nature, 349, 541.
72. Ray, W. J. and Puvathingal, J. (1985). Anal. Biochem., 146, 307.
73. Jurnak, F. (1986). J. Cryst. Growth, 76, 577.
74. Bello, J. and Nowoswiat, E. F. (1965). Biochim. Biophys. Acta, 105, 325.
75. Kuhlbrandt, W. (1988). Q. Rev. Biophys., 21, 429.
76. Lorber, B., Bishop, J. B., and DeLucas, L. J. (1990). Biochim. Biophys. Acta, 1023,
254.
77. Achstetter, T. and Wolf, D. H. (1985). Yeast, 1, 139.
78. Barrett, A. J. and Salvesen, G. (ed.) (1986). Research monographs in cell and
tissue physiology, Vol. 12, pp. 1-661. Elsevier, Amsterdam.
79. Calling, M. J. (ed.) (1986). Plant proteolytic enzymes, 2 Vol. CRC Press, Boca
Raton, FL.
80. Bond, J. S. and Butler, P. E. (1987). Annu. Rev. Biochem., 56, 333.
81. Arfin, S. M. and Bradshaw, R. A. (1988). Biochemistry, 27, 7979.
82. Wandersman, C. (1989). Mol. Microbiol, 3, 1825.
83. Howe, C. J. and Ward, E. S. (ed.) (1989). Nucleic acids sequencing: a practical
approach, pp. 1-254. IRL Press, Oxford.
84. Moras, D., Dock, A.-C., Dumas, P., Westhof, E., Romby, P., Ebel, J.-P., et al.
(1985). J. Biomol. Struct. Dyn., 3, 479.
85. Wold, F. and Moldave, K. (ed.) (1984). Methods in enzymology, Vol. 106,
pp. 1-574 and Vol. 107, pp. 1-688. Academic Press, London.
86. Kendall, R. L., Yamada, R., and Bradshaw, R. A. (1990). In Methods in
enzymology (ed. D. V. Goeddel), Vol. 185, pp. 398-08. Academic Press, London.
87. Rudd, P. M. and Dwek, R. A. (1997). Crit. Rev. Biochem. Mol. Biol., 32, 1.
88. Grosjean, H. and Benne, R. (ed.) (1998). Modification and editing of RNA. ASM
Press, Washington DC.
89. Lorber, B., Giege, R., Ebel, J.-P., Berthet, C., Thierry, J.-C, and Moras, D. (1983).
J. Biol. Chem., 258, 8429.
41
 B. Lorber and R. Giege
90. Van der Laan, J. M., Swarte, M. B. A., Groendijk, H., Hol, W. G. J., and Drenth,
J. (1989). Eur. J. Biochem., 179, 715.
91. Rhodes, D. G. and Laue, T. M. (1990). In Methods in enzymology (ed. M. P.
Deutscher), Vol. 182, pp. 555-66. Academic Press, London.
92. Lorber, B., Kern, D., Mejdoub, H., Boulanger, Y., Reinbolt, J., and Giege, R.
(1987). Eur. J. Biochem., 165, 409.
93. Andersen, J. S., Svensson, B., and Roepstorff, P. (1996). Nature Biotech., 14, 449.
94. Wuthrich, K. (1995). Acta Cryst, D51, 249.
95. Beynon, R. J. and Bond, J. S. (ed.) (1989). Proteolytic enzymes: a practical
approach, pp. 1-278. IRL Press, Oxford.
96. Beynon, R. J. (1989). In Protein purification methods: a practical approach (ed. E.
L. V. Harris and S. Angal), pp. 40-51. IRL Press, Oxford.
97. Gottesman, S. (1990). In Methods in enzymology (ed. D. V. Goeddel), Vol. 185,
pp. 119-29. Academic Press, London.
98. Berger, S. L. and Birkenmeier, C. S. (1979). Biochemistry, 18, 5143.
99. Scheele, G. and Blackburn, P. (1979). Proc. Natl. Acad. Sci. USA, 76, 4898.
100. Thompson, S. T., Cass, K. H., and Stellwagen, E. (1975). Proc. Natl. Acad. Sci.
USA, 72, 669.
101. Wierenga, K., Huizinga, J. D., Gaastra, W., Welling, G. W., and Beintema, J. J.
(1973). FEBS Lett., 31, 181.
102. Bergfors, T., Rouvinen, J., Lehtovaara, P., Caldentey, X., Tomme, P.,
Claeyssens, M., et al. (1989). J. Mol. Biol, 209, 167.
103. Kalisz, H. M., Hecht, H. J., Schomburg, D., and Schmid, R. D. (1990). J. Mol
Biol, 213, 207.
104. Stura, E. A., Chen, P., Wilmot, C. M., Arevalo, J. H., and Wilson, I. A. (1992).
Proteins, 12, 24.
105. Baker, H. M., Day, C. L., Norris, G. E., and Baker, E. N. (1996). Acta Cryst.,
DSO, 380.
106. Nitanai, Y., Satow, Y., Adachi, H., and Tsujimoto, M. (1996). J. Cryst. Growth,
168, 280.
107. Anderson, W. F., Boodhoo, A., and Mol, C. D. (1988). J. Cryst. Growth, 90, 153.
108. Bentley, G. A., Boulot, G., Riottot, M. M., and Poljak, A. (1988). J. Cryst.
Growth, 90, 213.
109. Spangler, B. D. and Westbrook, E. M. (1989). Biochemistry, 28, 1333.
110. Moreau, H., Abergel, C., Carriere, R, Ferrato, F., Fontecilla-Camps, J. C.,
Cambillau, C., et al. (1992). J. Mol. Biol, 225, 147.
111. Weber, W., Wenisch, E., Gunther, N., Marnitz, U., Betzel, C., and Righetti, P. G.
(1994). J. Chromatogr., A679, 181.
112. Kirsten, H. and Christen, P. (1983). Biochem. J., 211, 427.
113. Chen, M., Lord, R., Giege, R., and Rich, A. (1975). Biochemistry, 14, 4385.
114. Mozzarelli, A. and Rossi, G. L. (1996). Annu. Rev. Biophys. Biomol. Struct., 25,
3430.
115. Lorber, B. and Giege, R. (1985). Anal. Biochem., 146, 402.
116. Ford, R. C., Picot, D., and Garavito, R. M. (1987). EMBO J., 6, 1581.
117. Schuster, T. M. and Toedt, J. M. (1996). Curr. Opin. Struct. BioL, 6, 650.
118. Palmer, K. J., Ballantyne, M., and Galvin, J. A. (1948). J. Am. Chem. Soc., 70, 906.
119. Salemme, F. R., Genieser, L., Finzel, B. C., Hilmer, R. M., and Wendoloski, J. J.
(1988). J. Cryst. Growth, 90, 273.
42
 2: Biochemical aspects and handling of macromolecular solutions
120. Bode, W. and Schirmer, T. (1985). Biol. Chem. Hoppe-Seyler, 366, 287.
121. Cohn, E. J. and Edsall, J. T. (ed.) (1943). In Proteins, amino acids and peptides,
p. 370. Van Nostrand Reinhold, Princeton, NJ.
122. Cohen, G. and Eisenberg, H. (1968). Biopolymers, 6, 1077.
123. Matthews, B. W. (1985). In Methods in enzymology (ed. H. W. Wyckoff, C. H.
W. Hirs, and S. N. Timasheff), Vol. 114, pp. 176-87. Academic Press, London.
124. Westbrook, E. M. (1985). In Methods in enzymology (ed. H. W. Wyckoff, C. H.
W. Hirs, and S. N. Timasheff), Vol. 114, pp. 187-96. Academic Press, London.
125. Sewell, B. T., Bouloukos, C., and Von Holt, C. (1984). J Microsc., 136, 103.
126. Mikol, V. and Gieg6, R. (1992). In Crystallization of nucleic acids and proteins: a
practical approach (ed. A. Ducruix and R. Giege), 1st edn, pp. 219-40. IRL Press,
Oxford.

43
This page intentionally left blank
 3

Molecular biology for structural

biology
P. F. BERNE, S. DOUBLIE, and C. W. CARTER, JR

1. Introduction
The number of published 3D structures has increased exponentially in the last
decade and the resulting mass of structural data has contributed significantly
to the understanding of mechanisms underlying the biology of living cells.
However, these mechanisms are so complex that structural biologists face still
greater challenges, such as the study of higher-order functional complexes. As
an example, we can mention the protein complexes that assemble around
activated growth factor receptors to allow the transduction of extracellular
signals through the membrane and inside the cell (1).
Because of their diverse intrinsic properties, proteins exhibit variable diffi-
culty for structural biology studies. Before the rise of recombinant expression
methods, only a minority of protein structures were determined, representing
mainly favourable cases: proteins of high abundance in their natural source
which could be purified and crystallized, in contrast to rare proteins that
were often refractory to crystallization. The advent of methods for recom-
binant protein overexpression was a breakthrough in this area. It was
followed by an increasing number of publications describing the crystal-
lization of proteins, not under their native form, but in modified versions after
sequence engineering.
First we will consider the classical use of molecular biology applied to
optimize the expression system for a recombinant protein for structural
biology, without modification of its sequence. In the second part, we will deal
with molecular biology procedures aimed at engineering the properties of a
protein through sequence modifications in order to make its crystallization
possible. In the last part we will give an example where molecular biology can
help solve a crystallographic problem, namely that of phase determination by
introducing anomalous scatterers (e.g. selenium atoms) into the protein of
interest.
 P. F. Berne et al.

2. Molecular biology to optimize protein expression

systems
Whenever extraction of a protein from its natural source appears unsuitable for
structural studies, molecular biology resources can be brought in, initially
aiming at choosing and setting up an appropriate expression system. This initial
approach could involve comparing various expression hosts and vectors and
deciding if the protein is to be produced as a fusion to facilitate its purification.

2.1 Expression systems

Structural biology, whether NMR or crystallography, consumes large amounts
of protein (up to 100 mg or more for a complete study). When a protein of
major interest is discovered, it is often the case that the natural source does
not express it at a sufficient level to allow structural studies. If this protein
represents a small fraction of the total cellular proteins, important quantities
of cell extracts have to be treated, and multiple purification steps are
generally needed, therefore decreasing the overall yield of recovery.
Provided a gene encoding the protein to crystallize is available, expression
systems can be used for large scale production of recombinant proteins. By
'expression system' one should understand the host, vector, and all procedures
necessary to allow expression of a particular protein in a non-natural host. The
most frequently used expression host is the bacterium Escherichia coli, followed
by insect (Spodoptera frugiperda) cells infected by a baculoviral vector.
Mammalian cells (for example the CHO cell line) and yeast strains are used less
frequently and will be less detailed. The choice of an expression system (see
Table 1 for commercially available vectors) should take into account the nature
of the protein and the experimental constraints specific to each system.
2.1.1 Bacteria
The bacterial host is by far the easiest to manipulate, particularly when dealing
with large culture volumes, and it often yields a high level of overexpression
(up to 300 mg/litre of low density culture and 50% of the total cellular
proteins). The ease of manipulating bacterial expression systems facilitates
approaches involving several mutants, for example to engineer variants in
order to investigate the crystallization properties of a protein or to perform
structure/function studies. However, a major drawback of the E. coll system
lies in the fact that overexpressed proteins often misfold in the bacterium and
aggregate into inclusion bodies (see Section 2.3). Additionally, the lack of
post-translational modifications may be detrimental to protein activity.
2.1.2 Insect cells
The baculoviral system is used increasingly in many laboratories, facilitated
by the availability of commercial tools that make this eukaryotic expression
46
 3: Molecular biology for structural biology
system more user-friendly (2). In this system, a viral vector carrying the gene
of interest is used to infect insect cells. These host cells provide a more
favourable environment for the folding of most eukaryotic proteins than does
the cytoplasm of bacteria. In addition, cytoplasmic, membrane-localized,
transmembrane, or secreted proteins will each be directed to their respective
compartment, provided that the signals for localization are included in the
vector. Insect cells have the machinery to perform most of the usual post-
translational modifications, including the glycosylation of extracellular
domains. Expression levels can vary from one particular construction to
another, and overexpression yields of up to 50% of total proteins have been
reported. Tropomyosin is a typical example of high expression (30 mg purified
protein per litre of culture) in Sf9 insect cells that allowed crystallization (3).
However, compared to the bacterial system, baculovirus systems require
additional steps, such as the purification of a viral clone and the amplification
of the virus by several infection cycles.
2.1.3 Mammalian cells
Expression of a protein in mammalian cells for structural studies is not
frequent (4), because the set-up of a cell line overexpressing a given protein is
time-consuming and the level of expression is never very high. Nevertheless,
one should keep in mind that it could be the only system allowing expression
of a mammalian protein in a context where it will be correctly processed and
fully functional.
2.1.4 Yeast
Yeast can also be a potential host and has proved to be extremely efficient for
expressing certain secreted proteins (up to 4 g per litre of culture). However,
it should be limited to naturally secreted proteins because the vast majority of
other proteins would undergo abnormal glycosylation if secreted by yeast
cells. The Pichia pastoris strain has been advantageous for producing several
complex eukaryotic proteins (5) and a Pichia kit is commercially available
from Invitrogen. Yeast cells are less efficient tools for intracellular expression
and require special equipment to be disrupted (6). However, there are
examples of successful intracellular expression in P. pastoris, Saccharomyces
cerevisiae, or Schizosaccharomyces pombe (7) for structural studies.

2.2 Factors influencing the level of expression

Expression vectors dedicated to the transfer and expression of a particular
gene of interest into an expression host have been created, in order to
optimize the level of expression, by including a strong promoter and the
required elements for transcription and translation by the host machinery.
However, for a single expression vector, it is well known that expression levels
can vary from one protein to another, and there are no rules to predict these
variations.
47
 P. F. Berne et al.
For expression in E. coli, the signals required for the expression vector
(promoter, Shine-Dalgarno sequence for ribosome binding, transcription
terminator) are well characterized (8), but the influence of gene-dependent
factors is not fully understood (9). Codon usage is assumed not to play a
significant role (10), except when particularly rare codons for arginine and/or
isoleucine are repeated (11). The secondary structure of the messenger RNA
can be critical (9) for its stability and for initiation efficiency (accessibility of
the Shine-Dalgarno sequence). In addition, DNA hairpin structures can
result in premature transcription termination. It is not clear why large
proteins (i.e. > 50 kDa) are expressed on average at lower levels.
In E. coli expression vectors, the short number of nucleotides between the
Shine-Dalgarno sequence and the initiation codon has to be strictly re-
spected. This requires that the gene of interest be inserted using a restriction
site overlapping with the initiation codon (except when making N-terminal
fusion proteins). In contrast, there is no equivalent sequence in eukaryotic
vectors. The first ATG codon found downstream from the promoter is used as
the initiation codon. The sequence found around this codon seems to
influence to some extent the efficiency of translation (12).

2.3 Inclusion bodies in E. coli

Of the many specific problems encountered when recombinant proteins are
expressed in bacteria, the most frequent is the formation of inclusion bodies
(8, 13). Inclusions can be seen as aggregates of misfolded protein. Their
formation could result, at least in some cases, from the intrinsic insolubility of
the protein of interest in the cytoplasm of E. coli. This can be the case for
aberrant proteins, such as truncated variants or proteins mutated in struc-
turally important residues. In other cases, however, formation of inclusion
bodies may arise because the high rate of synthesis results in high concentra-
tions of transient states in the folding pathway. Because such folding inter-
mediates can expose surfaces that ultimately come together to stabilize the
native conformation, they can lead to intermolecular interactions and hence
to aggregation and lower stability. A fraction of the synthesized protein may
escape this aggregation, fold properly, and be perfectly soluble in E. coli. This
explains why some of the protein can be found soluble and active in the
cytoplasm of E. coli together with inclusion bodies.
Some 'recipes' used to reduce the fraction of protein trapped into inclusion
bodies in E. coli involve manipulating various factors that may contribute to
this phenomenon. Lowering the culture temperature (30°C, 25 °C, or even
20°C) is the most widely used procedure for this purpose (13). Lower
temperature reduces the rate of synthesis and may destabilize hydrophobic
interactions in the aggregated protein, increasing its stability and permitting a
higher fraction to fold properly. Improvements have also been obtained
occasionally by limiting the level of overexpression or by the use of additives
48
 3: Molecular biology for structural biology
in the culture medium (sorbitol or betain) (14). Lack of the appropriate
chaperones, post-translational modifications, or appropriate partner inter-
acting with the protein in the natural host suggest other possible remedies
including co-expression of the protein with chaperones (15, 16) or co-
expression of both subunits of a heterodimeric protein in the same host (17).
There are several examples for which none of the above procedures
prevented the formation of inclusion bodies. It is then possible to extract the
protein from the inclusion bodies using a denaturing agent, such as urea or
guanidinium chloride. This can be used as a selective purification step,
because inclusion bodies are often highly enriched in the desired protein, and
can be separated from the membranous fraction (18). Subsequent renatura-
tion of the protein works best if refolding is allowed to take place on a solid
support provided, for example, by a hydrophobic interaction resin (19). The
procedure in Protocol 1 was adapted by Sharon Campbell (personal com-
munication) and has been shown to be effective in a number of cases. The
specific hydrophobic column support is not critical; it apparently serves a
function analogous to that provided by chaperones in the natural host cell,
allowing an orderly progression of intermediate structures to form with
minimal intermolecular interactions.

Protocol 1. Renaturation of proteins isolated from inclusion

bodies

Equipment and reagents

• Hipropyl hydrophobic interaction column

Method
1. Resolubilize the insoluble protein in 6-8 M guanidinium hydrochloride.
2. Dilute it to ~ 100 ug/ml in a buffer with ~ 1.5 M NaCI, 4.5 M guani-
dinium hydrochloride, 10% glycerol, 10 mM DTT, 5% ethylene glycol,
and other components necessary for the activity of the protein. Use a
weakly buffered solution compatible with activity (i.e. ~ 40 mM
acetate).
3. Apply the mixture to a hipropyl hydrophobic interaction column pre-
equilibrated with the same buffer used in step 2, plus 2 M NaCI, 10-20
mM DTT, additional components as before, 10% glycerol, and 5%
ethylene glycol. For a 7.75 x 100 mm2 semi-preparative column, load
-1.0 mg of diluted protein. For a preparative 2.2 x 10 cm2 column,
load ~ 5.0 mg.
4. Elute with a gradient of the starting buffer mixed progressively with
the same buffer without NaCI.

49
 P. F. Berne et al.
Alternatives to renaturation trials, when a protein is produced in the
bacteria as a totally insoluble form, are to turn to other expression hosts,
to try to express the protein as a fusion protein, or to have it secreted by E.
coli, which will provide a different environment for the folding of the protein.
This is done by fusing the protein to an N-terminal signal peptide. One
commercially available vector is suitable for this purpose (Table 1).
The response surface methodology described in Chapter 4 offers a rational,
coherent way to evaluate and optimize the variety of factors influencing the
proportion of soluble protein.

2.4 Tagged proteins to facilitate purification

A protein can be modified in order to facilitate its purification for structural
studies. This is done by fusing its N- or C-terminus with another protein or
with a peptidic tail allowing for a purification by affinity chromatography (20).
Table 2 summarizes a selection of fusion tags commercially available. This
method can be applied, in principle, to proteins expressed in any host and is
especially useful in two cases: when the level of expression is low and when
there is a need to produce in parallel several variants of the protein for
crystallization. Although each of them could be preferred for specific reasons,
it is obvious from a literature survey of proteins successfully expressed and
crystallized in the last years that the polyhistidine tag has encountered a much
broader success than any other. This is probably due to the availability of
nickel chelate resins (Ni-NTA, Qiagen) which have higher capacities than
most other affinity resins. This allows for the relatively cheap purification of
10-100 mg of protein.
Besides facilitating its purification, the fusion of the target protein with a
partner protein, such as thioredoxin, may be done to allow the soluble ex-
pression of a protein otherwise insoluble (21). However this approach remains
largely empirical: in contrast to a generally accepted idea, fusion of a protein
with a very soluble bacterial protein, such as glutathione-S-transferase (GST),
does not always result in an improved solubility of the fusion product,
compared to its non-fusion counterpart (P. F. B., unpublished data).
A major drawback of working with fusion proteins is that recovery of the
native protein requires a cleavage step. When dealing with a short peptidic
tail, it can be envisaged to leave it for crystallization (22), but the tag could
affect the proper folding of the protein or prevent crystallization itself (23).
Fusions with proteins like GST (25 kDa) usually have to be dissociated by
proteolytic cleavage before crystallization, although there are exceptions.
Usually, the expression vectors for fusion proteins encode a cleavage site for a
specific protease between the two fusion partners. However, the cleavage step
happens to be costly for large quantities and can generate heterogeneity.
Indeed, it is difficult to control the proteolytic reaction to get a full cleavage at
the desired site without secondary cleavages. It should also be kept in mind
50
 3: Molecular biology for structural biology
that even after removal of the fusion partner, there remains generally a small
part of the linker at the extremity of the cleaved protein. This can be
minimized by an appropriate cloning strategy.

2.5 Use of molecular biology to design an appropriate

expression vector
In common language, molecular biology technology includes all the methods
for gene manipulation. They are currently used by an increasing number of
laboratories and this availability was greatly facilitated by the development of
standard techniques and the commercialization of the required enzymes and
reagent kits. In contrast to proteins, which exhibit a great diversity of physical
and chemical behaviour, DNA fragments are easily modified under standard
conditions, independently of base composition. Most basic techniques of
DNA manipulation, as well as the basic techniques for the expression of
proteins in E. coli and mammalian cells, can be found in the classical protocol
manual by Maniatis et al. (18).
Our purpose is to give strategies to construct new DNA vectors appropriate
to the expression of a target protein rather than detailed protocols. Using an
expression system requires, as a preliminary condition, the gene encoding the

Table 1. Selection of commercial expression vectors

Name (family) System Promoter Supplier

pET series E. coli T7 Novagen
pTrc99 E. coli trc Pharmacia
pLEX E. coli PL Invitrogen
pET12,pET20 E. coli secretion T7 Novagen
pVL1392 Baculovirus Polyhedrin Pharmingen
pAcSG2 Baculovirus Polyhedrin Pharmingen
pAcUW! Baculovirus p10 Pharmingen
pFastBac8 Baculovirus Polyhedrin Gibco BRL
pBluebac Baculovirus Polyhedrin Invitrogen
pBac Baculovirus Polyhedrin or gp64 Novagen
pBacPAK Baculovirus Polyhedrin Clontech
pAC5 Drosophila cells Actin 5c (constitutive) Invitrogen
pMT Drosophilace\\s Metallothionein (inducible) Invitrogen
pPIC Yeast P. pastoris Alcohol oxidase Invitrogen
pYEUra3 Yeast GAL1 Clontech
pcDNA3 Mammalian CMV Invitrogen
pZeoSV2 Mammalian SV40 Invitrogen
pMSG Mammalian MMTV-LTR Pharmacia
a
This baculoviral system is the only one where the recombination with the viral DNA takes place in
E. col!, instead of inside insect cells. Therefore, the purification and amplification of a viral clone is not
necessary, which shortens the delays, especially for large scale production.

51
 P. F, Berne et al.

Table 2. Fusion tags for purification

(a) Bacterial vectors

Vector name Fusion partner Affinity Supplier
pET-15, pET-28 Poly-His Metal chelate Novagen
pQE Poly-His Metal chelate Qiagen
pTrcHis Poly-His Metal chelate Invitrogen
pGEX Glutathione-S-transferase Glutathione Pharmacia
pCAL Calmodulin binding domain Calmodulin Stratagene
pMAL MalE Maltose New England
Biolabs
pTrxFus Thioredoxin Phenylarsine oxide Invitrogen
pET-32 Poly-His + thioredoxin Metal chelate Novagen
pET-29 S-tag S-protein Novagen
pEZZ Protein A domain IgG Pharmacia
pET34-38 Cellulose binding domain Cellulose Novagen
Impact system Chitin binding domain Chitin New England
Biolabs
FLAG system FLAG peptide Anti-FLAG Sigma
monoclonal antibody

(b) Baculovirus transfer vectors

Vector name Fusion partner Affinity Supplier
pBacPAKHis Poly-His Metal chelate Clontech
pAcHLT Poly-His Metal chelate Pharmingen
pFastBacHT Poly-His Metal chelate GibcoBRL
pBlueBacHis Poly-His Metal chelate Invitrogen
pAcGHLT Glutathione-S-transferase Glutathione Pharmingen
pBac7-11 Cellulose binding domain Cellulose Novagen

protein of interest. It is beyond the scope of this chapter to describe the

techniques available for cloning new genes (see, for example, ref. 18) and it is
assumed that the gene encoding the protein of interest is available to the
reader.
Given the gene of interest, the initial task may be to switch from one
expression vector and/or host system to another. Protocol 2 gives a general
'strategy for subcloning' a gene of interest and Table 1 contains a selection of
common, commercially available expression vectors. The procedure can be
applied to transfer a gene (or part of a gene if one wishes to reduce the size of
a protein) from one expression vector to another, including vectors to generate
tagged proteins. The use of PCR allows this work to be done independently of
the presence of compatible restriction sites on the vectors because any chosen
site may be introduced using PCR.
52
 3: Molecular biology for structural biology

Protocol 2. General subcloning strategy using PCR

Equipment and reagents

• Oligonucleotides (see Chapter 8) • Bacterial strains
• Appropriate restriction enzymes, plasmid, • PCR thermocycler and current molecular
expression vector biology materials
• Kits for DNA extraction, purification (e.g. • DNA sequencing facility
Qiagen or Promega)

Method
1. Perform a standard PCR reaction (25 cycles: 30 sec at 94°C, 1 min at
60°C, 1 min per kb amplified at 72°C) with the selected primers using
1 ng of a plasmid carrying your gene as template. Include control
reactions with individual primers.
2. Isolate the PCR product using a phenol:chloroform extraction,
followed by ethanol precipitation or using one of the commercially
available kits.
3. Digest the acceptor vector (1 ug) and your PCR product with the
selected restriction enzymes. Isolate the fragments of interest using
agarose gel electrophoresis. Under long UV illumination (365 nm), cut
out gel pieces containing the digested vector and PCR insert. Use one
of the commercial kits for the extraction of DNA out of the gel.
4. Perform a standard ligation reaction in the presence of about 1 nM of
the linearized vector and 2 nM of the PCR insert (concentration is not
critical). Include a ligation of the vector alone as control.
5. Transform competent bacteria (e.g. TG1) strain with 2 ul of ligation.
6. Prepare a small amount of DNA (commercial miniprep kit) of a few or
several clones and check for the presence of the insert using
restriction enzymes.
7. Sequence a few positive clones to make sure that no error has been
introduced by the PCR reaction. The complete region amplified by PCR
has to be checked.

2.5.1 Choice of restriction sites on the vector

Expression vectors contain host-specific elements required for efficient
expression, that have to be preserved, and a location delimited by restriction
sites (sometimes multiple ones) dedicated to the proper insertion of the target
gene. On a map of the vector, identify this location, and the promoter up-
stream from it. The coding sequence has to be inserted in the proper
orientation between an upstream site, closer to the promoter, and a down-
stream site. The cloning will be much facilitated if two different, protruding,
and non-compatible restriction sites are used.
53
 P. F. Berne et al.
Note that the upstream site of bacterial expression vector is normally an
imposed one, Ndel or Ncol, because the initiation codon has to be precisely
six or seven bases downstream from the Shine-Dalgarno sequence (to locate
on the map downstream from the promoter). In contrast, in eukaryotic
systems, there is no such sequence and protein synthesis starts at the first
ATG following the transcription start. Therefore, any of the multiple sites
could be chosen.
A particular case is the insertion of the target gene in a vector designed for
fusion proteins. These vectors often contain several cloning sites but it is only
by choosing the most upstream one (N-terminal tag) or the most downstream
one (C-terminal tag) that one minimizes the number of unrelated amino acids
at the junction.
In all cases, check that the intended sites are not present in the target gene.
If this is not the case, the vector has to be cut in two pieces for a ligation with
three partners (as described in Section 2.5.3) or another expression vector has
to be chosen.
2.5.2 Design of oligonucleotide primers
The sense (upstream) primer should contain, in the following order:
• a 4 base extension, for example TAGC, to make an overhang allowing
cleavage by the restriction enzyme
• the selected upstream restriction site
• the initiation codon ATG (overlapping with the restriction site in E. coli
vectors), except in case of fusion with an N-terminal tag
• 0 to 2 added bases to preserve the reading frame
• a 23 base sequence identical to the beginning of the sequence or, if one
wishes to reduce the size of the protein, of the chosen coding region.
The antisense (downstream) primer should contain, in the following order:
• a 4 base extension, for example TAGC, to make an overhang allowing
cleavage by the restriction enzyme
• the selected downstream restriction site
• a TTA, CTA, or TCA sequence, complementary to a stop codon, except in
case of fusion with a C-terminal tag
• 0 to 2 added bases to preserve the reading frame (in case of C-terminal tag)
• a 23 base sequence complementary to the end of the sequence, or, if one
wishes to reduce the size of the protein, to the end of the chosen coding
region.
2.5.3 Case of large inserts (> 1500 bp)
In principle Protocol 2 could function with any size of coding sequence.
However, the yield of the PCR amplification will decrease with the size of the
54
 3: Molecular biology for structural biology
fragment, and the risk of introducing mutations will increase. Also, the whole
PCR-amplified fragment has to be checked by sequencing. For these reasons,
there might be a better strategy in the case of a large insert (> 1500 bp) which
consists in amplifying only one part of the sequence, up to an internal
restriction site, and then constructing the vector by the ligation of three pieces
or, alternatively, in two subcloning steps.

2.5.4 From expression vector to expression strain

Once a new expression vector is constructed, it has to be introduced into the
host and tested for protein expression. The various stages of this process (cell
transformation, transfection or infection, clone selection, cell cultivation, and
induction of protein expression) are host- and vector-dependent and it is not
our purpose to detail them. We can recommend handbooks treating E. coli
(18), yeast (24), baculovirus (2), and mammalian (18) cell expression and to
refer to the technical tips accompanying commercial expression vectors.

3. Engineering physical properties of macromolecules

Direct modification of the properties of a protein by modifying its sequence is
another strategy for preparing suitable samples for 3D structure deter-
mination. Such an approach is generally not considered first, but rather when
difficulties arise. If this strategy is chosen, make sure that the biological
activity of the wild-type protein is conserved in the variants generated for
crystallogenesis.

3.1 Problems encountered and possible solutions

Various problems can be encountered at different stages of a structure
determination. They are generally detected by one of the following signs:
(a) During purification, the protein tends to aggregate or exhibits solubility
problems: these will likely become more pronounced when the protein is
concentrated for crystallization.
(b) The purified protein is heterogeneous, contaminated by slightly differing
variants, which are often proteolysis products (see Chapter 2).
(c) Changes in a protein's characteristics (e.g. solubility in the presence of
precipitating agents), during crystallization trials indicate protein in-
stability, typically the most critical case being the spontaneous formation
of irreversible precipitates in the concentrated protein solution.
(d) Diffraction patterns exhibit poor resolution, although crystals are treated
under optimal conditions.
(e) No useful heavy-atom derivative can be obtained by a traditional
screening approach.
55
 P. F. Berne et al.
Molecular biology may offer solutions in each of these cases. In most cases,
understanding the phenomenon may help in choosing the appropriate changes
in the protein sequence. First analyse the nature of the problem by all
available experimental tools: electrophoresis, isoelectric focusing, gel nitra-
tion, native electrophoresis, mass spectrometry, activity and stability assays,
dynamic light scattering (see Chapter 2).
Proteins can suffer various types of chemical degradation, and appropriate
alterations could reduce these instabilities. One frequently recognized
mechanism is oxidation of sulfhydryl groups, which can lead to formation of
oligomers, and/or to aggregation of the protein. In all cases, oxidation
produces some degree of heterogeneity in the protein solution, detrimental to
the crystallization efforts. Adding reducing agents is recommended to slow
down this phenomenon but, in some cases, it could be beneficial to mutate a
particularly sensitive cysteine residue to a serine in order to produce or
improve crystals.
When a protein exhibits a tendency towards aggregation, however, it is
generally not so simple to identify the residues responsible for this behaviour.
When the formation of disulfide bridges can be ruled out, it is often thought
that protein molecules associate through hydrophobic residues that are
located near the surface, or which become exposed under partially denaturing
conditions (freezing and thawing, warming, addition of chaotropes). De-
naturation being avoided, one can stabilize a protein by changing hydro-
phobic amino acids into more hydrophilic ones using mutagenesis. This type
of strategy was successfully used to allow the crystallization of the HIV
integrase (25). The authors faced a high tendency of the integrase to aggre-
gate. They systematically replaced every hydrophobic residue, except those
strictly conserved among the family of retroviral integrases, which where
supposed to play a critical structural role. As a general rule, the less conserved
hydrophobic amino acids are within a family of proteins, the better their
chance to be localized at the surface and be responsible for aggregation. An
alignment of hydrophobic cluster analysis (HCA) plots (26) can be used to
identify such residues, which are prime candidates for mutagenesis.
Current characterization methods, as mentioned above, allow detection of
the main problems of stability and heterogeneity. Mass spectrometry reveals
other potential sources of microheterogeneity, such as the presence of post-
translational modifications. An example of post-translational modification
that hampers crystallization is autophosphorylation of the protein, as found
with a tyrosine kinase (27). The authors decided to mutate the relevant
tyrosine residues into phenylalanine and succeeded in crystallizing the
protein.
There are, however, cases where lack of crystallization is not linked to a
problem of protein heterogeneity or stability, but to the impossibility of the
protein to establish the adequate intermolecular interactions required for
nucleation and crystal growth. Mutations can be created that favour these
56
 3: Molecular biology for structural biology
processes, as was done for fibronectin crystallization (28). Here, engineering
the protein to have an isoleucine at the C-terminus allowed its easy crystal-
lization and it was later observed that this isoleucine was involved in crystal
packing. However, our current understanding of the mechanisms of protein
crystallization does not allow predictions on which mutations will favourably
affect the interactions leading to crystallization.
Finally, when difficulties arise only in the phasing stage of structure deter-
mination, specific approaches can be employed, like the production of seleno-
methionyl proteins, which is the subject of Section 4, or cysteine mutagenesis,
recently discussed by Martinez-Hackert et al. (29).

3.2 Defining the optimal size of the molecule to be

crystallized
Although one should always first attempt to crystallize full-length proteins,
useful structural data have been obtained from systems that are hard to
crystallize by cutting out regions that interfere with crystal growth. This
approach may consist of limiting the study to a single functional domain
(typically a binding domain in the case of the crystallization of a complex) or
shortening the size by eliminating a dispensable extremity (e.g. dimerization,
nuclear localization, or membrane anchorage domains).
There are several methodological reasons for reducing the size of the
studied protein. Over-production of high molecular weight proteins is gener-
ally not very efficient. Truncating a protein often decreases its flexibility, a
property which is in general detrimental to crystallization. This is true for
proteins consisting of two or more functional domains joined by a flexible link
as well as for proteins carrying flexible segments at one extremity. An addi-
tional reason to eliminate flexible parts in proteins is that they are generally
good substrates for proteases. Therefore, a residual proteolytic activity
present in the sample could easily generate heterogeneity, as illustrated by the
crystallization of the interferon -y receptor (30), which required the deletion of
eight amino-terminal residues. A flexible extremity might even be responsible
for the poor diffraction quality of crystals (31).
Even though proteolysis under specific conditions has been used in some
cases for the preparation of a shorter variant for crystallization (32), genetic
engineering of a protein is a more secure way to obtain the same result. The
major difficulty of this approach consists in choosing the borders of the
domain to be expressed. Ideally, this domain should correspond to a struc-
tural and functional domain and be devoid of terminal parts dispensable for
activity. Identifying a functional domain in a protein can result from sequence
alignments although the corresponding structural domain is likely to
overhang on both sides the region of homology. Secondary structure and/or
other predictive methods (e.g. HCA diagrams) can be used to define the
position of the junction between two domains, which should correspond to a
57
 P. F. Berne et al.

TableS. Commonly used proteases

Protease Supplier Storage condition Incubation

(2-10mg/ml) buffer
Trypsin Sigma Ref.T8642 1 mM HCI, 20 mM Tris, 10 mM
20%glycerol,-80°C CaCI 2 pH8
Chymotrypsin Merck Ref. 2307 1 mM HCI, 20% glycerol 20 mM Tris, 10 mM
-80°C CaCI2 pH8
Subtilisin Boehringer Mannheim 20 mM Tris pH 8, 20 mM Tris, 10 mM
Ref. 165 905 10 mM CaCI2, CaCI2 pH8
20% glycerol, -80°C
Thermolysin Boehringer Mannheim 20 mM Tris pH8, 20 mM Tris, 10 mM
Ref. 161 586 10 mM CaCI2, CaCI2 pH8
20% glycerol, -80°C
Endoproteinase Boehringer Mannheim 20 mM sodium 25 mM sodium
Glu-C Ref. 791 156 phosphate pH 7.8, phosphate pH 7.8
20% glycerol, -80°C
Papain Boehringer Mannheim Suspension, 4°C 20 mM sodium
Ref. 108 014 phosphate pH 8

gap in the secondary structure. More reliable information should be obtained

from experimental data such as limited proteolysis (see Protocol 3) or
expression and characterization of truncated species (33, 34). Such a strategy
led to the successful crystallization of the kinase domain of the insulin
receptor (27) and of the catalytic domain of type II adenylyl cyclase (31).
From a practical point of view, it is quite easy for a molecular biologist to
modify an existing expression vector and to express a shorter form of a
protein. For this purpose, we suggest following the same general strategy as
described in Protocol2 (general subcloning protocol).

Protocol 3. Limited proteolysis assays

Equipment and reagents

• Proteases and protease inhibitors (see Table 3) • Equipment for SDS gel electrophoresis
• Water-bath . Polypeptide sequencing facility

Method
1. Prepare, as substrate for the assays, a stock solution of your protein
(2-10 mg/ml) in a buffer in which it is stable.
2. Select proteases exhibiting broad specificity (with several potential
sites inside proteins, e.g. trypsin, chymotrypsin, subtilisin, thermo-
lysin, papain, endoproteinase Glu-C). Prepare stock solutions (2-3
mg/ml) and store them as frozen aliquots in appropriate buffers (see
Table 3) containing 20% glycerol.a

58
 3: Molecular biology for structural biology
3. Prepare on ice serial dilutions (0.1 mg/ml down to 10 ng/ml) of the
proteases in their specific incubation buffers. In a first series of
experiments, serial tenfold dilutions might be done and the conditions
may then be refined in subsequent experiments.
4. Mix a small volume of the concentrated protein solution (2-5 ul) with
the diluted protease solution (40 ul for example) in order to reach a
final concentration of about 0.2 mg/ml.
5. Incubate for 1 h at room temperature or 37°C. Stop the reaction by
addition of a specific inhibitor. We recommend PMSF (1 mM final
concentration) for all the mentioned serine proteases and E-64 for
papain (cysteine protease) (see also Chapter 2).
6. Analyse the reaction products by migration on SDS-PAGE. Try to
identify the products of proteolysis at preferred sites. These products
correspond to a unique cleavage site at low protease concentration,
whereas a higher protease concentration leads to non-specific
cleavage at multiple sites.
7. For each protease, optimize the protease concentration that leads to
the limited cleavage at the preferred site. Once these conditions have
been established, perform a larger scale reaction in order to isolate a
larger amount of the proteolysis products.
8. Isolate the gel band of interest using PAGE followed by transfer on a
polyvinylidene difluoride (PVDF) membrane (Millipore). Give the
samples to a specialized service for N-terminal sequence analysis.6
9. The sequence information originating from various proteases may
pinpoint areas in the protein that are especially sensitive to proteo-
lysis. These positions probably correspond to connections between
various domains in the protein.

'Appropriate storage conditions have to be investigated for other proteases. Most importantly,
the activity has to be strictly reproducible from one experiment to another, and this is the
reason for using frozen aliquots (see also Chapter 2).
b
Alternatively, the protein species might be analysed by mass spectrometry, as in ref. 27.

3.3 Site-directed mutagenesis

Whatever the envisaged change in the amino acid sequence (amino acid
replacement, deletion, or insertion) it is possible to use a general and simple
method based on PCR amplification, named the overlap extension method
(35). The main requirement of this technique is the synthesis of four oligo-
nucleotides, used as primers in the PCR amplifications, two primers carrying
the appropriate DNA modification, and two 'external' primers. This is gener-
ally not a limitation, since synthesis of oligonucleotides has become cheap and
accessible to all laboratories.
59
 P. F. Berne et al.

Protocol 4. Site-directed mutagenesis using PCR

Equipment and reagents

• As for Protocol 2

Method
1. Design two mutagenic primers, a sense primer (primer a), encoding
the target sequence, and an antisense primer (primer b), strictly com-
plementary to primer a. The sense primer could include one or a few
modified bases, an insertion, or a deletion, provided that these modi-
fications are surrounded by two 13 base segments strictly com-
plementary to the starting sequence. These segments are necessary to
ensure that the primers will recognize the template in spite of the
sequence changes.
2. In order to design two external primers (20- to 22-mers with 50-60%
GC content), identify two unique restriction sites located each side of
the mutation, separated by 200-1500 base pairs. Choose a sense
primer (primer c) overlapping with or immediately upstream from the
upstream site and an antisense primer (primer d) overlapping with or
immediately downstream from the downstream site.
3. Perform two independent PCR reactions using the starting vectora as
template (10 ng) and standard PCR conditions. The first reaction uses
primers a and d, the second reaction uses primers b and c. In this way,
the PCR products will be two fragments overlapping over the segment
encoded by the mutagenic primers.
4. Isolate the two PCR products using agarose gel electrophoresis.
5. Perform a third PCR reaction using the two external primers (c and d)
and, as a template, a mixture of the two previous PCR products (~ 10
ng each). Using gel electrophoresis, check an aliquot for the presence
of the expected final PCR product.b
6. Follow Protocol 2, steps 3-7.

a
If available, use as template a vector containing your gene but different from the final vector.
That will facilitate the identification of the mutated versus wild-type clones. The template
vector should, of course, contain all the region between the two external primers and be
recognized by these primers.
b
During the hybridization step, some hybrids will appear carrying one strand of each fragment
annealed through their overlapping segment. These hybrids will be complemented to double-
strand DNA by the polymerase and will constitute an efficient template for further
amplification by the external primers.

60
 3: Molecular biology for structural biology
The overlap extension method is described in Protocol 4, adapted from ref.
35. Note that there are simpler strategies for mutagenesis in some particular
cases (e.g. mutations close to an existing restriction cleavage site). The
method proposed here has the advantage of being general. It is cheap and
relatively easy to perform, and one can envisage constructing several directed
mutations in parallel, if required. In this case, the external primers are
common to all mutations and only two oligonucleotides are required for
each specific mutant. Additionally, it is often possible to combine directed
mutagenesis with subcloning in another expression vector, as described in
Protocol 2, by designing appropriately the external primers.

3.4 Random mutagenesis

3.4.1 Background
As mentioned above, a great number of mutations at hydrophobic residues
were generated to obtain a more soluble variant of HIV integrase (25). It is
possible through random mutagenesis to generate an even broader range of
variants. This strategy can be used to improve a particular property of a
protein (level of expression, solubility, tendency towards aggregation). It is
especially suited when the mutations that have a chance to succeed cannot be
predicted. The less precisely the encountered problem can be localized in
terms of protein sequence, the larger will be the number of mutations that
have to be generated and tested.
3.4.2 Strategy
In comparison to classical chemical methods for random mutagenesis, a more
recent method (36) produces a broader spectrum of mutations. This method
relies on amplification of the gene — or piece of gene — of interest by Taq
DNA polymerase. This enzyme is usually employed in PCR reactions because
of its high thermostability. Although this enzyme is relatively accurate under
normal PCR conditions, it tends to introduce random mutations under special
conditions, i.e. in the presence of Mn2+. Protocol 5 (according to ref. 36) gives
a guideline for this procedure.

Protocol 5. Random mutagenesis

Equipment and reagents

• As for Protocol 2 • Taq DNA polymerase

Method
1. Choose the part of the protein DNA sequence that you want to
mutagenize randomly. It should be included between two unique
restriction sites. It may be necessary to first introduce these restriction
sites by directed mutagenesis.

61
 P. F. Berne et al.
Protocol 5. Continued
2. Design oligonucleotides as primers for a PCR amplification. Choose a
sense primer immediately upstream from the first restriction site and
an antisense primer immediately downstream from the second site.
3. Perform the PCR reaction under specific conditions in order to favour
the misincorporation of deoxyribonucleotides by Taq DNA polymer-
ase. The PCR reaction should include, in addition to the usual buffer,
0.5 mM MnCI2, 7 mM MgCI2, and modified concentrations of dNTPs
(0.2 mM of dATP and dGTP, 1 mM of dTTP and dCTP). Use a very low
initial concentration of the plasmid template (e.g. 1 pg in a reaction
volume of 50 ul), as the error rate of the PCR will increase if the
amplification factor is increased.a
4. Follow Protocol 2, steps 4-7.
5. Sequence several clones to determine the average number of muta-
tions per clone. An average rate of one per clone seems reasonable for
most applications. It might be necessary to sequence a series of clones
coming from PCR reactions starting with various amounts of DNA
template in order to identify the conditions that generate the
appropriate rate of mutation.b
6. Using the ligation mixture selected according to step 5, transform an
appropriate expression strain, e.g. BL21.
7. Grow small volume cultures of several clones to characterize potential
protein variants.
a
It might be necessary to adjust the amount of polymerase, the annealing temperature, the
length of elongation, the number of cycles, in order to get a good yield of amplification,
depending on the length of the amplified fragment and on the choice of the primers.
b
The rate of various mutation types might be adjusted by varying the nucleotide
concentration, as described by Fromant et al. (63).

3.5 Selection of variants following mutagenesis

When generating multiple protein variants in order to improve crystallization,
an obvious question is which criteria to choose to discriminate among
different constructs. The ultimate goal is to obtain diffraction-quality crystals.
However it is time-consuming to purify large amounts of several variants and
to carry on crystallization trials with each of them. There is therefore a need
for preliminary selection criteria that can be applied to many variants at an
early stage of the process.
Various criteria for the selection of variants can be retained, depending on
the nature of the problem encountered. Expression level or solubility in E. coli
can be tested easily in parallel on multiple clones using electrophoresis or,
even simpler if available, an immunodetection method. To assess the chances
62
 3: Molecular biology for structural biology
of succeeding in crystallizing a protein, one could utilize methods that analyse
its aggregation status. This can be done using native electrophoresis, analy-
tical gel filtration chromatography, or dynamic light scattering. These various
techniques require 10-100 ug of purified protein. In practice, the comparative
test of several variants using these techniques requires that the protein be well
overexpressed and purified selectively in one step. E. coli as expression host
and a fusion with a purification tag are, therefore, almost prerequisites.
Protocol 6 provides a guideline for screening variants generated by random
mutagenesis. This strategy has been used in our laboratory (P. F. B., personal
communication), but has to be adapted to other cases. The studied protein
exhibited multiple oligomerization states and we looked for a more homo-
geneous variant. The phenomenon was clearly detectable using native
electrophoresis. We took advantage of the high level of expression in bacteria
and of an easy and selective purification step. Growth of bacteria, lysis, and a
small scale purification could be performed in 96-wells plates, so that 2000
clones could be analysed in a few months. A few of them were stabilized as
dimers and looked homogeneous in native electrophoretic analysis.
Although there are still few reported cases of random mutagenesis to
engineer a protein for crystallization, it should be considered as a possible
approach for proteins particularly reluctant to crystallize. The development of
tools for the parallel micropurification of several proteins, such as Ni-NTA
magnetic agarose beads (Qiagen) for polyhistidine tagged proteins, will
facilitate such approaches.

Protocol 6. Screening for mutants derived from random

mutagenesis

Equipment and reagents

• As for Protocol 2 • Equipment for gel electrophoresis
• Lysozyme, DNase, ammonium sulfate solution

Method
1. Add 100 ul of a bacterial growth medium, supplemented with the
appropriate antibiotics, to each well of a microtitre plate.
2. With a tip, pick up colonies out of a Petri dish originating from a
transformation made the day before. Inoculate each well with a unique
colony and strike the tip on a new dish at an identified position to keep
a replicate of each clone.
3. Let the cultures grow by incubating the plate for 6 h at 25°C. In our
case, the gene was cloned into vector pET3®, which exhibited a
sufficient level of expression without IPTG under these conditions.
When working with a vector with a more tightly regulated promoter, it
could be necessary to induce the expression in a more specific way.

63
 P. F. Berne et al.
Protocol 6. Continued
4. Collect the cells by centrifuging the plate for 10 min at 2000 g (Sigma
centrifuges, for instance, have rotors that accept microtitre plates).
Suck in carefully the medium. The cells can be frozen at this stage for
future use.
5. This step and the following are done easily using a multichannel
pipette. Lyse the cells by resuspending the pellets in 100 ul of a buffer
containing 50 mM Tris, 20 mM NaCI, 0.1% Triton X-100, 0.5 mM EDTA,
10 ug/ml lysozyme pH 8.4. Incubate for 20 min at room temperature.
6. Add 10 ul of a DNase solution (10 ug/ml in the same buffer as above
plus 20 mM MgCI2) and homogenize each well. Incubate for 20 min at
room temperature.
7. Centrifuge the plate for 20 min at 2000 g. Carefully transfer the
supernatant to another plate.
8. Perform a purification step, which could be an incubation with an
affinity resin. In our case, the protein could be precipitated selectively
by the addition of 0.6 M ammonium sulfate and recovered after
centrifugation and elimination of the supernatant.
9. Analyse the protein by native PAGE electrophoresis. This gives a good
idea of the state of the protein under native conditions. Ideally, the
protein is expected to migrate as a unique and sharp band.

4. Preparation of selenomethionyl protein crystals

4.1 Background
Since the first edition of this book in 1992, the use of selenomethionyl proteins
for phase determination has increased dramatically. Selenomethionine sub-
stitution now accounts for about two-thirds of all structures solved by multiple
wavelength anomalous dispersion (MAD) (37). Use of MAD has been intro-
duced to circumvent non-isomorphism problems that can occur when using
heavy metal derivatization as a phasing method (38, 39). The MAD method
exploits the presence of anomalous scatterers, such as copper or iron in
metalloproteins, heavy metal derivatizing agents, or selenium in seleno-
methionyl proteins. All measurements relevant to determining a single phase
can be made on the same crystal so isomorphism is exact, and electron density
maps are very often of high quality (40, 41). With the increased number of
synchrotron beam lines dedicated to MAD experiments, the MAD method
has rapidly become a general method for phase determination.
The pioneering work of Hendrickson and co-workers showed that selenium
is a useful anomalous scatterer (42). Selenomethionine can totally replace
methionine in E. coli (43). Substitution of methionine by Selenomethionine
64
 3: Molecular biology for structural biology
thus offers a general method for introducing anomalous scatterers into cloned
proteins. Moreover, most selenomethionyl protein crystals are isomorphous
to their native counterpart. As a result, the difference of 18 e- between
selenium and sulfur can be used in a conventional isomorphous replacement
structure determination. In addition to its use as an anomalous scatterer and
isomorphous derivative, selenomethionine presents the additional advantage
of pinpointing methionines, therefore aiding model building.
Preparation and crystallization of selenomethionine-substituted proteins
are straightforward procedures (42, 44). In brief, one needs to:
• express the cloned protein in a strain auxotrophic for methionine or more
simply, block methionine biosynthesis
• ferment this strain in a medium in which methionine is replaced by
selenomethionine
• avoid oxidation during purification of the substituted protein
• crystallize the substituted protein under conditions similar to those used for
the native protein.

4.2 Expression and cell growth in a prokaryotic system

4.2.1 Expression in a methionine auxotroph strain
i. Transformation
Historically, the procedure used to engineer selenomethionyl proteins has
consisted of transforming an existing methionine auxotroph strain (mer) with
a plasmid containing the cloned gene of the protein of interest and growing
the resulting strain in a medium devoid of methionine and containing seleno-
methionine. E. coli met- strains differ in their tolerance to selenomethionine.
LeMaster studied the selenomethionine tolerance of several met~ strains and
constructed a strain, DL41, which grows well on selenomethionine-containing
media and therefore can be used as a general host for plasmid transformation
(42). Note: the strain DL41 can be obtained from the E. coli Genetic Stock
Center, Yale University School of Medicine, New Haven, CT 06510 USA
(http://cgsc.biology.yale.edu). Plasmid transformation is carried
out by standard procedures (18).
ii. Cell growth
Although selenomethionine is recognized almost equally well as methionine
by methionyl-tRNA synthetase (Doublie and Carter, unpublished observa-
tions), cells grow more slowly in selenomethionine and generally reach
stationary phase at a lower final cell density. Furthermore, cells grown in
selenomethionine tend to stay in stationary phase. A low percentage of LB in
the starter culture of a met- strain will provide sufficient methionine to revive
the cells from stationary phase, as well as thiamine and biotin. Hydrolysed LB
contains 5 mg/litre of methionine, which will be incorporated preferentially to
selenomethionine during fermentation (45). To minimize the amount of
65
 P. F. Berne et al.

Table 4. Starter medium for met" strains

Ingredients Concentration
1. Minimal mediuma Minimal medium Ab or M9 (18), with carbon
source 5 g/litre
2. All amino acids except methionine" 40 mg/litre
3. Selenomethioninec 20-60 mg/litre
4. LB (12) x%(v/v) to be determinedd
a
One can also use LeMaster's medium (60) instead of 1 and 2.
b
Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, J. A., and Struhl, K.
(ed.) (1987). Current protocols in molecular biology. Greene Publishing Associates and Wiley
Interscience, New York, NY.
c
L-selenomethionine can be purchased from Fisher/Acros or Sigma.
d
As an example, Yang et al. used a 100 ml starter medium containing 5% (v/v) LB for a 20 litre
fermenter (43).

Table 5. Fermentation medium for met strains

Ingredients Concentration
1. Minimal mediuma Minimal medium Ab or M9 (12), with carbon
source 5 g/litre
2. All amino acids except metnioninea 40 mg/litre
3. Selenomethionine 20-60 mg/litre
4. Thiamine 2 mg/litre
5. Biotin (if needed) 2 mg/litre
a
One can also use LeMaster's medium (60) instead of 1 and 2.
b
Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, J. A., and Struhl, K.
(ed.) (1987). Current protocols in molecular biology. Greene Publishing Associates and Wiley
Interscience, New York, NY.

residual methionine in the purified protein, the dilution factor for the final
inoculation must be adjusted according to the amount of LB used in the
starter inoculum. This amount will be a compromise between a better growth
rate and complete selenomethionine substitution. For each particular strain,
one will have to determine the amount of rich medium in the starter culture,
as well as the optimal concentration of selenomethionine throughout the
fermentation (see Protocol 7).

Protocol 7. Cell growth of met strains

Equipment and reagents

• 100 ml of starter medium (Table 4) • Fermentation medium (Table 5)

Method
1. Isolate single colonies by streaking an LB plate (18) supplemented with
antibiotics with strain of interest. Incubate at 37°C.

66
 3: Molecular biology for structural biology
2. Ferment the cells in medium containing all appropriate antibiotics.
(a) Inoculate 100 ml of starter medium with a single colony. Shake at
37°C.
(b) Inoculate a 10-20 litre fermentera containing pre-warmed ferment-
ation medium with the 100 ml starter inoculum in mid-log phase.
(c) Monitor cell growth in order to identify times for induction and
harvest.
3. Induce, if necessary.
4. Harvest the cells in mid- to late log phase.b Resuspend the harvested
cells in the appropriate lysis buffer and quick-freeze in dry ice or liquid
nitrogen. Store at-80°C.
a
Cell growth should be done in a fermenter because regulated temperature and pH improve
the yield.
b
We have noticed cell lysis shortly after cells reached late log phase. Care must be taken to
harvest cells as quickly as possible.

4.2.2 Methionine pathway inhibition

In prokaryotes, an alternative to using a met- strain is simply to block
methionine biosynthesis (46). Aspartate is a precursor in the biosynthesis
pathway of lysine, threonine, and homoserine (an intermediate in the form-
ation of methionine). Lysine and threonine block the methionine biosynthesis
pathway in E. coli by inhibiting the enzymes that phosphorylate aspartate
(aspartokinases). Moreover, phenylalanine and leucine are known to act in
synergy with lysine. One can therefore produce selenomethionyl protein by
growing a non-auxotroph E. coli strain in the absence of methionine but with
ample amounts of selenomethionine and of the amino acids known to block
methionine biosynthesis. This procedure has been successfully applied to a
number of proteins, among them FKBP12 (46), UDP-JV-acetylenopyruvyl-
glucosamine (47), T7 DNA polymerase (48), and human 3-methyladenine
DNA glycosylase (49). This procedure is more straightforward than the one

Table 6. Fermentation medium for methionine pathway inhibition

Ingredients Concentration
1. Minimal medium Minimal medium Aa or M9 (18), with carbon
source 5 g/litre
2. Thiamine 2 mg/litre
3. Biotin (optional) 2 mg/litre
4. Lysine, phenylalanine, and threonine 100 mg/litre
Isoleucine, leucine, and valine 50 mg/litre
5. Selenomethionineb 60 mg/litre
a
Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, J. A., and Struhl, K. (ed.)
(1987). Current protocols in molecular biology. Greene Publishing Associates and Wiley Interscience,
New York, NY.
b
L-selenomethionine can be purchased from Fisher/Acros or Sigma.

67
 P. F. Berne et al.
involving a met- strain and should be applicable to any prokaryotic strain (see
Protocol 8).

Protocol 8. Methionine pathway inhibition

Equipment and reagents

• Current molecular biology reagents • LB medium 08), minimum medium (Table 6)

Method
1. Isolate single colonies by streaking an LB plate (18) supplemented with
antibiotics with strain of interest. Incubate at 37°C.
2. Inoculate 1 ml of LB medium with a single colony. Grow overnight.
3. Spin down cells (2 min at 1300 g in a microcentrifuge) and resuspend
in 1 ml of supplemented minimum medium (items 1, 2, and 3 in Table
6). Inoculate 1 litre of the same, pre-warmed medium.
4. Add all seven amino acids at mid-log phase.
5. Induction is done 15 min after addition of the amino acids, if
necessary.
6. Harvest the cells in mid- to late log phase. Resuspend the harvested
cells in the appropriate lysis buffer and quick-freeze in dry ice or liquid
nitrogen. Store at -80°C.

4.3 Eukaryotes
Animal organisms are naturally auxotrophic for methionine and cells can
grow in a selenomethionine containing medium with good incorporation of
the modified amino acid. There are to date only a few examples of seleno-
methionyl protein production in eukaryotic cells. Selenomethionyl human
chorionic gonadotropin (hCG) was produced in insect cells (Sf9) (50, 51) as
well as in Chinese hamster ovary (CHO) cells (52, 53). The selenomethionyl
variant of a functional fragment of sialoadhesin was also produced in CHO
cells (54). The substitution rates reported for these three proteins range from
84-92%.

4.4 Purification
Introduction of selenomethionine into proteins has two consequences that
impact on purification. The altered chemistry of selenium makes substituted
proteins more sensitive to oxidation than natural proteins. Moreover, if
selenium atoms are solvent exposed, they can alter protein solubility and
behaviour on chromatography resins. These properties require the following
modifications to the normal purification as shown in Protocol 9.
68
 3: Molecular biology for structural biology

Protocol 9. Purification 'of selenomethionyl proteins

Equipment and reagents

• Chromatographic equipment • Ethylenediaminetetraacetic acid (EDTA)
• Dithiothreitol (DTT)

Method
1. Purify as quickly as possible, with modifications to avoid oxidation.
(a) Degas all buffers by boiling or evacuation.
(b) Include a reducing agent such as DTT and a chelator such as EDTA
to remove traces of metals that could catalyse oxidation (55) (see
also Chapter 2). Use 0.2-1 mM EDTA and 5-20 mM DTT.
2. Expect selenomethionyl proteins to be slightly less soluble than their
natural counterparts.
(a) Anticipate lower optimal ammonium sulfate concentrations in
trituration protocols.
(b) Anticipate increased retention in some chromatography procedures.
3. Store purified protein in an oxygen-free environment, at -80°C in the
presence of glycerol, or if possible, as frozen droplets at -180°C.
4. Mass spectroscopy is the most accurate method to quantitate seleno-
methionine incorporation. If this technique is not available, one can
undertake an amino acid analysis to check the percentage of subs-
titution. Selenomethionine is destroyed under the acid hydrolysis
conditions used in amino acid analysis, so that it is the disappearance
of methionine that is monitored.

4.5 Crystallization
Experience to date suggests that selenomethionyl proteins crystallize in con-
ditions that are very similar to those used with native proteins (42, 45, 48, 49).
As a consequence of the lowered solubility of selenomethionyl proteins,
either the protein or the precipitant concentration should be slightly reduced
to achieve comparable degrees of supersaturation. It is often the case that
growth of selenomethionyl protein crystals require microseeding with a
crushed wild-type protein crystal. Selenomethionine oxidation can lead to
aberrant X-ray fluorescence spectra in which the position and shape of the K-
edge are altered and the white line intensity decreased (37). This can be
avoided by maintaining the crystals in a solution containing DTT and EDTA.
Crystals should be stored in an oxygen-free environment such as an anaerobic
chamber if possible (see Chapter 5 for a crystallization method suitable for
oxygen-sensitive proteins). They should be irradiated as soon as possible or
69
 P. F. Berne et al.
flash-frozen and stored in liquid nitrogen while they await data collection.
Selenomethionine incorporation does not appear to alter diffraction limits
and selenomethionyl protein crystals are generally isomorphous with native
crystals. However, they can be more sensitive to radiation damage.
There are now several large protein structures (> 90 kDa) that have been
solved by the MAD method using solely the anomalous signal of selenium
(48, 56; J. L. Smith, personal communication). It is also clear that large
numbers of selenium sites (15 or more) can be readily located with direct
methods programs. This realization should increase even further the wide-
spread use of selenomethionyl proteins for phase determination. There are
also encouraging results regarding the incorporation of telluromethionine
into proteins (57, 58). Even though tellurium cannot be used as an anomalous
scatterer (its K-edge (0.389 A) and L-edges (> 2.5 A) correspond to wave-
lengths not usually reachable at synchrotron facilities), its 36 e~ difference
with sulfur has been successfully used in conventional isomorphous replace-
ment methods (58). This procedure will reach its true potential when telluro-
methionine becomes commercially available.

4.6 Warning
Selenium is an essential element for most animal and bacterial life (59), but
it is also a very toxic compound because of its ability to replace sulfur. In
mammals (e.g. protein crystallographers), ingested methionine is a source of
sulfur. As a result, selenomethionine can be harmful or even fatal if inhaled,
swallowed, or absorbed through the skin. Experiments should always be done
in a hood and the experimenter should be sure to wear gloves. Experimenters
should contact the Health and Safety office at their institution and inquire
about proper disposal of selenomethionine containing media.

5. Conclusion
Structural studies require homogeneous concentrated solutions of the protein
of interest. Some proteins cannot be obtained in this form and optimizing the
expression system can sometimes solve the problem. However, even an
apparently perfect protein solution might be reluctant to crystallize, perhaps
because of surface residues that could form destabilizing intermolecular con-
tacts. In some cases, such a protein can be stabilized by the addition of specific
agents like glycerol, detergents, zwitterions, or amino acids (61, 62) (see also
Chapter 2). An alternative approach consists in identifying residues respons-
ible for the instability of the protein and mutating them. Chances are that
these residues, being located at the surface, will not be crucial for keeping the
overall 3D folding of the protein. This chapter gives an overview of the genetic
engineering tools that one can exploit to optimize expression systems and to
design variants by systematic or random mutagenesis. These modifications
70
 3: Molecular biology for structural biology
can lead to improved crystallization or to novel physical properties useful for
structure determinations by X-ray diffraction studies. Such macromolecular
engineering strategies may find routine use in modern structural biology.

References
1. Pawson, T. (1995). Nature, 373, 573.
2. Gruenwald, M. D. and Heitz, M. S. (1993). In Baculovirus expression vector
system: procedures and methods. A manual published by Pharmingen.
3. Miegel, A., Sano, K. I., Yamamoto, K., Maeda, K., Maeda, Y., Taniguchi, H., et al
(1996). FEBS Lett., 394, 201.
4. Marchot, P., Ravelli, R. B. G., Raves, M. L., Bourne, Y., Vellom, D. C., Kanter, J.,
et al. (1996). Protein Sci., 5, 672.
5. Cregg, J. M., Vedvick, T. S., and Raschke, W. C. (1993). Biotechnology, 11, 905.
6. Kern, D., Dietrich, A., Fasiolo, F., Renaud, M., Giege, R., and Ebel, J.-P. (1977).
Biochimie, 59, 453.
7. Weijland, A., Williams, J. C., Neubauer, G., Courtneidge, S. A., Wierenga, R. K.,
and Superti-Furga, G. (1997). Proc. Natl. Acad. Sci. USA, 94, 3590.
8. Friehs, K. and Reardon, K. F. (1993). Adv. Biochem. Eng. Biotech., 48, 53.
9. Olins, P. O. and Lee S. C. (1993). Curr. Opin. Biotech., 4, 520.
10. Dale, G. E., Broger, C., Langen, H., D'Arcy, A., and Stuber, D. (1994). Protein
Eng., 7, 933.
11. Kim, K. K., Yokota, H., Santoso, S., Lerner, D., Kim, R., and Kim, S. H. (1998).
J. Struct. Biol., 121, 76.
12. Kozak, M. (1997). EMBO J., 16, 2482.
13. Schein, C. H. (1989). Biotechnology, 7, 1141.
14. Blackwell, J. R. and Horgan, R. (1991). FEBS Lett, 295, 10.
15. Caspers, P., Stieger, M., and Burn, P. (1994). Cell. Mol. Biol, 40, 635.
16. Yasukawa, T., Kanei-Ishii, C., Maekawa, T., Fujimoto, J., Yamamoto, T., and
Ishii, S. (1995). J. Biol. Chem., 270, 25328.
17. Mayer, M. P., Prestwich, G. D., Dolence, J. M., Bond, P. D., Wu, H., and Poulter,
CD. (1993). Gene, 132, 41.
18. Maniatis, T., Fritsch, E. F., and Sambrook, J. (ed.) (1989). Molecular cloning: a
laboratory manual, 2nd edn. Cold Spring Harbor Press, Cold Spring Harbor, NY.
19. Geng, X. and Chang, X. (1992). J. Chromatogr., 599, 185.
20. Ford, C. F., Suominen, I., and Glatz, C. E. (1991). Protein Expression Purification,
2, 95.
21. Stoll, V. S., Manohar, A. V., Gillon, W., Macfarlane, E. L. A., Hynes, R. C., and
Pai, E. F. (1998). Protein Sci., 7, 1147.
22. Zhang, F., Robbins, D. J., Cobb, M. H., and Goldsmith, E. J. (1993). J. Mol. Biol.,
233, 550.
23. Fields, B. A., Ysern, X., Poljak, R., Shao, X., Ward, E. S., and Mariuzza, R. A.
(1994). J. Mol. Biol., 239, 339.
24. Campbell, I. and Duffus, J. H. (ed.) (1989). Yeast: a practical approach. IRL Press,
Oxford.
25. Jenkins, T. M., Hickman, A. B., Dyda, F., Ghirlando, R., Davies, D. R., and
Craigie, R. (1995). Proc. Natl. Acad. Sci. USA, 92, 6057.
71
 P. F. Berne et al.
26. Woodcock, S., Moraon, J.-P., and Henrissat, B. (1992). Protein Eng., 5, 629.
27. Wei, L., Hubbard, S. R., Hendrickson, W. A., and Ellis, L. (1995). J. Biol. Chem.,
270, 8122.
28. Dickinson, C. D., Gay, D. A., Parello, J., Ruoslahti, E., and Ely, K. R. (1994).
J. Mol. Biol., 238, 123.
29. Martinez-Hackert, E., Harlocker, S., Inouye, M., Berman, H. M., and Stock, A. M.
(1996). Protein Sci., 5, 1429.
30. Windsor, W. T., Walter, L. J., Syto, R., Fossetta, J., Cook, W. J., Nagabhushan,
T. L., et al. (1996). Proteins: Structure, Function, Genetics, 26, 108.
31. Zhang, G., Liu, Y., Qin, J., Vo, B., Tang, W. J., Ruoho, A. E., et al. (1997). Protein
Sci., 6, 903.
32. Bergfors, T., Rouvinen, J., Lehtovaara, P., Caldentey, X., Tomme, P., Claeyssens,
M., et al. (1989). J. Mol. Biol, 209, 167.
33. Cohen, S. L., Ferre-D'Amare, A. R., Burley, S. K., and Chait, B. (1995). Protein
Sci., 4, 1088.
34. Pfuetzner, R. A., Bochkarev, A., Frappier, L., and Edwards, A. M. (1997). J. Biol.
Chem., 272, 430.
35. Ho, S. N., Hunt, H. D., Horton, R. M., Pullen, J. K., and Pease, L. R. (1989). Gene,
77, 51.
36. Cadwell, C. and Joyce, G. F. (1992). PCR Methods Applications, 2, 28.
37. Smith, J. L. and Thompson, A. (1998). Structure, 6, 815.
38. Hendrickson, W. A. (1988). In Crystallographic computing, Vol. 4 (ed. N. W.
Isaacs and M. R. Taylor), pp. 97-108. Oxford University Press, Oxford, UK.
39. Hendrickson, W. A. (1985). Trans. Am. Crystallogr. Assoc., 21, 11.
40. Hendrickson, W. A., Pahler, A., Smith, J. L., Satow, Y., Merrit, E. A., and
Phizackerley, R. P. (1989). Proc. Natl. Acad. Sci. USA, 86, 2190.
41. Yang, W., Hendrickson, W. A., Crouch, R. J., and Satow, Y. (1990). Science, 249,
1398.
42. Hendrickson, W. A., Horton, J. R., and LeMaster, D. M. (1990). EMBO J., 9, 1665.
43. Cowie, D. B. and Cohen G. N. (1957). Biochim. Biophys. Acta, 26, 252.
44. Doublie, S. (1997). In Methods in enzymology (eds C. W. Carter and R. Sweet),
Academic Press, London. Vol. 276, p. 523.
45. Yang, W., Hendrickson, W. A., Kalman, E. T., and Crouch, R. J. (1990). J. Biol.
Chem., 265, 13553.
46. Van Duyne, G. D., Standaert, R. F., Karplus, P. A., Schreiber, S. L., and Clardy, J.
(1993). J. Mol. Biol., 229, 105.
47. Benson, T. E., Filman, D. J., Walsh, C. T., and Hogle, J. M. (1995). Nature Struct.
Biol., 2, 644.
48. Doublie, S., Tabor, S., Long, A. M., Richardson, C. C., and Ellenberger, T. (1998).
Nature, 391, 251.
49. Lau, A. Y., Scharer, O. D., Samson, L., Verdine, G. L., and Ellenberger, T. (1998).
Cell, 95, 249.
50. Chen, W. Y. and Bahl, O. P. (1991). J. Biol. Chem., 266, 8192.
51. Chen, W. Y. and Bahl, O. P. (1991). J. Biol. Chem., 266, 9355.
52. Wu, H., Lustbader, J. W., Liu, R. E., Canfield, W., and Hendrickson, W. A. (1994).
Structure, 2, 545.
53. Lustbader, J. W., Wu, H., Birken, S., Pollak, S., Gawinowicz Kolks, M. A., Pound,
A. M., et al. (1995). Endocrinology, 136, 640.
72
 3: Molecular biology for structural biology
54. May, A. P., Robinson, R. C., Aplin, R. T., Bradfield, P., Crocker, P. R., and Jones,
E. Y. (1997). Protein Sci., 6, 717.
55. Scopes, R. K. (1982). Protein purification. Springer-Verlag, New York, NY.
56. Turner, M. A., Yuan, C. S., Borchardt, R. T., Hersfield, M. S., Smith, G. D., and
Howell, P. L. (1998). Nature Struct. Biol., 5, 369.
57. Boles, J. O., Lewinski, K., Kunke, M., Odom, J. D., Dunlap, B. R., Lebioda, L., et
al. (1994). Nature Struct Biol., 1, 283.
58. Budisa, N., Karnbrock, W., Steinbacher, S., Humm, A., Prade, L., Neuefeind, T.,
et al. (1997). J. Mol Biol., 270, 616.
59. Stadman, T. C. (1980). Annu. Rev. Biochem., 49, 93.
60. LeMaster, D. M. and Richards, F. M. (1985). Biochemistry, 24, 7263.
61. Sousa, R., Lafer, E. M., and Wang, B. C. (1991). J. Cryst. Growth, 110, 237.
62. Jeruzalmi, D. and Steitz, T. A. (1997). J. Mol. Biol., 274, 748.
63. Fromant, M., Blanquet, S., and Plateau, P. (1997). In Genetic engineering with
PCR (ed. R. M. Horton and R. C. Tait), p. 39. Horizon Scientific Press,
Wymondham, UK.

73
This page intentionally left blank
 4

Experimental design, quantitative

analysis, and the cartography of
crystal growth
C. W. CARTER, JR

1. Introduction
This chapter is about practical uses of mathematical models to simplify the
task of finding the best conditions under which to crystallize a macromolecule.
The models describe a system's response to changes in the independent
variables under experimental control. Such a mathematical description is a
surface, whose two-dimensional projections can be plotted, so it is usually
called a 'response surface'.
Various methods have been described for navigating an unknown surface.
They share important characteristics: experiments performed at different
levels of the independent variables are scored quantitatively, and fitted im-
plicitly or explicitly, to some model for system behaviour. Initially, one
examines behaviour on a coarse grid, seeking approximate indications for
multiple crystal forms and identifying important experimental variables.
Later, individual locations on the surface are mapped in greater detail to
optimize conditions. Finding 'winning combinations' for crystal growth can be
approached successively with increasingly well-defined protocols and with
greater confidence. Whether it is used explicitly or more intuitively, the idea
of a response surface underlies the experimental investigation of all multi-
variate processes, like crystal growth, where one hopes to find a 'best' set of
conditions. The optimization process is illustrated schematically in Figure 1.
In general, there are three stages to this quantitative approach:
(a) Design. One must first induce variation in some desired experimental result
by changing the experimental conditions. Experiments are performed
according to a plan or design. Decisions must be made concerning the
experimental variables and how to sample them.
(b) Experiments and scores. Each experiment provides an estimate for how
the system behaves at the corresponding point in the experimental space.
 C. W. Carter Jr

Figure 1. Three stages of a response-surface experiment aimed at locating an optimal

point in the solubility diagram for macromolecular crystal growth, (A) Design of the
experiment involves decisions regarding which variables to test, the resources (number
of individual tests, amount of protein, etc.) to be devoted, and explicit descriptions of the
experiments in the experimental matrix. (B) Performing the experiments involves
quantitative measurement of one or more 'responses'. (C) Fitting and testing the model
involves taking a generic model (the gauze) and fitting it to the observed data by
adjusting model parameters. Sections of the chapter devoted to each stage are indicated
in the grey labels.

When these estimates are examined together as a group, patterns often

appear. For example, a crystal polymorphism may occur only in restricted
regions of the variable space explored by the experiment.
(c) Fitting and testing models. Imposing a mathematical model onto such
patterns provides a way to predict how the system will behave at points
where there were no experiments. The better the predictions, the better
the model. Adequate models provide accurate interpolation within the
range of experimental variables originally sampled; occasionally a very
good model will correctly predict behaviour outside it (1). Quadratic
polynomial models are particularly useful for optimization, because they
can possess 'stationary points', where their gradient vanishes, and which
may represent optima (2).
It is increasingly important for structural molecular biology to establish the
76
 4: Experimental design and quantitative analysis
repertoire of different molecular conformations accessible to a given macro-
molecule, in order to understand their functional roles (3-6). Perhaps the
most valuable aspects of the quantitative procedures described in this chapter
have to do with building evidence about relationships between physico-
chemical parameters and crystal growth. Crystal growth depends on many
physico-chemical factors that also influence protein conformation (7-13).
Understanding relationships between conformational equilibria and crystal
growth conditions may benefit from establishing the reliability of anecdotal
evidence about crystal growth (14). Statistical analysis of models entails an
inherent management of 'signal' and 'noise' from the experiments, providing
an appropriate framework of confidence limits within which inferences about
such relationships can be drawn (15-18).

2. Response surfaces and factorial design

2.1 Mathematical models and inference
By quantitative analysis, is meant estimating and interpreting parameters for
an appropriate mathematical model for crystal growth that minimizes the sum
of squared differences between observed results, Q0bs' and predictions of the
model, Qcalc. A model provides a way to coordinate the sometimes complex
interplay between the experimental variables, making sense of how they affect
crystal growth. Sometimes, a good model can suggest novel ways to interpret
experimental behaviour. Models can be useful in three distinct phases of a
crystal growth project: screening, characterization, and optimization.
2.2 What is a response surface?
Response-surface models use a small number of parameters to describe
system behaviour; therein lies their economy. Good model parameters predict
approximate values for the system response at any set of values for the
independent, experimental variables. Linear one-dimensional response-surface
models, y = B1x + B0, are familiar in many contexts. Quadratic polynomial
models describe more complex behaviour, and hence can indicate the location
of optimal conditions.
A response-surface experiment works in the opposite direction, using the
system behaviour at a defined sample of points representing the experimental
space to estimate the parameters, {Bi}, which are the coefficients of the model.
This reciprocal nature of a response-surface model, endows it with dual
abilities to estimate parameters from experimental observations and to
predict experimental results based on the parameters (Figure 2).
2.3 Factorial experimental design
Because they coordinate the influence of all important experimental
variables, response-surface models are closely linked to factorial experimental
77
 C. W. Carter Jr

Figure 2. Mathematical models provide a two-way link between crystal properties and
experimental effects. (a) The mathematical relationships linking response-surface para-
meters and experimental scores via the design matrix, F[k. (scorek) is a row vector of the
N experimental scores, (B i ) T is a column vector of the model coefficients. (b) Schematic
presentation of the reciprocity between parameters and scores described in the text.

design. Factorial design makes the experimental dependencies explicit in the

form of an experimental matrix, denoted here by Fjk, whose elements, {F ik ),
indicate the level of experimental factor i used in the k' h experiment.
The 'causes' of particular processes, like crystallization of a macromolecule,
frequently comprise only a small number of particular factors. or combin-
ations of factors. These specific factors, the 'explanation' as well as the 'recipe'
for the effect, usually act somewhat independently of one another. Rarely
do all contributing factors interact so intimately as to require simultaneous
fine-tuning of each one. In the jargon of experimental design, the contributing
factors themselves are called main effects, while combinations of main effects
are called n-factor interactions. Thus a factorial design is a set of N experiments
intended to identify and map important main effects and interactions
simultaneously for M experimental variables or factors. The N X M matrix, Fik,
assigns different levels of each experimental variable to each experimental
test, simultaneously changing values systematically for each factor. When all
78
 4: Experimental design and quantitative analysis
possible combinations are tested, the design is called a full factorial and
contains

2.3.1 Simultaneous variation of several variables

Factorial designs change several variables simultaneously. An example is
given in f i g u r e 3, where a 6 X 3 experimental matrix and observed scores
are shown in reverse contrast. The levels are conveniently coded as numbers.
-l < [Fik] < 1. which are referred to as the assigned 'treatments'. [Fik] = 0
represents the mean value of the range studied, and the point [0,0,0....0]
represents the 'centre' of the experiment. Other choices (1 < [F ik ] < 2) can be
used. Unlike traditional 'one-at-a-time' designs, simultaneous variation of
several variables allows each experiment to contribute consistently to the
estimation of all effects and interactions, doing several jobs at the same lime.
Figure 3 shows how averaging three results at each pH can be achieved by
using two experiments at one temperature and one at the other temperature.
2.3.2 Contrasts
This example shows the rudimentary relationships between experimental
matrix, observed and calculated scores, and model parameters. This particular

Figure 3. Example of an incomplete factorial design matrix, showing how scores and
contrasts (model parameters) are connected to each other via the experimental matrix
(shaded). Here, six experiments are used to define a model with four parameters, which
include the constant term and the three main effect contrasts,

79
 C. W. Carter Jr
model states that the expected score for the kth experiment, Scorecalc, k, is given
by the average observed score, (30 = 19.33, plus or minus a constant value, pi,
from each of the three columns:

where (30 is the average score. Thus, for the first experiment, Scorecalc,1 = 19.33
+ [(-1 X 4.0) + (1 X -1.75) + (-1 X 7.75)] = 5.8. Similarly, the three
coefficients of the model, {ft}, are obtained from the average sum of the
products of each score and the matrix element from the appropriate column:

where the sums over experiments at the two levels are kept separately, in case
there are not the same numbers of experiments at each level. Equation 2 is
also called the contrast between the high and low levels for the ith factor.
Other contrasts can be calculated, and may be important. Most important
are those for the two-way interactions between pairs of factors. Consider the
data for the full-factorial design in Table 1, and in which the incomplete
factorial sample from Figure 3 is indicated by bold face. The effects of three
variables are each tested at two levels. Contrasts for all possible n-factor
interactions are shown, in addition to those for the main effects. There are
three two-factor interactions and one three-factor interaction. Entries for
each interaction are the products of the entries for the respective main effects.
Contrasts for higher-order interactions are generally smaller than those for
main effects.
In summary, the following aspects of the entries in Table 1 should be noted.
(a) All possible combinations for the three variables are tested at two levels
by the 8 (~ 23) experiments in the full design. This means that the

Table 1. Extended matrix and contrasts for a three-factor, two-level full-factorial design

Exp't PH Temp [Prt] pHx pHx Temp X pH x Temp Score

Temp [Prt] [Prt] x[Prt]
1 -1.000 -1.000 -1.000 .000 1.000 1.000 - .000 8.7
2 -1.000 -1.000 1.000 .000 -1.000 -1.000 .000 23.0
3 -1.000 1.000 -1.000 - .000 1.000 -1.000 .000 7.0
4 -1.000 .000 1.000 - .000 -1.000 1.000 - .000 22.0
5 1.000 - .000 -1.000 - .000 -1.000 1.000 .000 18.0
6 1.000 - .000 1.000 - .000 1.000 -1.000 - .000 34.0
7 1.000 .000 -1.000 .000 -1.000 -1.000 - .000 12.0
8 1.000 .000 1.000 1.000 1.000 1.000 1.000 30.0
Contrasts 4.16 - .59 7.91 -0.91 0.59 0.34 0.16 19.338

80
 4: Experimental design and quantitative analysis
experiments are necessarily uniformly distributed among the different
possible combinations.
(b) Each column is a different linear combination of the eight scores. This is
true both for the full-factorial design and the 6-experiment sample.
Designs for which any two Fij columns are the same are said to involve
confounding or aliasing of the effects denoted by identical columns. It is
impossible to distinguish which of the confounded columns is responsible
for the contrast in the experimental scores without additional experiments
specifically designed to distinguish between the multiple possibilities (19).
(c) The seven columns all have equal numbers of 1s and -1s. This is true for
the complete factorial and for all but the temp x [Prt] column of the 6-
experiment sample. Both the full design and the sampled design are
therefore balanced with respect to the main effects.

2.3.3 Balance
If each level is tested by the same number of experiments, the design is said to
be balanced. Balance is important. The standard deviation of an average
value, <x>, is given by

where Xi refers to an observation. If the design in Figure 3 had only two

experiments at the low pH and four at the high pH, then the standard
deviation of the average score for the latter four experiments would tend to
smaller by a factor of V(2-l)/(4-l) = V\J3 = 0.57. The estimate would tend
to be more precise, and probably more accurate than that for the other two
experiments. For this reason, balanced designs are preferred for quantitative
multivariate problems, because testing each level the same number of times
distributes both signal and noise as evenly as possible among the different
experiments.
2.3.4 Resolution
The dot products between all of the columns in the full (but not the in-
complete) factorial design all equal zero. Thus, they form a set of orthogonal
basis vectors for the experimental space. This property means that in a full-
factorial design, the experimental treatments for all effects and interactions
are completely uncorrelated, which enhances the ability to separate their
impact on the scores. This 'separability' property is related to what is called
the resolution of a sampling design.
The optimal balance, resolution, and freedom of confounding of full-
factorial experiments make them especially useful whenever a phenomenon is
already suspected to depend on the effects and interactions of a small number
81
 C. W. Carter Jr
of factors. An example was the demonstration of how pH, temperature, and
two low molecular substrates influenced the crystal growth of tryptophanyl-
tRNA synthetase (TrpRS) (2). TrpRS shows both pH- and ligand-dependent
conformational changes, so the main effects (pH) and two-factor interactions
(pH X substrate and substrate X substrate) identified from a quantitative
analysis of the crystal growth behaviour in a 24 factorial experiment were
biochemically significant. Strategies for sampling full-factorial designs dis-
cussed in Section 3 differ from one another in the degree to which each
compromises one or more of these properties.

3. Sampling appropriate subsets from a full-factorial

design
Sampling is always an issue. No experiment provides more than a sample of
how a system responds to changes in the conditions that vary from experiment
to experiment. A full-factorial design in M factors yields just enough observa-
tions to uniquely define all of the main effects and all n-factor interactions.
However, for M > 3-5, it becomes increasingly difficult to carry out all such
experiments even at only two levels per factor, and additional sampling is
required. Sampling sacrifices some of the advantages of the full-factorial
design in return for the economy of doing fewer experiments. The particular
selection of experiments actually performed can determine whether or not a
design achieves a desired goal, and the appropriate sampling strategy depends
on the experimental context.

3.1 Screening versus optimization

As the broad distinction between 'screening' and 'optimization' suggests, one
often proceeds by first identifying 'nodes' of the response surface where
crystallization occurs, and then mapping finer details of individual nodes.
Neither process is typically done with full-factorial designs. Different sampling
strategies can be used. The appropriate quantitative analytical methods are
generally similar, but use different types of models which, in turn, dictate how
the experimental sampling points should be selected. The sampling strategies
are illustrated schematically in Figure 4. Each sacrifices some of the strengths
of the full-factorial design, making different compromises with respect to
balance, confounding, and resolution.

3.2 Subsets for screening

3.2.1 Incomplete factorial sampling
Efficient covering of the entire experimental space is a key requirement for
screening. Incomplete factorial designs were introduced to detect the most
important factors and their interactions from screening experiments (20).
82
 4: Experimental design and quantitative analysis

Figure 4. Sampling strategies, (a) Screening. Incomplete factorial designs sample

parameter space randomly and uniformity is enforced on the sampling of main effects
and two-dimensional interactions. 'Sparse matrix' sampling involves an intentional bias
toward combinations that have worked previously. Orthogonal arrays impose strict
orthogonality on the columns of the experimental matrix, and are thus less flexible, (b)
Optimization. Hardin-Sloane design for minimum variance parameter estimation
distribute experiments non-uniformly, with one or two at the centre and the others at the
boundary of the search region.

Their use has been described previously (15-18). Factor levels are chosen
randomly and then balanced to achieve nearly uniform sampling (Figure 4).
Levels for each main effect are sampled the same number of times, and all
two-factor interactions are sampled as uniformly as possible. This strategy
preserves the ability to detect large main effects and two-factor interactions
with minimal confounding (18). This process leads to a very flexible sampled
factorial design which has given superior performance in a wide variety of
contexts (15,21-24).
The number of experiments in a design should be chosen relative to the
size, N, of the full-factorial design, which, in turn, depends on the number of
variables, and their levels according to Equation 1. Incomplete factorial de-
signs with a sampling density as coarse as roughly VM2 can be found that are
quite evenly balanced with respect to main effects and two-factor interactions,
and entirely free of explicit confounding. This rule of thumb is more useful for
83
 C. W. Carter Jr
larger factorial designs; designs with smaller numbers of factors are less
efficient, as they require more experiments for adequate signal-to-noise.

3.2.2 'Sparse matrix' recipes

Factorial design strongly influenced the development of 'sparse matrix kits'
now used for screening in an overwhelming majority of crystal growth efforts.
These unbalanced, pre-packaged samples from a broad, factorial space, emu-
late factorial design without actually capturing their essential features (Figure
4) (25-29). The conceptual basis for such designs is to bias the choices of
combinations in favour of conditions that have previously produced crystals
for other proteins, RNA, or membrane proteins, as found, for example, in the
Biological Crystallization Database (30) to select conditions that have pre-
viously produced crystals. A recent extension of this approach uses the biases
in that database to weight sampling strategies for particular classes of proteins
(31).
The success of kits, arising from their ease of use and commercial access-
ibility, speaks for itself, but comes at significant cost. Biasing is necessarily
done at the expense of balance. Thus, scoring is nearly useless for quantitative
model building and analysis. Biased searches may also miss useful crystal
polymorphs. Moreover, reinforcing existing biases during crystallization also
biases the database of solved macromolecular crystal structures in quite subtle
ways. For example, the high salt structure of haemoglobin A, known as the R
structure (32), differs from those obtained under near physiological conditions,
and is almost certainly an artefact of crystallization at high salt (33-36).
The attractions of sparse matrix kits are virtues only because the physical
chemistry of pure, homogeneous protein solutions strongly favours crystal-
lization over a considerable range of solubility. When kits work, one still must
optimize conditions suggested by preliminary screening. When they fail one
must re-screen, from new combinations of conditions. Since little can be
learned from quantitative analysis of results obtained using kits, the focus
here is on experimental designs that can be analysed.
3.2.3 Orthogonal arrays and fractional factorial designs
Other sampling strategies place a higher priority on separating effects of a
particular subset of factors as completely as possible; what was termed the
'resolution' in Section 2.3.4. Resolution is achieved by accepting a higher
degree of confounding between main effects and higher-order interactions,
which are usually less significant. For full-factorial designs, higher-order
interactions can provide estimates for the experimental variance. 'Fractional
factorial' designs (37) intentionally alias higher-order interactions with treat-
ments of additional main effects, thereby increasing the number of factors
tested. 'Orthogonal arrays' preserve much of the geometrical symmetry of
full-factorial designs and have been developed and exploited for screening
(38). Here, the treatment effects are completely uncorrelated and can hence
84
 4: Experimental design and quantitative analysis
be more completely resolved. Orthogonal subsets of the two level, three-
factor design in Table 1 each have four experiments. As noted in Section 6,
they provide poor quantitation that is improved substantially by increasing
the number of experiments from four to six by the use of an incomplete
factorial design.

3.3 Minimum variance sampling: Hardin-Sloane designs

Response-surface experiments are sampled by different criteria from those
used in screening (Figure 4b). Optimization requires a non-uniform sampling
strategy because the objective is to construct an accurate analytical approxi-
mation to how the system actually responds to the input variables. If one
already knows something about where the best result might be obtained, it is
no longer sensible to scatter the experimental test points uniformly through-
out the space. Rather, experimental points are selected for maximal impact on
the accuracy of the parameters of the response-surface function, and hence on
the coordinates of its stationary points. Designs that minimize errors in
parameter estimation are called 'minimum-prediction variance' designs (39).
The strategy of Hardin-Sloane designs can be visualized in a one-dimensional
example where the goal is to distinguish as accurately as possible between a
linear relation, y = ax + b, and a parabola, y = ax2 + bx + c. Three groups of
experimental points have maximal impact on this distinction: those near the
suspected maximum value of the parabola and those at upper and lower limits
of x for which the parabolic approximation may be appropriate. Distributing
several experiments near the suspected maximum value and the remaining
ones at low and high values of x gives the most accurate (averaged over
multiple experimental measurements) values for these crucial points of the
response surface. For multivariate response-surface models the same strategy
applies; a small number of experiments should be done near the centre of the
design and the remainder evenly distributed around the perimeter of the
experimental space (Figure 4b). Hardin-Sloane design matrices are given in
Tables 2 and 3.

3.4 Computer programs to generate designs for special

purposes
Because they require efficient sampling, screening and response-surface
designs are often selected from a large number of potential designs generated
by a computer program according to specific criteria. Such programs include
INFAC (http://russell.med.unc.edu/~carter/designs), a pro-
gram to generate incomplete factorial designs, and GOSSET (http: //www.
research.att.com/~njas/gosset), a more general and much more
powerful program to generate minimum-variance designs for response-
surface determination. Other programs for experimental design also have
85
 C. W. Carter Jr

Table 2. Hardin-Sloane minimum integrated variance design matrix for four factors, 20
experiments*

Exp't Variable 1 Variable 2 Variable 3 Variable 4

1 0.000 -0.056 0.000 -0.250
2 0.000 -0.056 0.000 -0.250
3 0.000 1.000 0.000 -0.250
4 0.000 -1.000 0.000 -1.000
5 1.000 -0.007 0.116 1.000
6 -1.000 -0.007 -0.116 1.000
7 0.210 0.108 -1.000 -1.000
8 -0.210 0.108 1.000 -1.000
9 -1.000 -1.000 1.000 -0.250
10 -1.000 1.000 -1.000 -0.250
11 1.000 -1.000 -1.000 -0.250
12 1.000 1.000 1.000 -0.250
13 0.492 -1.000 1.000 1.000
14 -0.492 -1.000 -1.000 1.000
15 -1.000 1.000 0.577 -1.000
16 1.000 1.000 -0.577 -1.000
17 0.669 1.000 -1.000 1.000
18 -0.669 1.000 1.000 1.000
19 -1.000 -1.000 -1.000 -1.000
20 1.000 -1.000 1.000 -1.000

aThis design was prepared specifically for use in the experiments reported here by N. J. A. Sloane,
using GOSSET (39). Matrix entries should be interpreted as: 0 = the centre, -1 = the low end, and 1 =
the high end of the variable range. The same design has been used repeatedly in different contexts, by
assigning the matrix entries to different parameters and/or ranges.

been described (40). INFAC designs have optimal coverage and minimal
aliasing of main effects with each other and with two-way interactions. The
program is entirely interactive, prompting for all necessary information.
The GOSSET interface is less intuitive and requires explicit description of
variable ranges and the type of response-surface function to be fitted. Much
can be obtained from the examples in the users manual, but the overview in
Protocol 1 of how to generate a design like that in Table 2 should be helpful.

Protocol 1. Generating a Hardin-Sloane design using GOSSET

Method

1. Once launched, GOSSET requests the user to select a working

directory. Description of the design is done without prompts, and must
be preceded by line numbers, e.g. 10 range x y z -1 1.

2. Enter the range to be considered for each of the variables, including

86
 4: Experimental design and quantitative analysis
those that will take on discrete values, e.g. 20 discrete T 4 14 21 can be
used to specify three discrete values for temperature.
3. Enter the model for which the design will be used, e.g. 30 model
(1+x+y+z+T) A 2 will specify a quadratic polynomial model in four
variables, x, y, z, and T. Such a model has (4+1)*(4+2)/2 = 15
parameters, so the design must have at least that number of
experiments.
4. Compile the above 'program' using: compile (from this point, no line
numbers are entered!).
5. Compute the matrix of 'experimental moments' using: moments
n=1000000. These moments are used internally to represent the
impacts of experimental points on the predicted variance (39).
6. Ask for the design. The command: 'design' will generate a design with
the minimum required number of experiments for the specified
model. Various modifications include:
(a) design runs=24 n=20. This forces the program to generate 24
experiments and to find the best design from 20 different starting
points.
(b) design type=I extra=5. This forces the program to generate an I-
optimal design (the default choice, which optimizes the Integrated
prediction variance) with 5 more than the minimum number of
experiments. The resulting design will have 20 experiments in this
case.
7. Generate a formatted file with the design: interp >20expt_xyzT.design.
This converts the output file into a formatted table in the file
20expt_xyzT.design.

4. Screening with factorial designs

Surprisingly little thought has been devoted to characterizing the dimensions
of 'crystallization space' for any protein. Nevertheless, recent studies of how
model proteins crystallize (41-50) have led to new insights (see Chapter 10)
that change how one should think about screening and optimization of crystal
growth. These insights provide a rudimentary, but essential guide to
experimental design.

4.1 Selecting experimental variables

Crystallization requires an appropriate balance of variables that influence:

materials: composition and homogeneity of the unit cell

87
 C. W. Carter Jr

Table 3. Hardin-Sloane minimum integrated variance design matrix for five factors, 30
experiments a

Exp't [Macromol] Supersaturation [Additive] PH Ligand

1 -1.00 -1.00 0.03 -1.00 -1.00
2 0.04 0.04 0.10 0.07 0.03
3 -0.82 -1.00 1.00 -1.00 0.86
4 -0.04 1.00 -1.00 -0.11 -1.00
5 1.00 -1.00 -1.00 1.00 1.00
6 -1.00 -0.45 -1.00 -0.18 -0.20
7 1.00 1.00 -0.12 -1.00 1.00
8 0.18 -0.17 0.03 -0.12 1.00
9 -1.00 1.00 -0.08 1.00 -1.00
10 1.00 0.00 -0.63 1.00 -1.00
11 -0.04 0.01 -0.07 1.00 0.05
12 -0.68 0.45 -1.00 -1.00 1.00
13 -1.00 1.00 1.00 -0.23 1.00
14 0.20 1.00 0.73 1.00 1.00
15 -0.61 -1.00 -1.00 1.00 -1.00
16 1.00 1.00 1.00 0.91 -0.94
17 1.00 -0.12 -1.00 -1.00 -1.00
18 1.00 -1.00 -0.02 -0.30 -0.55
19 1.00 1.00 -1.00 0.37 0.31
20 0.36 1.00 1.00 -1.00 -1.00
21 0.21 -1.00 1.00 0.29 -1.00
22 -1.00 -0.87 1.00 1.00 0.19
23 -1.00 0.07 1.00 -0.11 -1.00
24 0.04 0.04 0.10 0.07 0.03
25 -1.00 -1.00 -0.22 0.44 1.00
26 1.00 -0.20 1.00 -1.00 0.42
27 0.37 -1.00 -1.00 -1.00 0.39
28 1.00 -1.00 1.00 1.00 1.00
29 -1.00 1.00 -1.00 1.00 1.00
30 -1.00 1.00 0.04 -1.00 -0.16

aThis five-variable design was prepared using the GOSSET program (39). It was designed to
compensate for the failure of one of the experiments to produce a score, and is called J-optimal. It
contains 30 experiments, which are nine more than the 21 experiments required to estimate
parameters for a five-variable quadratic model. Although variable names have been suggested, this
matrix can be used in any desired context.

relative interaction potentials and free energies: interparticle potentials,

solubility, and supersaturation
relative rates of various processes: nucleation, diffusive transport, and
interfacial deposition.
Screening therefore must sample four different types of factors that inter-
vene between an experimental design and crystal growth (Figure 5), including

88
 4: Experimental design and quantitative analysis

Figure 5. Crystal growth variable space. Variables are organized hierarchically, from
those directly under explicit experimental control on the left, to those that dictate crystal
growth behaviour on the right. To challenge one experimental design in screening and/or
optimization of crystal growth is to find an efficient parameterization of variables in this
space that will permit post-hoc quantitative analysis.

those under direct experimental control (column I) and composite properties

that ultimately determine the physical system evolution (column IV). These
variables can influence both macromolecule and solvent (column II), changing
the conformation, net charge, and quaternary structure of what is to be
crystallized while reducing solubility and promoting crystal growth. Proteins
have significant net charge, except at their isoelectric points (45), and are
accompanied by stoichiometric amounts of counterions, whose concentration
increases with the protein concentration (51,52) and these ions may incorpor-
ate in, and influence crystal properties. Solvent properties, notably the activity
of water and the dielectric constant, are also sensitive to variables in column I.
The balance between thermodynamic and kinetic effects, including rates of
equilibration (53,54), will determine the extent of growth and hence the
crystal size. They also influence the rate at which impurities are incorporated
(50, 55-57). The challenges in designing experiments to study crystallogenesis
arise from the need to sample the effects suggested in Figure 5 as effectively as
possible.

4.2 Preparing the experimental matrix

The experimental matrix specifies how each individual experiment or 'test' is
to be carried out. Protocol 2 summarizes a procedure for constructing an
incomplete factorial experimental matrix.
89
 C. W. Carter Jr

Protocol 2. Preparation of the experimental matrix for an

incomplete factorial design

Method
1. Gather information about the protein to be crystallized from all in-
volved with expression and purification; from previous screening
experiments; and from databases (Biological Macromolecule Crystal-
lization Database): (http://ibm4.carb.nist.gov:4400/bmcd/
bmcd.html) (30, 58).
2. List factors that might influence crystal growth.
(a) Estimate the variation of net charge with pH from the amino acid
composition (Chapter 10). Verify the pi by isoelectric focusing. Use
a pH range with values above and below the pi.
(b) Factors required for stability and monodispersity (59, 60).
(c) Ligands and other factors likely to influence the conformation of
the macromolecule.
4. Choose from this list the factors to be screened in the current
experiment and the levels to test.
5. Choose the number of tests. In general, this number should be
somewhat more than the number of factors to be screened. There is
no hard and fast rule; experience from diverse sources (20, 61)
suggests using ~ VN/2 tests if the full-factorial design requires N
tests.
6. Compile the experimental matrix itself (the computer program INFAC
http://russell.med.unc.edu/~carter/designs will do steps
a-e interactively).
(a) Choose factor levels at random, working down each column, and
from column to column.
(b) Balance each column, readjusting levels to equilibrate the
numbers of tests at each level.
(c) Balance each two-factor interaction by compensating readjust-
ments to two columns. In this case, it is adequate to ensure that
each combination is represented by at least one test.
(d) Verify the balance of all columns and two-factor interactions.
(e) Examine the experimental treatments for possible confounding.
Confounding is indicated whenever two different effects have

90
 4: Experimental design and quantitative analysis
identical patterns of level assignments. If a confounded effect
turns out to be large, this knowledge is useful in further
experimentation to distinguish between the two possibilities.
(Designs generated by INFAC are selected to minimize
confounding.)

4.3 'Floating' variables, initial values, and sampling

intervals
Crystal nucleation can be achieved by changing either the solubility or the
macromolecular concentration, or both, via changes in [crystallizing agent],
temperature (62, 63), or pH. One of these must be chosen as a 'floating
variable'; the appropriate choice is system-dependent (45). Temperature and
pH (net protein charge) have reduced influence on solubility at high ionic
strength, and ionic strength has reduced influence on solubility close to the
isoelectic point. Thus, for example, proteins that crystallize near their
isoelectric points will almost certainly be insensitive to ionic strength, leaving
pH or temperature as possible floating variables. Alternatives include finding
an ion that interacts with the protein, changing its pI.
4.3.1 Scanning for supersaturation
Since solubility behaviour is difficult to determine a priori, bootstrap methods
must be used in order to exploit prior knowledge of solubility behaviour in
choosing concentrations of protein and crystallizing agent. An upper limit to
the solubility in the absence of crystallizing agents can be estimated by
concentrating the protein in its usual buffer to the greatest extent possible for
storage. Screening should sample fractions of this value, say 1.0, 0.5, and 0.3,
and crystallizing agent concentrations should reduce solubility to ~ 10-30%
of its initial value. A range of crystallizing agent concentrations can be
scanned using the same sample by dialysis. Vapour diffusion uses less protein,
permitting tests at different [crystallizing agent]. This is the basis of the
'footprint' method (64, 65), in which four concentrations are tested for six
reagents in a Linbro plate. Footprinting can be effectively combined with
streak seeding (66) (Chapter 5) because it provides protein samples
equilibrated with supersaturated, but non-nucleated solutions, which are ideal
for seeding. Moreover, coverslips can be transferred to reservoirs of lower
vapour pressure, permitting simultaneous increases of [macromolecule] and
[crystallizing agent]. The simplex procedure described in Section 7.1 offers
another way to find the approximate macromolecule and crystallizing agent
concentrations for the nucleation zone.
Supersaturation is so critical to controlling crystal nucleation that an
approach called 'reverse screening' (65), presumes that crystals should grow
91
 C. W. Carter Jr
with almost any agent if the appropriate supersaturation can be found, and
examines solubility behaviour for two or three crystallizing agents. The
importance of solubility data, and the ad hoc nature of screening reinforce the
importance of utilizing early screening experiments to construct experimental
solubility databases when working with a new problem. Mother liquors
should be separated routinely from all solid phases, and assayed for the
concentration of soluble macromolecule. Parameters of the Cohn-Green
equation, In Csol = 3 - Ks [crystallizing agent] can be estimated from two or
three different determinations of the protein concentration in the liquid
phases of initial experiments. Ks changes significantly with the net charge of
the protein, the position of the anion in the Hofmeister series, as indicated in
Section 4.3.
4.3.2 Sampling intervals
How finely sampled a design should be depends on its purpose. For screening,
the high level should 'titrate' out any specific binding interactions that might
influence crystal growth. For optimization experiments, which require scores
for nearly all experiments, the following guidelines may be useful:
(a) Variables whose effects are related to chemical potentials (e.g. pH) are
logarithmic; changes of ~ ± 0.5 will effect a tenfold variation in
equilibrium concentrations.
(b) Protein concentration and supersaturation exert higher-order effects on
nucleation rates, proportional to n ln[protein], where n is the (generally
unknown) order of the nucleation step (67, 68). Finer sampling of
In[protein], say by ± 0.5/n, should change nucleation rates by an order of
magnitude and may be more appropriate.

4.4 Design for initial screening of variables for crystallizing

a new protein
The experimental variables in Figure 5 suggest that screening designs can be
made more effective in producing crystals and more useful sources of
evidence about the factors governing crystal growth. The design in Table 4 is

Table 4 . Incomplete factorial screening design in ten variables

Exp't A B C D E F G H I J
1 2 2 4 1 1 3 2 3 1 1
2 4 5 5 1 2 3 1 1 1 2
3 5 1 4 2 1 1 3 1 2 2
4 1 2 3 1 2 2 3 3 2 1
5 3 5 2 3 2 1 1 2 1 2
6 4 3 1 1 1 3 1 3 2 1
7 5 4 1 3 3 1 1 2 2 2
8 1 4 3 3 1 2 2 2 2 2
9 2 1 2 1 3 2 2 3 1 1

92
 4: Experimental design and quantitative analysis

Table 4. continued
Exp't A B C D E F G H I J
10 3 3 5 3 3 3 3 3 1 2
11 1 1 5 2 3 2 1 1 2 1
12 2 1 1 2 1 1 3 2 1 1
13 3 1 2 2 2 3 2 2 2 1
14 5 1 4 3 2 2 3 1 1 1
15 4 5 3 2 3 1 2 1 1 2
16 5 2 2 3 1 1 3 1 2 2
17 4 4 2 2 2 2 3 2 1 2
18 2 5 1 3 3 1 2 3 2 1
19 3 5 4 2 1 2 2 2 1 1
20 1 4 3 2 3 2 1 2 2 2
21 5 5 5 1 1 3 1 3 2 2
22 2 4 5 3 2 1 1 1 1 2
23 1 3 4 3 1 3 2 1 2 1
24 3 4 1 2 2 3 3 3 1 1
25 4 2 1 1 3 3 3 3 1 2
26 5 3 3 1 2 3 2 3 2 2
27 4 1 3 3 3 3 1 2 2 1
28 3 2 5 2 1 1 3 2 1 1
29 2 3 5 2 3 1 1 1 1 1
30 1 2 3 1 1 2 2 1 2 2
31 5 4 4 1 3 2 1 3 2 1
32 3 3 1 3 2 2 3 1 2 2
33 2 2 4 3 1 3 1 1 1 2
34 1 5 2 1 2 1 2 2 1 2
35 4 3 2 1 1 2 3 1 1 1
36 5 2 5 1 2 1 1 2 2 1
37 3 4 2 2 1 1 3 3 1 1
38 4 5 2 3 3 1 2 2 2 1
39 2 1 5 3 2 3 2 3 2 2
40 1 3 3 3 3 3 3 2 1 2
41 4 2 2 1 3 2 2 1 2 2
42 2 5 5 3 2 2 1 3 1 1
43 5 1 1 2 2 3 3 1 2 1
44 3 2 1 1 2 2 2 3 1 2
45 1 4 5 1 3 3 2 1 1 2
46 1 2 4 2 3 1 3 2 1 1
47 5 4 1 1 3 1 1 2 1
48 4 3 2 2 2 1 2 2 2 2
49 3 3 4 3 1 1 1 2 1 2
50 2 4 3 2 3 2 3 3 2 2
51 3 1 2 3 2 1 3 2 2 1
52 5 5 4 2 1 2 1 3 1 1
53 2 3 1 3 3 1 1 1 1 2
54 1 5 1 2 3 3 2 3 2 1
55 4 1 4 1 3 2 1 2 2 2
56 3 5 3 3 1 2 3 3 2 1
57 4 4 1 1 2 2 2 2 1 1
58 2 2 3 1 1 1 2 1 1 1
59 5 3 5 2 1 3 3 1 2 2
60 1 1 3 2 2 3 1 3 1 2

93
 C. W. Carter Jr
motivated by that goal. It provides a balanced incomplete factorial screen
covering a similar sample of the conditions from Figure 5 to that represented
in the Hampton kit(s) (25), and is informed by the underlying physical
chemistry. It can be used either with a standard set of pHs or centred on a
known or estimated pi value.

4.4.1 Physico-chemical parameterization

As far as possible, this design associates each variable with a single physico-
chemical property from column II in Figure 5. The Hofmeister series (42, 44,
45) likely reflects the kosmotropy or lyotropy of the salt. Its effect on
solubility and hence on crystal growth is likely to be related, via the surface
tension (69) to solubility and to the relative interactions of the various ion
pairs in the protein-solvent interface (70). The series of polyols with
increasing molecular weight samples a range of excluded volumes. Different
glycerol concentrations sample osmotic pressure exerted on water molecules
in intramolecular crevices and hence possible conformational changes (71)
are screened explicitly. The design was implemented in Protocol 3 according
to these and similar considerations.

4.4.2 The 'organic moment'

A distinctive feature of the Jancarik and Kim screen (25) is that organic
crystallizing agents like PEG are combined with significant amounts of salts.
One rationale for doing this is that different proteins may have different pro-
portions of polar and non-polar surface textures, whose solubilities depend in
different ways on volume exclusion mediated by PEG and electrostatic
shielding by ionic crystallizing agents. An extreme example, for which there is
considerable literature (27, 72), are the integral membrane proteins, which
explicitly demand simultaneous manipulation of ionic and polymer excluding
crystallizing agents. The design in Table 4 allows for the interaction of salts
with organic and/or polymeric crystallizing agents at three levels, e.g. 0%,
35%, 70%.
If solubility constants are available, the interaction can be quantified
explicitly by a 'moment',

M approximates the relative impact on solubility for each crystallizing agent,

via the molar concentration of each together with its solubility coefficient, Ks,
which usually must be determined (Section 4.3.1). If these are unavailable, M
can be approximated for screening purposes using molar concentrations alone
for all agents except the high molecular weight PEGs, whose Ks values are
generally an order of magnitude greater than those for salts and low
molecular weight organic reagents.
94
 4: Experimental design and quantitative analysis

Protocol 3. Example: incomplete factorial screening design

(Table 4)

Method
Identify columns in Table 4 with factors such as the following:

A. pH: five levels, 4.5, 5.5, 6.5, 7.5,8.5, cover the range normally observed
for crystallization of proteins. Alternately, if the pi is known five levels
can be centred on the pI.
B. Use five ionic crystallizing agents, with the following anions: sulfate,
phosphate, acetate, chloride, nitrate from the Hofmeister series.
C. Organic crystallizing agent: five levels, isopropanol, methylpentane-
diol, PEG 4000, PEG 8000, PEG 20000.
D. Protein concentration: three levels, Max, Max/2, Max/3. Three levels are
chosen because this variable may not be linear, and may have a maxi-
mum. Some effort is placed on estimating the curvature of its behaviour.
E. Temperature: three levels, for instance 4°C, 14°C, 22°C.
F. 'Moment' of non-polar reagent, M: 0%, 35%, 70%, as defined in
Section 4.4.2.
G. Divalent cation: three levels, none, Mg2+, and either Ca2+, Cd2+, or
Mn2+. Salts with acetate, nitrate, sulfate, and chloride are sufficiently
soluble. All are problematical with phosphate. This is a constant
weakness of all such plans. Phosphate experiment(s) calling for level
three of this variable can use Mn2+.
H. Additive: three levels, none, arginine, (JOG. These suggested choices
are based on the selection of additives that either stabilize proteins
(arginine ~ 50-100 mM) (59, 60) or destabilize weak, non-specific
contacts (00G 0.1-0.3%, w/v).
I. Glycerol: two levels, 0%, 10% (v/v). This variable samples additional
osmotic pressure exerted by the solution over and above that
produced by the crystallizing agent(s) themselves.
J. Ligand: two levels, presence or absence of a substrate or inhibitor.

4.5 Experimental set-ups

Experiments in the design should be carried out by systematically approach-
ing supersaturation, manipulating the floating variable to ensure that each
experiment is taken to completion. This should be done repeatedly for
95
 C. W. Carter Jr
conditions producing precipitates, to verify that precipitation was not due
simply to excessive precipitant concentrations. Microdialysis buttons have
obvious advantages from this point of view, but vapour diffusion can be used
if desired. Vapour diffusion is essential for some experimental conditions, such
as organic crystallizing agents and polymers. Factorial experiments preclude
using the same buffer for any two experiments, so some thought should be
given to rational preparation of stock solutions. We have found no good
alternative to simply making up each buffer separately, in order to keep the
ionic compositions consistent with design requirements.

5. Quantitative scoring
Previous quantitation and statistical analysis of crystal growth experiments (1,
2, 17, 73) are sufficiently compelling to justify a more intensive investigation
of practical requirements and procedures for using quantitative analysis of
crystal growth more routinely. We presume throughout this chapter that
experiments can be quantitatively scored in a variety of different ways. This
section addresses problems associated with scoring crystallization experi-
ments, and suggests protocols for solving those problems.

5.1 Hierarchical evaluation: interrogating nature by

experimental design
Quantitative scoring can be integrated into the study of crystal growth most
effectively in the context of a hierarchical framework of specific questions
(Protocol 4). Figure 6 shows how we 'interrogate' crystal growth experiments
progressively, first establishing which kind of quantitation may be appropriate.
Several qualitatively different 'scores' give purpose to the interrogation,
increasing the likelihood of success by directing attention to the appropriate
context and question, before attempting to assign scores. Further statistical
analysis is appropriate for many of the scores.
Scores should reflect the consequences of changing the experimental vari-
ables. Several pre-conditions help assure that they do, and should be satisfied
first; the questions in the scheme provide a guide to assuring that these are
met before trying to build models for any particular scores. Above all, it is
essential to have a sufficient'number of scores for analysis. Hence, the initial
question concerns how many tests actually produced crystals. The answer to
this question determines whether to pursue the left-hand or right-hand
branches of Figure 6.
5.1.1 Scoring crystal properties
If most tests produced crystals, then two kinds of scores representing crystal
properties can be developed, depending on whether or not the crystals
correspond to the same or to different polymorphs. Screening experiments
provide qualitatively more diverse information about the location of station-
96
 4: Experimental design and quantitative analysis

Figure 6. A hierarchical scheme for evaluating designed crystallization experiments. A

nested set of questions is posed of the experiments in the design. Answers to these
questions determine the appropriate action to take. Questions in lightly shaded boxes
can be answered categorically, either with 'yes' or 'no' or comparative answers.
Questions in mid-grey boxes can be addressed by obvious subsequent experiments.
Directives to 'score' a particular parameter imply that a quantitative score can be used,
leading to multivariate regression of those scores against experimental factors for
purposes suggested in rectangles with large grey borders.

ary points than do optimization experiments. In general, the solid phases of

serecning experiments may be microcrystalline or amorphous precipitates,
and there may be different crystalline polymorphs in the ensemble. The
scoring system should therefore make use of this broad range of information.
Polymorphism frequently occurs in crystallization, and probably reflects, in
part, underlying structural polymorphisms that may he associated with
function. One should distinguish cases where the different polymorphs appear
in different tests from a screen from those where the polymorphism is evident
within the same sample. In the former case, subsequent experiments should
be directed toward optimizing the different nodes. The latter situation often
arises in the early stages of optimization from a screen that produced con-
ditions where response surfaces for more than one polymorph overlap.
Scoring the proportions of each polymorph in mixtures can help resolve
conditions under which the different polymorphs can be grown uniquely (73)
(Section 8). Finally, if all crystals belong to the same polymorph, it makes
sense to determine if their sizes and habits vary enough to justify oplimizing
those properties by response-surface methods (1).

5.1.2 Troubleshooting super-saturation

When few of the experiments in a design produced crystals, as depicted on the
right-hand side of Figure 6, other types of questions can lead to alternative
97
 C. W. Carter Jr
scoring aimed at finding a 'sense of direction' from the current position. This
is the case when seeking initial conditions with a screening experiment. One
searches in this case for an unknown point in a multidimensional solubility
diagram where discrete crystal nucleation occurs, but similar considerations
underlie the interrogation. Unless one has the happy result of generating
many crystals from an initial screen, it is necessary to make sense of variations
in the type of phase separation. Prior binary decisions are nevertheless useful
here, too. If few experiments have a solid phase, the culprit is likely that the
experiments are not sufficiently supersaturated. In this case, the score can be
simply the presence or absence of a phase separation.

Protocol 4. Scoring crystallization experiments (see Figure 6)

Method
1. Determine and score how many experiments in the design have some
kind of phase separation.
(a) If a majority are still clear, adjust the experiments to increase
supersaturation. Regression of this first score against the experi-
mental variables can reveal which changes to make in experi-
mental conditions to increase the proportion of experiments
producing a solid phase.
(b) If a majority have a solid phase, proceed to step 2.
2. Determine and score, according to Table 5, how many experiments in
the design have crystals.
(a) If most experiments have crystals, how many different morpho-
logies are represented? If there is only a single crystal habit,
proceed to step 4. Otherwise, proceed to step 5.
(b) If most experiments have precipitates, oils, and spherullites,
streak seeding (Chapter 7) can be helpful.
3. Examine a 'scatterplot matrix' showing how each score depends on
each independent variable for obvious trends and/or non-linearities.
4. Measure crystal sizes and shapes.
(a) Two dimensions, and occasionally a third can usually be measured
using a microscopic ruler. Enter dimensions for representative
samples from each experiment.
(b) Calculate sizes (surface area or volume) and shapes (aspect ratio,
or width/length).

98
 4: Experimental design and quantitative analysis
(c) Determine the maximum, minimum, mean value, for each score.
Note significant variations.
5. Count the numbers of crystals with the same habit in each experiment.
Enter separate scores for each distinct morphology. These scores can
help resolve polymorphs (Section 8).

Table 5. Scale of crystal quality

Result Score, Q
Cloudy/amorphous precipitates 1.0
Gelatinous/particulate precipitates 2.0
Oils 3.0
Spherulites 4.0
Needles 5.0
Plates 6.0
Prisms 7.0

5.2 Rating different solid phases

When no proper crystals are obtained, sense often can be made from vari-
ations in the quality of the solid or denser phase. Precipitates differ somewhat
from one another. Fluffy, cloudy, or filamentous precipitates have little
likelihood of being crystalline, but uniform, granular, and/or particulate
precipitates often are microcrystalline. Similarly, among different crystalline
samples, one can readily distinguish spherulites (radially symmetric aggre-
gates of microscopic needles) from needles, plates, and prisms, with obvious
preference for the latter over the former. An intuitive sense of this pro-
gression (15) provides a basis for the quantitative scoring protocol in Table 5.
Exploration of the lysozyme phase diagram into regions where a liquid-liquid
phase separation occurs (49) provide a solid rationale for scoring of patho-
logical behaviour in this way. Under appropriate conditions even lysozyme
can be induced to produce all of the pathological solid phases observed with
many proteins which seem refractory to crystal growth. The authors proposed
a 'generic phase diagram' applicable to all or many proteins, in which the
types of solid phases listed in Table 5 (spherullites, oils, and various kinds of
precipitate) assume a coherent progression as one approaches some optimal
value of supersaturation. Relationships between these and more optimal
locations are probably somewhat reproducible from protein to protein.
Further support for quantitative scoring came when detailed examination
of the behaviour of monoclinic tryptophanyl-tRNA synthetase (1) provided
evidence for a 'sweet-spot' in the solubility diagram where crystal volume and
99
 C. W. Carter Jr
shape were optimized simultaneously, suggesting that growth in all three
dimensions progressed unimpeded and at similar rates. Interestingly,
secondary nucleation, which had been a difficult problem, was also minimal
for the same, optimal conditions, whereas various pathologies, including
elongation (needles) and multiple nucleation appeared away from the optimal
supersaturation, which was higher than that suggested by intuition.
A problem with the scale in Table 5 is the tendency to confuse micro-
crystalline and amorphous precipitates (16, 17). The following methods can
help confirm the rank-ordering of scores:
(a) Streak seed from precipitates to assay for microcrystallinity (Chapter 7).
(b) Examine precipitates for birefringence in a glass depression slide to avoid
plastic/air interfaces, with reservoir solutions in the depressions surround-
ing the sample and a large glass slide as a coverslip over the entire dish to
avoid crystallization of the crystallizing agent.

5.3 Size and shape

Optimization experiments show how the properties of a particular crystal
form depend on smaller variations of the important parameters. Thus, the
scoring must reflect this variation. Crystal dimensions can be scored directly
by microscopic inspection.

5.4 Scoring the best result or the 'average' from a given

test?
Intuition and experience suggest different answers to this question. Intuitively,
averaging seems safer, because it helps to insulate the scores from fluctua-
tions, giving a truer account of given experimental conditions. On the other
hand, an exceptional crystal among many smaller or misshapen crystals is
often an 'existence proof that one is close to an optimum. Averaging pro-
perties of the good crystal with those of the many poorer samples will tend to
dilute the influence of that one experimental combination, submerging the
signal in the analysis. Our experience is that using the properties of the best
crystals in a drop has been considerably more successful than focusing on
average properties. Replication (see Section 7.2.3) helps to separate spurious
from meaningful results.

6. Regression, the analysis of variance, and analysis of

models
Key to all factorial response-surface methods is the quantitative analysis of
the different behaviours observed as experimental conditions are varied.
100
 4: Experimental design and quantitative analysis
Quantitative analysis begins with some quantitative 'observed score', given
here by Qobs. Two related kinds of analysis can be done (74):
(a) Estimation of main effect and interaction contrasts.
(b) Construction and testing of linear models for calculating the scores, Qcalc.
Before using these procedures, carry out a preliminary analysis as given in
Protocol 5 to assess which scores and/or variables deserve more careful
analysis.

Protocol 5. Preliminary analysis of scores

Method
1. Prepare a file with the experimental matrix and all scores. Most
statistics programs provide separate modules for data entry, statistical
model building and analysis, and presentation graphics. For factors
whose levels are attributes (e.g. different ions), create a separate
column for each attribute, giving values of 0 or 1 for its presence
or absence. It may also be useful to approximate monotonic behaviour
like that of the ions of the Hofmeister series in a separate column.
2. Determine the maximum, minimum, mean value, for each score (this
may be programmed in EXCEL; ,in SYSTAT, use the Stats/Stats
command). Identify scores with significant variation.
3. Examine the data graphically. A matrix of two-way scatterplots for
each score versus each experimental variable is a useful, visual
presentation of the data, showing obvious correlations.
4. Try to express such trends as linear models explaining the variation in
scores (Protocol 6).

6.1 Analysis of contrasts

Contrasts (Section 2) compare averages of specific, balanced sets of experi-
ments treated at different levels of a particular factor for information about
whether or not that factor is a significant source of the variation observed in
the results. The null hypothesis is that these averages, also called the
treatment means, are equal within the error of the measurements. Unless all
variables are converted to 'standard' variables with all means = 0 and
standard deviations = 1, the contrasts themselves can be misleading because
the units of different variables may be on quite different scales. Thus, as the

101
 C. W. Carter Jr
italicized phrase suggests, the simple contrast sum calculations presented in
Section 2.3.2, Equation 2, can be misleading, and should be verified by a full
analysis of variance whenever possible.

6.1.1 Averaging results from replicated experiments

Averaging is a well-understood source of confidence in making inferences.
The signal-to-noise of a repeated experimental result increases as the square
root of the number of times it is repeated. For example, a 24-experiment
design provides 12 experiments at each level of a two-level variable,
increasing the signal-to-noise by ~ 2.5 times over that from two replicates.

6.1.2 Averaging experiments that are not exact replicates

The efficiency of factorial experiments comes from contrasts between en-
sembles of experiments that are not exact replicates of one another.
Interestingly, the standard deviations of averages from different experiments
in a design depend strongly only on the precision of individual measurements
and only weakly on the number of different factors being varied. Balanced
designs average-out the effects of other variables, highlighting the bona fide
sources of variation in scores from all experiments. In these cases, the average
of all experiments treated at one level of a particular factor, relative to the
average of experiments treated at another level remains sensitive to the
influence of that factor.

6.2 Analysis of models by multiple regression and the

analysis of variance
Multiple regression provides (3j values that minimize the sum of squared
n
differences 2(Qobs,i - QPred,i)2- This predictive model can provide an estimate,
i=l
Qcalc, for the experimental result, based on contributions from the different
factors. If there are K adjustable parameters in the model, they can be
estimated by minimizing the sum of the squares of differences between Qobs
and Qcalc over all the experiments in a design of N > K experiments.
Analysis of designed experiments is a broad and well-established field, and
no attempt will be made here to summarize what can be found in excellent
monographs on the subject (19,75-79). Procedures necessary for rudimentary
statistical and graphic analyses are available in standard data analysis
programs (EXCEL, KALEIDAGRAPH) for personal computers. Regression
analysis of linear models is available only in the more comprehensive stat-
istical packages. Particularly useful are the two programs SYSTAT (80) (now
SPSS SYSTAT) and IMP (80). MATHEMATICA is a useful adjunct for
102
 4: Experimental design and quantitative analysis
examining linear models, locating stationary points, and plotting two-
dimensional level surfaces (81). Templates are available from the author at
http: //russell. med. unc. edu/ —carter/designs.
The type of design matrix and mathematical model used for fitting the
observed data are associated with specific objectives typically encountered in
a crystal growth or other experimental project. Three types of models are
summarized in Table 6, together with definitions for some terms used in the
text, and how they are most often used to identify suitable conditions,
characterize effects, and optimize. Models (II) and (III) are derived from the
first by adding specific terms, which are indicated in bold face. The linear
model used in our original study (15), included only main effects (Table 6-I).
For the full-factorial design in Table 2 we added all the multi-factor inter-
action terms (Table 6-II). Quadratic models (Table 6-III) supplement the
general linear model with all possible squared terms and two-factor inter-
actions. By substituting new variable names for the squared and interaction
terms, these models become special cases of the general linear model (I).
For all models, e is the residual error to be minimized, and the constant value,
PQ, is the mean score for all experiments in the design. Each /J; coefficient in
the linear model (I) is the average amount by which the presence of factor, Fi,
raises or lowers the score from the overall average. Higher-order /^ and /?„
coefficients have similar meaning for models II and III. Since many important
variables have non-linear effects, it is not surprising that many processes,
including crystal growth, can be modelled more effectively by multivariate

Table 6. Designed experiments for different contexts

Factorial design N experiments with simultaneous variation of M < N factors

Objective Design/model

Detection of important main Incomplete factorial

effectsand interactions.

Verification of these Replicated, full factorial

inferences.

Optimization of crystal Response surface

growth conditions.

103
 C. W. Carter Jr
quadratic functions. These are the simplest functions that assume maxima,
minima, and saddle points, within the range of independent variables. Once a
model has been fitted, its stationary points can be determined analytically by
partial differentiation with respect to all the independent variables and
equating the gradient to zero. Stationary point coordinates provide estimates
for the factor levels giving the best result. This is the basis of the response-
surface method (19).

6.2.1 Selecting, fitting, and evaluating models

Protocol 6 outlines the search for a good regression model. Statistical analysis
is used to gauge how much of the scatter in scores can be attributed to
experimental changes, relative to what must be attributed to noise. It is worth
repeating that if all experiments have nearly the same scores, then one has a
smaller chance of identifying significant gradients. A full quadratic response-
surface model with N variables takes the general form of a polynomial with
(N + 1)(N + 2)12 coefficients (Table 6, model III).

Protocol 6. Data analysis: finding a good regression model

Method
1. Examine the mean value, range, and standard deviation statistics for
each score. Identify scores with significant variations.
2. Examine scatterplot matrices of each score for each independent
variable.
(a) Look for and note any obvious trends.
(b) Identify any categorical variables that might usefully be re-
ordered. These include cases where scores are clustered within a
category, but where their mean values show no pattern. Re-order
numerical assignments for categories, if doing so would create a
physically sensible monotonic or parabolic series.
(c) Rationalize patterns created in step 2(b) in terms of physically
reasonable effects.
3. Decide whether or not the variation in mean scores (step 1) and
suggested patterns (step 2) are sufficient to support meaningful
regression against the experimental variables.
4. Build and test trial models as described in Protocol 7, testing all main
effects first, then including two-way interactions and quadratic terms.

104
 4: Experimental design and quantitative analysis
The analysis involves two interdependent tasks: model selection and
parameter estimation. The terms of a model represent the calculable
effects of the experimental factors, pjFj, their interactions, PyFiFj, and their
squares, 3HF2i, plus the intrinsic variation or noise, e, associated with the
experimental set-up and scoring. It is essential to discard terms that do not
contribute significant information about the response, and use the extra
degrees of freedom to improve the estimate for the residual error, thereby
reducing the variances of parameter estimates. Incorrect fitting of question-
able parameters can lead to model bias. Similarly, finding the best 'inter-
pretation' for a response-surface experiment can be haphazard because co-
efficients and their statistical significance change when the model itself
changes. Choosing which coefficients should be retained in the response-
surface model is therefore a challenging task (19, 82). Once the best set of
predictors has been identified, their coefficients are estimated by multiple
regression least squares.
These tasks are the job of a full statistics program. We have used two such
programs. SYSTAT (80) (now SPSS/SYSTAT) and JMP (80). Both provide a
powerful multiple regression module, with the appropriate statistical calcula-
tions, and graphing tools. The following illustrations present output from the
SYSTAT MGLH (Multiple regression, General Linear Hypothesis) module.
This module cannot evaluate partial derivatives or solve for stationary points,
but it is easy to use, intuitive, and fast, and the former tools are available in
Mathematica (81). The JMP user interface is well-developed, and JMP may be
easier to use.

Protocol 7. Identifying and fitting model parameters

Method
1. Define (on a command line or in a dialog box) a linear model for a
single dependent variable (the score) as a function of a set of
independent variables.
2. Identify the best subset of terms by using 'stepwise multiple re-
gression'. As the name suggests, this algorithm is an automated
procedure for choosing the best subset of coefficients. Generally, start
with a complete model and gradually eliminate insignificant contrib-
utors. It is sometimes useful to work forwards, finding the most
significant terms first. Stepping can be carried out with a variety of
different tolerance and threshold criteria for including or eliminating
terms, as appropriate, until the model appears stable and sensible.
Depending on the algorithm, terms may be recycled if they appear to
regain or lose significance as the stepping proceeds.
105
 C. W. Carter Jr

Protocol 7. Continued
3. Think about what the model is saying; add/delete coefficients that
might/might not make sense.
4. Compare different models according to three types of criteria:
(a) The squared multiple correlation coefficient, R2. This indicates the
percentage of the variation that can be 'explained' by the model. It
should be as high as possible.
(b) The probability of the F-ratio test under the null hypothesis that
the model has no predictive value. This value should be very
small.
(c) Individual Student t-test probabilities for each coefficient. These
should be as small as possible.
5. Verify coefficients indicating two-way interactions; calculate and
examine the average scores for all four of the combinations (—, +-,
-+, ++) in the two-way matrix.
6. Expect that useful models of optimum behaviour will have positive
linear and negative quadratic coefficients. Look for models with these
characteristics.
7. Plot two-dimensional projections of the model surface. Super-
imposing these plots onto the observed scores is a good way to get a
feel for what the model has to tell about the system.
8. Verify that plots actually reflect the data. Discrepancies usually mean
errors in entering data, scores, and/or model coefficients, but can point
to unexpected effects.
9. Generally, the residuals, ([Qobs,i - Qcalcj]), can be saved and analysed in
the same ways used for the scores (Protocols 5-8). Examine them for
clues about where the model may be deficient.

Statistics programs output a summary table with the relevant information.

Models first of all should make physical sense. Different models then are
compared using two statistical properties indicated by bold face in Table 7,
which summarizes a model based on the data in Table 1 and Figure 3. First is
how well they account for the deviation of scores from their mean value (the
'variation'). The squared multiple correlation coefficient, R2, represents the
percentage of this variation predicted by the model, and should be as close as
possible to 1.0. By itself, however, it is insufficient to discriminate between
good and bad models. An incomplete, but valid model may have a low value
for R2, (0.3-0.7). High values also can result from models with too many
parameters.
106
 4: Experimental design and quantitative analysis

Table 7. Multiple regression and analysis of variance for the illustration in Figure 3

A. Data from the full-factorial experiment from Table 1

Squared multiple R: 0.994 F-ratio = 128.7 P = 0.001
Variable Coefficient Std error T P(2 tail)
Constant 19.4 0.402 48.1 0.00002
PH 4.2 0.402 10.4 0.002
Temp -1.6 0.402 -4.0 0.029
Protein 7.9 0.402 19.7 0.00029
pH*Temp -0.9 0.402 -2.3 0.108

B. Data from incomplete factorial subset in Figure 3

Squared multiple R: 0.999 F-ratio = 443.1 P = 0.036
Variable Coefficient Std error r P(2 tail)
Constant 18.9 0.22 87.2 0.007
PH 4.0 0.25 16.0 0.040
Temp -1.8 0.25 -7.0 0.090
Protein 7.8 0.25 31.0 0.021
pH*Temp -1.4 0.22 -6.4 0.099

C. Data from orthogonal array A

Squared multiple R: 0.951 F-ratio = 9.78 P = 0.221
Variable Coefficient Std error T P(2 tail)
Constant 19.175 2.175 8.816 0.072
PH 3.825 2.175 1.759 0.329
Protein 8.825 2.175 4.057 0.154

D. Data from orthogonal array B

Squared multiple R: 0.986 F-ratio = 34.6 P = 0.119
Variable Coefficient Std error T P(2tail)
Constant 19.500 1.000 19.500 0.033
pH 4.500 1.000 4.500 0.139
Protein 7.000 1.000 7.000 0.090

A second kind of statistic, the 'P values', give the probability of obtaining
equally good models by random processes, i.e. under the 'null hypothesis' that
the variation is uncorrelated with changes in the experimental variables. This
information is available for the Student t-tests of individual coefficients and
the overall F-ratio. The Student t value is the ratio of a coefficient to its
standard error, so it is a statistic about the signal-to-noise of an estimated
coefficient. The overall F-ratio is the squared distance between calculated and
average scores for all experiments divided by the squared distance between
calculated and observed scores. It can be considered an estimate of the signal-
107
 C. W. Carter Jr

to-noise of the model. Useful models can have P-values as high as 10-3, our
best models are better than 10-11. Protocol 8 and the associated statistical
criteria apply equally to the investigation of any model.
Try models using main effects one or two at a time, to see how much of the
variation is explained (multiple R2), and at what cost in terms of the F-ratio
probability. How much better does the model get by adding another factor?
Since this procedure is the most difficult, it requires some intuitive feel, which
comes from practice. This approach is illustrated for data from Table 1 in
Table 7, which illustrates the trade-off that necessarily accompanies sampling.
All models have quite high R2 values, and satisfy the criterion of predictive
power. In other respects, however, the quality of the models depends on the
amount of available data. The complete ensemble of eight experiments
affords the best model by all criteria: it supports a model with more significant
parameters, while at the same time giving the best F-ratio and t-test prob-
abilities. The incomplete factorial subset performs nearly as well, giving
nearly the same parameter estimates. Their statistical significance, however, is
degraded by about two orders of magnitude. Deleting additional data, as with
the two different orthogonal arrays, degrades the models well beyond the point
where they are useful. None of the parameters is statistically significant.
Usually, the default stepwise regression will produce a reasonable repre-
sentation of what is in the data. That model can occasionally be improved
using different tricks. Check individual t-tests and try deleting the worst one
(with the highest probability). Omit the constant term only when its t-test is
poor (has a high probability). It is recommended to retain the main effect in
any model that involves a higher-order interaction, even if this makes the
model worse. There are exceptions to all of these guidelines. A trade-off must
always be made between the decreased variance of the model parameters,
achieved by reducing their number, and the potential loss of real information
about the response surface that occurs when a 'true' parameter with a large
variance is deleted from the model. A more detailed description, with
examples, is provided in ref 83). The ultimate test is a model's usefulness.

6.2.2 Stationary points

Quadratic and higher-order polynomial models are selected, fitted, and
evaluated in the same way. However, these models can be sufficiently curved
to possess maxima and minima, and require another level of analysis con-
cerning their stationary points. Stationary points of a function occur where its
gradient goes to zero with respect to all the independent parameters, as
illustrated for a one-dimensional case in Figure 7 and in Figure 8b. They are
solutions to the simultaneous equations obtained by equating the partial
derivatives to zero, and are determined as in Protocol 8. Partial derivatives, by
definition, estimate changes in a dependent variable (the response) induced
by small fluctuations in the independent variables (the experimental con-
108
 4: Experimental design and quantitative analysis

Figure 7. Stationary points of polynomial functions and experimental reproducibility.

Stationary points occur where the partial derivatives of a function all vanish. A one-
dimensional parabola is illustrated to show the relationship between its derivative and its
maximum value, and to emphasize that far from the maximum value, the function has
increasingly steep slopes. These steep slopes translate into experimental fluctuations,
which have greater impact far from the stationary point.

dilions). If the function describes the system behaviour adequately, the

stationary point coordinates will specify an optimal set of experimental
conditions.

Protocol 8. Identifying stationary points

A. Determination and characterization of stationary point coordinates

These steps can be programmed in MATHEMATiCA. Templates are avail-
able at http: //russell.rned.unc .edu/-carter/designs).

1. Calculate the partial derivatives of the model function with respect to

each of the independent variables. The set of partial derivatives
constitutes the gradient of the function.
2. Solve the simultaneous equations for the stationary point by equating
the partial derivatives to zero.
3. Check that the stationary point coordinates correspond to experi-
mentally sensible values.
109
 C. W. Carter Jr
Protocol 8. Continued
4. Determine the nature of the stationary point by examining the signs of
the second derivatives, and/or by plotting two-dimensional level
surfaces with constant values for all but two variables (1).

B. Verification of the behaviour at and near a stationary point

1. Set up duplicate experiments under conditions given by the stationary
point.
2. Compare the experimental results with the predictions that they
should be 'optimal'.

6.2.3 Analysis and verification

Identification of a stationary point does not guarantee optimality. One must
first examine the behaviour nearby to determine whether it corresponds to a
maximum, a minimum, or to a saddle point. This can be done by evaluating
the second partial derivatives: negative curvature in all variables implies a
local maximum, whereas positive curvature implies a minimum, and mixed
second partial derivatives imply a saddle point. An accessory strategy in such
cases is to examine two-dimensional level surface plots in all subspaces.
Examples of these level surfaces are shown in Figure 8.
Frequently, the dominant feature of a response surface is not a stationary
point, but a 'ridge', along which the value of the function increases, but
normal to which it decreases (Figure 8a). It is hard to overemphasize the
importance of using this kind of feedback to iterate the search for
improvements; an illustration is provided in Section 7.
A model is only as good as its valid predictions. Consequently, any pre-
dictions regarding optimum behaviour of the system must be verified with
replicated experiments at the stationary point.

6.2.4 Advantages of working with stationary points

Stationary points may be determined for any desirable crystalline property
which can be 'scored' precisely enough for optimization, including volume,
shape, diffraction limits, stability, and relative freedom from secondary
nucleation (1). Finding and using conditions close to stationary points of
analytical response surfaces has several important advantages.
(a) Optimization. The first real benefit is the obvious one: conditions at a
convex stationary point produce crystals that are in some sense optimized
(84,85).
(b) Reproducibility. Stationary points are not only optima; they also represent
points where the response is most reproducible. Even the most carefully
110
 4: Experimental design and quantitative analysis

Figure 8. (a) Ridges observed in the temperature x concentration level surface for
response surfaces determined for two different TrpRS crystal polymorphs (1). The
ridgeline represents combinations of temperature and [protein] which produce the same
supersaturation level, suggesting that supersaturation is a natural search variable, (b)
Level surfaces for a third TrpRS polymorph determined using as one of the search
variables an approximation to supersaturation (87). Here, the three level surfaces
involving [protein] all show optima.

performed experiments can suffer from a frustrating level of irrepro-

ducihility. One source of variability is intimately connected to the partial
derivatives of the underlying, multidimensional response surface (Figure
7). Working at stationary points helps insulate crystal growth from
experimental errors in pH determination, pipetting errors affecting
concentrations of various components, temperature fluctuations, and so
on,
(c) Insight. Examining plots of the response surface is often a powerful aid to
the interpretation of crystal growth experiments. We discovered a useful
new search dircclion for response-surface experiments when graphs of
two-dimensional level surfaces (Figure 8a) revealed ridges, where the
same result was obtained for many combinations of factors. Ridges were
111
 C. W. Carter Jr
conspicuous in the temperature X concentration level planes. Tempera-
ture and concentration affected the protein's solubility coordinately, the
ridge corresponding to a constant value of supersaturation. An important
inference was that it would be better to sample simultaneously for the
effects of protein concentration and supersaturation. Using [protein] and
[protein] X [crystallizing agent] eliminated the ridges in the resulting
level surfaces (Figure 8b). This observation led to the approximations to
supersaturation described in Section 7.2.1.
Empirical response surfaces provide scientific documentation about crystal-
logenesis that is otherwise difficult to achieve, including conclusions with
important and interesting biochemical relevance (2). Particularly interesting
as evidence is the statistical support for the stationary point coordinates.

7. Optimization
The term 'optimization' is used frequently in discussions of crystal growth.
Usually, it refers to variation of some experimental variables with the aim of
empirically finding 'better' crystals from among the conditions tested. Given
sufficient time and materials, and a fortunate choice of experimental variables
to explore, any search strategy can lead to better crystals. Often, however,
both time and materials are limited. In such cases, there are two ways in which
the search for optimal conditions can be made more systematic and efficient.
One uses either a line search or a more elaborate variant called 'simplex'
optimization (86), the other uses response surfaces. The two approaches are
complementary; the former may actually be more appropriate if existing
conditions are far from an optimum.

7.1 Steepest ascent and simplex optimization

Conceptually, these two closely related methods provide the most intuitive
algorithms for optimization. Sensing improvement along a direction, one
marches toward that direction which gives the greatest improvement. This
single search direction is specified by the gradient of the function to be
optimized, which may depend on several variables. Generally, this approach
is appropriate far from a stationary point, where experiments are available to
fit a plane or ridge function. An example is given in Section 8. A simplex is a
polytope having one more point than the dimension of the space in which it is
conceived; the two-dimensional simplex is a triangle. Simplices provide a
model-independent basis for an optimization search. One deletes the worst
point from the current simplex, and reflects that point across the face to which
it forms a perpendicular, forming a new simplex. The gradient of the under-
lying surface is detected indirectly, by comparing scores at each point of the
simplex (86).
112
 4: Experimental design and quantitative analysis

7.2 Optimization using quadratic polynomial models

A more powerful optimization procedure is to fit quadratic polynomial
models to data from a Hardin-Sloane response-surface design. In our experi-
ence, quadratic models with statistical significance orders of magnitude better
than the familiar 95% confidence limit can be obtained for a variety of differ-
ent types of scores (1, 24, 83). Protocol 9 outlines the set-up for a response-
surface experiment.

Protocol 9. Setting up a response-surface crystal growth

optimization

Method
1. A least four variables should be sampled simultaneously to use
quadratic polynomial models.
2. Two variables should always represent the solubility diagram,
irrespective of the other variables.
(a) Protein concentration and supersaturation are more nearly ortho-
gonal than protein concentration and precipitant concentration,
facilitating sampling of the nucleation zone (Figure 9).
(b) A product, either [protein] x [crystallizing agent] (87) or In
[protein] + [crystallizing agent] are useful approximations to super-
saturation or ln[supersaturation] when the latter are unknown.
3. Choose additional variables based on any available prior information.
Regression analyses of an incomplete factorial screening experiment
often provide indications of the most significant main effects (24).
4. Centre the experiment close to the best known set of conditions.
(a) Exploitation of response-surface experiments is most successful
when quantitative and reliable scores have been obtained for all
or nearly all experiments.
(b) Designs not centred on conditions known to produce crystals are
effectively screening experiments, and should be carried out using
qualitatively different experimental matrices.
5. Choose sufficiently large ranges for each variable to induce significant
variation in the score without losing the response itself. Follow the
guidelines in Section 4.3.

7.2.1 Identifying critical variables

It is essential before optimizing a system that the critical variables be
identified. This is one of the most important reasons to analyse screening
113
 C. W. Carter Jr

experiments quantitatively. Two of the most important search variables can

usually be chosen at the outset: the protein concentration and some variable
representing supersaturation. These two variables 'orthogonalize' the
irregularly shaped nucleation zone of the solubility diagram (Figure 9). We
have used the product ([protein] X [crystallizing agent]} to approximate this
variable (87), Logarithmic sampling based on Green's approximation to
solubility: S = [protein ]-exp(k [crystallizing]), leads to InS —[In[protein]
k[precipitating agent]}, which gives a more manageable range of [crystallizing
agent], and should be preferred. Other important variables include the pH
(implicitly the net charge), the position of the anions in the Hofmeister series
(43,45), and the presence of any important ligands. The range and mean values
for these values are often critical to success: too narrow a range leads to too
small a variation in the results for the regression to attribute effects to anything
but fluctuations. By the same token, the ranges must be small enough that all or
most experiments actually produce crystals. Some guidelines are given in
Section 4.3,2,

7.2.2 Replication
The variances of replicate experiments done at a random combination of
parameters near a stationary point should vary inversely with the distance of
that point from the stationary point (Figure 7). Our experience suggests that
this is indeed the case. An important consequence is that the estimates of the
variance obtained from an ensemble of unrcplicated experiments may fail to
capture the information about stationary points that can be obtained from
replicated sets of experiments, which would provide an approximate map of
the variances over the experimental space. For this reason, it is useful to carry
out each experiment in a design twice.

Figure 9. Orthogonalization of the nucleation zone for sampling purposes, (a) The
nucleation zone is shown as the curved figure bounded by the metastable and
precipitation zones, (b) The product [macromolecule] x [crystallizing agent] is constant
along the rectangular hyperbolae in (a). Thus, the nucleation zone becomes a rectangle
which can be sampled evenly on a regular grid (87).

114
 4: Experimental design and quantitative analysis

7.2.3 An example
A response-surface experiment for monoclinic tryptophanyl-tRNA synthetase
(TrpRS) crystals is presented step-by-step to illustrate the process of building
and analysing the model. This crystal form had always previously given long
and thin crystals that had to be grown bigger by repeated reseeding (84, 85).
Intense efforts to improve the size of the initial crystals by systematic
variation of protein concentration had led only to modest improvements.
After some initial successes with other forms of TrpRS, we proceeded as
follows with the monoclinic form (1):
(a) We selected [protein], {[prot] X [ppnt]}, temperature, and the con-
centration of the additive, PEG 400, as the independent variables. A
range of values (+\- 9-23%) was chosen to surround the best conditions
we had previously achieved. Values of these four variables were assigned
to 20 experiments of the Hardin-Sloane matrix in Table 2. These experi-
ments were each done twice and scored using three different criteria:
volume, the ratio of the smallest dimension to the largest, and a subjective
assessment of their uniformity and freedom from satellite crystals. These
scores are more objective and quantitative than the subjective scale we
used previously to score screening experiments and should be easier to
use. They were input to a SYSTAT data file together with the H-S matrix.
(b) Using the MGLH (Multiple Regression, General Linear Hypothesis)
module in the statistics program SYSTAT for the Macintosh (82), tests
were carried out first using a model containing all 15 terms of equation
(HI) (a constant, four each of the main effects and the quadratic terms,
and the six two-factor interactions). The initial and subsequent models
were evaluated according to criteria described in Section 6.2.1. This
complete model had an F-ratio probability of10-4.R2 was 0.87, indicating
that all but about 13% of the variation in observed scores could be
attributed to the model. Nevertheless, in several respects the model
needed adjustment. In particular, three coefficients had t-test probabili-
ties > 0.05, indicating that they were without significance and should be
removed.
(c) The full model was pruned by backward stepwise regression, eliminating
three of the 15 parameters. The final model was obviously very significant;
its F-ratio probability, P, was 10-11, R2 was 0.95; and t-test probabilities of
the 11 coefficients of the model were nearly all below the 5% confidence
limit (1). Two were around 0.1 and of questionable significance. However,
removing them resulted in a serious deterioration, causing P to decrease
by an order of magnitude, and four additional factors with significant
t-tests in the best model became completely insignificant.
Prediction and verification. Partial derivatives of the calculated score with
respect to all variables were evaluated from the model expression. Equating
115
 C. W. Carter Jr
the tour derivatives to zero and solving for the coordinates ([protein]opt.
[ p r o _ p t j l j p l , [PEG]l)pt, and Temp0|,,) of the optimum predicted that crystals
grown at these values would he better any of those observed in the H-S
experiment. Crystals grown at the optimum point were two orders of
magnitude larger and of sufficient volume for diffraction experiments. Similar
analysis for two other scores (volume and uniformity) showed that the three
optima were essentially in the same place, simultaneously optimizing all three
scores (1).

8. Resolution of polymorphs
A recurring problem with screening experiments is that they sample con-
ditions that may he far from optimal for a particular crystal form. Attempts to
optimize such cases using Hardin-Sloane designs can give rise to surfaces like
that in Figure JO, where instead of an optimum, the surface represents a
saddle. Sometimes, this problem is confounded by the appearance of multiple
crystal forms in the same experiments. Indeed, the appearance of saddle points
is often diagnostic of the superposition of overlapping response surfaces for
different poymorphs. Since different crystal polymorphs rarely, if ever, have
the same dependence on all experimental variables, finding optimal stationary
points for different polymorphs can be a useful way to 'purity' them away
from one another.
Figure II shows how a line search helped relocate the centre for a Hardin-

Figure 10. Response surface determined for E. colicytidine deaminase at the neighbour-
hood of a 'hit' from a kit (73) showing that the conditions are nearly at a saddle point and
hence far from an optimum. The dot indicates experiment number 28 of Jancarik and Kim
125).
UK
 4: Experimental design and quantitative analysis

Figure 11. Use of steepest ascent (line search) to resolve polymorphs of E. colicytidine
deamlnase crystals, (a) The proportion of the desired form (Form I) fitted to the ridge
function shown above the figure. Tests selected to fall on the line as described in the text
showed that this proportion increased along that search, until at the X, only Form I was
observed, (b) A Hardin-Sloane design was then centred on the X, giving the optimum
shown (73).

Sloane design away from the saddle point in Figure 10, ultimately locating
optimal conditions for a new polymorph of E. coli cytidine deaminase (73).
Experiments at higher temperature in the original Hardin-Sloane design had
variable amounts of a second polymorph. The proportion of this second
polymorph was used as a score and fitted to the ridge surface in Figure 11a.
The initial conditions from the Hampton screen lay near the bottom of a
steeply sloped ridge in two variables, temperature and [Na Acetate]. The
gradient at that point indicated that the proportion of this form increased by
0.114 for every degree the temperature was raised, and by 2.07 as the con-
centration of acetate was reduced by 1 mole. Combining the two indications
gives that for each degree of temperature increase, the acetate concentration
should be reduced by 0.114/2.07 = 0.055 M. Experiments stepped out along
this gradient showed the expected increase in the fraction of the desired
crystal form. When the proportion was equal to 100%, a new Hardin-Sloane
design was performed, giving the surface in Figure 11b.
The first step in separating out the polymorphs was to identify the variables
critical to the different polymorphs. This was inferred from how the poly-
morphs behaved with respect to changes in all variables in the first Hardin-
Sloane design. A subset was identified by model fitting and analysis, and the
surface used to identify the gradient for a steepest ascent search to resolve the
different polymorphs from one another. Subsequent, re-centred response-
surface experiments can locate the stationary points associated with each
117
 C. W. Carter Jr

polymorph. The potential advantages of using response surfaces in this way

also include the evidence provided about the influence of each of the factors
on the polymorphism.

References
1. Carter, C. W., Jr and Yin, Y. (1994). Acta Cryst., D50,572.
2. Carter, C. W., Jr., Doublie, S., and Coleman, D. E. (1994). J. Mol. Biol., 238,346.
3. Schutt, C. E., Lindberg, U., Myslik, J., and Strauss, N. (1989). 7. Mol. Biol., 209,
735.
4. Anderson, C. M., Zucker, F. H., and Steitz, T. A. (1979). Science, 204,375.
5. Sack, J. S., Saper, M. A., and Quiocho, F. (1989). J. Mol. Biol., 206,171.
6. Bullough, P. A., Hughson, F. M., Skehel, J. J., and Wiley, D. C. (1994). Nature,
371, 37.
7. Colombo, M. F., Rau, D. C., and Parsegian, A. V. (1992). Science, 256, 665.
8. Rand, R. P., Fuller, N. L., Butko, P., Francis, G., and Nicholls, P. (1993).
Biochemistry, 32, 5925.
9. Rand, R. P. (1992). Science, 256,618.
10. Reid, C. and Rand, R. P. (1997). Biophys. J., 72,1022.
11. Parsegian, V. A., Rand, R. P., and Rau, D. C. (1995). In Methods in enzymology
(eds M. L. Johnson and G. Ackers), Academic Press, Inc., London. Vol. 259,
p. 43.
12. Leikin, S. and Parsegian, V. A. (1994). Proteins: Structure, Function, Genetics, 19,73.
13. Royer, W. E., Pardnamani, A., Gibson, Q. H., Peterson, E. S., and Friedman, J. M.
(1996). Proc. Natl. Acad. Sci. USA, 93,14526.
14. Ladner, J. E., Finch, J. T., King, A., and Clark, B. F.C. (1972). J. Mol. Biol., 72,99.
15. Carter, C. W., Jr. and Carter, C. W. (1979). J. Biol. Chem., 254,12219.
16. Carter, C. W., Jr., Baldwin, E. T., and Frick, L. (1988). 7. Cryst. Growth, 90, 60.
17. Carter, C. W., Jr. (1990). Methods: a companion to methods in enzymology, 1,12.
18. Carter, C. W., Jr. (1992). In Crystallization of proteins and nucleic acids: a practical
approach (ed. R. Giege and A. Ducruix), p. 47. IRL Press, Oxford University
Press, Oxford, UK.
19. Box, G. E. P., Hunter, W. G., and Hunter, J. S. (1978). Statistics for experimenters.
Wiley Interscience, New York.
20. Carter, C. W. (1995). Qual. Eng., 8,181.
21. Bell, J. B., Jones, M. E., and Carter, C. W., Jr. (1991). Proteins: Structure, Function,
Genetics, 9,143.
22. Belts, L., Frick, L., Wolfenden, R., and Carter, C. W., Jr. (1989). J. Biol. Chem.,
264, 6737.
23. Doublie, S., Xiang, S., Bricogne, G., Gilmore, C. J., and Carter, C. W. J. (1994).
Acta Cryst A50,164.
24. Yin, Y. and Carter, C. W., Jr. (1996). Nucleic Acids Res., 24,1279.
25. Jancarik, J. and Kim, S.-H. (1991). J. Appl. Cryst., 24, 409.
26. Doudna, J. A., Grosshans, C., Gooding, A., and Kundrot, C. (1993). Proc. Natl.
Acad. Sci. USA, 90,7829.
27. Song, L. and Gouaux, E. (1997). In Methods in enzymology, ibid ref. 54. Vol. 276,
p. 60.
118
 4: Experimental design and quantitative analysis
28. Kundrot, C. E. (1997). In Methods in enzymology, Vol. 276, p. 143.
29. Scott, W. G., et al. (1995). J. Mol. Biol, 250, 327.
30. Gilliland, G. L., Tung, M., Blakeslee, D. M., and Ladner, J. E. (1994). Acta Cryst.,
D50,408.
31. Rosenberg, J. M., et al. (1997). In Annual Meeting 56 (American Crystallographic
Association, St. Louis).
32. Shaanan, B. (1983). J. Mol. Biol., 175,159.
33. Smith, F. R., Lattman, E. E., and Carter, C. W., Jr. (1991). Proteins: Structure,
Function, Genetics, 10,81.
34. Janin, J. and Wodak, S. J. (1993). Proteins, 15,1.
35. Srinivasan, R. and Rose, G. D. (1994). Proc. Natl. Acad. Sci. USA, 91,11113.
36. Smith, F. R. and Simmons, K. C. (1994). Proteins: Structure, Function, Genetics,
18, 295.
37. Plackett, R. L. and Burman, J. P. (1946). Biometrika, 33,305.
38. Kingston, R. L., Baker, H. M., and Baker, E. N. (1994). Acta Cryst., D50,429.
39. Hardin, R. H. and Sloane, N. J. A. (1993). J. Stat. Plan. Inference, 37, 339.
40. Audic, S., Lopez, F., Claverie, J.-M., Poirot, O., and Abergel, C. (1997). Proteins:
Structure, Function, Genetics, 29,252.
41. Ferrone, F., Hofrichter, J., and Eaton, W. A. (1985). /. Mol. Biol., 183,611.
42. Ducruix, A. F. and Ries-Kautt, M. (1990). Methods: a companion to methods in
enzymology, 1,25.
43. Ries-Kautt, M. and Ducruix, A. (1992). In Crystallization of nucleic acids and
proteins: a practical approach (ed. A. Ducruix and R. Giege), p. 195. IRL Press,
Oxford, UK.
44. Carbonnaux, C., Ries-Kautt, M., and Ducruix, A. (1995). Protein Sci., 4,2123.
45. Ries-Kautt, M. and Ducruix, A. (1997). In Methods in enzymology ibid ref. 54.
Vol. 276, p. 23.
46. Ataka, M. and Tanaka, S. (1986). Biopolymers, 25,337.
47. Ataka, M. and Michihiko, A. (1988). J. Cryst. Growth, 90,86.
48. Rosenberger, F. (1996). J. Cryst. Growth, 166, 40.
49. Muschol, M. and Rosenberger, F. (1997). J. Chem. Phys., 107,1953.
50. Rosenberger, F., Muschol, M., Thomas, B. R., and Vekilov, P. G. (1996). J. Cryst.
Growth, 168,1.
51. Retailleau, P. (1997). Thesis Ph. D. Universite Paris XI Orsay.
52. Retailleau, P., Ries-Kautt, M., and Ducruix, A. (1997). Biophys. J., 73,2156.
53. Luft, J. R, et al. (1994). J. Appl. Cryst., 27,443.
54. Luft, J. R. and DeTitta, G. T. (1997). In Methods in enzymology (eds C. W. Carter
and R. Sweet), Academic Press, Inc., London. Vol. 276, p. 110.
55. Rosenberger, F., Vekilov, P. G., Lin, H., and Alexander, J. I.D. (1997).
Microgravity science and technology 10,29.
56. Vekilov, P. G., Thomas, B. R., and Rosenberger, F. (1998). J. Phys. Chem., 2, 5208.
57. Vekilov, P. G. and Rosenberger, F. (1998). J. Cryst. Growth, 186,251.
58. Gilliland, G. and Bickham, D. M. (1990). Methods: a companion to methods in
enzymology, 1,6.
59. Guilloteau, J. P., et al. (1996). Proteins: Structure, Function, Genetics, 25,112.
60. Timasheff, S. N. (1992). In Pharmaceutical biotechnology (ed. T. J. Ahera and
M. C. Manning), p. 265. Plenum, New York.
61. Bricogne, G. (1993). Acta Cryst., D49, 37.
119
 C. W. Carter Jr
62. Rosenberger, F. (1986). J. Cryst. Growth, 76, 618.
63. Rosenberger, F. and Meehan, E. J. (1988). J. Cryst. Growth, 90,74.
64. Stura, E. and Wilson, I. (1990). Methods: a companion to methods in enzymology,
1,38.
65. Stura, E. A., Satterthwait, A. C., Calvo, J. C., Kaslow, D. C., and Wilson, I. A.
(1994). Acta Cryst., D50,448.
66. Stura, E. (1998). In Crystallization of nucleic acids and proteins: a practical
approach (ed. A. Ducruix and R. Giege). IRL Press, Oxford, UK.
67. Hofrichter, J., Ross, P. D., and Eaton, W. A. (1974). Proc. Natl. Acad. Sci. USA,
71,4864.
68. Hofrichter, J., Ross, P. D., and Eaton, W. A. (1976). Proc. Natl. Acad. Sci. USA,
73,3035.
69. Collins, K. D. and Washabaugh, M. W. (1985). Q. Rev. Biophys., 18,323.
70. Collins, K. D. (1997). Biophys. J., 72,65.
71. Sousa, R. (1996). In Methods in enzymology Vol. 276, p. 131.
72. Scarborough, G. (1994). Acta Cryst., D50,643.
73. Kuyper, L. and Carter, C. W., Jr. (1996). J. Cryst. Growth, 168,155.
74. Edwards, A. L. (1979). Multiple regression and the analysis of variance and
covariance. W. H. Freeman and Company, San Francisco.
75. Milliken, G. A. and Johnson, D. E. (1989). Analysis of messy data: Volume 2,
nonreplicated experiments, pp. 1-199. Chapman and Hall, New York.
76. Milliken, G. A. and Johnson, D. E. (1992). Analysis of messy data: Volume 1,
designed experiments, pp. 1-173. Chapman and Hall, New York.
77. Neter, J. and Wasserman, W. (1974). Applied linear statistical models, pp. 1-842.
Richard D. Irwin, Homewood, IL.
78. Morrison, D. F. (1990). Multivariate statistical methods. McGraw Hill, New York.
79. Mardia, K. V., Kent, J. T., and Bibby, J. M. (1979). Multivariate analysis.
Academic Press, London.
80. SAS Institute, I. (1989). SAS Institute, Inc., Gary, NC.
81. Wolfram, S. (1994). Mathematica, the student book. Addision Wesley, Reading,
MA.
82. Wilkinson, L. (1987). SYSTAT, Inc., Evanston, IL 60601.
83. Carter, C. W., Jr. (1997). In Methods in enzymology Vol. 276, p. 74.
84. Thaller, C., et al. (1981). J. Mol. Biol., 147, 465.
85. Thaller, C., et al. (1985). In Methods in enzymology Vol. 114, p. 132.
86. Wilson, L. (1998). In SpaceBound 1997 (ed. J. Sygusch). Canadian Space Agency,
Montreal, Quebec, Canada.
87. Carter, C. W., Jr. (1996). Acta Cryst., 53,647

120
 5

Methods of crystallization
A. DUCRUIX and R. GIEGE

1. Introduction
There are many methods to crystallize biological macromolecules (for reviews
see refs 1-3), all of which aim at bringing the solution of macromolecules to a
supersaturation state (see Chapters 10 and 11). Although vapour phase equilib-
rium and dialysis techniques are the two most favoured by crystallographers
and biochemists, batch and interface diffusion methods will also be described.
Many chemical and physical parameters influence nucleation and crystal
growth of macromolecules (see Chapter 1, Table 1). Nucleation and crystal
growth will in addition be affected by the method used. Thus it may be wise to
try different methods, keeping in mind that protocols should be adapted (see
Chapter 4). As solubility is dependent on temperature (it could increase or
decrease depending on the protein), it is strongly recommended to work at
constant temperature (unless temperature variation is part of the experi-
ment), using commercially thermoregulated incubators. Refrigerators can be
used, but if the door is often open, temperature will vary, impeding repro-
ducibility. Also, vibrations due to the refrigerating compressor can interfere
with crystal growth. This drawback can be overcome by dissociating the
refrigerator from the compressor. In this chapter, crystallization will be
described and correlated with solubility diagrams as described in Chapter 10.
Observation is an important step during a crystallization experiment. If you
have a large number of samples to examine, then this will be time-consuming,
and a zoom lens would be an asset. The use of a binocular generally means the
presence of a lamp; use of a cold lamp avoids warming the crystals (which
could dissolve them). If crystals are made at 4°C and observation is made at
room temperature, observation time should be minimized.

2. Sample preparation
2.1 Solutions of chemicals
2.1.1 Common rules
Preparation of the solutions of all chemicals used for the crystallization of
biological macromolecules should follow some common rules:
• when possible, use a hood (such as laminar flux hood) to avoid dust
 A. Ducruix and R. Giege
• all chemicals must be of purest chemical grade (ACS grade)
• stock solutions are prepared as concentrated as possible with double
distilled water.
Solubility of most chemicals are given in Merck Index. Filter solutions with
0.22 um minifilter. If you use a syringe, do not press too hard as it will enlarge
the pores of the filter. Filters of 0.4 um will retain large particles whereas
0.22 um filters are supposed to sterilize the solution. Label all solutions
(concentration, date of preparation, initials) and store at 4°C. Characterize
them by refractive index from standard calibrated solutions. Use molar units
(mole per litre) in preference to percentage. This avoids confusion between
weight to weight (w/w), weight to volume (w/v), and volume to volume (v/v).
Quite often crystallization articles refer to percentage without any inform-
ation, making the results difficult to reproduce. As an example, a 20% (w/v)
stock solution twice diluted will give a 10% solution whereas this would not
be the case if starting from a 20% (w/w) solution.

2.1.2 Buffer
The chemical nature of the buffer is an important parameter for protein crystal
growth. It must be kept in mind that the pH of buffers is often temperature-
dependent; this is particularly significant for Tris buffers. Buffers, which must
be used within one unit from their pK value, are well described in standard
text books (4).

2.1.3 Purification of PEG

PEG is available in a variety of polymeric ranges; the most commonly used
are compounds with mean molecular weights of 2000, 4000, and 6000. These
are polydisperse mixtures and their composition around the mean may vary
from one producer to the other; it is better to always use the same brand.
Molecular weights higher than 10000 are rarely used because of excessive
viscosity of their solutions. Reproducibility and quality of crystals may
depend on PEG molecular weight.
The optimal range of PEG concentration for crystallization of a given
protein depends on PEG molecular weight and may be very narrow (about
1.5%, w/v). As the viscosity of PEG solutions may lead to pipetting errors, it
is better to routinely verify the concentration of PEG in reservoirs for dialysis
or vapour diffusion by measuring the refractive index with a Abbe refracto-
meter. A reference curve is established on known amounts of PEG dissolved
in the same buffer. From a practical point of view, commercial PEG does
contain contaminants, either ionic (5) or derived from peroxidation. Re-
purification as shown in Protocol 1 is strongly recommended before use (6).

122
 5: Methods of crystallization

Protocol 1. Purification of PEGa

Equipment and reagents

• Bio-Rad AG501X8 . PEG
. Na2S2O4.5H20

Method
1. Pour a column (2.5 x 10 cm2) with a mixed-bed strong ion exchange
resin in the H+-OH- form (e.g. Bio-Rad AG501X8). Wash with 300 ml
methanol:water (3:7, v/v) then with 500 ml water.
2. Dissolve 200 g PEG in water (500 ml final volume). Measure the
refractive index of the solution. Degas for 30 min under vacuum (water
aspirator) with gentle magnetic stirring. Add 1.24 g Na2S2O4.5H20 and
let stand for 1 h.
3. Pass the solution through the column at a flow rate of 1 ml/min.
Discard the first 30 ml and collect the following eluate.
4. Check the concentration of PEG by refractometry and store frozen in
small aliquots at -20°C.
5. Before use, an antioxidant can be added (para-hydroxyanisole, stock
solution in isopropanol, 1.3 mg/ml; add 1 ul per ml PEG stock
solution).

" See Chapter 2, Table 2 for another method.

2.1.4 Mother liquor

Mother liquor is defined as the solution containing all crystallization chemicals
(buffer, salt, crystallizing agent, and so on) except protein or nucleic acid at
the final concentration of crystallization.

2.2 Preparing samples of biological macromolecules

2.2.1 Removing salts
Proteins and nucleic acids often contain large amount of salts of unknown
composition when first obtained. Thus it is wise to dialyse a new batch of a
macromolecule against a large volume of well-characterized buffer of given
pH, to remove unwanted salts and to adjust the pH. Starting from known
conditions helps to ensure reproducibility.
Commercially available dialysis tubing are generally composed of cellulose
or polyacetate. They should be prepared as described in Protocol 2. Molecular
cut-off (i.e. the pore size limit) is given by manufacturers for spherical
particles. As most proteins and nucleic acids are better described as ellipsoids
or cylinders, a cut-off far enough from the molecular weight should be chosen.
123
 A. Ducruix and R. Giege
As an example, a 12 000 cut-off is not appropriate for lysozyme (Mr 14 305)
and if used will allow the protein to leak slowly through the membrane, thus
diluting it in the external chamber. You can check the impermeability of the
membrane towards the protein by placing a dialysis tube containing the
macromolecule at a given concentration in a beaker (outer reservoir) con-
taining a small volume (a few millilitres). After 24 h, check the biological
macromolecule concentration in the outer reservoir. In practice, it should not
contain more than 1% of the amount of the macromolecule.
Depending on the choice of the commercial membrane, it should be either
prepared and demetallized (following Protocol 2) or washed with distilled
water. The membrane must be kept cold. If kept for a long period (even in the
form of dry tubes), contamination problems may arise leading to leaks of
protein. Dialysis membranes are fragile and it is quite easy to puncture them
with nails; so do wear plastic gloves, remembering to rinse them because they
are often treated with talc.

Protocol 2. Preparation of dialysis tubing

Equipment and reagents

• Dialysis tubing • Solution A: 5% (w/v) NaHCO3 (50 g/litre)
• Bunsen burner and 50 mM EDTA (18.6 g/litre)

Method
1. Boil the tubing for 30 min in solution A. Avoid puncture at this stage
when mixing with glass rods or magnetic stirrers.
2. Rinse several times with distilled water.
3. Store in 50% (v/v) ethanol solution.
4. Check each tubing integrity for possible puncture.
5. Prior to crystallization, rinse membranes several times with distilled
water then with buffer.

2.2.2 Concentration
Whatever the crystallization method used, it requires high concentrations of
biological macromolecules as compared to normal biochemistry conditions.
Before starting a crystallization experiment, a concentration step is generally
needed. Keep pH and ionic strength at desired values, since pH may vary
when the concentration of the macromolecule increases. Also, low ionic
strength could lead to early precipitation (see Chapter 2 for further practical
advice). It could be very frustrating when the macromolecule precipitates
irreversibly or adsorbs on concentration apparatus membrane and/or support.
Many commercial devices are available; they are based on different principles
and operate:
124
 5: Methods of crystallization
(a) Under nitrogen pressure.
(b) By centrifugation (e.g. Centricon).
(c) By lyophilization (because it may denature some proteins, test first on
a small amount). Non-volatile salts are also lyophilized and will
accumulate.
Choice of the method of concentration depends on the quantity of macro-
molecule available. Dialysis against high molecular PEG proved to be
successful in our hands. We use a dialysis chamber (volume 50-500 ul), the
top of which is covered by a glass coverslip which is sealed to the plastic
chamber with grease. Figure 1 describes the apparatus. This allows for an easy
access from the top of the dialysis chamber.

Figure 1. Dialysis apparatus used for the concentration of biological macromolecules.

Prepare a solution of 20% (w/v) of PEG 20000 in an appropriate buffer and dialyse your
biological macromolecule against it. Check the macromolecule concentration using
optical absorbance, colorimetric or enzymatic assay on a small aliquot.

2.2.3 Removing solid particles

Before a crystallization experiment, solid particles such as dust, denatured
proteins, and solids coming from purification columns (beads) or lyophil-
ization should be removed. This could be achieved by centrifugation or
filtration, depending on the available quantity.
2.2.4 Measuring concentration of biological macromolecules
The most common method to measure macromolecular concentrations is to
sample an aliquot, dilute it with buffer, and measure absorbance at 280 nm or
260 nm (for proteins or nucleic acids, respectively) within the linear range of a
spectrophotometer. Proper subtraction with the reference cell should be
made especially when working with additives absorbing in the 260-300 nm
wavelength range. When working with enzymes, an alternative method to
measure the concentration of protein is to perform activity tests, otherwise,
colorimetric methods can be used. This can be done either by a modification
(7) of the assay described by Winterbourne (8) or by the modification (9) of
the reagent assay developed by Bradford (10). See also Chapter 2, Section 4.2
and Chapter 8, Section 2.2.
125
 A. Ducruix and R. Giege

3. Crystallization by dialysis methods

3.1 Principle
These methods allow for an easy variation of the different parameters which
influence the crystallization of biological macromolecules. Different types of
dialysis cells are used but all follow the same principle. The macromolecule is
separated from a large volume of solvent by a semi-permeable membrane
which gives small molecules (ions, additives, buffer, and so on) free passage
but prevents macromolecules from circulating. The kinetics of equilibrium
will depend on the membrane cut-off, the ratio of the concentration of
crystallizing agent inside and outside the protein chamber, the temperature,
and the geometry of the cell.

3.2 Examples of dialysis cells

3.2.1 Macrodialysis
The most simple technique is to use a dialysis bag. Large crystals are
occasionally obtained. It is very convenient for successive recrystallization.
Commercially available dialysis tube such as Spectrapor of inner diameter
2 mm can be used to limit the amount of protein. However, each assay
requires about 100 ul at least per sample.
3.2.2 Microdialysis
i. Zeppenzauer cells
Crystallization by dialysis was first adapted to microvolumes by Zeppenzauer
(11). The miocrodialysis cells are made from capillary tubes closed either by
dialysis membranes or polyacrylamide gel plugs. Those cells require only 10
ul or less of macromolecule solution per assay. A modified version of the
Zeppenzauer cell was described by Weber and Goodkin (12).
ii. Dialysis buttons
Commercially available, microdialysis cells are made of transparent Perspex
(Figure 2a). They can be obtained from your local workshop. The protein
chamber should be filled so that it forms a dome. The membrane is then
placed over the button and held by an O-ring of appropriate diameter.
Installing the membrane has a reputation of difficulty, and beginners often
trap air bubbles between the protein solution and the membrane. To avoid
the problem, one can use either a piece of plastic (Figure 2b) having the same
diameter as the button, or when working with flat buttons use a plastic cork.
The cell is then immersed in a vial and an inexpensive way is to use
transparent scintillation counting vials.
Observation with a binocular or microscope through the membrane is easy.
However, if you use cross-polarizer, the membrane will depolarize light. For
crystal mounting, O-ring and membrane should be removed gently. Problems
126
 5: Methods of crystallization

Figure 2. Dialysis button. (a) Diameter of the buttons (A) generally varies between 10-20
mm and volume of the biological macromolecule chamber is 5-350 ul. (b) To install a
dialysis membrane, use a pipetter tip of diameter adapted to the concave shape of the
dialysis button.

occur when crystals stick to the wall of the chamber. In this case a whisker can
be used to gently free the crystal.
iii. Microcap dialysis
The technique, described in Figure 3 and adapted from ref. 13, was useful for
membrane proteins (see Chapter 9). Although it is more difficult to observe
crystal growth with this method it is very convenient for storing, and micro-
caps are disposable. The method is quite easy to use (Protocol 3) and you can
play with the ratio of the diameter versus height of the microcap to influence
the kinetics of crystallization. It should be noted that when the macro-
molecule does not entirely fill the microcap chamber, the presence of air
(which is compressible) allows osmotic pressure to develop, and thus modifies
the macromolecule concentration.

Protocol 3. Crystallization by microcap dialysis

Equipment and reagents

• Microcaps • Low melting wax
• Hamilton syringe • Dialysis membrane cut in small squares
• Tygon tubing of 1.3 and 3 mm diameter • Low temperature soldering
• 1.5 ml Eppendorf tubes

Method
1. Commercial microcaps are cut with a glass saw (for instance a 50 ul
cap is cut in three parts to fit in an Eppendorf tube of 1.5 ml).
2. Wrap a piece of dialysis membrane around one end (the one which is
smooth) and secure with a piece of tubing of diameter 1.3 mm.

127
 A. Ducruix and R. Giege

Protocol 3. Continued
3. Load the biological macromolecule with a Hamilton syringe.
4. Shake the assembly to bring the biological macromolecule solution in
contact with the membrane.
5. Seal the free microcap end with wax molten by soldering bit.
6. Split a second ring of tubing of diameter 3 mm and superpose it to the
first one. The aim is to prevent the membrane from touching the
bottom of the Eppendorf tube which would limit the exchange with the
reservoir.
7. Insert in an Eppendorf tube (volume 1.5 ml) containing 1 ml of the
crystallizing solution.
8. Close cap and wrap top of Eppendorf tube with Parafilm (American
Can Company).

Figure 3. Crystallization by microcap dialysis. (a) Place the dialysis membrane on the
microcap; (b) secure with Tygon ring; (c) load the protein; (d) close the extremity with
wax; (e) fill up a 1.5 ml Eppendorf tube with crystallizing agent and insert microcap.

128
 5: Methods of crystallization

3.2.3 Double dialysis

The purpose of double dialysis (14) is to reduce the rate of equilibration and
therefore to provide a better control of crystal growth. The apparatus is
shown in Figure 4. Large crystals of delta toxin of Staphylococcus aureus were
obtained this way (14). Equilibration time is rather long (could be several
weeks) as the gradient concentration of crystallizing agent is low. It is there-
fore more geared toward production of large crystals than screening. Protocol
4 describes the methodology. One can manipulate the different parameters
(membrane cut-off, distance between dialysis membranes, relative volumes,
and so on) to optimize crystallization.

Figure 4. Double dialysis set-up (adapted from ref. 14). Macromolecule is contained in a
conventional dialysis button placed in a second dialysis set-up. The equilibration rate
depends upon the volumes of buffers in the different compartments.

Protocol 4. Crystallization by double dialysis

Equipment and reagents

• Dialysis button Dialysis membrane cut in small squares
• O-ring

Method
1. Prepare the dialysis burton as in Section 3.2.2 with a solution of
crystallizing agent at a concentration in which the biological macro-
molecule is undersaturated. This is called the 'inner compartment'.
2. Insert the conventional dialysis burton in a vial (about 10 ml) called the
'middle compartment' containing a solution of crystallizing agent at
a concentration in which the biological macromolecule is super-
saturated.
3. Cover with a dialysis membrane maintained by an O-ring.
129
 A. Ducruix and R. Giege
Protocol 4. Continued
4. Place in a larger vial (for instance a beaker of 50 ml) containing the
solution of crystallizing agent at a concentration in which the bio-
logical macromolecule will precipitate completely. This is the 'outer
compartment'.
5. Cover with Parafilm or a stopper.

4. Crystallization by vapour diffusion methods

Among the crystallization micromethods, vapour diffusion techniques are
probably the most widely used throughout the world. They were first used for
the crystallization of tRNA (15).

4.1 Principle
The principle of vapour diffusion crystallization is indicated in Figure 5. It is
very well suited for small volumes (down to 2 ul or less). A droplet containing
the macromolecule to crystallize with buffer, crystallizing agent, and addi-
tives, is equilibrated against a reservoir containing a solution of crystallizing
agent at a higher concentration than the droplet. Equilibration proceeds by
diffusion of the volatile species (water or organic solvent) until vapour
pressure in the droplet equals the one of the reservoir. If equilibration occurs
by water exchange from the drop to the reservoir, it leads to a droplet volume
decrease. Consequently, the concentration of all constituents in the drop will
increase. For species with a vapour pressure higher than water, the exchange
occurs from the reservoir to the drop. In such a 'reverse' system, the drop
volume will increase as well as the concentration of the drop constituents.
This last solution, less widely used, has led to the crystallization of tRNAAsp
(16) and of several proteins (17, 18). The same principle applies for hanging
drops, sitting drops, and sandwich drops.
Glass vessels in contact with macromolecular solutions should be treated

Figure 5. Schematic representation of hanging drop, sitting drop, and sandwich drop.

130
 5: Methods of crystallization

in a way to obtain an hydrophobic surface. Coaled glass coverslips are

commercially available but can be prepared following Protocol 5.

Protocol 5. Preparation of glass coverslipsa

Equipment and reagents

• Coverslips . Soap solution
• Temperature controlled water-bath • Distilled water
• Toluene • Ethanol
• Dimethyldichlorosilane

A. Silanization
1. Place coverslips in a bath of toluene containing 1% dimethyl-
dichlorosilane at 60°C for 10 min.
2. Coverslips are then washed with soap solution and rinsed with
distilled water and ethanol.
3. The same procedure is used for Pyrex plates.
4. Dry overnight at 120°C to sterilize vessels.

B. Siliconization
1. This can be achieved with commercially available reagent solutions
(e.g. Sigmacoat). Coverslips are washed in the solution, and dried
overnight at 120°C.
"All operations can be performed under vacuum when dealing with narrow vessels like
capillaries.

A device shown in Figure 6 helps to treat coverslips. If made in Perspex, the

device is only used for drying coverslips; if Teflon made, it can be used for alt
the silanization process.

Figure 6, A device for treating coverslips (Perspex or Teflon made). The set-up displayed
(in Teflon) can be manufactured in the laboratory workshop. Coverslips are held in the
threading of the two bottom axis; a smaller unthreaded axes secures the coverslips. The
set-up shown is about 20 cm long and permits handling of about 60 coverslips.

131
 A. Ducruix and R. Giege

4,2 Experimental set-ups

4.2.1 Hanging drops in Linbro boxes
Commercially available Linbro boxes arc plastic boxes (Figure 7) normally
used for tissue culture (they may be replaced by Costar or VDX plates,
supplied by Hampton Research). Plastic boxes will depolarize light; so unless
for a particular orientation of a crystal, no birefringence will be observed.
Boxes contain 24 wells labelled A, B, C, D vertically, and 1-6 horizontally. It
is convenient, to avoid confusion, to always use them in the same orientation.
Each box must be carefully labelled (date, experiment number, operator, and
so on). Each well has a volume of approximately 2 ml and an inner diameter
of 16 mm. There is a small rim which will be used for sealing the system. Each
well will be covered by a glass coverslip of 22 mm diameter treated as
described in Protocol 5.

Figure 7, A Linbro box for vapour diffusion crystallization in hanging drops. The photo-
graph shows the box with its cover which is held by Plasticine in the corners. For a better
display, two drops were prepared with dyes (at the left).

Drops are set up following Protocol 6. Most of the people use a 'magic'
ratio of two between the concentration of the crystallizing agent in the
reservoir (well) and in the droplet. This is conveniently achieved by mixing a
droplet of protein at twice the desired final concentration with an equal
volume of the reservoir at the proper concentration (other ratios can be used
as well). Avoiding local over-concentration can be achieved by placing the
two drops (protein and reservoir) on each side of an Eppendorf tube and
vortexing it quickly.

Protocol 6. Crystallization in Linbro boxes

Equipment and reagents

• Silicone grease • Pair of brussel
• Linbro boxes • Plasticine
• Coverslips

132
 5; Methods of crystallization
Method
1. Grease rims with silicone grease.a
2. Fill up reservoir with 1 ml of filtered (0.22 um) crystallizing agent.
3. Spray glass coversllp with antidust.
4. Mix a 2-10 ul drop of filtered (0.22 um) biological macromolecule
solution with an equivalent volume of reservoir.
5. Layer the drop on the 22 mm diameter coverslip (do not touch the
coverslip with the extremity of the tip of the pipettor or it will spread)
so that a nearly hemispherical drop is formed.b
6. Return the coverslip with a pair of brussel (or fingers). First train
yourself with water!
7. Set on the grease rim and gently press to seal the well with the
grease. Do not press too firmly, otherwise the coverslip will break.
8. Check the sealing by inspecting the rim in an azimuthal way. If sealing
is not properly done, the drop will concentrate as well as the
reservoir. Crystallization will occur eventually, but will be very
difficult to reproduce.
9. Adjust glass coverslips tangent to each other otherwise they overlap.
10. Put Plasticine in the corners to avoid the contact between the cover
and grease, otherwise slips get stuck at the cover.
a
To dispense the grease, fill up a syringe and replace the needle by an Eppendorf yellow tip.
b
You may layer several microdrops on a coverslip.

When no crystal or precipitate is observed, either supersaturation is not

reached or one has reached the metastable region (see Chapter 10 for
definition). In the latter case, changing the temperature by a few degrees is
generally sufficient to initiate nucleation. For the former, the concentration of
crystallizing agent in the reservoir must be increased. In the former case,
gently rotate the coverslip in the plane of the rim to ease the grease (which
becomes 'stiff' with time), then lift it, suck the reservoir entirely, and replace
it by a more concentrated solution. More grease is added and the coverslip
sealed. The volume of the drop will decrease again and all constituent
concentrations in the drop increase.

i. Problems
For membrane proteins (see Chapter 9), the presence of detergent tends to
spread out the drops and lower the surface tension. In all cases gravity will
tend to sink the drops containing the macromolecule in the reservoir for
volumes exceeding 25 ul. Shaking Linbro boxes will give the same results.
Boxes must be transported horizontally and carefully. It is always painful for
beginners to ruin an experiment when transporting a box. If you prepare
133
 A. Ducruix and R. Giege
boxes at room temperature and transfer them to 4^°C, condensation will
occur on the surface of the coverslip. Water droplet will surround the
macromolecule drop. If it mixes, the protein will dilute and probably stay in
an undersaturated state. To avoid this problem, set up boxes at the final
experiment temperature and cover boxes with polystyrene sheets.
It is possible to recrystallize macromolecules using hanging drops as
described in Protocol 7.

Protocol 7. Recrystallization of macromolecules

Equipment and reagents

• Minifilters • Crystallizing agent
• Coverslips • Grease

A. For purification
1. Remove mother liquor.
2. Wash crystals with fresh buffer.
3. Redissolve in fresh crystallizing agent solution.
4. Recrystallize.
B. For crystal growth
1. Redissolve crystals by replacing crystallizing agent in the reservoir by
buffer. Depending on the biological macromolecule, it may take a few
hours or a few days. The drop volume will increase; so control the
process, otherwise the drop will sink.
2. Filter the drop in a minifilter (e.g. Costar, Millipore) by centrifugation.
Warning: the dead volume is at least 5 ul.
3. Place the drop on a clean glass coverslip.
4. Add some grease to the rim.
5. Fill the reservoir with crystallizing agent and recrystallize.

4.2.2 Crystallization with ACA CrystalPlates®

The American Crystallographic Association (ACA) sponsored a vapour
diffusion plate called CrystalPlate® (manufactured by ICN Flow). Although
dedicated for use with automated systems (19), it is equally useful for manual
crystallization. As shown in Figure 8, it may be used to set up crystallization
by hanging drops, sitting drops, or sandwich drops, depending on the
thickness of the lower glass slip and the volume of the drop (Protocol 8). Each
box contains 15 wells. The glass coverslips should be prepared as described in
Protocol 5. If desired, the breakaway plug in the bottom of a reservoir — see
134
 5: Methods of crystallization

HANGING OR STANDING DROP SANDWICHED DROP

Figure 8. ACA CrystalPlate® (courtesy of ICN Flow). This is a versatile system for vapour
diffusion crystallization allowing individual experiments on sitting, hanging, or
sandwiched drops under classical or automated conditions.

tank in Figure 8 — may be removed with pliers and a rubber septum inserted
so that the reservoir solution may be changed with an hypodermic syringe.
A crystallization set-up based on the same versatile concept as the
CrystalPlates® is the so-called Q-Plate™ (supplied by Hampton Research).
i. Advantages
Because of the glass windows, when looking at drops under polarized binocular,
birefringence of crystals can be observed. Drops with macromolecules are no
longer above the reservoir thus eliminating sinking. Large drops can be
prepared with the sandwich method.
Warning: if you use a rubber septum, plates stick when you translate them
during observation, eventually provoking disasters.

Protocol 8. Crystallization with ACA CrystalPlates®

Equipment and reagents

• ACA CrystalPlates® and glass slips • Grease

Method
1. Prepare the plate by filling up the upper and lower troughs of each
well with ordinary hydrocarbon vacuum pump oil or grease.
2. To dispense oil or grease, fill up a syringe and replace the needle by
an Eppendorf yellow tip.
3. Put 0.5 ml of crystallizing agent into each reservoir.
4. Position one of the 14 x 14 mm2 glass coverslips over the hole in each
well. The coverslips should seal quickly if there is enough oil or grease.
(a) For hanging or sitting drops use 14 x 14 x 0.2 mm3 glass
coverslips.

135
 A. Ducruix and R. Giege,
Protocol 8. Continued
(b) For sandwich drops (5-25 ul) use 14 x 14 x 1 mm3 glass cover-
slips.
(c) For sandwich drops (15-75 ul) use 14 x 14 X 1.5 mm3 glass cover-
slips.
5. Put a drop of the biological macromolecule solution in the centre of
the lower (for sitting or sandwich drops) or upper (for hanging drops)
glass coverslips and then set one of the 24 x 30 x 1 mm3 glass
coverslip in position on the upper trough.

4.2.3 Other systems

Among the first used sitting drop set-ups are systems using Pyrex plates (e.g.
Corning Glass 7220) with three or more depressions placed in different type
of boxes, with the reservoir solution below the plate (20, 21). Cryschem
MVD24 plates are commercially available (manufactured by Cryschem Inc.
or C. Supper Company), Each well contains a plastic post in the centre to hold
the protein sitting drop. The cup has been designed (Figure 9) to provide
maximum surface area for free diffusion during equilibration. The reservoir
solution is held within the narrow moat surrounding the support port. The
plates are sealed with an adhesive tape which is supplied by the manufacturer.
An ingenious way to seal the plates with the tape and to set hanging drops is

Figure 9. Crystal growth multi-chamber plate for vapour diffusion crystallization on

individual sitting drops (courtesy of C, Supper Company).

136
 5: Methods of crystallization
to use the HANGMAN framework. Here the protein droplets are first in-
stalled on the tape which is than inverted in a second step over the plate (22).
A variety of other set-ups have been designed in many laboratories, allow-
ing for instance Linbro or VDX boxes to be used for sitting drop experiments
(with the depression on a small plastic bridge (23) or on a glass rod as on the
Oxford or Perpetual Systems Corporation set-ups, respectively); or doing
vapour phase equilibration in capillaries (24, 25), or even directly in X-ray
capillaries as was described for ribosome crystallizations (26) or in the gel
acupuncture method. For extremely fragile crystals, when transfer from
crystallization cells to X-ray capillaries (see Chapter 14) can lead to internal
damage and mechanical cracks of crystals, this last method may be well
adapted.

4.3 Varying parameters

Although unique in this respect, vapour diffusion methods permit easy
variation of physical parameters during crystallization, and many successes
were obtained by modifying supersaturation by temperature or pH changes
(27, 28) (see also Section 5). With ammonium sulfate as the crystallizing
agent, it has been shown that the pH in the droplets is imposed by that of the
reservoir (29). Consequently, varying the pH of the reservoir permits gentle
adjustment of that in the droplets. From another point of view, sitting drops
are well suited for attempting epitaxial growth of macromolecule crystals on
appropriated mineral matrices (30).
In vapour diffusion crystallizations, the contamination by micro-organisms
can be prevented by placing a small grain of thymol, a volatile organic
compound, in the reservoir (see also Chapter 2). Thymol, however, can have
specific effects on crystallization, as shown with glucose isomerase (31), and
may thus represent an useful additive to assay in crystallization screenings.

4.4 Kinetics of evaporation

4.4.1 Final concentrations
Calculating the final concentration of constituents in an equilibrated drop is
often a source of misunderstanding in protocols. Sometimes it refers to the
final concentration in the drop before the vapour diffusion process is initiated,
sometimes it describes the final concentration in the drop at equilibrium at the
end of the process. So, except for the species of vapour pressure higher than
that of water, at equilibrium, if the ratio of crystallizing agent concentration
between reservoir and drop is two, final concentrations and volumes are as
follows:
• final drop volume = 1/2 initial volume
• final concentration of all constituents of the drop (protein, additive, and so
on) equals twice the initial concentration.
137
 A. Ducruix and R. Giege
Many other ratios can be used. Varying the volume of the droplet will
influence the kinetics of crystallization and so the protein crystal size.
4.4.2 Equilibration kinetics
The kinetics of water evaporation determines the kinetics of supersaturation
and accordingly affects nucleation rates. Evaporation rates from hanging
drops have been determined experimentally in the presence of ammonium
sulfate, PEG, MPD (32), and NaCl (33-35) as crystallizing or dehydrating
agents. The main parameters which determine the rate of water equilibration
are temperature, initial drop volume (and initial surface to volume ratio of the
drop and its dilution with respect to the reservoir), water pressure of the
reservoir, and the chemical nature of the crystallizing agent (Figure 10a).
Theoretical modelling (35) has shown in addition the pivotal role of the drop
to reservoir distance (d), but effect of this parameter is negligible in classical
set-ups, e.g. in Linbro boxes (32), and becomes only noticeable in special
experimental arrangements (34, 35) when d > 2 cm (Figure 10b). Note: the
presence of macromolecule does not seem to affect the water evaporation
rate.
From the practical point of view, the time for water equilibration to reach
90% completion can vary from about 25 hours to more than 25 days, the
fastest equilibration occurring in the presence of ammonium sulfate, that in
the presence of MPD being slower, and that in the presence of PEG by far the
slowest (Figure 10a). Estimates of the minimal duration of equilibration
under several standard experimental conditions can be obtained from an
empirical model (36). Equilibration rates are significantly slowed down by
increasing appropriately the distance between the drop and the reservoir
(Figure 10b). This can be done using the Z/3 plate design (34) or just simple
test-tubes as reservoirs (35). An alternative solution to decrease equilibration
rates is to layer oil over the reservoir (37).
The particularly slow equilibration rates observed with PEG may explain
crystallization successes using this precipitating agent (38). Indeed crystal
growth may be favoured when supersaturation is attained very slowly. This
fact is corroborated by independent experiments in which the terminal crystal
size was significantly increased by reducing the vapour pressure of the
reservoir, i.e. the evaporation rate, as a function of time (37, 39, 40).

5. Crystallization by batch methods

5.1 Classical methods
The biological macromolecule to be crystallized is mixed with the crystallizing
agent at a concentration such that supersaturation is instantaneously reached.
This can be achieved with all methods previously described. For hanging drop
or sitting drops, the reservoir no longer acts to concentrate the drop but is
138
 5: Methods of crystallization

Figure 10. Water evaporation kinetics in the presence of ammonium sulfate (AS), MPD,
PEG, and NaCI as dehydrating agents. (a) Measurements done in drops set in Linbro
boxes with AS, MPD, and PEG. The data also show the influence of protein (in AS
experiment), and of initial drop volume (in MPD experiment) on final drop volume. V0 is
the initial volume of the drop; experiments were conducted with a concentration of
crystallizing agent in the reservoir twice that in the drop, time at zero. Adapted from ref.
32 (b) Measurements done on hanging drops (24 ul) set over test-tubes with distances
between the drop and reservoir varying from 7.6-78.3 mm. Experiments were conducted
at 22.9°C with 1.0 M NaCI as initial concentration in the drop and 2.0 M NaCI in the
reservoirs. Adapted from ref. 35.

only present to maintain constant vapour pressure. Because one starts from
supersaturation, nucleation tends to be too large. However, in some cases
fairly large crystals can be obtained when working close to the metastable
region. This is illustrated in Chapter 10. If supersaturation is too high, a
precipitate may develop in a batch crystallization vessel; do not discard such
experiments because crystals can eventually grow from the precipitate by
Ostwald ripening (41). Notice, however, that the growth kinetics under such
circumstances are decreased.
An automated system for microbatch macromolecule crystallizations and
screening has been described (42) allowing the set up of samples of less than
2 ul. Reproducibility of experiments is guaranteed because samples are
139
 A. Ducruix and R. Giege
dispensed and incubated under paraffin oil, thus preventing evaporation and
uncontrolled concentration changes of the components in the micro-droplets.
Using silicone oils that are slightly soluble in water, or appropriate mixtures
of paraffin and silicone oils, results in gradual protein concentration in the
droplets like in vapour diffusion experiments (43).
A variation of classical batch crystallization is the sequential extraction
procedure of Jakoby (44), based on the property that many proteins (not all)
are more soluble in concentrated salt (e.g. ammonium sulfatc) when lowering
the temperature. The method can be adapted for microassays and was
successfully applied for the crystallization of a proteolytic fragment of
methionyl-tRNA synthetase from Escherichia coli (45).

5.2 Advanced methods

Solubility and supersaturation of proteins is influenced by hydrostatic
pressure. Advantage has been taken of this fact to crystallize proteins at high
pressure (46, 47). Effects become significant for pressures higher than 50 Mpa
(500-fold atmospheric pressure) as shown for lysozyme crystals that can be
grown in the range 50-250 Mpa. Such crystals exhibit habits different from
controls grown at atmospheric pressure and diffract at high resolution (47).
Experimental set-ups that can contain up to 24 samples of 80 ul crystallizing
solution each and that can be pressurized up to 400 Mpa have been designed
(47). Although the method has only been occasionally used in the macro-
molecular field, in the future it may represent an interesting alternative to
obtain new crystalline forms of proteins.
The floating drop method enables crystallization of biological macro-
molecules under conditions where the crystallizing solution has no contact
with the container walls (48, 49). Drops (5-100 ul) are placed at the interface
between two layers of inert and non-miscible silicone fluids contained in square
glass or plastic cuvettes. The density of the fluids can be such that drops
containing the most common crystallizing agents can be floated (Figure II).

Figure 11. Crystallization in floating drops. Droplets (40 ul) of four different crystallizing
agent solutions placed at the interface of two silicone fluids are displayed. The two
silicone fluids at 20°C (800 ul each with low density fluid PS037 layered over high density
PS181 from Huls America, Inc.) [silicone oils may be provided by Hampton Research) are
placed in square glass or polystyrene spectrophotometer cuvettes. Adapted from ref. 48.

140
 5: Methods of crystallization
Several proteins and a spherical plant virus were crystallized in the
temperature range 4-20°C using this method (48). Its main advantage is to
reduce the nucleation rate. Thus crystallization in floating drops provides a
means to obtain a small number of larger crystals in an homogeneous liquid
medium. Because drops are not in contact with air, the method may be
convenient to crystallize proteins sensitive to oxidation. Further, when im-
plemented in a thermostated device, the method provides a simple and
convenient way for kinetic measurements of macromolecule crystal growth
(48).
Other advanced methods useful to prepare crystals for diffraction studies
(e.g. in gelified media, under microgravity, and the gel acupuncture method)
are described in Chapter 6.

6. Crystallization by interface diffusion

This method was developed by Salemme (50) and used to crystallize several
proteins. In the liquid/liquid diffusion method, equilibration occurs by dif-
fusion of the crystallizing agent into the biological macromolecule volume. To
avoid rapid mixing, the less dense solution is poured very gently on the most
dense (salt in general) solution. Sometimes, the crystallizing agent is frozen
and the protein layered above to avoid rapid mixing.
One generally uses tubes of small inner diameter in which convection is
reduced. This could be achieved more easily by using gels as described in
Chapter 6. It follows the same diffusion method without the inconvenience of
the metastability of two liquids sitting on the top of each other. This method
gained new attention because of microgravity experiments (see Chapter 6).

7. Correlations with solubility diagrams

Even if it is not possible to determine the solubility (or phase) diagram for
each biological macromolecule, it is important to understand the correlation
between solubility diagrams and the method used to reach supersaturation
and crystallization, using schematic diagrams (for more details see
Chapter 10).

7.1 Dialysis
In the case of dialysis buttons, if one considers that stretching of the
membrane is negligible, the macromolecule concentration will remain con-
stant. However, if the macromolecule solution does not fill the chamber
entirely, leaving room for air, it is no longer exactly true since the macro-
molecule concentration may vary (increase or decrease depending on the
situation). The initial concentration of the crystallizing agent in the reservoir
(this could be buffer) leaves the macromolecule in an undersaturated state.
141
 A. Ducruix and R. Giege

Figure 12. Schematic solubility diagram and correlation between macromolecule and
crystallizing agent concentrations in a crystallization experiment using a dialysis set-up.
Ci is the initial concentration of crystallizing agent and C the constant protein con-
centration. The area between the precipitation and solubility curves is the supersaturated
region where crystallization can occur. Precipitation and solubility curves can be
determined experimentally, although for the latter one crystals should be obtained first.
For more details see Chapter 10.

As shown in Figure 12, when increasing the crystallizing agent concentration,

the supersaturation state is reached after passing through point S which is the
equilibrium point on the solubility curve. Then, depending on the final
crystallizing agent concentration, it will crystallize or precipitate.

7.2 Vapour diffusion

In a classical case when the concentration of crystallizing agent in the reservoir
is twice the one in the drop, the protein will start to concentrate from an
undersaturated state A (at concentration Ci) to reach a supersaturated one B
(at concentration Cf) with both protein and crystallizing concentrations
increasing by a factor two. Two hypothetical cases are represented in Figure
13a and 13b corresponding to experiments not leading (Figure 13d) or leading
(Figure 13b) to crystals. Since no crystals are obtained in Figure 13a, the
equilibrium at point B will be located in the metastable region; when the first
crystals appear (at the break of the arrow) the trajectory of equilibration is
more complex. In that case the remaining concentration of protein in solution
will converge towards point C located on the solubility curve.

7.3 Batch crystallization

In batch crystallization using a closed vessel, three cases can be considered as
shown in Figure 14. If the protein concentration is such that the solution is
undersaturated (point A), crystallization will never occur (unless another
parameter such as temperature is varied). The protein concentration may
belong to the supersaturated region between solubility and precipitation
142
 5: Methods of crystallization

Figure 13. Schematic solubility diagram and correlation between macromolecule and
crystallizing agent concentrations in crystallization experiments using vapour diffusion
set-ups. Situation without (a) and with (b) crystallization. See legend to Figure 12 and text
for further explanations.

Figure 14. Schematic solubility diagram and correlation between macromolecule and
crystallizing agent concentrations in crystallization experiments using a batch method (in
closed vessels). See legend to Figure 12 and text for further explanations.

curves (point B). In that case the arrow describes the variation of the
remaining concentration of protein in solution. In the last case (point C) the
protein will precipitate immediately because supersaturation is too high. In
some cases, however, crystals may grow from the precipitates (41).

8. Practising crystallization
It is a good exercise to train oneself with a cheap easy accessible protein.
Lysozyme, thaumatin, thermolysin, and BPTI are good candidates which are
commercially available from various manufacturers and which crystallize
readily. Examples of crystallization using various methods in hanging drops
are given in Protocols 9 and 10.
143
 A. Ducruix and R. Giege

Protocol 9. Testing lysozyme crystallization using hanging drops

Equipment and reagents

• Linbro box • 40 mg/ml lysozyme in 50 mM acetate pH 4.5
• 50 mM sodium acetate pH 4.5 • 0.22 um microfilter
• 3 M NaCI • Coverslips

Method
1. Prepare stock solutions of 3 M NaCI and 40 mg/ml (2.74 mM)
lysozyme in 50 mM acetate pH 4.5 and buffer stock solution (50 mM
sodium acetate pH 4.5). Filter all solutions with a 0.22 um microfilter.
2. Prepare a Linbro box as described in Section 4.2.1.
3. Fill up reservoirs of row A with solutions of NaCI ranging from
0.5-1.5 M in steps of 0.2 M.
4. On a coverslip, mix 4 ul of protein stock solution with 4 ul of
reservoir. Flip it and set it on the greased rim.
5. Fill up reservoirs of row B with solutions of NaCI ranging from
0.8-1.8 M in steps of 0.2 M. Repeat the experiment on row B after
diluting the protein stock solution by a factor two to obtain a new one
of 20 mg/ml (1.37 mM).
6. Fill up reservoirs of row C with solutions of NaCI ranging from 1.5-2.5
M in steps of 0.2 M. Repeat the experiment after diluting the protein
stock solution by a factor two.
7. Use row D for duplicate or testing particular parameters (e.g. volume
of drops to see the influence of kinetic effects on growth).
8. Store the experiments at 18°C.
9. Observe the experiments once a day for a week.
10. Train yourself in mounting crystals (see Chapter 14).

Protocol 10. Testing thaumatin crystallization using a dialysis

button

Equipment and reagents

• Dialysis membrane • Linbro box
• Dialysis button • Stock solutions (see Protocol 9)

Method
1. Prepare stock solutions as in Protocol 9.
2. Fill the protein chamber of a dialysis button with protein stock solution

144
 5: Methods of crystallization
diluted to 10 mg/ml. The solution must form a dome above the entry.
Install the dialysis membrane of appropriate cut-off as described in
Section 3.2.2.
3. Fill up with 1 ml of reservoir solution the reservoir of a Linbro box and
drop the button in it with the aperture of the protein chamber on the
top.
4. Store the experiments at 18°C.
5. Increase the concentration in the reservoir by 200 mM of crystallizing
agent every day.
6. Observe the next day.
7. Repeat steps 5 and 6 until crystallization occurs.

9. Concluding remarks
In this chapter we have described the most common crystallization methods;
all of them have advantages and drawbacks. Most of crystallographers favour
vapour phase diffusion which provides an easy way to practise crystallization.
It is also the method of choice for robotics (51). Dialysis presents the
advantage that macromolecular concentrations remain constant, so that only
one parameter varies at a time and nature of buffer or crystallizing agent can
be changed easily. It differs with a classical vapour phase equilibrium crystal-
lization experiment where all constituents in the drop are concentrated.
Beside the classical and advanced crystallization methods described in this
chapter, less standard methods may be useful in particular cases. These can be
methods based on old ideas not yet well explored, as pulse-diffusion of pre-
cipitant combining dialysis and free diffusion in capillaries (52), or combina-
tions of dialysis and electrophoresis (53). Also, crystallization in particular
environments should be considered, such as under levitation (54), in centri-
fuges (55, 56), or in magnetic (57, 58) or electric (59) fields. Particular
attention should be given to crystallization methods where convection is
reduced; e.g. the gel acupuncture method, crystallization in gels which is
becoming popular in the macromolecule field, and crystallization in micro-
gravity (Chapter 6). Methods based on temperature diffusion, which are
widely used in material sciences (60), may be adapted under certain con-
ditions for macromolecule crystallization (61, 62). Finally, the use of novel
types of crystallization cells may represent an interesting alternative for
growing better crystals. In particular, cells based on principles developed for
microgravity experiments may be appropriate (see refs 63-65). It is our hope
that the methods and ideas discussed in this chapter will help readers, not only
to solve their crystallization problems, but also to improve existing methods,
and even to develop new crystallization methodologies.
145
 A. Ducruix and R. Giege

References
1. McPherson, A. (1985). In Methods in enzymology, (eds Wickoff, H. W., Hirs,
C. H., and Timasheff, S. N.), Academic Press, London. Vol. 114, p. 112.
2. Giege, R. (1987). In Crystallography in molecular biology (ed. D. Moras,
J. Drenth, B. Strandberg, D. Suck, and K. Wilson), NATO ASI Series, Vol. 126,
pp. 15-26. Plenum Press, New York and London.
3. McPherson, A. (1990). Eur. J. Biochem., 189, 1.
4. Perrin, D. D. and Dempsey, B. (1974). Buffer for pH and metal ion control.
Chapman and Hall Ltd., London and New York.
5. Jurnak, F. (1986). J. Cryst. Growth, 76, 577.
6. Ray, W. J. and Puvathingal, J. (1985). Anal. Biochem., 146, 307.
7. Chayen, N., Akins, J., Campbell-Smith, S., and Blow, D. M. (1988). /. Cryst.
Growth, 90, 112.
8. Winterbourne, D. J. (1986). Biochem. Soc. Trans., 14, 1179.
9. Mikol, V. and Giege, R. (1989). J. Cryst. Growth, 97, 324.
10. Bradford, M. M. (1976). Anal. Biochem., 72, 248.
11. Zeppenzauer, M. (1971). Methods in enzymology, ibid ref. 1. Vol. 22, p. 253.
12. Weber, B. H. and Goodkin, P. E. (1970). Arch. Biochem. Biophys., 141, 489.
13. Pronk, S. E., Hofstra, H., Groendijk, H., Kingma, J., Swarte, M. B. A., Dorner, F.,
et al. (1985). J. Biol. Chem., 260, 13580.
14. Thomas, D. H., Rob, A., and Rice, D. W. (1989). Protein Eng., 2, 489.
15. Hampel, A., Labananskas, M., Conners, P. G., Kirkegard, L., RajBhandary, U. L.,
Sigler, P. B., et al. (1968). Science, 162, 1384.
16. Giege, R., Moras, D., and Thierry, J.-C. (1977). J. Mol. Biol., 115, 91.
17. Richard, B., Bonnete, F., Dym. O., and Zaccai, G. (1995). In Archaea, a laboratory
manual, pp. 149-54. Cold Spring Harbor Laboratory Press.
18. Jeruzalmi, D. and Steitz, T. A. (1997). J. Mol. Biol., 274, 748.
19. Zuk, W. M. and Ward, K. B. (1991). J. Cryst. Growth, 110, 148.
20. Kim, S.-H. and Quigley, G. (1979). In Methods in enzymology, Vol. 59, p. 3.
21. Dock, A.-C., Lorber, B., Moras, D., Pixa, G., Thierry, J.-C., and Giege, R. (1984).
Biochimie, 66, 179.
22. Luft, J. R., Cody, V., and DeTitta, G. T. (1992). J. Cryst. Growth, 122, 181.
23. Harlos, K. (1992). J. Appl. Crystallogr., 25, 536.
24. Phillips, G. N., Jr. (1985). Methods in enzymology, ibid ref. 1. Vol. 104, p. 128.
25. Luft, J. R. and Cody, V. (1989). J. Appl. Crystallogr., 22, 386.
26. Yonath, A., Mussig, J., and Wittmann, H. G. (1982). J. Cell. Biochem., 19, 145.
27. McPherson, A. (1985). In Methods in enzymology, Vol. 114, p. 125.
28. McPherson, A. (1995). J. Appl. Crystallogr., 28, 362.
29. Mikol, V., Rodeau, J.-L., and Giege, R. (1989). /. Appl. Crystallogr., 22, 155.
30. McPherson, A. and Shlichta, P. (1988). Science, 239, 385.
31. Chayen, N. E., Lloyd, L. F., Collyer, C. A., and Blow, D. M. (1989). J. Cryst.
Growth, 97, 367.
32. Mikol, V., Rodeau, J.-L., and Giege, R. (1990). Anal. Biochem., 186, 332.
33. Luft, J. R., Arakali, S. V., Kirisits, M. J., Kalenik, J., Wawzzak, I., Cody, V., et al.
(1994). /. Appl. Crystallogr., 27, 443.
34. Arakali, S. V, Easley, S., Luft, J. R., and DeTitta, G. T. (1994). Acta Cryst., D50, 472.

146
 5: Methods of crystallization
35. Luft, J. R., Albrigth, D. T., Baird, J. K., and DeTitta, G. T. (1996). Acta Cryst.,
D52, 1098.
36. Fowlis, W. W., DeLucas, L. J., Twigg, P. J., Howard, S. B., Meehan, E. J., and
Baird, J. K. (1988). J. Cryst. Growth, 90, 117.
37. Chayen, N. E. (1997). J. Appl. Crystallogr., 30, 198.
38. McPherson, A. (1976). J. Biol. Chem., 251, 6300.
39. Gernert, K. M., Smith, R., and Carter, D. C. (1988). Anal. Biochem., 168, 141.
40. Przybylska, M. (1989). J. Appl. Crystallogr., 22, 115.
41. Ng, J., Lorber, B., Witz, J., Theobald-Dietrich, A., Kern, D., and Giege, R. (1996).
J. Cryst. Growth, 168, 50.
42. Chayen, N. E., Shaw Stewart, P. D., Maeder, D. L., and Blow, D. M. (1990).
J. Appl. Crystallogr., 23, 297.
43. d'Arcy, A., Elmore, C., Stihle, M., and Johnston, J. E. (1996). J. Cryst. Growth,
168, 175.
44. Jakoby, W. B. (1971). In Methods in enzymology, Vol. 22, p. 248.
45. Waller, J. P., Risler, J.-L., Monteilhet, C., and Zelwer, C. (1971). FEBS Lett., 16,
186.
46. Visuri, K., Kaipainen, E., Kivimaki, J., Niemi, H., Leissla, M., and Palosaari, S.
(1990). Biotechnology, 547.
47. Lorber, B., Jenner, G., and Giege, R. (1996). J. Cryst. Growth, 158, 103.
48. Lorber, B. and Giege, R. (1996). J. Cryst. Growth, 168, 204.
49. Chayen, N. E. (1996). Protein Eng., 9, 927.
50. Salemne, F. R. (1972). Arch. Biochem. Biophys., 151, 533.
51. Zuk, W. M. and Ward, K. B. (1991). J. Cryst. Growth, 110, 148.
52. Koeppe, R. E., Stroud, R. M., Pena, V. A., and Santi, D. V. (1975). J. Mol. Biol.,
98, 155.
53. Chin, C.-C., Dence, J. B., and Warren, J. C. (1976). J. Biol. Chem., 251, 3700.
54. Rhim, W.-K. and Chung, S. K. (1990). Methods: a companion to methods in
enzymology, 1, 118.
55. Barynin, V. V. and Melik-Adamyan, V. R. (1982). Sov. Phys. Crystallogr., 27, 588.
56. Lenhoff, A. M., Pjura, P. E., Dilmore, J. G., and Godlewski, T. S., Jr. (1997).
J. Cryst. Growth, 180, 113.
57. Sazaki, G., Yoshida, E., Komatsu, H., Nakada, T., Miyashita, S., and Watanabe,
K. (1997). J. Cryst. Growth, 173, 231.
58. Ataka, M., Katoh, E., and Wakayama, N. I. (1997). J. Cryst. Growth, 173, 592.
59. Taleb, M., Didierjean, C., Jelsch, C., Mangeot, J.-P., Capelle, B., and Aubry, A.
(1998). J. Cryst. Growth,
60. Feigelson, R. S. (1988). J. Cryst. Growth, 90, 1.
61. Lorber, B. and Giege, R. (1992). /. Cryst. Growth, 122, 168.
62. DeMattei, R. C. and Feigelson, R. S. (1993). /. Cryst. Growth, 128, 1225.
63. Stoddard, B. L., Strong, R. K., and Farber, G. K. (1988). J. Cryst. Growth, 110,
312.
64. Bosch, R., Lautenschlager, P., Potthast, L., and Stapelmann, J. (1992). J. Cryst.
Growth, 122, 310.
65. McPherson, A. (1996). Crystallogr. Rev., 6, 157.

147
This page intentionally left blank
 6

Crystallization in gels and related

methods
M.-C. ROBERT, O. VIDAL, J.-M. GARCIA-RUIZ, and F. OTALORA

1. Introduction
From the first studies showing the feasibility of macromolecular crystal
growth in gels (1), an increasing attention has been paid to applications of gel
techniques to the domain of biological macromolecules. Confidence in these
techniques is such that kits of crystallization in gels are now commercially
available (Hampton Research, Laguna Hills, CA, USA).
Basically, the protein crystallization process consists of two consecutive
steps:
• first, the transport of growth units towards the surface of the crystals
• second, the incorporation of the growth units into a crystal surface position
of high bond strength.
The whole growth process is dominated by the slowest of these two steps
and is either transport controlled or surface controlled. Avoiding convection
in the growth environment will increase the possibility of growing the crystal
under slow diffusive mass transport providing that the surface interaction
kinetics are faster than the characteristic diffusive flow of macromolecules (in
the range of 10-6 cm2/sec for proteins). The ratio between transport to surface
kinetics, which can be tuned by either enhancing or reducing transport pro-
cesses in the solution, has been shown (2) to control the amplitude of growth
rate fluctuations (which is thought to reduce crystal quality). These are the
main reasons why gels (as well as capillaries and microgravity conducted
experiments), if correctly designed, are expected to enhance the quality of
crystals. This quality enhancement (3), as well as the possibility of getting
crystals when conventional solution techniques failed (4), have been experi-
mentally demonstrated. However, up to now, gel methods have been used on
a rather empirical basis, as a simple transposition of solution techniques, and
recent fundamental studies of nucleation and growth in gels show that the
situation is not as simple as first expected (5, 6).
After summarizing the main characteristics of crystal growth in gels, we will
 M.-C. Robert et al.
examine what are the best conditions using a gel method. Recipes for the
preparation of different gel growth experiments will be given. Considering gel
growth as a possible simulation of experiments under reduced gravity, recent
results of space experiments will be reviewed. Mention will also be made to
growth under hypergravity conditions.

2. General considerations
Gels used for crystal growth are hydrogels with a growth solution soaking a
polymeric network. For physical gels like gelatin or agarose, sol-gel transition
is obtained by decreasing the temperature (physical parameter variation).
Polymerization corresponds to the formation of weak bonds and this process
is reversible with some hysteresis (~ 50°C for agarose). For chemical gels,
such as polyacrylamide, polymerization corresponds to strong bonding and is
not reversible. Although formation of silica gels also results from a chemical
reaction, it rather corresponds to an intermediate case between chemical and
physical gels: as a matter of fact, the chemical reaction leads to the formation
of dense beads which further aggregate by weak bonding (7). As far as we
know, only agarose and silica gels have been successfully used for macro-
molecule crystal growth.

2.1 Formation and structure of gels

2.1.1 Agarose gels
Agarose is a polysaccharide extracted from seaweed. Its basic repeat unit is
agarobiose (Figure la) with different substituents like O-methyl or O-sulfate
groups which vary according to the agarose origin and subsequent chemical
treatment. The sol-gel transition temperature varies according to the nature
and content of these different substituents so that one can find commercially
available (e.g. Sigma) agaroses in the 15-40°C gelling temperature range.
The gelling process is not yet fully understood; it has been thought for a long
time that polysaccharide chains first associate to form double helices, then
aggregate to form fibres. Recent studies (8) contradict the existence of double
helices but suggest that agarose strands form ordered lateral associations inter-
connected through disordered junction zones. (Figure 1b). In any case, such
associations leave large voids through which very large molecules can migrate.
That is, for 0.6% (w/v) agarose gels, the pore size is around 7.000 A, and for gels
less than 0.3% (w/v), their average size is larger than 10.000 A. Indeed, gel
media are largely used for electrophoresis of biological macromolecules.
2.1.2 Silica gels
Silica gels result from the polycondensation of silicic acid which occurs either
by neutralization of sodium metasilicate:

150
 6: Crystallization in gels and related methods

Figure 1. (a) Basic repeat unit of agarose chains: the agarobiose. (b) Schematic structure
of the agarose gel showing the disordered junction zones.

or by hydrolysis of a siloxane, like tetramethoxysilane (TMOS) (or tetra-

ethoxysilane, TEOS):

The waste products of the reactions are, in the first case a sodium salt and in
the second case an alcohol (if necessary, the waste products can be washed
off, after gel setting).
The silicic acid tends to polymerize according to the reaction:

Polymerization proceeds via formation of rings to which monomers add to

form dense particles (silica beads) leaving OH groups outside (Figure 2a).
The size and charge of these particles are pH-dependent. Above the iso-
electric point (at pH 2.0), the particles are negatively charged. Gelation easily
occurs in the pH 2-7 range; according to the pH value and the salt content, the
silica beads either grow (Figure 2b) or aggregate (Figure 2c and 2d) to form
branched chains and then a 3D network. This process is accelerated by
increasing the temperature. A detailed description of the whole process is
given in . 7.
151
 M.-C. Robert et al.

Figure 2. (a) The polymerization of the silicic acid gives rise to colloidal silica beads
whose internal structure is due to siloxane bonding, leaving silanol groups on the
surface, (b) The growth of beads depends on the chemical conditions: (c) network
obtained at high pH, (d) network obtained at low pH.

2.2 Gel properties related to crystal growth

2.2.1 Diffusion of species
Entrapping a growth solution by a gel network prevents the onset of con-
vection which is unavoidable on earth in ungelled solutions: a growing crystal
is surrounded by a depletion zone which is less dense than the bulk so that
density driven convection shifts the lighter layers upwards. Such movements
are not possible in a gel network where mass transfer only proceeds by
diffusion. The diffusion coefficients of macromolecules like lysozyme in light
gels (i.e. 0.2% (w/v) agarose or 0.4% (w/v) silica) are not significantly changed
with respect to diffusion in a gel-free medium. This is not the case for species
the dimensions of which come close to the pore size of the gel network.
Furthermore an adsorption process can be superimposed to the diffusion
process when the diffusing species interacts with the gel.
The diffusion properties of the different gel media are illustrated on
interferograms taken during growth of lysozyme crystals in an agarose gel
(Figure 3a) or a silica gel (Figure 3b) (9). Mass transfer proceeds by diffusion
from the solution to the crystals and the fringe patterns allow the solute
152
 6: Crystallization in gels and related methods

Figure 3. Interferograms taken during growth of lysozyme tetragonal crystals (450

microns) (a) after 24 h in a 0.15% (w/v) agarose gel and (b) after 48 h in a 0.81% (w/v)
silica gel. The conditions are the following: 25 rng/ml of HEWL with 0.4 M NaCI in 0.1 M
acetate buffer solution (pH 4,5) at T -. 17°C for the agarose gel case and T = 11°C for the
silica gel case.

concentration profile around each crystal to he calculated. In silica gels, there

are fewer fringes: growth of crystals of similar size does not cause the same
concentration decrease. Indeed, by decreasing supersaturation, molecules
fixed on the silica gel progressively desorb which limits the solution depletion.
Application of diffusion properties are developed in Section 3,2.

2.2.2 Suspension of crystals

Crystals growing in gel do not sediment as they do in free solution; they
develop at the nucleation site, sustained by the gel network. For small
molecule crystals grown in silica gel, the gel often fissures and forms cusp-like
cavities around crystals (10), and a thin liquid film, that reduces contamin-
ation risk, separates the crystal from the gel. Such cavities have not been seen
in macromolecule crystals.
Recent studies have even shown that silica gel can be incorporated in the
crystal network ( 1 1 ) almost without disturbing the crystal lattice (Figure 4).
Such crystals, that still diffract to high resolution, are mechanically reinforced
(they can be manipulated with tweezers or even with the fingers) and are
more resistant to dehydration, because the silica gel framework embedded in
the crystal lattice slows down water loss due to its hygroscopic properties.
This can be beneficial for X-ray diffraction and crystal growth experiments
(Figure 5) as well as for the future use of protein crystals as a material for
technological applications.
153
 M.-C. Robert et al.

Figure 4. Rocking curve and topography (inset) acquired from a tetragonal HEW lysozyrne
crystal grown in high concentration gel. Although the gel network is embedded into the
crystal lattice, the crystal quality is preserved.

High resolution diffraction has been observed for crystals grown from
rather firm agarose gels (12). However the use of light gels is advisable, except
for the special cases discussed above, first to make removal of crystals out of
the gel matrix easier, and secondly to minimize the gel contamination.
2.2.3 Nucleation inside the gel
Although seeding can be used, it appears that most of the gel-grown crystals
arc obtained by spontaneous nucleatkm inside a macroscopically homo-
geneous gel. When the gel well adheres to the walls of the container (without
intercalated liquid film), no nucleation occurs on the cell walls, neither on dust
or fibres which have been embedded in the gel. So heterogeneous nucleation
is strongly reduced, if not suppressed. Another type of nucleation, namely
secondary nucleation, is due to attrition of a previous crystal by the solution
flux. It is quite clear that, in gels, this type of nucleation is prevented. One can
154
 6: Crystallization in gels and related methods

Figure 5, Growth of a tetragonal HEW lysozyme followed by interferometry. This experi-

ment, performed inside a FID APCF reactor (protein chamber is 5 mm wide), was
intended to study the depletion zone around a growing crystal. In this experiment, the
presence of overlapping depletion zones around several small crystals degrades the data
quality, so we implemented a seeding set-up using a large reinforced crystal grown in
high concentration silica gel as seed. The seed crystal was glued to one of the walls of the
reactor, benefiting from their improved mechanical properties. This image corresponds
to a preliminary on-ground experiment during the preparation of the mission. The
depletion zone around the crystal in on-ground experiments was not observed because
of the homogenization of the solutions inside the reactor by buoyancy convection. To
solve this problem, low concentration agarose gel was set inside the growth chamber
after gluing the seed. Using this assortment of gel techniques we were finally able to
follow on-ground the evolution of a large diffusive depletion zone around an isolated
large protein crystal. The inset at left shows a picture of the experiment in which the seed
(dark part inside the crystal) and the overgrown volume Iclear part around the seed) can
be clearly seen.

even take advantage of llie absence of convection to apply feeding techniques

(13): a simple one consists of putting on the gelled droplet containing a
growing crystal a small droplet of protein solution at a concentration higher
than inside the droplet. In solution, such a procedure gives rise to a shower of
secondary nuclei which is not the case in gels where a unique crystal keeps
growing.
When nuclealion occurs inside the gel, one observes thai all the crystals
appear at the same time and consequently have about the same size; they are
homogeneously distributed in the whole volume (Figure 6). This is not the
155
 M.-C. Robert ct al.

Figure 6, Example of a nucleation of tetragonal lysozyme crystals in agarose gel; the

crystals are homogeneously distributed. The capillary diameter is 2 mm.

case in gel-free solutions. One generally observes that, in silica gels, the
number of nucleated crystals decreases by increasing the gel content. This has
been shown with a large variety of macromoleeules (14). In agarose gels, on
the contrary, the number of crystals increases by increasing the gel content
(15.16).
The influence of the gel media on the nucleation rate cannot be explained
by changes of solubility values as far as a solubility curve relates equilibrium
between a crystalline phase and a solution. However, supersaturation can be
lowered as observed in silica gels where protein molecules ean adsorb on the
gel surface. Small angle neutron scattering (SANS) spectra of proteins,
differing markedly from those corresponding to the gel-free solution, account
for this effect. So, with HEW lysozyme solutions at pH 4,5, part of the protein
molecules are adsorbed on the gel through electrostatic and H-bond inter-
actions, which reduces the content of free molecules remaining in solution.
The concentration of protein adsorbed increases by increasing the total
protein content (until binding sites on the gel surface are covered) and by
increasing the supersaturation (through protein-protein interaction) (see e.g.
Figure 5 in ref. 6). The latter process is reversible. Thus, reduction of the
nucleation rates is simply explained by a reduction of the actual super-
saturation. One can counterbalance this effect by using mixed silica gels
(TMOS and MeTEOS) (Section 3.1.1), which increases the nucleation rate.
In agarose gels, the initial free protein concentration is the same as in gel-
free solution. Differences in the SANS signals of gelled and gel-free HEW
156
 6: Crystallization in gels and related methods
lysozyme solution are only visible in the very small scattering vector range.
These signals are related to the presence of aggregates (in the 100 nm range):
when the protein solution is trapped in agarose gel the signal is enhanced (5).
For macromolecular crystals, it is assumed that nucleation could occur via
restructuring of amorphous aggregates (17, 18). Here the concomitant observ-
ations of enhanced aggregation and enhanced nucleation, both increasing
with the gel content, supports this hypothesis. As large aggregates cannot
sediment in gel, they are maintained in the whole bulk as shown by refractive
index measurements as a function of time during the nucleation process (19).
An opposite effect is visible in solution under normal gravity or hypergravity
conditions (Section 5.2).
From a practical point of view, use of either silica or agarose gels is inter-
esting because situations exist where, in solution, nucleation is either too
abundant or too scarce.
2.2.4 Parameters influenced by the presence of a gel
Taking into account the above considerations, one can select from the list of
parameters influencing biological macromolecule crystal growth (see Chapter 1,
Table 7) those which are influenced by the presence of a gel structure. They
are parameters either related to the supply of reactants or related to the
mechanical behaviour of the solid or liquid phases.
A priori, gels are not expected to improve crystal growth with regards to
biochemical parameters like purity or degree of denaturation of the macro-
molecule. However, this assertion is refuted by a comparative study (in gel
and in solution) on the effect of contamination by a parent molecule (20, 21).
Indeed, it was shown that crystals of good quality can be obtained, even when
contamination levels are much higher in the gel than in free solution. In
current crystallization conditions, impurities are either rejected or incorpor-
ated in the crystal. Consequently, the impurity concentration nearby the
interface increases (or decreases) with respect to the concentration in the
bulk. In free solution, on earth, convections provoke fluctuations of the im-
purity content at the interface resulting in time-dependent incorporation of
impurity in the crystal (growth striations). In gels, due to the supply of solute
by diffusion, such fluctuations are damped; furthermore, with a slow growth,
foreign molecules or molecules having a distorted conformation can be
rejected from the growth interface instead of being buried in the crystal
network. In silica gels, the adsorption-desorption process could also act as a
purification process, assuming that ill-folded molecules are more strongly
bound to the gel surface.

3. Practical consideration
Gel growth is a particular case of solution growth so that it must always be
considered downstream with respect to classical solution growth. It results
157
 M.-C. Robert et al.
that the same chemical components would be chosen among those which have
given the best results in solution. Then using gels, one tries to offset the
drawbacks encountered during these first trials. A first possibility is to
nucleate and grow the crystals inside the gel. This implies that the protein
solution is either gelified or brought into a gel previously set. The different
procedures are detailed in Section 3.2.1. A second possibility is to use the gel
as a diffusion medium to monitor the supply of reactants, the crystal growing
outside the gel. This technique, known as gel acupuncture method, will be
described in Section 3.3.1.

3.1 Gel preparation

3.1.1 Silica gel
i. Preparation of silica sols
Silica sols can be prepared either from hydrolysis of a siloxane or by
neutralization of sodium metasilicate.

Protocol 1. Preparation of 2% (w/v) silica sol

Equipment and reagents

• Tetramethoxysilane (TMOS) • 2 M HCI solution
• Methyltriethoxysilane (MeTEOS) • pH meter
• 0.5 M sodium metasilicate solution • Magnetic stirrer

A. Hydrolysis of siloxane
1. Tetramethoxysilane or tetraethoxysilane are liquids very soluble in
alcohol but not much in water. Add drop by drop 1 ml TMOS to 20 ml
buffer solution (e.g. 0.1 M acetate buffer solution). Dissolution occurs
through a vigorous stirring of siloxane droplets in water (Figure 7a);
this emulsifying provides a large contact surface, which allows the
dissolution of a small amount of siloxane. The reaction proceeds
according to Equation 2 (Section 2.1.2) and methanol is progressively
released which makes easy the complete siloxane dissolution. This
process consumes 12.5 ml water and releases 54.5 ml methanol per
litre of solution. This must be taken into account to know the final
growth solution composition.
2. The homogenization step must be achieved as quickly as possible
because it competes with the polymerization reaction step, which
begins as soon as monomers are available in the medium. It is
possible to delay the polymerization step by keeping the mixture in a
water/ice bath (Figure 7b).
3. The lower the pH, the more rapid is homogenization.
4. The mixture first looks like an oily emulsion in water, then, when no
parasitic reaction occurs, it becomes clear and homogeneous.
158
 6: Crystallization in gels and related methods
5. One can reduce the interactions between protein and silica gel by
adding some amount (~ 30% in weight) of methyltriethoxysilane,
MeTEOS, (C2H5O)3Si-CH3 (for which the hydrolysis reaction is similar)
to the TMOS so that Si-CH3 groups are substituted to Si-OH groups on
the gel surface (6).
Siloxanes are corrosive liquids. Careful protection of skin and eyes are
recommended. All vessels in contact with them must be thoroughly
rinsed with alcohol prior to water cleaning.

B. Neutralization of sodium metasilicate

1. Fill a burette with 0.5 M sodium metasilicate set over a container
containing 2 M HCI. Add drop by drop the metasilicate up to the
desired pH (control the pH with a pH meter or with some coloured
indicator on a test sample). A careful and rapid cleaning of the pH
electrode is recommended to avoid plugging by silica.
2. Under these conditions, the neutralization reaction releases NaCI at a
concentration of 0.66 M.
3. If other Na salts are preferred as crystallizing agents, e.g. nitrate, the
corresponding acid must be used for neutralization in the following step.

ii. Polymerization
After having added an aliquot of the silica sol to the solution at the required
composition, the preparation is thoroughly mixed (Figure 7c). It must remain
homogeneous (no flocculation). The mixture is poured in clean, dried
crystallization containers and allowed to gelify without mechanical disturb-
ances. The gel must stick to the container cell walls. It can look somewhat
opalescent, but without macroscopic heterogeneities such as fissures.
Dehydration of the gel surfaces must be avoided, either by sealing them with
a minimum air volume enclosed or by closing them in a vessel containing a
reservoir of solution giving the suitable vapour pressure (Figure 7d).
Gelation time depends on many parameters such as concentration of gelling
agent, nature and concentration of species in solution, pH, and temperature.
So, at room temperature it can vary from a few minutes at pH 7.0 to several
hours at pH 4.0. With thermally stable solutions, one can shorten this time by
increasing temperature (typically 40°C for further use at room temperature).
The gel can be considered as set when it resists pouring, though it undergoes
some further evolution as shown by light scattering techniques.

3.1.2 Agarosegel
Commercially available agaroses are powders which can be dissolved in water
as described in Protocol 2. Homogeneous preparations at concentrations
higher than 2% (w/v) are difficult to achieve.
159
 M.-C. Rnhnrt et al.

Figure 7. Two-step procedure for crystal growth in a CG silica gel. After setting the silica
gel at T - 40°C, the protein solution diffuses into the gel. Then, the crystal growth occurs
under the final concentrations of protein CP, salt Cs, and buffer CB/2.

160
 6: Crystallization in gels and related methods

Protocol 2. Preparation of a 1% (w/v) agarose sol

Equipment and reagents

• Agarose powder • Water-bath
• Magnetic stirrer

Method
1. Add progressively 0.1 g agarose to 10 ml water at room temperature
(not the reverse which could leads to agglomerates) with a slow
stirring (Figure 8a).
2. Keep stirring for a couple of hours at ambient temperature.
3. Raise the temperature to 100°C (possibly in a water-bath to avoid
overheating) and maintain the slow stirring. The mixture must rapidly
become as clear and transparent as water (Figure 8b).
4. Keep it on the hot plate (80°C) until use.

3.2 Gel methods

3.2.1 Crystallization inside the gel: batch method
In order to gelify a protein solution one can either add the protein solution to
the agarose (or silica) sol and then carry out the sol-gel processing (one-step
procedure) or gelify a protein-free sol and then bring the protein into the gel
by diffusion (two-step procedure). The two procedures are equivalent with
agarose gels, because there are no significant interactions between protein
and agarose as shown by small angle neutron scattering (SANS) (5). Here, the
presence of protein does not perturb the gelation process. This is not the case
for silica gels, because protein molecules may adsorb on the gel surface
(Section 2.2.3). In that case, all components (protein, salt) and the pH of the
growth solution influence the gelation process, and therefore the two-step
procedure is advised.

i. Case of silica gels

Due to the strong interactions between protein and silica gel in the pre-
sence of salt, the protein solution will preferably be brought into a gel already
set.

Protocol 3. Preparation of growth solution with silica

Equipment and reagents

• Glass capillaries • Water-bath
• Silica sol (see Protocol 1)

161
 M.-C. Robert et al.

Protocol 3. Continued
Method
1. The silica sol is prepared as explained in Protocol 1 in a buffer at
concentration CB. It is kept in a water/ice bath to avoid a premature
gelation. The different growth settings are presented Figure 7.
2. Prepare a growth solution with a protein concentrations CP, salt
concentration Cs, gel concentration CG.
3. Mix equal volumes of stock salt solution (at concentration 2 Cs) and
silica sol (at concentration 2 CG) (Figure 7c).
4. Suck a few microlitres of this preparation in capillaries (or prepare
droplets of this preparation on coverslips). Place the capillaries
opened at one end (or the coverslip) in a closed vessel containing a
reservoir of salt at concentration Cs to avoid dehydration during
gelling. Gelling can be accelerated by setting the closed vessel in an
incubator at 40°C (Figure 7d).
5. (a) For hanging drops. When the gel is set, pour carefully a volume V
of protein solution at concentration Cp on the gel surface taking
care not to touch it with the pipette (Figure 7e).
(b) For capillaries. Set again the capillaries (or droplets) in the closed
vessel. Diffusion of protein in the gel matrix and dehydration of
the liquid droplet occurs simultaneously so that the final result is a
gelled droplet with buffer, salt, and protein at concentrations CB,
Cs, and Cp respectively.
6. Use the droplets as in usual hanging drop or sitting drop techniques.
Seal the capillaries and proceed as for a batch technique, i.e. by
keeping them at constant temperature. One can also apply a regulated
temperature variation to increase supersaturation.

ii. Case of agarose gel

In the protocol presented on Figure 8, the protein is added to the agarose sol
before gelation; so, the different components of the sol are kept at a
temperature above the gelling point TG before use. To prepare a growth
solution containing gel, protein, salt, and buffer at final concentrations,
respectively CG, CP, Cs, CB, one starts from (warm) stock solutions (Figure 8c)
whose compositions are:

gel: 10 CG
protein: 2 Cp and CB
salt: 4 Cs
buffer: 5 CB
162
 6: Crystalization in gels and related methods

Figure 8. One-step procedure for crystal growth in a CG agarose gel. The volume
proportions are given in order to have final concentrations of protein CP, salt CE, and
buffer CB, For calculating Xg, take Cg in %. Agarose sol uptake needs reverse pipetting.

163
 M.-C. Robert et al.

Protocol 4. Preparation of growth solution with agarose

Equipment and reagents

• Thermostated water-bath • Agarose sol

Method
1. Prepare a crystallizing agent/agarose solution by thoroughly mixing
salt, buffer, water, and gel in the proportions indicated in Figure 8d.
This solution is kept at T > TG with the protein solution (Figure 8e).
2. Sample equal volumes of protein and crystallizing agent/agarose solu-
tions in an Eppendorf tube and gently mix with a Pipetman (Figure 8ft.
3. Use this preparation as you do for classical solution growth tech-
niques (see Chapter 4).
4. Decrease the temperature under TG to allow the gel to set.

To avoid a temperature-induced denaturation of the protein, one has to

select an agarose of medium gelling point. Due to the 50°C hysteresis asso-
ciated with the gel-sol process, once the gel has set, one can use it at a
temperature higher than TG without altering the gel structure. For example,
one can grow orthorhombic HEW lysozyme crystals at 40°C with an agarose
of 36°C gelling point.
3.2.2 Counter-diffusion
All techniques currently used to grow protein single crystals (see Chapter 5)
can be implemented using gels instead of free solutions in order to minimize
convective flow and to avoid movements of the growing crystals, including
sedimentation. Among these techniques, those in which there is a continuous
change of supersaturation in space (through the growth reactor) and time
(during the experiment) are of interest. This is particularly the case when they
are forced to work out of equilibrium allowing the self-search for the best
crystallization conditions. Because of the geometry and mass transport
involved, they are termed counter-diffusion techniques.
In counter-diffusion techniques, the interacting solutions are placed one in
front of the other either in direct contact (free interface diffusion) or sep-
arated by a membrane (dialysis) or by an intermediate chamber working as a
physical buffer. By definition, counter-diffusion techniques require avoiding
convection. In addition to using gels, two other ways are known to reduce
convection: to perform the experiments under microgravity conditions in
space and/or to perform the experiment inside narrow volumes, for instance
glass capillaries. As shown in Figure 9, all the three implementations share the
same geometry. Note that the term 'buffer chamber' for the intermediate
chamber has a physical meaning, i.e. a chamber that slows down the transport
164
 6: Crystallization in gels and related methods

Figure 9. Counter-diffusion arrangements for protein crystallization using three different

implementations sharing the same geometry. In the gel acupuncture method the physical
buffer has a length two times the punctuation depth as this is the length of the diffusive
path separating salt and protein reservoirs.

165
 M.-C. Robert et al.
process, being filled either with a chemical buffer, with a porous network or
with any other fluid such as water. Batch crystallization into capillary volumes
was used by Feher et al. (22) to illustrate diffusional transport in protein
crystallization. The use of capillaries for counter-diffusion methods dates
back to the work of Zeppezauer et al. (23), and others reviewed by Phillips
(24) (see also Chapter 5). It seems that these methods using capillaries were
designed with the aim of reaching the critical supersaturation for nucleation
very slowly, looking for a single nucleation event. Attempts to make use of
Otswald ripening processes were also considered (25). Unlike these, recent
studies have tried, starting from conditions far enough from equilibrium, to
search for multiple nucleation events under conditions progressively approach-
ing equilibrium. To illustrate the method we can use a simple technical im-
plementation, the gel acupuncture technique, stressing that our discussion
also applies for counter-diffusion arrangements using gelled protein solutions
and microgravity experiments.
3.2.3 Crystallization outside the gel: gel acupuncture method
The gel acupuncture method is based upon the properties of gels, which are
used to act as the mass transport medium for the precipitating agent and also
to hold capillaries containing the ungelled protein solution. The experimental
set-up is as simple as shown in Protocol 5 (see ref. 26 for specific recipes to
crystallize several proteins).

Protocol 5. A recipe to grow crystals of lysozyme by the gel

acupuncture technique

Equipment and reagents

• Sodium silicate • 1 M acetic acid
• Agarose

Method
1. Mix the sodium silicate as commercially supplied with four parts of
water. Mix slowly under continuous stirring 12.5 ml of this solution
with 10 ml of 1 M acetic acid. For this step, you can use a 50 ml vessel.
In a few hours the silica gel is set with a pH about 5.8-6.
2. Prepare the following solutions while the gel sets down:
(a) 20% (w/v) sodium chloride: pour 4.5 g of salt into 22.5 ml of water
and stir until complete dissolution.
(b) 100 mg/ml protein solution: weigh 100 mg of lysozyme into a
small tube (e.g. an Eppendorf tube) and pour 1 ml of
water. Stir gently until complete dissolution.
3. Fill the capillaries with the protein solution. Introduce one of the ends
of the capillary into the protein solution. You will see that the solution

166
 6: Crystallization in gels and related methods
flows up by capillarity. Once it reaches a level 1 cm from the other end
of the capillary, remove it from the solution. The solution holds inside
the capillary. Then seal the upper end of the capillary with a small
piece of Plasticine. The next step is to punch the capillary in the gel
layer (but be sure that the gel is set!). Insert the capillary into the gel
about 0.5-1 cm, enough to maintain it straight.
4. Pour the solution of salt (22.5 ml) onto the gel layer and cover the
experiment with a large vessel turned upside down.

In a typical experiment, the crystallizing agent and the protein start to

counter-diffuse through the porous gel network used to hold the capillary
(Figure 10). As the diffusion constant of macromolecules is one or two orders
of magnitude smaller than that of the small molecules of the crystallizing
agent (1), the latter reaches the open side of the capillary and starts to diffuse
up into it while almost all the protein molecules are still in the capillary. This
creates a set of supersaturation conditions s(x, f) that changes in time t for
every location in the capillary x. Eventually, at a given (x, t), the critical
supersaturation value s* for nucleation is achieved and protein starts to
precipitate inside the capillary. The development of supersaturation in time
and space is rather complex and needs to be studied by numerical methods
(see ref. 27 for a computer simulation of the problem), but the basic
behaviour can be briefly outlined.
As the experiment evolves (in time), the salt diffuses up through the capil-
lary. Then the system experiences precipitation phenomena which take place
at different values of both supersaturation and rate of supersaturation testing
a wide range of plausible precipitation conditions along the capillary. The
nucleation events take place under conditions very far from equilibrium
(amorphous precipitation) at the entry of the capillary and then under con-
ditions slightly closer to equilibrium (polycrystalline precipitation) further
along the capillary because of a lower gradient of the reactants. Thus, as time
advances, the precipitation system slowly approaches equilibrium iteratively,
experiencing nucleation events, which yield successively fewer and larger
crystals. In the middle and upper part of the capillary, few large and well-
faceted single crystals form and, under optimum starting conditions, a single
crystal completely filling the capillary diameter can be obtained (28) (Figure
11). In short, as soon as the protein molecules start to precipitate, the system
behaves like a saturation (B = C/CS, see Chapter 11) wave leading to complex
precipitation events in space and time. In fact, the typical output of any
counter-diffusion arrangement is a patterned precipitation in space and
time, the archetype pattern being the so-called Liesegang rings precipitation
(10).
For practical purposes, the technique can be used to grow protein crystals
with most of the classical crystallizing agents (including PEG) (29) and it
167
 M.-C. Robert et al.

Figure 10. Protein crystallization by the gel acupuncture method. The diffusive path of
the salt towards the capillary filled with protein solution is illustrated as well as the
crystal size distribution obtained along the capillary. Picture at right shows an actual
ferritin crystallization experiment using the gel acupuncture method.

consumes reasonable amounts of protein (5-50 mg). The main variables

affecting the crystallization behaviour in the gel acupuncture method are;
• punctuation depth
• initial protein concentration
• initial crystallizing concentration,
168
 6: Crystallization in gels and related methods

Figure 11. Rod-shaped crystal of tetragonal HEW lysozyme obtained by gel acupuncture
method. The isolated crystal was grown until it completely filled the capillary and then
growth continued at both ends of the cylinder. Capillary diameter 0.3 mm. Protein
concentration 100 mg/ml. Salt concentration 10% (w/v) pH 4.5.

In the search for single crystals by this method, large protein concentrations
are recommended to start with. From the point of view of classical crystal-
lization methods, this can be surprising. However, it should be realized that in
the get acupuncture method the precipitation system itself searches for the
best crystallization conditions. Thus the idea, when using large protein con-
centrations, is to trigger the nucteation of an initial amorphous precipitate in
the lower part of the capillary and then leave the system to search for optimal
growth conditions which typically occurs in a time seale of days. The crystal-
lizing agent concentration is another important variable. A very high initial
concentration will exhaust the protein in the capillary with amorphous pre-
cipitation, while a very low initial concentration will produce a batch-type
precipitation behaviour due to the small salt gradients inside the capillary.
Thus, an intermediate concentration, in the range known to precipitate the
protein, is recommended. Finally, the punctuation depth affects the waiting
time and the overall salt gradient inside the capillary. The suggested value of
8 mm has been experimentally found to be a good compromise between
waiting time (longer for higher punctuation depth values) and mechanical
stability of the capillary.
The gel acupuncture technique has been demonstrated for proteins of
different types and diverse molecular weights. To work properly, the tech-
nique must avoid sedimentation of the crystals, as well as the buoyancy driven
convection created as soon as the protein concentration falls in the lower part
of the capillary. It is recommended to use, as the reservoir to be filled with gel.
169
 M.-C. Robert et al.
rectangular boxes made of two glass plates separated by a rubber frame hold
with clips (30, 31). With this simple arrangement, the capillaries can be
orientated perpendicular to the gravity field. Nevertheless, very large (up to
10 mm) crystals have already been obtained with this technique, showing a
high diffraction resolution limit and very low mosaicity (1.2 A and 9 arc
second, respectively, for tetragonal HEW lysozyme crystals) (32, 33). These
results are expected to be enhanced in future with in situ measurements
during the growth process, ensuring mechanical stability (34). Finally, use of
the technique for preparation of heavy-atom derivatives and search of crystal-
lization conditions by pH variation (a method advocated by McPherson) (35)
also seems promising.

4. Crystal preparation and characterization

Crystals grown in capillaries are ready for use but, to reduce diffuse scatter-
ing, it is recommended to pump out the gel surrounding the selected crystals
using micropipettes or some filter paper wick. This is especially easy with light
gels, which behave as viscous media.
For crystals grown in droplets or in dialysis buttons, the same procedure
can be followed after introducing a capillary in the gel to suck up the selected
crystal together with its surroundings. This avoids any direct contact of the
crystal with the capillary.
The crystalline quality of different biological macromolecules crystals
grown in agarose and silica gels has been characterized by measuring their
resolution limit and mosaic spread. These studies concern lysozyme crystals
(3, 11, 12, 36), but also crystals of higher molecular weight substances (37).
As to resolution, it appears that gel-grown crystals are, on average, better
than solution-grown ones. In a few cases, comparison was made with space-
grown crystals and plots of average I/s(I) values versus resolution show that
the best diffraction data collected from gel-grown crystals lie in between those
of earth-grown and space-grown crystals (3, 36). As to mosaic spread,
differences have also been evidenced. In particular, misorientations between
the different domains of a same crystal are generally less important in gel-
grown crystals. Up to now, the best results have been obtained with silica gel.

5. Related methods
5.1 Microgravity
Space experiments share with growth in gels the ability to reduce buoyancy-
driven convection, to reduce impurity concentration on the crystal surface,
and to avoid sedimentation of crystals as well as the secondary nucleation of
3D protein clusters. In addition, the microgravity scenario removes the
plausible chemical interaction of the gel with the reactants used in the chem-
170
 6: Crystallization in gels and related methods
ical protocol, including the protein itself. In short, microgravity conducted
experiments may be considered as 'clean' gel experiments.
As with gels, all techniques used in protein crystallization can be im-
plemented under microgravity. A diverse range of facilities, covering vapour
diffusion and liquid counter-diffusion techniques, are currently offered by
several Space Agencies to grow single protein crystals under microgravity
conditions (see refs 38-40 for a full description of facilities).
Because of the youth of microgravity science and the limited number of
opportunities to fly experiments, space crystallization of biological macro-
molecules still faces a number of technical and conceptual problems, which
need to be solved.
After ten years of microgravity-conducted experiments, some improve-
ments of crystal size and crystal quality have been reported (41-43, see ref. 38
for a review), and in a few cases reduced mosaicity was reported. Often,
however, the quality of the space-grown crystals, as evaluated by Wilson-type
plots or mosaicity measurements, does not show a dramatic increment of
quality, especially in terms of limit of resolution.
Current evaluation of space crystallization is basically performed by com-
paring space-grown crystals and crystals grown on earth using the same type
of reactor and the same crystallization conditions. Considering that the typical
dimensions of the reactors are large enough to allow density-driven and
thermal convection on earth, it is evident that space-grown crystals would be
expected to be of higher quality. In any case, what needs to be explained is
why in some cases the reverse result was found. In the future, space crystal-
lization has to face comparison with on-ground techniques emulating micro-
gravity conditions, that is crystal growth within gels and/or capillary volumes.
The similarity between the geometry and mass transport properties of the
gel acupuncture technique and the microgravity facilities is strong enough
(Figure 9) to permit an extrapolation of the above discussion on super-
saturation spatiotemporal evolution (see Section 3.2.3). The immediate
advantage of microgravity conducted experiments on capillary methods is
that, for the same path length, larger volumes of protein solutions can be
used, which could yield larger crystals. Unfortunately, the typical linear
dimensions in the direction of the diffusion path in the facilities currently
available are too short (15 mm for PCF or 8 mm for APCF) to exploit the
advantage of convection free counter-diffusion in large volumes. Therefore
the typical and useful spatial heterogeneity of the counter-diffusion tech-
niques is lost and only a limited set of the wide range of crystallization
conditions that could be reached is tested in practice. This enforces the use of
lower concentrations in all the experiments performed to date, converting the
free interface diffusion and dialysis experiments into batch experiments
because the characteristic time for the nucleation and transport processes are
similar (44, 45). It is clear therefore that the use of longer protein chambers
will permit the exploration of a larger set of local growth conditions.
171
 M.-C. Robert et al.
An early criticism of space crystallization (46) was that the limited number
of opportunities to fly experiments makes it impossible to employ the 'trial
and error' methodology used so far on earth in the search for crystallization
conditions. However, this restriction has, in fact, been a major driving force
for the development of the current trend to rationalize protein crystal growth,
as the advance in the understanding of fundamental aspects of protein crystal-
lization (including the knowledge of the growth environment and crystal
quality evaluation) directly derived from space-induced research has been
substantial. Today, it seems evident that such a rationale will be obtained only
by coupling on-ground and space crystallization research. In particular, the
numerical simulation of mass transport and precipitation phenomena (27,
44-50), the development of growth techniques emulating microgravity con-
ditions on-ground, the characterization of nucleation phenomena (51, 52),
growth mechanisms, and surface kinetics (53) will be required in the future. In
addition, the search for a relationship between growth history and crystal
quality requires the improvement of the existing X-ray characterization tools
(54, 55).
Finally, to evaluate properly the future of protein crystallization in space, it
should be considered that beside producing crystals of improved perfection
(with reduced mosaicity), another current interest is basically motivated by
the need for high quality crystals larger than the size (tenths of millimetre)
required for structural studies. The crystallization of biological macro-
molecules will face in the future the optimization of the final crystal size for
purposes not only related to structural biology (e.g. neutron diffraction
measurements) but also the characterization of their physical properties
because their use for technological application is still an unexplored and
exciting field. To grow large macromolecular single crystals, microgravity
offers the best scenario without the limitations of capillary volumes, but first
we need to learn how to properly control impurity distribution and its effect
on the cessation of growth.

5.2 Hypergravity
Contrary to microgravity experimentation, kilogravity experimentation in
centrifuges is a rather unexplored field, although successful results were
reported as early as 1936 for tobacco mosaic virus (56).
In a review on this subject (57), Schlichta emphasized that gravity might be
regarded as a variable in crystal growth and material processing. Besides
convection and sedimentation which are drastically altered by increasing
gravity forces, one has also to consider the increased compressive and
shear stresses, as well as hydrostatic pressure (which can induce solubility
variations) (58).
All these factors influence directly or indirectly the growth parameters. For
example, due to forced sedimentation effect, supersaturated zones appear in
172
 6: Crystallization in gels and related methods
an initially undersaturated solution, leading to crystal nucleation and growth
(59). However, the role played by the gravity field may promote additional
effects:
(a) Centrifugation can separate the molecules having different molecular
weights (foreign macromolecule, impurities) or different conformer.
(b) Centrifugation will progressively modify the spatial distribution of the
different populations of oligomeric species or aggregates or interacting
monomers contained in a supersaturated macromolecular solution. Thus,
one could set some local solution compositions suitable for nucleation.
The devices needed for hypergravity experiments are not necessarily
complicated: the crystal growth presented in ref. 59 required a centrifuge
currently available in any biochemistry laboratory.
More fundamental studies need ultracentrifuges equipped with observation
set-ups such as Schlieren optics. As a matter of fact, the development of
centrifugal crystal growth is related to the development of basic studies. A
current feeling is that crystal grown under kilogravity would suffer from
plastic deformation and would not diffract at very high resolution. This does
not seem valid, at least considering the few examples found in the literature
(57, 60-61).

References
1. Robert, M.-C. and Lefaucheux, F. (1988). /. Cryst. Growth, 90, 358.
2. Vekilov, P. G., Alexander, J. I. D., and Rosenberger, F. (1996). Phys. Rev. E, 54,
6650.
3. Miller, T. Y. and Carter, D. C. (1992). J. Cryst. Growth, 122, 306.
4. Sica, F., Demasi, D., Mazzarella, L., Zagari, A., Capasso, S., Pearl, L. H., et al.
(1994). Acta Cryst., D50, 508.
5. Vidal, O., Robert, M.-C., and Boue, F. (1998). J. Cryst. Growth, 192, 257.
6. Vidal, O., Robert, M.-C., and Boue, F. (1998). J. Cryst. Growth, 192, 271.
7. Her, R. K. (1979). The chemistry of silica. Wiley. Interscience, New York.
8. Ramzi, M. (1996). PhD Thesis, University de Strasbourg.
9. Vidal, O. (1997). PhD Thesis, Universite de Paris.
10. Henisch, H. K. (1988). Crystal growth in gels and Liesegang rings. Cambridge
University Press.
11. Garcia-Ruiz, J.-M., Gavira, J. A., Otalora, F., Guasch, A., and Coll, M. (1998).
Mat. Res. Bull., 33, 1593.
12. Vidal, O., Robert, M.-C., Arnoux, B., and Capelle, B. (1999). J. Cryst. Growth, 196,
559.
13. Bernard, Y., Degoy, S., Lefaucheux, F., and Robert, M.-C. (1994). Acta Cryst.,
D50, 504.
14. Cudney, B., Patel, S., and McPherson, A. (1994). Acta Cryst., D50, 479.
15. Provost, K. and Robert, M.-C. (1991). J. Cryst. Growth, 110, 258.
16. Thiessen, K. J. (1994). Acta Cryst., D50, 491.
173
 M.-C. Robert et al.
17. McPherson, A., Malkin, A. J., and Kuznetsov, Y. G. (1995). Structure, 3, 759.
18. Georgalis, Y., Umbach, P., Raptis, J., Saenger, W. (1997). Acta Cryst., D53, 691
and 703.
19. Vidal, O., Bernard, Y., Robert, M.-C., and Lefaucheux, F. (1996). J. Cryst.
Growth, 168, 40.
20. Provost, K. and Robert, M.-C. (1995). / Cryst. Growth, 156, 112.
21. Hirschler, J. and Fontecilla-Camps, J. C. (1996). Acta Cryst., D52, 806.
22. Feher, G. and Kam, Z. (1985). In Methods in enzymology (eds H. W. Wyckoff,
C. H. W. Hirs, and S. N. Timasheff), Academic Press, Inc., London, Vol. 114,
p. 77.
23. Zeppezauer, M., Eklund, H., and Zeppezauer, E. S. (1968). Arch. Biochem. Anal.,
126, 564.
24. Phillips, G. N. (1985). In Methods in enzymology (eds. H. W. Wyckoff, C. H. W.
Hirs, and S. N. Timasheff) Academic Press, Inc., London. Vol. 114, p. 128.
25. Weber, P. (1991). Adv. Protein Chem., 41, 16.
26. Garci'a-Ruiz, J.-M., Moreno, A., Otalora, F., Vieoma, C., Rondon, D., and
Zautscher, F. (1998). J. Chem. Educ., 75, 442.
27. Otalora, F. and Garcfa-Ruiz, J.-M. (1996). J. Cryst. Growth, 169, 361.
28. Moreno, A., Rondon, D., and Garcia-Ruiz, J.-M. (1996). J. Cryst. Growth, 166,
919.
29. Garci'a-Ruiz, J.-M. and Moreno, A. (1994). Acta Cryst., D50, 484.
30. Garci'a-Ruiz, J.-M., Lopez Martinez, C., and Martin-Vivaldi Caballero, J. L.
(1985). Cryst. Res. Techn., 20, 1615.
31. Garcia-Ruiz, J.-M. and Moreno, A. (1997). J. Cryst. Growth, 178, 393.
32. Otalora, F., Garcia-Ruiz, J.-M., and Moreno, A. (1996). J. Cryst. Growth, 168, 93.
33. Otalora, F., Capelle, B., Ducruix, A., and Garcia-Ruiz, J.-M. (1999). Acta Cryst.,
D55, 644.
34. Otalora, F., Gavira, J. A., Capelle, B., and Garcia-Ruiz, J. M. (1999). Acta Cryst.,
D55, 650.
35. McPherson, A. (1985). In Methods in enzymology (eds. H. W. Wyckoff, C. H. W.
Hirs, and S. N. Timasheff), Academic Press, Inc., London. Vol. 114, 125.
36. DeLucas, L. J., Long, M. M., Moor, K. M., Rosenblum, W. M., Bray, T. L., Smith,
C., et al (1994). J. Cryst. Growth, 135, 183.
37. Lorber, B., Sauter, C., Ng, J. D., Zhu, D. W., Giege, R., Vidal, O., Robert, M.-C,
Capelle, B, et al. (1999). J. Cryst. Growth, in press.
38. Giege, R, Drenth, J., Ducruix, A, McPherson, A, and Saenger, W. (1995). Prog.
Cryst. Growth Charact., 30, 237.
39. McPherson, A. (1996). Cryst. Rev., 6, 157.
40. Pletser, V, Stapelmann, J, Potthast, L., and Bosch, R. (1999). J. Cryst. Growth,
196, 638.
41. Koszelak, S., Day, J, Leja, C, Cudney, R., and McPherson, A. (1995). Biophys. J.,
69, 13.
42. Ng, J. D, Lorber, B., Giege, R., Koszelak, S, Day, J, Greenwood, A, et al.
(1997). Acta Cryst., D53, 724.
43. Snell, E. H, Weisgerber, S., and Helliwell, J. R. (1995). Acta Cryst., D51, 1099.
44. Otalora, F. and Garcia-Ruiz, J.-M. (1997). J. Cryst. Growth, 182, 141.
45. Garcia-Ruiz, J.-M. and Otalora, F. (1997). J. Cryst. Growth, 182, 155.
46. Leberman, R. (1985). Science, 230, 370.
174
 6: Crystallization in gels and related methods
47. Wagner, G. and Linhardt, R. (1991). J. Cryst. Growth, 110, 114.
48. Fowlis, W. W., DeLucas, L. J., Twigg, P. J., Howard, S. B., Meehan, E. J., and
Baird, J. K. (1988). J. Cryst. Growth, 90, 117.
49. Huo, C, Ge, P., Xu, Z., and Zhu, Z. (1991). J. Cryst. Growth, 114, 486.
50. Savino, R. and Monti, R. (1996). J. Cryst. Growth, 165,308.
51. Malkin, A. J., Cheung, J., and McPherson, A. (1993). J. Cryst. Growth, 126, 544.
52. Georgalis, J., Schiller, P., Frank, M., Soumpasis, D., and Saenger, W. (1995). Adv.
Colloid Interface Sci., 58, 57.
53. Land, T. A., Malkin, A. J., Kuznetsov, Yu. G., McPherson, A., and Yoreo, J. J.
(1995). Phys. Rev. Lett., 75, 2774.
54. Helliwell, J. R. (1988). J. Cryst. Growth, 90, 259.
55. Chayen, N. E., Boggon, T. J., Casseta, A., Deacon, A. Gleichmann, T., Habash, J.
et al. (1996). Q. Rev. Biophys., 29, 227.
56. Wyckoff, R. W. G. and Corey, R. B. (1936). Science, 84, 513.
57. Schlichta, P. J. (1992). J. Cryst. Growth, 119, 1.
58. Lorber, B., Jenner, G., and Giege, R. (1996). J Cryst. Growth, 158, 103.
59. Pitts, J. E. (1992). Nature, 355, 117.
60. Behlke, J. and Knespel, A. (1996). J. Cryst. Growth, 158, 388.
61. Barynin, V. V. and Melik-Adamyan, V. R. (1982). Sov. Phys. Crystallogr., 27, 588.

175
This page intentionally left blank
 7

Seeding techniques
E. A. STURA

1. Introduction
A seed provides a template for the assembly of molecules to form a crystal
with the same characteristics as the crystal from which it originated. Seeding
has often been used as a method of last resort, rather than a standard practice.
Recently, these techniques have gained popularity, in particular, macro-
seeding, used to enlarge the size of crystals. Seeding has many more applica-
tions, and the use of seeding in crystallization can simplify the task of the
crystallographer even when crystals can be obtained without it. We will
explore the various seeding techniques, and their applications, in the growth
of large single crystals and the methods by which we may attempt to obtain
crystals that diffract to higher resolution.
Crystallogenesis can be divided into two separate phases. The first being
the screening of crystallization conditions to obtain the first crystals, the
second consisting of the optimization of these conditions to improve crystal
size and quality. Seeding can be used advantageously in both these situations.
The first stage in crystallogenesis consists of the discovery of initial crystals,
crystalline aggregates, or microcrystalline precipitate. This may result from a
standardized screening method (1, 2), a systematic method (3), an incomplete
factorial search (see Chapter 4 and refs 4 and 5), or by extensive screening of
many conditions. This may be bypassed by starting with seeds from crystals of
a related molecule that has been previously crystallized. Molecules that have
been obtained by genetic or molecular engineering of a previously crystallized
macromolecule fall in this category. This method is termed cross-seeding. It
has been used to obtain crystals of pig aspartate aminotransferase starting
with crystal from the chicken enzyme (6) and between native and complexed
Fab molecules (7).
Whatever the method used to obtain the initial crystals, seeding may pro-
vide a fast and effective way to facilitate the optimization of growth conditions
without the uncertainty which is intrinsic in the process of spontaneous
nucleation. The streak seeding technique can be used to carry out a search
quickly and efficiently over a wide range of growth conditions. Later the use
 E. A. Stum
of macroseeding and microseeding methods can be used to grow large crystals
with a high degree of reproducibility.

2. Seeding
2.1 Supersaturation and nucleation
Details on the use of precipitants, together with general and theoretical
considerations, and practical methods for macromolecule crystallization, are
to be found in other chapters and in other publications (8-12). Here we will
consider some of the aspects of crystallization that directly affect the
application of seeding techniques. It is useful to separate the events leading to
the spontaneous formation of a crystal nucleus and those conditions that
allow a crystal or nucleus to grow. While both events depend on the degree of
supersaturation of the protein, the physical processes involved are very dif-
ferent, and the degree of supersaturation required for nucleation is generally
higher than that required for growth onto an existing crystal plane. In the case
of spontaneous nucleation a new seed must be generated while other events
are taking place and is driven by the requirement to lower the free energy of
the supersaturated state. These other events involve the aggregation of
molecules into various phases. During aggregation, reversible and irreversible
processes are at work simultaneously. The formation of ordered nuclei may
be competing for protein with irreversible processes that produce amorphous
aggregates (such as precipitates and protein skins) and hence constantly lower
the degree of supersaturation of the macromolecule. As supersaturation is
decreased the chance of forming a stable nucleus is reduced. Since the
occurrence of spontaneous nucleation depends on the relative rates at which
these various competing events take place, crystals might never form even
under conditions which might otherwise support crystal growth. The principle
that there is a lower energy requirement in adding to an existing crystal
surface than in creating a new nucleus has important consequences. If the
inverse were true, the protein would partition into a very large number of
small nuclei and large crystals would never grow. Instead, in many cases it is
possible to grow large crystals in the absence of seeding. Since spontaneous
nucleation is a statistical phenomenon, whose probability increases with
increasing degree of supersaturation, the nucleation and growth of crystals is
a process with negative feedback. As a nucleus is formed its growth reduces
the degree of supersaturation of the solution, and hence decreases the prob-
ability that other nuclei will form. To take advantage of this, supersaturation
must be achieved slowly. The degree of supersaturation should be just
sufficient to obtain a small number of nucleation centres. With the proper
choice of crystallization conditions and good control over the environmental
factors, it is often possible to fulfil all of the above conditions. Determining
the appropriate conditions can require many crystallization trials and is con-
sequently time-consuming involving the use of many milligrams of macro-
178
 7: Seeding techniques
molecule and other materials. Seeding can be used efficiently and effectively
during the crystallization trials to minimize the quantity of protein required
for this analysis.

2.2 Crystal growth

When crystal seeds are added to an equilibrated protein solution, the protein
partitioned between the soluble phase and irregular aggregate phases will
redistribute. The final equilibrium will be achieved only when crystal growth
has ceased. From this we may understand why seeding can be used not only in
situations where the protein remains in the soluble phase, but also in
situations where most of the protein has precipitated. In fact, as crystals grow
the degree of supersaturation is reduced and protein may be transferred from
the various (reversible) amorphous phases to the soluble phase, and from this
phase it can accrete onto crystal surfaces. Several factors affect the quality of
the crystals obtained, including the rate of growth, the internal order of the
initial nucleus, and the purity of the sample. For some macromolecules, the
initial growth following nucleation may be too fast, because of the high degree
of supersaturation, resulting in the incorporation of crystal defects, which may
eventually lead to the premature termination of crystal growth and to poorly
formed crystals. In certain cases, the nuclei which are generated spontane-
ously may be polycrystalline in nature. Growth from such nuclei may result in
the formation of crystal clusters rather than individual single crystals. Recent
studies have shown that lysozyme crystals have impurity rich cores of the
order of 20-30 um and that seeding may provide a method to avoid such
problem (13, 14). By decoupling crystal growth from nucleation seeding pro-
vides a means by which growth conditions may be tailored for crystal growth
rather than nucleation and lead towards the production of large, regular
crystals. Seeding allows the experimenter to control not only the number of
seeds but also reduce the supersaturation level of the protein and to decrease
the incorporation of defects detrimental to crystal quality. Seeding provides a
preformed, regular crystal surface onto which further molecules may aggre-
gate in an orderly fashion. The seeds to be used in the seeding experiments
can be selected from the best crystals previously grown, this will lead to the
best results.

2.3 Seeding techniques

Seeding consists of three stages: a pre-seeding stage, an analytical stage to
refine crystal growth conditions, and the final production stage using the
refined conditions to produce large single crystals. The pre-seeding stage is
essential for seeding to work in a consistent and reproducible manner. It can
be separated into four important aspects:
• the environment, and the associated precautions necessary for seeding
• pre-equilibration of the protein solution to be seeded
179
 E. A. Stum
• the determination of the appropriate supersaturation level for seeding
• choice of the crystals from which to obtain seeds and their preparation.
The design of the experiment is important since during seeding super-
saturation may increase resulting in spontaneous nucleation leading to a
shower of small crystals.
Seeding methods can be separated into:
(a) Microseeding which involves the transfer of microscopic crystals from a
seed source to a non-nucleated protein solution.
(b) Macroseeding in which pre-grown crystals are washed and introduced
individually into a pre-equilibrated protein solution. This method has
been widely applied to tackle the problem of enlarging small crystals into
crystals of a suitable size for X-ray diffraction studies (15, 16).
These methods are common to both homogeneous or heterogeneous
seeding. Homogeneous seeding involves duplicating the three-dimensional
lattice of the crystal using an identical macromolecular solution. Hetero-
geneous seeding is somewhat more complex and can be divided into:
(a) Cross-seeding, a form of seeding in which seeds come from crystals of a
protein or macromolecular complex that is different from that being
crystallized. A closely related crystal form is generally expected.
(b) In epitaxial nucleation, a regular surface rather than a three-dimensional
lattice is the template for the growth of new crystals. It is this type of
seeding that is exploited to grow more tightly-packed crystals that in
some cases diffract to higher resolution than the original crystals. The
nucleation of protein crystals on cellulose fibre impurities, which often
end up in protein solutions, is another case of epitaxial nucleation.
In the latter of these methods the lattice dimensions of the crystals obtained
is different from those of the seeds from which they grew. Heterogeneous
seeding should be followed by homogeneous seeding as the core is likely to
maintain some of the characteristics of the original seed.

3. Crystallization procedures
Crystallization by vapour diffusion is a relatively simple technique (see
Chapter 5). This section deals with the establishment of crystallization
procedures which are suited to the application of seeding.

3.1 Pre-seeding: sitting drop vapour diffusion

The crystallization procedures used in conjunction with seeding techniques
may be different from those that would be used otherwise. The design of the
crystallization experiment must allow for the introduction of seeds at some
180
 7: Seeding techniques
stage in the equilibration phase. This must be done with the minimum
disruption to the crystallization environment. A compact variation of the
sitting drop vapour diffusion method has been used successfully in the seeding
of many proteins (17-20).
Vapour diffusion provides a controlled and relatively slow method of
equilibration by the transfer of vapour between the protein and precipitant
solutions.

i. Temperature
To achieve success in crystallization and seeding it is important to control
the overall environment of the set-up. This must include temperature.
Temperature regulation can be achieved by the combined use of a sitting drop
vapour diffusion set-up using a glass pedestal and a constant temperature
incubator. The sitting drop environment provides better temperature control
than its hanging drop counterpart because of heat conduction between the
reservoir solution and the protein solution in the inverted glass pot (see
Figure 1). The problem of condensation on the glass coverslip, caused by
temperature gradients and convection currents in the sealed set-up, are less
likely to affect a sitting drop experiment where the protein drop is situated
close to the surface of the reservoir. In contrast, in the hanging drop environ-
ment, the drop is effectively in thermal contact with the outside air. The thin
coverglass absorbs the heat of condensation and dissipates it to the outside.
Short-lived changes in temperature, such as opening the door of the constant
temperature incubator containing the experiment, will rapidly vary the
temperature of the hanging drop but not that of the reservoir because of its
higher heat capacity. During a rise in temperature, vapour will distil away
from the drop, increasing the degree of supersaturation, which may result in a
shower of microcrystals. This will be more common for crystals that are grown
at high salt concentrations. When the temperature decreases more vapour
condenses onto the drop diluting the protein solution. It is not uncommon in
low salt, or in crystallization trials using hanging drop vapour diffusion under
low PEG concentrations, to observe an increase rather than a decrease in the
volume of the protein-precipitant drop. The use of the sitting drop method
reduces these problems.

ii. Multiwell sitting drop vapour diffusion plate

The set-up shown in Figure 1 has been designed and fabricated in our
laboratory to shield against short-lived temperature fluctuations, other
similarly compact set-ups are now in use in other laboratories. Its compact
size makes efficient use of incubator space. The sitting drop method gives
easier access to the protein drop and the reservoir than hanging drop set-ups.
This allows the precipitant concentration to be varied, when required, during
the course of the experiment. By lifting the coverglass the protein-precipitant
drop can be seeded, and the plate design enables minimum disturbance to the
181
 E. A. Stura

Figure 1. Schematic illustration of the sitting drop vapour diffusion plates. (A) A glass cup
is held between long sized forceps over a Bunsen burner with the closed end down
towards the flame. (B) The cup is heated until the glass becomes soft. (C) When the bottom
is malleable the pot is inverted and a depression is made in it with a glass plunger. (D) A
ring of silicone vacuum grease is placed in the bottom of each of the wells of the microtitre
plate. Pots are placed onto top of the ring and pressed down. (E) Each depression is

182
 7; Seeding techniques

siliconized, washed repeatedly with distilled water, and the coat baked in an oven. The
siliconized pots can now be placed in each of the wells in the multiwell cluster on top of
the silicone grease ring which holds them in position. (G) The rim of the individual well
is smeared with petroleum jelly to seal the well once the coverglass is placed on top. (H)
The precipitant solution is placed around the inverted pot, the protein in the depression
and the desired amount of precipitant mixed with it before sealing the experiment.

183
 E. A. Stura
seeding environment. Such features can greatly enhance growth of quality
single crystals. The tray consists of a 24-well Costar tissue culture plate from
Hampton Research combined with 0.6 ml glass cups from Fisher Scientific.
The wells are significantly smaller than those in the more commonly used
Linbro plates, which use 22 mm circular coverglass slips instead of the 18 mm
required for the Costar plates. The plates are made as described in Protocol 1
and illustrated in Figure 1.

Protocol 1. Making multiwell sitting drop plates

Equipment and reagents

• 0.6 mm cups • Plastic syringe
• Bunsen burner • Forceps
• Silicone grease • Pipette
• Glass plunger • 18 mm round coverglass

Method
1. Make a depression in the bottom of the 0.6 ml cups by heating over a
Bunsen flame and pressing down on the cylindrical base of each of the
inverted cups with a rounded-end glass plunger.
2. Hold the resulting cup in place at the centre of each well of the tissue
culture plate with Corning silicone vacuum grease. The open end of
the cup sits on the bottom and the depression in the cup is at the
volumetric centre of each individual well (sitting drop rods with
depressions, made by Perpetual System Corporation, and micro-
bridges are available from Hampton Research).
3. Siliconize the cavity created by the depression. Up to 100 ul of protein-
precipitant mixture can be used with this set-up. The reservoir
solution, typically 1 ml, occupies an annulus around the inverted cup.
4. Smear the edges of each well in the Costar tissue culture plate (avail-
able from Hampton Research) with petroleum jelly. Silicone grease
can be used instead of petroleum jelly to give a better seal, although
this will make covers harder to remove for seeding.
5. Place a 18 mm round microscope coverglass on top of each well
ensuring that an airtight seal is achieved.

Since this method is similar in many ways to the hanging drop method,
crystallization conditions determined for hanging drop experiments require
little modification for implementation with sitting drop vapour diffusion.
Another added advantage is that larger volumes can be used. Other sitting
drop methods provide similar advantages although glass pedestals are
recommended for temperature stability.
184
 7: Seeding techniques

3.2 Analytical seeding

It is important to first determine under what conditions seeding will be
effective. This is done by the use of an analytical seeding method such as
streak seeding.

3.2.1 Streak seeding technique

i. Making the probe
A probe for analytical seeding is easily made with an animal whisker mounted
with wax to the end of a pipette tip. The tip is then mounted on a wooden rod.
Since the cross-section of the whisker varies along its length it is possible to
obtain several probes from the same whisker, by repeatedly cutting the
whisker to lengths of 5-20 mm from the end of the wax. The probes so
obtained will be of different strength and thickness.
ii. Cleaning the probe
To clean the probes prior to their use the fibres are degreased using ethanol or
methanol, and then washed in distilled water and wiped dry. Probes can be
used several times by cleaning them with distilled water and tissue paper in
between experiments. After a period of time the whiskers need to be trimmed
as the end becomes frayed. A good probe should be able to transfer seeds to
six to twelve drops consecutively without being dipped in seed solution. A
good probe is best for titrating the number of seeds; old whiskers deposit
many seeds in the first two drops and virtually none in further drops. Cat
whiskers are generally used for making probes, although other animal
whiskers have also been found suitable.
Hi. Seeding
The end of the fibre is then used to touch an existing crystal and dislodge
seeds from it (Figure 2A). Gentle friction against the crystal is normally
sufficient. Some of the dislodged seeds remain attached to the fibre. The
probe in now used to introduce seeds into a pre-equilibrated drop by rapidly
running the fibre in a straight line across the middle of the protein-precipitant
drop (Figure 2B).
Sitting drop set-ups are preferable since hanging drops tend to dry out
when exposed to the ambient air, even in the short time interval between
collecting the seeds on the fibre and streaking the new drop, typically 5-30
seconds. The precipitant collected from the first drop increases in con-
centration as it travels to the next drop though the air, and this can affect the
conditions in the seeded drop. Therefore, the distance the whisker has to
travel should be kept to a minimum, in most circumstances, less than 10 cm.
Both the source well and the receiving well are resealed immediately after the
transfer. Seed nucleated crystals grow along the streak line; any self-nucleated
seeds will occur elsewhere in the drop. The pre-incubation time and the range
185
 E. A. Stum

Figure 2. Schematic drawing of the stages in the application of the streak seeding
technique for analytical seeding. (A) A probe made from a cut pipette tip mounted on a
wooden shaft. On the end of the tip a short segment of an animal whisker is attached
with molten wax. The probe so constructed is used to pick up seeds from an existing
crystal, precipitate, or other ordered aggregate by simply touching it and displacing
seeds from it. (B) The seeds remain attached to the whisker and can be transferred to a
pre-equilibrated drop by running the end of the probe across the drop. Some seeds are
deposited along the path, where they will either grow into large crystals or dissolve into
the solution. (C) Growth of crystals along the streak line indicates that the conditions may
be suitable for the application of other techniques such as micro- or macroseeding. Self-
nucleated crystals will appear away from the streak line.

186
 7: Seeding techniques
of supersaturation which allows for sufficient crystal growth without self-
nucleation is determined experimentally by observing the growth (or lack of
growth) of seeded crystals along the streak line (Figure 2C).

3.2.2 Protein pre-equilibration

In batch crystallization the precipitant concentration is slowly increased until
the protein solution turns cloudy; further solvent is then added until it is clear
again. The protein solution is continuously stirred until seeds are added.
When seeding small volumes, it is best to avoid producing spontaneously
nucleated seeds, as could be produced by exceeding the solubility threshold,
since it is difficult to control their number, or ensure that they will later
dissolve. By introducing the seeds well before sufficient supersaturation is
reached, additional nuclei which might form spontaneously are prevented.
However, if the protein is not suitably pre-equilibrated, the seeds will dissolve.
Streak seeding conditions should be optimized by repeating the procedure
after different pre-equilibration times, using different drops, although if seeds
dissolve, the same drop can be seeded again.

3.2.3 Determining the degree of supersaturation for seeding

Initially, during the screening phase of a crystallization, we want to obtain
results quickly in order to determine the many parameters that control the
growth and morphology of the crystals. In the production phase, when con-
ditions are optimized, we want to slow down the growth rate. As long as
protein is not being lost to amorphous phases, drops are generally allowed to
equilibrate fully before seeding, unless conditions for crystal growth and
nucleation are tightly coupled. To optimize the conditions, drops are set up
under conditions which vary only slightly from those previously determined in
the fast growth experiments. Drop size, which was kept to a minimum to save
material, should be increased at this stage, while simultaneously reducing the
precipitant-protein ratio, and the precipitant concentration in the reservoir.
This will have the effect of slowing down the rate of equilibration and the
desired state of supersaturation will be approached more slowly. The final
conditions will be established at reduced precipitant concentration but at a
compensating, higher protein concentration. The precise protein-precipitant
ratio and precipitant concentration are determined experimentally by allow-
ing the drops to equilibrate, usually three to five days, and then streak seeded.
The streak will not appear in drops where the concentration of the precipitant
in the reservoir is too low, and where the precipitant concentration is too high
crystals will initially appear along the streak line, followed later by the
formation of others away from the streak line. Caution should be used when
making this determination, as some seeds may drift away from the line along
which they were deposited and sink down to the bottom of the drop, crystals
that appear at the drop/air/glass interface are generally an indication that the
187
 E. A. Stura
well was left open too long during seeding rather than an indication that the
seeding conditions were unsuitable. The experimental conditions where crystals
grow only along the streak line determine the precipitant concentration range
for production seeding. This method is also well suited for testing minor
changes to growth conditions, such as adding a co-precipitant, testing new
additives, or simply finely analysing the pH range. Small changes at this stage
can result in significant improvements in the quality of the crystals obtained
and can be essential for growing suitable crystals for high resolution X-ray
structure analysis.

3.2.4 Assaying for microcrystallinity

Microcrystalline precipitates are often indistinguishable from their amorph-
ous counterparts. Streak seeding can be used to distinguish between these two
possibilities by using particles from an uncharacterized precipitate as a source
of seeds. For example, the initial crystallization trials of a complex between
the Fab' fragment from an anti-peptide murine antibody B13I2 with its 19
amino acid peptide antigen (myohaemerythrin residues 69-87) gave
a precipitate composed of round or oval particles of roughly equal size
(Figure 5A) (19). No indication of microcrystallinity could be deduced from
microscopic observations of the precipitate since it was not appreciably
birefringent. Three adjacent drops in which protein had not precipitated were
streaked with this precipitate. Hexagonal-shaped crystals appeared in one of
the drops as a result of the streak seeding experiment (Figure 5B). These
crystals were then used for macroseeding onto other drops. After an
adjustment in the crystallization conditions from 1.5 M sodium citrate pH 6.0,
to 1.6-2.0 M mixed sodium and potassium phosphate pH 5.0-6.0 it was
possible to grow crystals (Figure 5C) that diffracted 2.6 A resolution (19). The
presence of the peptide in the solution is essential to obtain these crystals and
cross-seeding from the Fab'-peptide complex onto native Fab' solutions does
not produce crystals or precipitate.

4. Production seeding methods

4.1 Microseeding
In microseeding microscopic crystal fragments are introduced into a prepared
protein solution. By using an analytical seeding technique (Section 3.2) the
supersaturation threshold can be accurately determined by scanning a range
of precipitant concentrations. The streak seeding technique was initially
developed for this purpose. Microseeding is composed of three stages:
• preparation of the seed stock
• repeated dilution of the seeds
• the seeding itself.
188
 7: Seeding techniques
4.1.1 Preparatory steps
The two first stages in microseeding are described in Protocol 2.

Protocol 2. Microseeding

Equipment and reagents

• Tissue homogenizer • Vortexer
• Test tube • Seeding probe
• Precipitant solution • Pipette
• Small crystals • Microtest tubes for diluted seeds

A. Preparation of the seed stock

1. To produce seed stock wash three or four small crystals in a slightly
dissolving solution to remove defects or amorphous precipitate from
the crystal surfaces.
2. Stabilize the washed crystals in an appropriate precipitant solution
and transfer them to a glass tissue homogenizer in which they are
crushed (Figure 3).
3. Wash the crushed seeds from the sides of the homogenizer into the
bottom by adding further solution. Transfer the solution from the
homogenizer to a test-tube. This is the seed stock, which for most
proteins can be stored in a constant temperature incubator for future
use.

B. Repeated dilution of the seeds

The seed stock is normally diluted with further solution as it contains too
many nuclei to be useful in the nucleation of only a small number of
crystals.
1. Dilute the seeds, typically in the range 10-3 to 10-7, vortexing the tube
containing the microseeds between dilutions to evenly distribute the
seeds. The reservoir solution is normally well suited for the initial
dilutions as some of the smaller microseeds dissolve to provide a
residual protein background. Further dilutions are likely to require
extra precipitant, usually 10% or more above that of the reservoir to
prevent the seeds from dissolving. Maintain the temperature constant
as seeds may dissolve either on heating or cooling. One of the diluted
seed solutions should be suitable to supply a small number of seeds
into each drop.
2. Test the seed stock produced by streaking or by adding a measured
amount of seed solution to several drops. The results should be visible
in one to two days. If there is a sudden drop off in the number of

189
 E. A. Stura
Protocol 2. Continued
crystals obtained, inconsistent with the a tenfold dilution, extra
precipitant should be added to the stabilizing solution. If nucleation is
independent of the dilutions, buffer should be added to the seed
solution to reduce the precipitant concentration. Also test the drops by
streak seeding from a crystal (Protocol 3) to ensure that the reservoir
solution is appropriate for seeding.

4.1.2 Streak seeding as a microseeding technique

Since the seeds that are transferred from crystal to drop in the use of the
streak seeding technique are microscopic, the technique is technically a
microseeding method. But, while in the analytical streak seeding technique
the deposition of many seeds along the path of the whisker is essential for the
subsequent visualization, when growing large crystals for X-ray structure
analysis only a few seeds should be deposited, as seeds compete with each
other for the available protein. When changing the use of streak seeding from
an analytical to a production seeding mode we must find a way of diluting the
seeds quantitatively in a reproducible manner. The same probe is thus used in
each repeat experiment to ensure that the volume of liquid and the number
and size of the seeds which are transferred from one drop to another, remains
constant. Thicker whiskers transfer more liquid containing microseeds and
can potentially carry larger size seeds.

Protocol 3. Streak seeding

Equipment and reagents

• Seeding probe • Vacuum grease
• Seeds • Cover glass
• Preequilibrated protein drops • Crystallization tray forceps
• Microscope

Method
1. By changing the angle at which the whisker is drawn out of the we can
affect the size and number of seeds loaded onto the fibre. This should
be kept constant for reproducibility. The whisker is lifted vertically
upwards, maintaining it perpendicular to the drop's surface to mini-
mize seed retention, seeds are scooped up to maximize size and the
number of seeds picked up.
2. Pre-equilibrate the drops before seeding under conditions previously
determined analytically by streak seeding.
3. Streak seed subsequent drops without loading the probe with new
seeds to achieve seed dilution. To obtain greater dilutions the probe

190
 7: Seeding techniques
can be dipped in and out of the reservoir in between streaks, to allow
some seeds to drop into the precipitant solution.
4. Reduce the time the probe spends in the air by opening all the
chambers to be seeded just before picking up the seeds. Streak
the drops sequentially, as speed is important to prevent drying of the
solution on the fibre.
5. Cover all the chambers without delay to reduce evaporation from the
seeded drops.

The seed stock and seed dilutions prepared as previously described, can be
used in streak seeding reliably by dipping the whisker into each of the diluted
solutions including the seed stock and applying the seeds to new drops
(Figure 3). It is common to start by dipping the probe in the most dilute
solution first, streaking one drop, then progressing up the dilution series to the
seed stock. Typically the results are analysed two days to one week after
streaking.

4.2 Macroseeding
In macroseeding a single crystal is introduced into a suitably pre-equilibrated
solution. Single prismatic crystals, which are free from twinning or any other
crystallites, are most suitable for this technique. As in other seeding protocols
it is important to take steps to maintain constant conditions, as even a slight
dehydration of the drop being seeded could temporarily change the state of
supersaturation and induce unwanted nucleation. Performing the experiment
in a very humid environment and by using large drops can reduce dehydra-
tion. A beaker with a filter paper cylinder soaked in distilled water is such an
environment. However it is more practical to use sitting drop multiwell trays,
which have been used with high success in our laboratory. Macroseeding is
done under a dissecting microscope where the small amount of heat gener-
ated from the microscope stage light bulb may actually be slightly beneficial in
increasing the humidity level around the drop. The heat raises the tempera-
ture of the reservoir faster than that of the drop increasing the rate of evapor-
ation from the reservoir, and counteracting evaporation into the room. Seeds
are washed in a slightly dissolving solution to remove the top layer of protein,
which contains possible defects, from the surface of the seed without causing
excessive etching or cracking. They are then transferred to a stabilizing
solution to re-equilibrate the crystals (Figure 4). Older seeds benefit the most
from this treatment, whereas freshly grown crystals may be put directly
through a series of washes in stabilizing solution. From the final wash solution
each seed is then transferred to the protein-precipitant drop to be seeded
(Figure 4).
191
Figure 3, Diagrammatic illustration of the steps involved in microseeding. (A) Crystals of
good morphology are crushed in a glass tissue homogenizer. The resulting seeds are
washed into the bottom of the tube and stored in a test-tube. (B) The seed stock is diluted
to produce a dilution series. (C) Seeds can be picked up from the diluted solutions by
using a probe, or precipitant can be added to these, so that they may be mixed with
protein solution (D). The wells are seated by replacing the coverglass and the seeds are
allowed to grow for several days.

192
 7: Seeding techniques

Figure 4. Illustration of the steps involved in macroseeding. (A) A single crystal is picked
up from a drop. Crystals should be of good morphology and free from defects. (B) A
series of washes is performed by repeatedly transferring the crystal from one depression
to another, taking care not to damage the seed. (C) The seed is finally transferred to a pre-
equilibrated drop for further enlargement.

193
 E. A. Stura

4.2.1 Details of crystal handling in macroseeding

The handling of crystals is especially important to avoid the generation of
microseeds or unwanted nuclei in the transfer to the drop being seeded.

Protocol 4. Handling of crystals

Equipment and reagents

• Glass syringe • Forceps
• C-flex tubing •Seeding tray
• X-ray capillary • Seeding probe
« Microscope

Method
1. Connect a glass or quartz capillary to a 1 ml glass syringe with a short
piece of rubber tubing such as c-flex (Fisher, 14-169-5c) which gives an
excellent seal.
2. Snap open the end of the capillary with tweezers or scissors, and
siliconize if the experimental situation can benefit from diminished
adhesion of the solution to the glass capillary. After siliconizing it
should be extensively washed.
3. Pick up the crystals under a dissecting microscope, using a magnifica-
tion of x 10 to x 100.
Crystals from hanging drops should first be washed into a secondary
vessel; in a sitting drop vapour diffusion set-up it is possible to do
this directly from the depression in which the crystals have been
growing.
After the coverglass sealing the vapour diffusion chamber is lifted off,
the tip of the capillary is inserted into the drop and a crystal is drawn
into the capillary.
If the crystals are adhering to the well, by withdrawing liquid from the
drop and gently ejecting it onto a chosen crystal, it is often possible to
dislodge the crystal. Unfortunately some crystals have severe
adhesion problems and cannot be dislodged without breaking them.
To reduce this problem, the depressions in glass pots should be
coated with a thin film of Corning vacuum silicone grease before the
protein-precipitant drop is added. The seeds will remain suspended on
top of the grease, and the final crystals are mounted for X-ray
diffraction use without the recurrence of this problem.
4. Once in the capillary the crystal is brought to the middle, and then
allowed to sink and adhere sufficiently to the inside wall of the
capillary so that the liquid can be moved over the crystal.

194
 7; Seeding techniques
(a) For crystals that fail to sink to the bottom and adhere to the
capillary wall, a long hair or whisker may be wedged against the
crystal to stop movement while liquid is drawn out.
(b) For soft crystals there is the danger that during this procedure
microseeds may be dislodged from them by the hair with obvious
consequences. A series of washes will minimize the number of
microseeds that will be transferred but not eliminate the risk that
one or more will still be present in the solution which is
transferred together with the crystal.
5. Once the crystal has been separated from the bulk of the mother
liquor, the hair withdrawn and the mother liquor ejected from the
capillary and returned to the original drop, the crystal should remain in
a small pool of liquid inside the capillary. Removing more of the
remaining solution from around the crystal may help diminish the
number of microseeds and aggregated material, and since the
solution may now be at a higher precipitant concentration due to
evaporation during the handling, transferring less of this solution may
avoid creating conditions which are unsuited to the seeding.

The crystal can be repeatedly washed in a stabilizing solution (typically the

reservoir solution is used) prior to transferring it to the new drop as described
in Protocol 4.

Protocol 5. Washing crystals

Equipment and reagents

• Multiwell sitting drop plate • Microscope
• Distilled water • Tray with source crystals
• Stabilizing solution • Tray to be seeded

Method
1. Fill four depressions of a multiwell sitting drop plate with about 100 ul
of solution from the reservoir where the seeds originated (or prepare a
solution identical to this).
2. Fill the reservoirs around these drops with 1.5 ml of distilled water to
maintain a high degree of moisture around the solution (Figure 4).
3. The crystal is repeatedly transferred and picked up from each of these
stabilizing solution drops until finally, it is picked up into the capillary.
Because the addition of stabilizing solution to the new drop would
unnecessarily modify the equilibrium, or dilute the equilibrated
protein-precipitant solution, it is best to minimize the amount of liquid
that remains around the crystal.
195
 E. A. Stura
Protocol 5. Continued
4. Remove the excess liquid with a small thin strip of filter paper or a
very thin capillary.
5. Resuspend the crystal in new mother liquor drawn in from the drop to
be seeded and return it to the well for equilibration and further growth.
Alternatively, after the series of washes, the crystal is allowed to sink
towards the open end of the capillary, so that when the capillary
touches the solution of the drop being seeded, the crystal falls directly
into this solution with little transfer of wash solution.

4.2.2 Macroseeding of needles

Macroseeding using needles as seeds is more complicated since needles have a
tendency to bend while being transferred from the original growth solution to
the new solution. The stresses created in the crystals during this process may
result in defects at each stress point, and each of these points may act as a
nucleation site for growth of new needles. By breaking the needles with a
sharp object (glass or metal) into smaller segments, the resulting fragments
can be used for macroseeding. The sharper the instrument the less damage
will be done to the seeds. A small number of these needle fragments are then
transferred to a stabilizing solution. From this point the procedure is
essentially the same as for prismatic crystals, as each is then transferred from
this solution to another container with the same stabilizing solution, and then
to a third and a fourth, to wash away any microseeds that may be transferred
to the pre-equilibrated growth medium.

5. Heterogeneous seeding
The principle that there is a lower energy requirement in adding to an existing
surface than in creating a new nucleus (Section 1.2) holds for many surfaces.
Such aggregation onto surfaces may be considered more of a problem than an
advantage. However, regular surfaces may offer a charge distribution pattern
which is complementary to a possible protein layer and could provide a
suitable starting point for the nucleation of new crystals. The work of
McPherson (21) with various inorganic minerals provides strong support for
the idea that regular planes are able to catalyse the nucleation of crystals of
macromolecules, even if the lattice dimensions of the crystalline minerals
differ from those of the resulting protein crystals. In these studies, nucleation
occurred preferentially on the mineral substrate at a lower degree of super-
saturation than was required for the same crystals to nucleate in the absence
of the minerals. Crystals of related macromolecules can also be used to induce
nucleation of proteins; the resulting crystals may maintain some, but not all,
of the lattice dimensions or symmetry axes of the initial seeds. In such cases,
196
 7: Seeding techniques
where the protein in the crystals from which the seeds are obtained is related
to the protein in the solution being seeded, the operation is termed cross-
seeding. When crystals of the same macromolecule are used to induce a
related crystal form under different crystallization conditions, an epitaxial
jump (by analogy with quantum jumps) has been achieved.

5.1 Cross-seeding
5.1.1 Cross-seeding between Fab-peptide complexes
In X-type light chain dimers, the dimers pack so as to form an infinite B-sheet
maintaining one cell dimension in common, 72.4 (± 0.2) A, along one of the
21 axes (22). Such packing in preferred planes for certain classes of protein
molecules, may indeed provide suitable surfaces for nucleation for other
members of that class. A similar observation was made in the course of our
work with different anti-haemagglutinin monoclonal Fab-peptide complex
crystals, where it was noticed that these have a common crystal lattice plane
with cell dimensions 73.0 (± 1.0) A along the 21 axis and 66.4 (± 2.5) A along
one of the other axes (19, 23, 24).
Within the description of the three-dimensional structure of the complex of
Fab 26/9 that recognizes the same six residue epitope of an immunogenic
peptide from influenza virus haemagglutinin (HA1 75-110) as Fab 17/9, it was
possible to understand the hierarchy in the crystal contacts responsible for the
differences and similarities between these crystal forms (25). In brief, 26/9 and
17/9 antibodies are very similar, but their interaction with the peptide are
slightly different. Structural and sequence analysis suggests that amino acid
differences near the peptide binding site are responsible for altering slightly
the specificity of 26/9 for three peptide residues. Since the peptide is essential
for one of the crystal interactions, we can understand the influence of peptide
length on the crystallization and the similarity in crystallization between these
antibodies. Cross-seeding, using the streak seeding method can bridge the gap
between the various peptide complexes of 26/9. Initial crystals were obtained
by spontaneous nucleation with a nine residue peptide (HA1 100-108). The
quality of these crystals was improved by using the streak seeding technique
as a microseeding method. Seeds from these crystals were used to search for
growth conditions of complex crystals of Fab 26/9 with longer peptides, for
which no conditions for spontaneous nucleation had been found. Both the 13-
mer (HA1 98-110) and the 23-mer (HA1 88-110) peptide-Fab mixtures
responded positively to the seeding. Seeds obtained from this cross-seeding
were used to seed repeat experiments to dilute out the effect of the hetero-
geneous seeds and optimize the crystallization conditions for the new
complexes. The crystals obtained from this second seeding diffract to 2.5 A
resolution.
A third anti-peptide antibody 21/8 also belonging to this same panel but
their heavy chains belong to different classes; Fab 21/8 is derived from an
197
 E. A. Stum
IgG
2b, while 26/9 is cleaved from an IgG2a. Nevertheless, seeds from Fab
26/9-13-mer complex crystals have been used to induce crystallization in
solutions of Fab 21/8-13-mer mixtures, under identical crystallization con-
ditions. The Fab 21/8-13-mer crystals first obtained by the streak seeding
experiment (Figure 5D) were very thin needles and although optimization of
the conditions resulted in better crystals, the real breakthrough was achieved
from the refined conditions which yielded large crystals of an unrelated form
by spontaneous nucleation.
Cross-seeding does not need to be carried out by streak seeding as, for
example, large crystals from the chicken mitochondrial aspartate amino-
transferase were used as seeds in the cross-seeding of the pig enzyme (6)
(initial cross-seeded crystals of the chicken enzyme were badly twinned but
were improved to X-ray quality in a second cycle of macroseeding). However
streak seeding can substantially increase the speed with which crystals can be
obtained.
5.1.2 Cross-seeding from native Fab to Fab-peptide complex
Anti-peptide Fab 50.1, that recognizes an epitope of the gp120 surface glyco-
protein of HIV-1, is an example where native crystals were used to seed the
Fab-peptide complexes. The native crystals can be grown in three different
morphologies. Spontaneous nucleation has not been observed for any of the
peptide complexes tested. The peptide lengths vary from 13-40 residues.
Crystals of the Fab-13-mer complex have now been obtained by streak
seeding with the native Fab crystals in 12-24% PEG 10000 pH 5-8 (26). In
the Fab-peptide solution most of the protein is found partitioned in a gel
phase covering the bottom of the drop. On addition of native seeds, crystals
grow by acquiring protein from the gel phase surrounding the seeds. The
morphology of the Fab-peptide complex crystals differs substantially from
that of native crystals (Figure 5E). It is also worth noting that while the
crystals of the native Fab are very mosaic, data have been collected on well

Figure 5. Photomicrographs of the results of crystallization experiments mentioned in

the text. (A) The initial precipitate obtained in the crystallization of the complex of anti-
peptide Fab' B13I2 with the 19 residue peptide corresponding to the C-helix of
myohaemerythrin. (B) When this precipitate was streaked onto drops under similar
conditions to those that yielded the precipitate crystal were obtained. (C) These crystals
were used in macroseeding experiments to yield X-ray quality crystals. (D) Cross-seeding
between anti-haemagglutinin Fab 26/9 and Fab 21/8. (E) Crystals obtained by
spontaneous nucleation from 21/8 under the optimized conditions for the cross-seeded
crystals. (F) Crystals of anti-HIV-1 Fab 50.1 complexed with a 13 residue peptide from the
glycoprotein gp120 sequence of the MN isolate. The crystals were obtained by cross-
seeding from native crystals (G) that grow spontaneously under similar conditions. (H)
Epitaxial nucleation on a cellulose fibre. The first crystals for this Fab nucleated from a
drop without fibres after a period of six months. The seeded crystals were obtained in
less than a week. Notice the number of crystals nucleated on the fibre compared to the
number nucleated separately.

198
 7: Seeding techniques
ordered crystals of the seeded complex that diffract to better than 2.8 A
(Figure 5F) sufficient for X-ray structure determination (27).

5.1.3 Cross-seeding chemically modified proteins

Because solubility changes as a result of modifications, proteins which have
been altered either by chemical means or by site-directed mutagenesis, may
not yield crystals spontaneously in cryslallization trials. Selenolsubtilisin, in

199
 E. A. Stura
which serine 221, the active serine, of the bacterial protease (28-30), is con-
verted into a selenolcysteine (31) is one such example. Even after extensive
efforts to better purify the engineered enzyme (32) crystals could only be
obtained by cross-seeding from crystals of the commercially available native
subtilisin. Although the quality of these seed crystals, obtained from com-
mercial grade enzyme was rather poor, the crystals of the selenolsubtilisin
obtained from the cross-seeding experiment were of good morphology and
size. The crystallization conditions under which the cross-seeding was done
were similar to those for the native subtilisin. After further optimization it
was possible to obtain good quality crystals, which were used for the structure
determination of the modified enzyme (32). Under the optimized conditions
some preparations nucleated spontaneously.

5.2 Epitaxial nucleation

Epitaxial nucleation is a particular instance of adhesion where the regularity
of the surface facilitates nucleation. Many substrates mediate adhesion of
proteins, and precautions may need to be taken to avoid this interaction.
Glass surfaces are generally siliconized, but even after this treatment crystals
are still found to be preferentially attached to the siliconized surfaces. The
strength of the interaction can be stronger than the forces that bond the
crystalline lattice. In crystallization trials it is possible to find many instances
in which crystals or microcrystals can be nucleated on cellulose fibres which
are accidentally present in the protein-precipitant drop (Figure 5H). Because
of the regularity of the fibres this can be considered a case of epitaxial
nucleation. In most cases microcrystals are also observed nucleating spon-
taneously away from foreign particles. The nucleation of crystals from
aggregates and oils may also be due to epitaxial nucleation. Here ordered
surfaces may be present within the random aggregation of macromolecules
which come out of solution as oils and precipitates. These surfaces may
provide platforms suitable for macromolecular nucleation, and may possibly
support three-dimensional crystal growth. The streak seeding technique can
be used not only with small unusable crystals but also with any promising
aggregate or precipitate, to test for the possibility that ordered planes within
such aggregates may be able to stimulate the growth of crystals, or that such
aggregates may be polycrystalline.
Seeds can be used to seed supersaturated solutions equilibrated under
conditions quite dissimilar from those from which the seed crystals originated.
The resultant crystals may result in a lattice similar but not identical to that of
the seeds. Epitaxial jumps can be induced by increasing the precipitant
conditions of the reservoir, without any manual seeding.
Crystals of the tissue factor-factor VII-5L15 complex (33) were obtained
using a combination of Protocol 5B and A. The epitaxial jump was accom-
plished by streak seeding a drop, equilibrated at the same precipitant con-
200
 7: Seeding techniques
centration as that from which the seeds were grown but at lower protein
concentration. Subsequently, the precipitant concentration of the well was
increased by adding NaCl to the reservoir solution causing a rise in the
precipitant and protein concentrations in the drop. Initially there was minimal
growth of form 1 crystals (because of the low protein concentration) later as
the precipitant concentration was raised, a more compact form nucleated off
the crystals of the first form. The crystals so obtained were thicker and larger
than the initial form 1 crystals, but similar to form 1 crystals grown under high
protein concentration. In this case the first crystal form (P321; a = b = 127.2
A, c = 110.7 A; Vm = 3.4 A3/Da) diffracted to 7 A and could not be
distinguished from the second form (P321; a = b = 67.2 A, c = 314.8 A; Vm =
2.7 A3/Da) that diffracted to 3.2 A by its morphology.

Protocol 6. Epitaxial jumps

Equipment and reagents

• Forceps • Seeding syringe
• Seeding probe • Pipette
• Precipitant • Microscope

A. Jumps without manual seeding

1. Set up drops under the usual crystallization conditions. Wait for
crystals to nucleate and grow. After growth is completed the level of
supersaturation of the drop should be substantially lower.
2. Increase the concentration of the reservoir. This may consist of a
series of gradual additions of precipitant to the reservoir or one large
jump in concentration. Observe the crystals after each addition of
precipitant for new crystals. These are likely to nucleate off the original
crystals and grow in a different morphology and direction from the
original crystals. An epitaxial jump may have occurred. Use streak
seeding to make use and propagate the new crystals.

B. Jumps by streak seeding

1. Set up drops under any desired supersaturated conditions. Different
protein and precipitant concentrations should be tried. To try to obtain
more tightly-packed crystals, the protein concentration should be
halved and new supersaturated conditions established at higher
precipitant concentrations.
2. Streak seed drops after equilibration with any crystal form obtained
for the same macromolecule. The seeds may not be able to enlarge
under the new conditions, typically characterized by a higher pre-
cipitant, lower protein concentrations, but may be able to provide a
template to enable a switch to a more compact crystal form.

201
 E. A. Stura
Protocol 6. Continued
3. Check for the development of a line. If a line develops use the crystals
obtained to seed other drops. If growth is slow, increase the
precipitant concentration.

C. Propagation of epitaxially grown crystals

1. Use seeds from either part A or part B to seed new drops. Once
crystals can be grown large enough for X-ray diffraction, these should
be checked for increased diffraction limit. Crystals which have been
grown in this manner and have a lower solvent content are termed
'squeezed' crystals.

Crystals of class I deoxyribose-5-phosphate aldolase from Escherichia coli

were originally obtained for the unliganded enzyme and in complex with its
substrate, 2-deoxyribose-5-phosphate P212121 with cell dimensions a = 183.1
A, b = 61.4 A, c = 49.3 A and a = 179.2 A, b = 60.5 A, c = 49.1 A,
respectively (34). In one instance, after these crystals were mounted the
coverglass was replaced and the chamber sealed as it is routinely done.
During the subsequent months the reservoir solution slowly increased be-
cause of evaporation through the petroleum jelly seal. When the experiment
was viewed months later, new small crystals had appeared. This is often the
case, except that the new crystals were morphologically different from those
which had been previously obtained. Seeds from these crystals were streak
seeded into pre-equilibrated drops and then macroseeded to obtain crystals
large enough for X-ray studies. The resulting crystals were even more stable
in the X-ray beam than the original crystal form and diffracted to 2.1 A
surpassing those of the first form which diffracted to 2.6 A. The space group
for the new form is P1, a = 49.2 A, b = 51.3 A, c = 54.0 A, a = 77.23°, B =
78.4°, y = 77.8°. A third example is the anti-testosterone Fab AN (35). In this
case an epitaxial jump was achieved by streak seeding from crystals grown in
PEG 600 to drops equilibrated in the same concentration of PEG 4000. The
results were obvious (Figure 5G); the first form consisting of thin needles (P21;
a = 55.1 A, b = 70.6 A, c = 66.3 A, B = 106.3°) the second of more prismatic
crystals (P312l; a = b = 72.0 A, c = 156.0 A).

6. Crystallization of complexes
6.1 Considerations in the crystallization of complexes
When crystallizing complexes, such as a receptor-ligand, an enzyme-
inhibitor, or an Fab-antigen complexes, it is important to consider the
resultant heterogeneity of the system. Both members of the complex will be
somewhat heterogeneous, and the resulting mixture will be composed of
202
 7: Seeding techniques
complexed and uncomplexed molecules in different ratios depending on the
molar ratio of the two molecules in the solution and the dissociation constant.
To reduce the heterogeneity it is important to optimize the number of
complexed versus the uncomplexed molecules.
For protein complexes with small ligands it is possible to increase the
number of complexed protein molecules by adding an excess of ligand.
Theoretically, the excess ligand that is necessary to achieve the desired ratio
of bound to unbound may be calculated from the dissociation constant if
known. In practice it is best to set up experiments at different protein:ligand
ratios, typically 1:1 to 1:20. Larger excesses are generally unnecessary and
may even inhibit crystal growth.
Co-crystallization of enzyme substrate complexes presents a different level
of complexity. Catalysis of the substrate into product will result in a mixture
of free enzyme, enzyme-product complex, and enzyme-substrate complex.
Such experiments are best attempted by using non-productive substrate
analogues, inhibitors that mimic the transition state, and undissociable end-
products. Triphosphate nucleotides, such as ATP, are easily hydrolysed,
hence, non-hydrolysable analogues such AMP-PNP, adenosine diphosphate
y-S (ADPyS), and their analogous guanosine derivatives are now commonly
used for crystallization instead of ATP or GTP. Vanadate, molybdate, and
tungstate are commonly used as phosphate mimics (36).
When the complex consists of two or more macromolecules of comparable
size the addition of an excess of one increases rather than decreases the
heterogeneity of the system. In such cases, if the affinity between the
macromolecules is 108 or better, the complex can be purified, otherwise it is
best to mix the macromolecules in the appropriate stoichiometric ratio. When
the stoichiometry of the system is not known it is best to set up the
crystallization of the complex at different receptor:ligand ratios. Uncom-
plexed molecules are likely to adhere to the lattice of the complex crystal, and
interfere with the growth of such crystals. In order for the crystal to continue
its growth without defects the unbound molecule must either become com-
plexed while still maintaining its lattice contacts, else it must break all lattice
bonds and diffuse away from the crystal surface to be replaced by a
complexed molecule. The energy involved in each of the lattice interactions
that must be broken will determine the inhibitory effect of the uncomplexed
molecule with respect to the growth of the complex crystal. Assuming that the
number of lattice bonds is proportional to the surface area of the molecule,
we expect that the inhibiting effect of an uncomplexed molecule will be
roughly proportional to molecular weight. Hence a small excess of the smaller
molecule is expected to be less damaging to the crystallization of the complex
than an excess of the larger molecule. When the crystallization proceeds
slowly, the crystal is less likely to incorporate unbound molecules as defects. If
we consider the deleterious effect of incorporating a defect into the lattice the
203
 E. A. Stum
absence of the larger molecule will carry a greater energy penalty, and hence
it will be less likely to occur. Therefore, it is better to have a slight excess of
the smaller macromolecule when growing complex crystals, except during the
initial search, when suitable growth conditions have not been established, and
the most ordered nucleus is likely to occur with a slight excess of the larger
macromolecule.
When the affinity between the molecules in the complex is not high, and the
off rate is substantial, we must screen for possible crystallization conditions
where the relative solubility of the complex is lower than that of the
uncomplexed molecules. It is also important to use high concentrations of
both macromolecules to push the equilibrium in favour of the complex,
although the use of lower concentrations will allow for a larger number of
trials. The number of trials can be minimized by using a screening approach as
described in ref. 3. When crystals are obtained, streak seeding can be used to
determine whether the complex or either macromolecule has crystallized.

6.2 Use of streak seeding in protein complex

crystallization
The optimal ratio of the two molecules in the complex for the nucleation of
crystals is not necessarily the optimal ratio for crystal growth. By using streak
seeding, it is easy to determine the optimal ratio and concentrations of
macromolecules experimentally from the response along the streak line.
Initial trials should be started as soon as possible and success in obtaining
microcrystals, microcrystalline aggregates, or even crystals can often be
achieved with less than absolutely pure macromolecules, as determined by
SDS-PAGE and isoelectric focusing (IEF) gels. In later stages, it is important,
as in all crystallizations, to attempt to obtain higher sample purity in order to
grow X-ray quality crystals (see Chapters 2 and 3). The use of affinity columns
ensures that the only macromolecules that can form complexes are present in
the crystallization trials. Such method can be advantageous as long as the
subsequent elution conditions do not affect adversely the macromolecules
being crystallized. The ability of the unliganded macromolecules to crystallize
generally correlates well with purity and their ability to also crystallize
complexes.

6.3 Analytical techniques for determination of crystal

content
6.3.1 Protein-protein complexes
SDS-PAGE (37) can be used to determine the composition of crystals of a
putative complex. Because oligomerization through the formation of
disulfides may occur, reduced and non-reduced gels should be used.
204
 7: Seeding techniques

Protocol 7. Preparing crystals for SDS-PAGE analysis

Equipment and reagents

• Multiwell sitting drop plate • Microscope
• Seeding probe • Washing solution
• Syringe with capillary • SOS-PAGE apparatus
• Cryo-loop

Method
1. Separate several crystals from the mother liquor from which they have
been growing, and wash them to remove residual mother liquor. The
procedure described for the handling of crystals for macroseeding
may be used here. Alternatively, the crystals can be lifted from the
drop using the same probe used for steak seeding or a loop as for
cryo-crystallography (Chapter 13).
2. If a skin has formed first remove the skin by running the probe around
the drop. Skins are often correlated with oligomerization through the
formation of disulfides (it is advisable to analyse the sample under
both reducing and non-reducing conditions). Place the probe under
the crystal and lift it out from the solution. Repeated attempts are
generally needed.
3. Dissolve the crystals in distilled water typically with a final volume of a
few microlitres. If a probe is being used, just touch the end of the
probe with the crystal on top and the crystal will drop into the water
and dissolve. More crystals can then be picked up if the crystals are
small.
4. Check under a microscope that the crystals have dissolved. Crystals
that do not dissolve under these conditions may have to be dissolved
under more acidic, more basic, or higher salt conditions, or by adding
urea or SDS directly to the crystals. When using urea or high salt to
dissolve the crystals, the urea or the salt will have to be dialysed out
before running the polyacrylamide gel. Silver staining of the gel may
be needed if the crystals are small.
5. Run the solutions used for crystallization in one of the lanes. Some
oligomerization will occur as a result of boiling in SDS under non-
reducing conditions. This should be more accentuated for the control,
which will be at higher protein concentration than the dissolved
crystals.

The comparison between the control and the dissolved crystals should be
able to determine the macromolecular content of the crystals.
205
 E. A. Stum

6.3.2 Complexes with small ligands

For small charged ligands a native polyacrylamide gel electrophoresis (native
PAGE) (38, 39) using a PhastGel Gradient 8-25 (Pharmacia, Piscataway, NJ),
or other gel system, can be run with separate lanes for the complexed and
uncomplexed macromolecules, and for the dissolved crystals (Protocol 6).
The comparison may allow for the determination of the crystal content. This
has been successful for Fab-peptide complexes. The bound peptide modifies
the mobility of the Fab, and usually shows as an identifiable shift in the
position and distributions of the protein bands.
6.3.3 Chemical reactions in the crystal
The channels in protein crystals are typically large enough to allow for the
diffusion of many small molecules throughout the lattice. Heavy-atom
derivatization relies on this fact (Chapter 13). If a chemical reaction can be
done on the compounds which are presumed bound to the protein in the
complex, so that colour or fluorescence is developed, the reaction can be tried
on the crystals. For example, in the crystallization of steroid complexes of the
anti-progesterone Fab DB3 (40), the presence of the steroids was verified
before collecting X-ray data. Since progesterone has a free ketone group at
position 3 on the A-ring, it reacts with 2,4-dinitrophenyl hydrazine. The
reagent was diffused into the crystals, and for 5 min dilute HC1 was added. As
the outside of the crystals dissolved, a brownish red precipitate developed,
demonstrating the presence of the steroid. As controls, the same reaction was
repeated with crystals of 'uncomplexed' Fab DB3 and of an unrelated Fab.
Those of the unrelated Fab remained clear. Those of the 'uncomplexed' Fab
DB3, did not dissolve, but developed a light yellow colour. Subsequently it
was determined that the Fab preparations contained 10-20% progesterone.
Crystals of truly uncomplexed DB3 could be grown from steroid-free
preparations, where the antibody was produced in cell culture rather than
ascites.

7. Concluding remarks
The application of seeding methods in macromolecular crystallization has
proven invaluable for obtaining high resolution X-ray quality crystals when
conventional methods have failed. It provides a means of analysing many
conditions without requiring large amounts of protein solution. Streak seed-
ing is particularly valuable as it provides a fast method of analytical seeding
with easy visualization of the results. Cross-seeding is a powerful tool to
crystallize a given protein with seeds from a related protein. The application
of micro- and macroseeding methods can result in the production of large
single crystals for X-ray structure determination. Without such methods many
projects would not have been viable.
206
 7: Seeding techniques

Acknowledgements
The contribution of Dr Ian Wilson to the editing of the previous version as
well as the support provided through his grants by National Institutes of
Health Grants AI-23498, GM-38794, and GM-38419 is here acknowledged.
The French Atomic Energy Commission (CEA) has provided support for the
revisions.

References
1. Stura, E. A., Nemerow, G. R., and Wilson, I. A. (1992). /. Cryst. Growth, 122, 273.
2. Jancarik, J. and Kim, S.-H. (1991). J. Appl. Cryst., 24,409.
3. Stura, E. A., Satterthwait, A. C, Calvo, J. C, Kaslow, D. C., and Wilson, I. A.
(1994). Acta Cryst, D50, 408.
4. Carter, C. W., Jr. and Carter, C. W. (1979). J. Biol Chem., 254, 12219.
5. Betts, L., Frick, L., Wolfenden, R., and Carter, C. W., Jr. (1989). J Biol. Chem.,
264, 6737.
6. Eichele, G., Ford, G. C., and Jansonius, J. N. (1979). J. Mol. Biol, 135, 513.
7. Wilson, I. A., Rini, J. M., Fremont, D. H., and Stura, E. A. (1990). In Methods in
enzymology (ed. Langone, J. J.), Academic Press, London. Vol. 203, p. 270.
8. McPherson, A. (1982). Preparation and analysis of protein crystals. Wiley, New
York.
9. Blundell, T. L. and Johnson, L. N. (1976). Protein crystallography. Academic
Press, New York.
10. Arakawa, T. and Timasheff, S. N. (1985). Methods in enzymology, (eds Wychoff,
H. W., Hirs, C. H. W., and Timasheff, S. N.), Academic Press, London, Vol. 114,
p. 49.
11. Feher, G. and Kam, Z. (1985). In Methods in enzymology, ibid ref. 10. Vol. 114,
p. 77.
12. McPherson, A. (1985). In Methods in enzymology, ibid ref. 10. Vol. 114, p. 112.
13. Vekilov, P. G., Monaco, L. A., Thomas, B. R., Stojanoff, V., and Rosenberger, F.
(1996). Acta Cryst., D52, 785.
14. Rosenberger, F., Vekilov, P. G., Muschol, M., and Thomas, B. R. (1996). J. Cryst.
Growth, 168,1.
15. Thaller, C., Weaver, L. H., Eichele, G., Karlsson, R., and Jansonius, J. (1981).
J. Mol. Biol., 147,465.
16. Thaller, C., Eichele, G., Weaver, L. H., Wilson, E., Karlsson, R., and Jansonius,
J. N. (1985). In Methods in enzymology, ibid, ref 10, Vol. 114, p. 132.
17. Stura, E. A. and Wilson, I. A. (1991). J. Cryst. Growth, 110, 270.
18. Stura, E. A. and Wilson I. A. (1990). Methods, 1, 38.
19. Stura, E. A., Stanfield, R. L., Fieser, T. M., Balderas, R. S., Smith, L. R., Lerner,
R. A., et al. (1989). J. Biol. Chem., 264, 15721.
20. Stura, E. A., Johnson, D. L., Inglese, J., Smith, J. M., Benkovic, S. J., and Wilson,
I. A. (1989). /. Biol. Chem., 264, 9703.
21. McPherson, A. (1989). Sci. Am., 260, 3, 62.
22. Schiffer, M., Chang, C.-H., and Stevens, F. J. (1985). J. Mol. Biol., 186,475.
207
 E. A. Stum
23. Schulze-Gahmen, U., Rini, J. M., Arevalo, J. H., Stura, E. A., Kenten, J. H., and
Wilson, I. A. (1988). J. Biol. Chem., 263,17100.
24. Wilson, I. A., Bergmann, K. F., and Stura, E. A. (1986). In Vaccines '86 (ed. R. M.
Channock, R. A. Lerner, and F. Brown), pp. 33-7. Cold Spring Harbor
Laboratory, Cold Spring Harbor, NY.
25. Churchill, M. E., Stura, E. A., Pinilla, C., Appel, J. R., Houghten, R. A., Kono, D.
H., et al. (1994). J Mol. Biol., 241, 534.
26. Stura, E. A., Stanfield, R. L., Rini, J. M., Fieser, G. G., Silver, S., Roguska, M., et
al (1992). Proteins, 14,499.
27. Rini, J. M., Stanfield, R. L., Stura, E. A., Salinas, P. A., Profy, A. T., and Wilson, I.
A. (1993). Proc. Natl. Acad. Sci. USA, 90, 6325.
28. Markland, F. S., Jr. and Smith, E. (1971). In The enzymes, 3rd edn (ed. P. D.
Boyer), Vol. III, pp. 561-608. Academic Press, New York.
29. Kraut, J. (1971). In The enzymes, 3rd edn (ed. P. D. Boyer), Vol. III, pp. 547-60.
Academic Press, New York.
30. Neidhart, D. J. and Petsko, G. A. (1988). Protein Eng., 2, 271.
31. Wu, Z.-P. and Hilvert, D. (1989). J. Am. Chem. Soc., III, 4513.
32. Syed, R., Wu, Z. P., Hogle, J. M., and Hilvert, D. (1993). Biochemistry, 32, 6157.
33. Stura, E. A., Ruf, W., and Wilson, I. A. (1996). J. Cryst. Growth, 168, 260.
34. Stura, E. A., Ghosh, S., Garcia-Junceda, E., Chen, L., Wong, C.-H., and Wilson, I.
A. (1995). Proteins, 22, 67.
35. Stura, E. A., Charbonnier, J. B., and Taussig, M. J. (1999). J. Cryst. Growth, 196,
250.
36. Arvai, A. S., Bourne, Y., Williams, D., Reed, S. I., and Tainer, J. A. (1995).
Proteins, 21,70.
37. Laemmli, U. K. (1970). Nature, 227, 680.
38. Andrews, A. T. (1981). In Electrophoresis. Theory, techniques and biochemical
and clinical applications (ed. A. R. Peacocke and W. F. Harrington), p. 63.
Clarendon Press, Oxford.
39. Hames, B. D. (1981). In Gel electrophoresis: a practical approach (ed. B. D.
Hames and D. Rickwood). IRL Press Limited, London, Washington DC 4.
40. Stura, E. A., Arevalo, J. H., Feinstein, A., Heap, R. B., Taussig, M. J., and Wilson,
I. A. (1987). Immunology, 62, 511.

208
 8

Nucleic acids and their complexes

A.-C. DOCK-BREGEON, D. MORAS, and R. GIEGE

1. Introduction
At first glance crystallizing nucleic acids poses the same problems as crystal-
lizing proteins since most of the variables to investigate are alike. It is thus
astonishing that crystallization data banks (1) that describe so many successful
protein crystallizations are so poor in information on nucleic acids. This relies
on the physico-chemical and biochemical characteristics of nucleic acids dis-
tinguishing them from proteins. The aim of this chapter is to underline features
explaining the difficulties often encountered in nucleic acid crystallization and
to discuss strategies that could help to crystallize them more readily, either as
free molecules or as complexes with proteins. Other general principles, in
particular for RNA crystallization, are discussed in ref. 2.
Among natural nucleic acids only the smaller ones provide good candidates
for successful crystallizations. Large DNAs or RNAs can a priori be excluded
because of their flexibility that generates conformational heterogeneity not
compatible with crystallization. Thus the smaller RNAs with more compact
structures (with 75-120 nt), especially transfer RNAs (tRNAs), but also 5S
RNA, were the first natural nucleic acids to be crystallized (3, 4). At present
attempts are being made with other RNA systems, such as ribozymes and
introns, fragments of mRNA, viroids, viral and other tRNA-like RNAs,
SELEX-evolved RNAs, and crystallization successes leading to X-ray struc-
ture determinations were reported for RNA domains of up to 160 nt long,
with the resolution of the P4-P6 domain of the self-splicing Tetrahymena
intron (5).
The recent excitement in nucleic acid crystallography, and particularly in
RNA crystallography, have partly been due to technological improvements in
the preparation methods of the molecules. Advances in oligonucleotide
chemical synthesis provide opportunity for making large amounts of pure
desoxyribo- and more recently of ribo-oligomers of any desired sequence.
This led to the crystallization of a number of DNA and RNA fragments and
was followed by the co-crystallization of complexes between proteins and
such synthetic fragments. Transcription methods of RNAs from synthetic
DNA templates were also essential for rejuvenating the structural biology of
 A.-C. Dock-Bregeon et al.
RNAs. In the case of complexes of proteins with RNAs, the main difficulty
was to purify large quantities of homogeneous biological material with well
defined physico-chemical properties. The problem has now been overcome in
many cases and problems of larger complexity are now addressed such as
improved crystallizations of ribosome. Virus crystallization will also be briefly
discussed.

2. Preparation of nucleic acids

General aspects on the characterization of nucleic acids and advice for the
preparation of homogeneous samples aimed at crystallogenesis are also
outlined in Chapter 2.

2.1 Synthetic nucleic acid fragments for crystallogenesis

2.1.1 Design of DNA fragments
Because of their large size and plasticity, genomic DNA molecules cannot be
crystallized as such. Thus the first problem concerns the choice of the
appropriate DNA fragments amenable to crystallization. It can be dictated by
the biological significance of the fragment, methodological aspects related
with its preparation, or considerations on its crystallizability potency.
Critical evaluation of results on DNA oligonucleotide crystallizations have
shown that these molecules crystallize in a limited number of crystal packing
families (6), the molecules adapting their conformation according to sequence
and crystallization conditions. For example, in the B-form of the d(CGCG-
AATTCGCG)2 dodecamer duplex (7), the terminal CGCxxxxxxGCG boxes
led to crystallization in the orthorhombic P2 ] 2 1 2 1 space group by favouring
formation of specific hydrogen bonds between the minor grooves of the
staggered duplexes, whereas the presence of C residues at particular positions
(i.e. positions 3 and 6) would allow a major groove-backbone interaction and
crystallization in the trigonal R3 space group (8). The practical consequence is
the possibility to design B-DNA molecules containing packing driving boxes
that will guide crystallization in well defined crystal lattices (6).
The biological relevance of the sequence and the stability of the nucleic
acid moiety (and for DNA:protein complexes, the stability of the complex)
have to be taken into account. Also the size of the oligonucleotide and the
way it is terminated, blunt or extended end, can both have advantages accord-
ing to packing or stability. The best bet is to first try the sequence that is
closest to the biological one. That can be obtained from sequence compari-
sons (9) and from structural probing data in solution (10) designed to find the
most stable structures, closing helices by, e.g. C-G rather A-U base pairs.
When crystallization is not successful the design can be more difficult; try to
cut out what seems to provide instability and try to add or eliminate
symmetry.
210
 8: Nucleic acids and their complexes

2.1.2 Design of RNA fragments

Except in the few cases where small RNAs have rather globular and compact
structures (e.g. tRNAs, hammerhead ribozymes), RNAs are large, flexible,
and multidomain molecules. Such characteristics are detrimental for crystal-
lization and make RNA crystallization challenging. Therefore use of bio-
logical knowledge gained for instance by comparative computer analysis of
RNA sequences (9) and/or structural solution studies of these molecules with
chemical or enzymatic probes (11,12) followed by computer modelling (13) is
mandatory to define compact domains which in a second step can be prepared
by chemical or in vitro transcription methods. RNA domains with pre-defined
function can also be generated by combinatorial SELEX-type methods (14).
Such RNA domains often contain motives, as internal loops, bulges, non-
Watson-Crick pairs, pseudo-knots, which may be prepared per se for crystal-
lization purposes. Examples include a synthetic RNA 12-mer that folds in a
duplex structure containing two G(anti)-A(anti) base pairs (15), a 34-mer
ribozyme with a 13-mer DNA inhibitor (16), an RNA 12-mer containing the
Escherichia coli Shine-Dalgarno sequence (17), an RNA helix incorporating
an internal loop with G-A and A-A base pairing (18), and a duplex RNA
mimicking the amino acid acceptor stem of E. coli tRNAAla (19). A promising
approach consists of engineering crystal contacts by introducing appropriate
structural modules in RNA domains. This was done by using the GAAA
tetraloop:tetraloop receptor interaction and allowed crystallization of group II
introns and hepatitis delta virus ribozyme RNA constructs. The method led to
crystals diffracting at 3.5 A resolution (20). Knowledge-based design of RNA
motifs was also essential for the crystallization of RNA:protein complexes (21).
2.1.3 Chemical synthesis of DNA fragments
Synthesis of DNA has been automated and DNA synthesizers are commer-
cially available. At present it is straightforward to prepare DNA fragments
(up to 100 nt) and most molecular biology institutes provide this facility. Due
to technical advances, production costs of deoxyoligonucleotides is decreasing
and now it is often advantageous to order the oligonucleotides from industrial
producers. However, large scale synthesis of DNA fragments for struc-
tural studies remains expensive and requires 10 umol solid support resins (i.e.
~ 20 mg of a 20-mer can be synthesized in such a way).
After cleavage from the resin by ammonia, the final product is generally
purified by HPLC. The expected DNA is contaminated by shorter molecules
that result from incomplete reaction within a cycle of synthesis when the
previous product was not fully deprotected. Other contaminants result from
side-reactions, like depurination that may occur during the acid treatment for
the removal of the 5'-protecting group at the end of each cycle. Purification
can be performed before or after the complete deblocking of the protecting
groups by ammonia (22).
211
 A.-C. Dock-Bregeon et al.
The purification procedure calls for the different properties of oligo-
nucleotides which are electrolytes with some hydrophobic regions. The resins
are generally adapted to HPLC. These are ion exchangers or reverse-phase
chromatography (RPC) columns. They are generally made of silica particles
modified with functional groups in C4, C8, or C18.They can be used in two
different ways, according to the counterion choice. With ammonium ions,
retention of oligonucleotides is mainly due to hydrophobicity and this method
is used to separate oligonucleotides of similar length but of different sequences.
With triethylammonium ions, the ion pairing phenomena comes into play; the
oligonucleotides are adsorbed to the stationary phase via their counterions. In
this case the strength of the interaction is dependent on the hydrophobicity of
the counterion and also on the length of the oligonucleotide, since the strength
of the interaction is proportional to the charge of the oligonucleotide (22). In
both cases, the oligonucleotides are eluted by increasing concentrations of
acetonitrile. An advantage of the second method is that triethylammonium
salts are volatile, so that products are easy to recover by lyophilization.
2.1.4 In vitro transcription methods for RNA preparation
Another way of producing large amounts of RNA is by in vitro transcription
with phage polymerases. The most used is T7 RNA polymerase, which is
cloned and over-produced (23). The yield of in vitro transcription can reach
several hundred moles of transcripts per mole of template. When used for
RNA production, the template is made according to the strategy described in
Protocol 1. Note that the procedure is simplified when short molecules (12-35
nt) are synthesized since in this case the template can be synthesized directly.
There are, however, a number of drawbacks that can limit the use of in vitro
RNA synthesis:
(a) The yield depends on the sequence of the 5'-end of the RNA product, i.e.
the +1 to +6 region that is part of the promoter of the polymerase. Thus,
the polymerase works more efficiently when residue +1 is a G (23, 24).
For poorly transcribed sequences, transcription yield may be improved by
adding synthetic polyamines in incubation mixtures (25) or by optimizing
experimental conditions by statistical methods (26). An alternative method
consists of transcribing precursor RNAs with strong promoters that are
processed at the desired position by RNase H after hybridization of the
precursor with an appropriate DNA oligonucleotide (27).
(b) In vitro transcription produces RNAs which lack modified nucleosides. In
the case of tRNAs, most transcripts are correctly recognized by their
cognate synthetases but have less stable tertiary structures (28).
(c) A major problem is transcription termination, since polymerase some-
times adds one or two nucleotides at the 3'-end.
For small RNAs resulting from in vitro transcription, the desired molecule
can be separated from the template and the nucleotide monomers by gel
212
 8: Nucleic acids and their complexes
filtration. Transcripts can be further purified and resolved in individual
species by electrophoresis in polyacrylamide gels. HPLC methods also give
good results. Other methods, of more general use for large scale preparations
of virtually any homogeneous RNA sequence, are based on transcription of
the RNA together with flanking ribozyme sequences, so that the desired
RNA is yielded after self-cleavage, and the 5'-promoter sequence as well as
the heterogeneous 3'-end eliminated (29).

Protocol 1. In vitro synthesis of RNA by a molecular biology

strategya

Equipment and reagents

• DNA synthesizer (DNA oligonucleotides • Standard reagents for molecular biology of
may be of commercial origin) RNA
• Gel electrophoresis and HPLC equipment • T7 RNA polymerase

Method
1. Construct an insert ending with restriction sites, and containing the T7
promoter and the RNA sequence. This is made by ligation of synthetic
DNA oligomers chosen to hybridize unambiguously in tandem so as to
give the correct, double-stranded, sequence.
2. Insert this synthetic gene into a plasmid, digested with the appropriate
restriction enzymes.
3. Amplify the plasmid by cell culture.
4. Extract the DNA.
5. Linearize the DNA template at the restriction site.
6. Transcribe the DNA template in an appropriate medium containing the
polymerase and the nucleotide monomers.
7. Remove the RNA from the transcription medium, and if needed purify
by gel electrophoresis or HPLC.

aDetailed experimental procedures are described in refs 23-25.

2.1.5 Chemical synthesis of RNA fragments

Chemical synthesis is the method of choice for the preparation of short
RNAs. RNA, however, is more difficult to make than DNA, because of the
need to protect the 2'-OH group of the ribose. Several protecting groups have
been designed and the automated solid phase method has been adapted for
RNA synthesis (e.g. ref. 30). Commercial DNA synthesizers can be used with
RNA monomers with the same facilities as for DNA synthesis, but giving
somewhat lower yields. The great advantage of chemical upon enzymatic
synthesis is that modified nucleosides, or even deoxynucleotides, can be
213
 A.-C. Dock-Bregeon et al.
introduced at a specific position (30), and that any sequence can be designed
(the disadvantage is the lack of synthons for many of the modified nucleotides
one would like to insert in synthetic RNAs). Homogeneous preparations
of milligram quantities of chemically synthesized RNA, devoid of false
sequences and of incompletely deprotected material, and suitable for
crystallogenesis, can be obtained by anion exchange HPLC (31).
Chemical synthesis of the tetradecamer U(UA)6A in the 10 mg range has
led to the crystallization of this short RNA (32). Other examples concern
crystallization of a mispaired RNA double helix (33) and that of 15-mer and
19-mer RNA sequences corresponding to a hairpin or a helix with an internal
loop and to an RNA pseudo-knot (31).

2.2 Preparation of natural small RNAs

2.2.1 Sources
These preparations concern essentially tRNAs, the decoding molecules of the
genetic message, and 5S RNAs, a ribosomal constituent. 5S RNA can be
prepared from ribosomes by a phenol extraction and is separated from
tRNAs and other RNAs by molecular sieving (34). Crystals have been
obtained with 5S RNA from Thermits thermophilus (35) and E. coli (4).
Bulk tRNA from yeast Saccharomyces cerevisiae and E. coli is com-
mercialized (e.g. by Boehringer or Sigma). Purification of a single species of a
yeast tRNA is better done starting from brewer's yeast bulk tRNA because of
the structural integrity of their -CCA at 3'-end, which it not the case for
tRNAs from baker's yeast (36). Other sources need a phenol extraction from
the cells (Protocol 2). Thermophilic bacteria provide RNAs of potential great
interest for crystallization, since these are more stable at higher temperature
(37). Halophilic organisms may represent another interesting alternative for
preparation of RNAs or nucleic acid-protein particles intended for improved
crystallization, as was exemplified in the case of ribosomes (38).

Protocol 2. Preparation of small RNAs by phenol extraction from

cells

Equipment and reagents

• Vortex and centrifuge • Extraction buffer: 0.1 M Tris-HCI pH 7.5,
20 mM
• Phenol Mg(Ac)2, 1 mM EDTA
• Ethanol and ether

Method
1. Prepare phenol by adding 50% (w/w) water. When melted add a few
drops of 1 M KOH to bring the pH of the supernatant around 7.0.
2. Suspend the cells (5 ml/g) in extraction buffer.

214
 8: Nucleic acids and their complexes
3. Add to the cell suspension the same volume of phenol. Shake
vigorously for 30 min at room temperature and centrifuge 5 min at
3000 g (room temperature). Recover the upper phase.
4. Add 10-20% of the initial volume of extraction to the phenol phase,
shake vigorously, and recover again the upper phase. Mix the aqueous
phases.
5. Repeat steps 3 and 4 (with some cells, like T. thermophilus, it is
advisable to add 0.1% SDS in the aqueous phase for the second
extraction). Phenol may be removed from aqueous phase by ether
extraction. Caution: ether is volatile and easily flammable.
6. Precipitate RNAs with ethanol (see Protocol 3).

Advances in genetics has made cloning a possibility when a specific tRNA is

chosen. This procedure was first used for E. coli tRNAGln which was crystal-
lized with glutaminyl-tRNA synthetase (39). Cloning allows over-production
of a single isoacceptor and then proper purification. However, this can lead to
under-modification of the tRNA, and in turn to microheterogeneous tRNA
samples, when modifying enzymes become limiting for processing of large
quantities of overexpressed tRNAs.
2.2.2 Purification of transfer RNAs
With natural tRNAs the purification problem is complicated since bulk tRNA
contains about 60 different species with similar structures. Also the quantities
of purified tRNA species needed for crystallization projects have to be in the
5-50 mg range, which excludes purification by effective micromethods (e.g.
2D gel electrophoresis).
Countercurrent fractionation is a powerful first step in a tRNA purification
procedure, but is no longer operating in most laboratories because of the com-
plexity of the instrumentation. Selectivity relies upon differential solubility of
individual tRNAs in two solvents and of their distribution between an aqueous
and an organic phase. The method allows large quantities material to be
handled, typically 5 g, and enrichments can be excellent (40).
Different chromatographic supports interacting with the negatively charged
tRNAs have been used, such as DEAE-Sephadex (41) or hydroxyapatite (42).
Ionic interactions take place between phosphates of the tRNA and positively
charged groups of the DEAE matrix or calcium ions of the hydroxyapatite
crystals. Additional weak interactions, sensitive to the presence of Mg2+ ions
or urea, to pH and temperature, are tuned by the tRNA structures. The resolu-
tion of such columns is limited since electrostatic forces, related to the number
of accessible phosphates are poor discriminators of tRNA species. BD-
cellulose provides greater resolution; it is a DEAE-cellulose modified by the
addition of benzoyl groups (43). The tRNAs are sorted by electrostatic inter-
actions between phosphate and DEAE groups, and hydrophobic interactions
215
 A.-C. Dock-Bregeon et al.
between the accessible bases and the benzoyl moieties. Resolutive RPC
systems consisting of an inert support coated with a quaternary ammonium of
high molecular weight not miscible with water were also employed (44).
Several of the classical ionic exchange methods were adapted to FPLC or
HPLC systems (45). We obtained excellent results with monoQ columns
(Pharmacia or equivalent) which can purify to homogeneity a tRNA species
starting from 20% enriched preparations.
Hydrophobic interaction chromatography (HIC) is efficient for tRNA
purification. It was first used with a Sepharose 4B support and a reverse
gradient of (NH4)2SO4 (46), so that tRNAs are separated according to their
solubility. The advantages of HIC are its high resolution, the possibility to
process large quantities of material (1 ml of Sepharose 4B can adsorb 8 mg of
unfractionated tRNA) (46), and its adaptation to HPLC (HIC has improved
with the n-alkylated silica supports available for medium and high pressure
chromatography). A resolutive HPLC system using C4-bonded silica gel has
been reported (47). A disadvantage of HIC is the presence of salt at high
concentration in the enriched or pure tRNA fractions. Getting rid of
(NH4)2SO4 is necessary, especially if ethanol precipitation is the following
step. It can be done either by dialysis or by buffer exchange in a concentration
set-up (see Protocol 4).
Purification of a single tRNA species results from the combination of these
different methods. For crystallization, the need of large quantities of pure
material (> 5 mg) directs the choice of the first steps to methods that can
handle large quantities of material; e.g. fractionation on hydrophobic matrices
with elution by reverse (NH4)2SO4 gradients. The last steps will be the most
resolutive (RPC, HIC).
2.2.3 General principles for RNA handling
Beside the problem of sensitivity towards RNases which requires work in
sterile conditions, RNAs are sensitive to alkaline hydrolysis (48) and there-
fore alkaline pH should be strictly avoided. The cleavage of the polyribo-
nucleotide chain is also favoured by some metal ions, of which the most
effective is lead (49). A chelating agent, like EDTA, is generally introduced
into buffers, at a concentration 0.1-0.5 mM, to complex the traces of heavy
metals. At the end of the purification the RNA preparation must be checked
for its integrity. The simplest method is electrophoresis in a denaturing
polyacrylamide-urea gel.
Concentrations are obtained from optical density measurements at 260 nm
with

where e = 25 ml/mg (this e value applies for most RNAs, but for exact
measurements it may be necessary to determine it experimentally) (50), / is
the optical path in cm, and c the concentration in mg/ml.
216
 8: Nucleic acids and their complexes
Two ways described in Protocol 3 and Protocol 4 can be used to concentrate
RNAs. The method described in Protocol 4 is convenient to change the
solvent. When the RNA is concentrated to a small volume, dilute it in the new
solvent and concentrate again. Repeat several times.

Protocol 3. Concentration of nucleic acids (20-120 nt) by ethanol

precipitation

Equipment and reagents

• Refrigerated centrifuge • Ethanol
• Deep freezer

Method
1. Prepare the solution. It should contain Mg2+ ions (> 2 mM) and a Na
salt such as Na acetate (> 10 mM, generally at pH 6.0). For good
recovery, the RNA (e.g. tRNA) solution should be > 0.1 mg/ml. If not,
raise the Na acetate concentration to > 100 mM. Take care that the
solution does not contain too much salt (i.e. after a chromatography,
dialyse first in water).
2. Add two or three volumes of ethanol (best quality). The precipitate
forms.
3. Leave to precipitate completely at -20°C (2 h or more) or at -80°C (20
min or more). For the shortest oligonucleotides or low concentrations,
use the lowest temperature.
4. Centrifuge at the lowest possible temperature, 10 min at > 5000 g
should be sufficient.
5. Dry the pellet under vacuum in the presence of solid KOH.
6. Dissolve the pellet in the desired amount of buffer.

To gain more homogeneity, the RNA solution to be crystallized is dialysed

thoroughly in a buffer at low concentration of Mg2+. For tRNA crystallization
one can use a 2 mM MgCl2 and 10 mM Na cacodylate buffer at pH 6.0. The
same result can be obtained by buffer exchange. The RNA samples can be
stored frozen in such solution at -20 °C or -80 °C.

Protocol 4. Concentration of nucleic acids on a membrane

Equipment and reagents

• Amicon or Centricon-type dialysis concen- • Dialysis membranes
trators • Compressed nitrogen

217
 A.-C. Dock-Bregeon et al.
Protocol 4. Continued
Method
1. Prepare a set-up of the type Amicon (for large volumes) or Centricon
(for volumes of a few millilitres). Use membranes of correct cut-off
(usually 10000).
2. Concentrate by pushing the solvent through the membrane under
nitrogen pressure (for the Amicon set-up) or by centrifugation (for
Centricon). The RNA (or DNA oligonucleotide) concentrates on the
membrane.
3. Recover the solution when the desired volume is obtained.

3. Crystallization of nucleic acids

Several examples (arbitrarily chosen) of oligonucleotide crystallizations are
given in Table 1. Notice that some oligonucleotides have palindromic self-
complementary sequences. Such sequences are favourable for crystallo-
genesis and were among the first to be crystallized. Additional data are in refs

Table 1. Some examples of crystallization conditions of oligonucleotides

Sequence Temp (°C) Oligo Concentration Crystals Ref.

Precipitant Buffera Spermine Mg2+b
DNA A-form
GGCCGGCC
MPD 30% - 1.2 mM 25 mM 0.6 mM 3.0 mM P4321g2 54
pH7.0 2.25 A
CTCTAGAG
MPD7 vs50% 18 1.2mM 60mM 1 mM 25 mM P41212 55
pH 6.8 or P43212
2.15A
GTACGTAC
MPD 5 vs30% 20 2.0 mM 14 mM 8mM 15 mM P43212 56
pH6.0

DNA B-form
CGCATATATGCG
MPD 10vs40% - 0.5 mM - 0.4 mM 22 mM P212121 57
Mg(Ac)2 2.2 A
CCAAGATTGG, with G:A mismatch
MPD 45% 4 3.0 mM - None 0.7 M C2 58
1.3 A
5'-ACCGGCGCCACA
TGGCCGCGGTGT-5'
MPD 40% 4 1.0 mM 50 mM 1.2 mM 18 mM R3 8
pH 6.0 Mg(Ac)2 2.8 A
Me
CCAGGC CTGG
MPD 40% 4 2.0 mM 20 mM 0.0 mM 50 mM P6 59
microdialysis pH 7.5 2.25 A

218
 8: Nucleic acids and their complexes

Table 1. Continued

Sequence Temp (OC) Oligo Concentration Crystals Ref.

Precipitant Buffera Spermine Mg2+b
DNA Z-form

CGCGCG
Isopropanol 5% 2 mM 30 mM 10 mM 15 mM P21212, 60
pH 7.0 0.9 A
5 5
m CGTAm CG
MPD 8 vs 50% - 4 mM 30 mM 7 mM 10 mM P212121 61
pH 7.0 1.2 A
(5BrCG)3
MPD 10 vs 60% 18 or 37 0.5 mM 20 mM - 200 mM P212121 62
pH 6.5 NaCI 1.4 A
m5CGUAm5CG
MPD8.5 vs 30% Room 4mM 28 mM - 15mM P212121 63
pH 7.0 1.3 A

Four-stranded intercalated DNA

cccc
MPD 20% - 2.7 mM 100 mM - - I23 64
pH 5.5 2.3 A

RNA:DNA hybrid

r(GCG)d(TATACGC)
MPD 40% - 1.5 mM 30 mM 8 mM 15 mM P212121 65
pH 6.0 1.9 A

RNA
U(UA)6A
MPD 35% 35 4 mM 40 mM None 0.4 M P212121 32
pH 6.5 2.25 A
5'-GGCC(GAAA)GGCC-3', with internal loop
PEG 400 30% Room 2 mM 50 mM - 5 mM P6522 18
Tris MnCI2 2.3 A
pH 7.5 + 20mM
NaCI
5'-GGGGCUAc 25 1 mM 12.5 mM 1 mM 50 mM C2 19
CCUCGAU-5' pH 6.5 MgSO4 1.7 A
MPD 6 vs 35-45%

DNA:drug complexes
CGCG + ditercalinium Room 0.7 mM 16.8 mM 0.3 mM 0.8 mM P41212 66
MPD6 vs 30% + 0.2mM pH 6.0 +14mM 1.7A
drug NH4Ac
CGCGAATTCGCG 5 3 mM 10 mM None 30 mM P2,2,2, 67
+ berenil + 2 mM pH 7.0 2.5 A
MPD 20 vs 50% drug

a
When the crystallization medium is buffered, the buffer is always sodium cacodylate.
b
Most often, MgCI2; in other cases, the salt is specified.
c
Amino acid accepting stem of tRNAAla with G:U mismatch.

219
 A.-C. Dock-Bregeon et al.
51 and 52. For tRNAs, a compilation of crystallization conditions is given in
ref. 3. Other general ideas on RNA crystallization can be found in refs 2 and
53.

3.1 General features

3.1.1 Crystallizing agents and concentration of nucleic acids
The more widely used are alcohols, and especially methyl 2,2 pentane diol
(MPD) which is not volatile and therefore easy to handle. It is used in the
range of 10% (v/v) in the case of tRNAs and 30% in that of oligonucleotides.
Isopropanol has also given good results with tRNAs, and especially with
tRNAphe (68). For tRNAs another successful crystallizing agent is (NH4)2SO4.
It gave good results with yeasttRNAfMet(69) and yeast tRNAAsp (70). Poly-
ethylene glycol (PEG) precipitates tRNAs at concentrations of a few per cent
for a medium sized PEG (Mt 4000-8000) and a different crystal form of yeast
tRNAASP could be obtained with PEG (3). Crystals of the Z-DNA hexamer
d(CG)3 were obtained with smaller sized PEG (61). A similar observation
came from the crystallization of RNA oligomers (71). Mixtures of PEG or
MPD with NaCl or NH4C1, or (NH4)2SO4 are also interesting possibilities; the
salt acts as an electrostatic shield and modulates the interaction between
RNA and additives.
For tRNAs, the crystallization is generally tried in the order of 5-20 mg/ml,
i.e. 0.2-0.8 mM. Higher molar concentrations are generally used for oligo-
nucleotides (Table 1).

3.1.2 Temperature, pH, and buffers

The temperature stability of nucleic acids allows examination of a large range
of temperatures, from 4°C (usual cold room temperature) to 30 °C or 35 °C (in
a bacteriologic incubator or an oven). The 35 °C assays bring new paths
towards crystallization especially when mixed precipitants are tried since it
modifies phase partitions.
The pH appears to play a smaller role in nucleic acid than in protein
crystallization where the overall charge of a protein, and then its capacity of
packing in a certain way, may be tuned through pH variations. The situation is
quite different with nucleic acids, which are negatively charged polyelectrolytes.
At pH 4.0-5.0 cytidines are protonated, and such a pH range can therefore
promote crystallization, when there is an accessible cytidine, by introducing a
potential additional interaction. A too low pH could, however, induce local
structural artefacts. Taking pH into account is also of importance for mis-
matched oligonucleotides. In the case of RNA the problem of degradations
forbids use of alkaline pH.
The buffer is often Na cacodylate (pH range 6.0-7.0) which pH is rather
temperature-insensitive and has the additional advantage of preventing
bacterial growth (a problem in PEG). In (NH4)2SO4, the buffer concentration
220
 8: Nucleic acids and their complexes
must be high enough (i.e. 100-300 mM) to maintain pH against variations due
to ammoniac evaporation (72) (Chapter 5). In low ionic strength media (PEG
or MPD) the buffer itself can introduce an electrostatic shield, and variations
of its concentration may modulate the electrostatic interactions between
nucleic acids and additives.

3.2 Specific features: additives

Nucleic acids are polyelectrolytes and therefore the counterions are import-
ant additives for crystallization. Two families of cations are generally used,
polyamines and divalent cations. Their role in crystallization differs subtly
and parallels their structural effects.
3.2.1 Polyamines
Polyamines are involved in many biological processes including DNA con-
densation and protein synthesis (73). Some examples of natural polyamines
are given in Table 2. They often enter in crystallization media but their
presence is not always needed (Table 1). Spermine is the most used. It is a
linear molecule with four positive charges at neutral pH and became popular
because of its key role in the production of the first crystals of a tRNA (74).
After growth of high diffracting yeast tRNAPhe crystals (reviewed in ref. 3),
spermine was systematically tried with nucleic acids, including oligonucleo-
tides. Spermidine, which is an asymmetric molecule bearing three positive
charges at neutral pH, was also reported to promote crystal growth, but with
less success. Positive effects, including resolution improvements, of several
synthetic cyclic polyamines on tRNAphe crystallization were recently reported
(75).
Spermine binds in the grooves of nucleic acids. The refinement of the
structure of yeast tRNAphe has identified two spermine molecules (76). One is
coiled in the deep major groove of the anticodon arm of the tRNA, at the
junction of D- and T-stems. It is H-bonded to four phosphates on both sides
of the groove. The second spermine molecule interposes a string of positive
charges between the extended polynucleotide chain of the variable region and
the P9-P10 sharp turn. Spermine has also been identified in crystals of Z-
DNA oligonucleotides (60, 77) but was not found yet in crystals of A-type
oligonucleotides.

Table 2. Natural polyamines used in nucleic acid crystallization

Putrescine H2N-(CH2)4-NH2
Cadaverine H2N-(CH2)5-NH2
Spermidine H2N-(CH2)3-NH-(CH2)4-NH2
Thermine H2N-(CH2)3-NH-(CH2)3-NH-(CH2)3-NH2
Spermine H2N-(CH2)3-NH-(CH2)4-NH-(CH2)3-NH2

221
 A.-C. Dock-Bregeon et al.

3.2.2 Divalent cations

Divalent cations, and especially Mg2+ ions, are involved in the stabilization of
nucleic acids structures and play an important role in their functions.
Crystallography has given a first insight into the structural effect of Mg2+ on
the conformation of tRNA. Preferential Mg2+ sites are located mostly in the
non-helical regions of the tRNA molecule and appear to stabilize loops and
bends of the tertiary structure. Some of these Mg2+ sites are of interest for
crystallization, since they are bridging tRNA molecules and therefore seem to
stabilize the crystal packing (e.g. one Mg2+ in the D-loop of the refined
tRNAphe structure) (78). Structures of oligonucleotides in the A- or B-helical
forms, refined to better resolution than tRNAs, unfortunately have brought
little additional information about the preferred co-ordination of Mg2+ ions.
The Z-structures, on the contrary, give generally more details about ion bind-
ing, a consequence of their better resolution. Examples of Mg2+ binding to Z-
DNA can be found in the structure of d(m5CGTAm5CG) (61). One Mg2+ ion
is surrounded with six oxygen atoms, one of which is a phosphate oxygen of
the backbone and the others are water molecules. Other examples are found
in the structure of d(CG)3 (79) or d(CGTACGTACG) (80) where intermol-
ecular Mg2+ sites are described. The presence of such sites confers probably,
with the H-bonding possibilities, an increased stability to the crystal packing,
and may explain why Z-DNA crystals often diffract to higher resolution.
Other divalent cations can be used instead of Mg2+, or in addition to it. For
tRNA crystallization, different divalent cations have been tried, like manga-
nese, calcium, cobalt, nickel, barium, mercury. Care must be taken, however,
since some metal ions may induce hydrolysis of the phosphodiester bonds in
RNA, especially lead (49). Crystallographic structures of mono- or dinucleo-
tides give an insight on the mode of binding of several ions to nucleic acids,
e.g. calcium binding to ApA (81). These ions sometimes provide stabilization
of local structures or new packing possibilities. More complex ions can also be
tried, of which cobalt hexamine is an interesting case. It stabilizes Z-DNA
with an efficiency that is five orders of magnitude greater than Mg2+. Cobalt
hexamine favoured crystallization of d(CG)3 (79) and d(CGTACGTACG)
(80) in the Z-form. Cobalt hexamine was also identified as an helix-stabilizing
agent in the case of tRNAphe (82).
3.2.3 Monovalent ions
The example of the cluster of ions in d(m5CGTAm5CG) (61) has shown that
Na+ ions can also play a role in helix stabilization, and therefore can favour
the crystallization of nucleic acids. Na+ only, without Mg2+, was used for the
crystallization of d(Br5CG)3 in the Z-form (62), and the structure shows how
Na+ bridges two neighbouring molecules in the crystal. Compared to Mg2+,
the octahedral co-ordination of Na+ is less precisely defined and in certain
cases can be accommodated more easily.
222
 8: Nucleic acids and their complexes

3.2.4 Concentration of the counterions

A important parameter is the relative concentration of spermine and Mg2+, as
well as the ratio spermine or Mg2+/nucleic acid molecule. For magnesium a
'rule of thumb' for first trials is 0.5-1.0 Mg2+ ion per phosphate (e.g. for a
tRNA at 0.4 mM corresponding to a phosphate concentration of 30 mM, try
Mg2+ concentrations of 15 mM and 30 mM). Smaller or larger concentrations
may be tried if results are disappointing; e.g. in this range the tetradecamer
U(UA)6A did crystallize readily, but crystals showed poor diffraction (with
maximal resolution of ~ 7 A). The best crystals were obtained at a Mg2+
concentration of 400 mM (32). For spermine the 'rule of thumb' is one
spermine molecule for 10-12 bp (e.g. for a tRNA at 0.4 mM the spermine
concentration to try is 3 mM). Since spermine and Mg2+ act as counterions,
the ionic strength of the medium has to be taken into account; in (NH4)2SO4
solutions or when monovalent salts are added, the concentrations of Mg2+
and spermine have to be somewhat higher than in PEG or MPD. The relative
concentration of spermine and Mg2+ is also to be considered; at higher Mg2+
concentration, higher spermine concentrations can be tested since Mg2+
brings its own shielding effect. An excess of spermine, especially at low ionic
strength, often produces crystals which do not diffract. Some assays without
spermine should also be tried (examples in Table 1).

3.3 Crystallization strategies

3.3.1 Design of crystallization conditions
As for proteins, parameters influencing crystallization of nucleic acids are
numerous. As seen above a number of them are of a special type, like those
mediated by polyamines or metal ions, and should therefore be always
assayed. Nevertheless an extensive screen of possible parameters remains
demanding in terms of macromolecular material needed. Therefore a
factorial design of crystallization experiments is advised (Chapter 4). A more
pragmatic approach, like in the protein field, is the use of crystallization
condition sparse matrices. Several such matrices, primarily designed for RNA
crystallization trials, have been described (83, 84). Commercial screening kits
designed for crystallizing nucleic acids and their complexes are available
(Nucleic Acid Mini Screen or Natrix™, Hampton Research, Laguna Hills,
CA) and could be used as a start. However, design of new or refined sparse
matrices, based on the increasing knowledge of experimenters, should not be
forgotten.

3.3.2 Refinements of crystallization conditions

Formation of crystals of poor quality is a often encountered with nucleic
acids. This is often the case of crystals obtained after rapid screening with
sparse matrices. For small duplexes, this may be due to the geometry of the
223
 A.-C. Dock-Bregeon et al.
helices, which can pack easily despite rotational disorder. The answer is to
play with additives, temperature, pH, trying to find a way of introducing
structural change or to bind additional small molecules which could act as a
lever promoting lattice building.
3.3.3 Engineering crystallization and heavy-atom derivatives
Sequence variations in nucleic acid oligomers may be a more powerful
strategy for obtaining high quality crystals than variations in crystallization
conditions. This mostly applies to structural characteristics of particular
nucleic acid sequences that favour ordered assembly of the molecules. For
instance overlapping sequences in DNA duplexes that allow H-bonding with
the neighbouring molecules (85) or anticodon/anticodon and stacking inter-
actions in tRNA (86) have been shown to trigger crystallization. Such effects
have been rationalized for the crystallization of B-DNA oligomers (6) and
were discussed for that of RNA domains (20, 87).
Thus in case of unsuccessful crystallizations it may be advantageous to
engineer the nucleic acid sequence by introducing structural elements that
favour packing interactions (20) or stabilize the nucleic acid structure. For
RNAs it can be advised to remove CpA sequences that are preferential
hydrolytic cleavage points (88) or to introduce stable tetraloops (20).
Preparation of heavy-atom derivatives often remains another bottleneck
for structure determinations, especially in the nucleic acid field, since the
polyanionic nature of nucleic acids offers too many possibilities of metal
chelations. However, the synthetic procedures for nucleic acid preparations
allow incorporation into DNA or RNA oligomers of brominated or iodinated
nucleotides at well designed positions which in principle can serve for phase
determinations. This engineering strategy was used to prepare heavy-atom
derivatives of the large P4-P6 160 nt domain of Tetrahymena intron (89).

3.4 The special case of DNA:drug co-crystallizations

The pharmacological importance of DNA-.drug complexes is obvious and
explains the interest of structural biologists to crystallize them. As compared
to free oligonucleotides, crystallization of drug complexes does not show any
particular features (52). Two typical examples are displayed in Table 1 and
concern an intercalating (66) and a minor groove binder (67) drug.

4. Co-crystallization of nucleic acids and proteins

Many basic biological mechanisms and particularly those regarding storage
and expression of the genetic message involve interactions between proteins
and nucleic acids. This promoted a need for 3D structural knowledge. When
the nucleic acid moiety of the complex of interest is a small ligand, like
nucleotides or short oligonucleotides, it is sometimes possible to diffuse it into
224
 8: Nucleic acids and their complexes
a 'receptor crystal' (e.g. dT4 in Klenow fragment of DNA polymerase I from
E. coli) (90). With larger substrates that cannot penetrate, or when much
conformational changes occur, co-crystallization is a necessity.
A first practical advice for newcomers is to consider crystallization of a
protein:nucleic acid complex as a new problem, different from the crystal-
lization of the protein or nucleic acid alone. Nevertheless, knowledge on
solubility and other behaviours of the free components can guide the search
of crystallization conditions for the complex.
From the early 1980s to now more than 60 different DNA binding proteins
have been co-crystallized with DNA oligonucleotides of biological signifi-
cance (91), and ~ 20 RNA binding proteins with natural (tRNAs) or synthetic
RNA substrates (see Tables 3 and 4). When looking at the successful attempts,
one is impressed by the increasing number of high quality crystals of com-
plexes with synthetic deoxyoligonucleotides with diffractions that can pass the
limit of 2 A resolution. Noticeable, in several cases crystallization of the
complex was more straightforward than that of the protein alone. Similarly,
RNA:protein complexes often crystallize more readily than the individual
components. The reason for that is the conformational stabilization of the
nucleic acid and/or protein components in the complex. After the first
successes of crystallized DNA:protein complexes of prokaryotic origin, like
restriction enzymes or phage repressors, the tremendous improvement of
protein over-production and purification methods (Chapters 2 and 3) enabled
structural biologists to tackle new challenges in the eukaryotic world. Thus
complexes of DNA fragments with nuclear receptors or transcription factors
could be crystallized. For RNA:protein complexes, a number of successes
were obtained with components originating from thermophilic organisms.
The higher stability of the proteins and RNAs from such organisms, certainly
is the key factor explaining the improved crystallizability of the complexes.

4.1 General features of nucleic acid:protein

co-crystallization
4.1.1 Co-crystallization or crystallization of pre-existent complexes
When dealing with complexes, a variety of situations can be encountered and
the crystallization strategy must adapt to each particular problem. Some
complexes can be isolated and purified from natural sources (e.g. viruses,
ribosomes, nucleosome) while many others are transient and require
independent purification of each component. Sometimes, a purification step
of the reconstituted complex is advisable.
The heterologous nature of complexes introduces additional problems.
Protein nucleic acid recognition involves specific interactions between macro-
molecules of different electrostatic properties, and for a given protein the
binding areas are adapted to this complementarity. In order not to be a
competitive site, the other part of the molecule will act as a repellent to the
225
 A.-C. Dock-Bregeon et al.
nucleic acid substrate. In many crystal structures of complexes the crystal
packing is built upon contacts between macromolecules of the same type, i.e.
protein:protein or nucleic acidrnucleic acid interactions.
4.1.2 Stability
Difficulties may arise regarding the stability of the complex under crystal-
lization conditions (pH, ionic strength, and so on) and it is important to
ascertain the physical existence of the particle in such conditions. A Kd value
of 10-5 M is an upper limit of stability for a crystallizable complex. The time
scale of experiments creates another problem when nucleic acids are sub-
strates of enzymatic reactions. If RNA is the substrate it is now possible to
chemically synthesize a mixed nucleic acid with the reactive ribo- being
replaced by a deoxyribonucleotide (30). When DNA is the substrate various
solutions have been found, like pH changes or removal of the cations
necessary for the enzymatic reactions; e.g. omit Mg2+ in the co-crystallization
of EcoRI (92) or add a chelating agent for Klenow fragment (90). In special
cases, as with DNase I, the cleaved oligonucleotide was co-crystallized with
the enzyme (93).
4.1.3 Homogeneity of samples, stoichiometry, and purity
A problem which is specific to natural samples containing large nucleic acids
like nucleosome and ribosomes, is the heterogeneity of the nucleic acid part
of the samples. Even if the size of the nucleic acid component can be defined
with some accuracy, the random dispersion of the nucleotide sequence is a
major problem. In the case of nucleosome core particle, this problem was one
of the major limitation to the formation of high resolution diffracting crystals.
Chemical synthesis or in vitro genetic engineering techniques, which enables a
large scale preparation of long oligonucleotide sequences up to a hundred
nucleotides, can provide solutions. Rendering DNA homogeneous in
nucleosome particles enabled quality crystals to be produced and to solve
their structure at 2.8 A resolution (94).
A slight excess of substrate is a general trend of all experiments. For the
complex between yeast tRNAASP and aspartyl-tRNA synthetase, where
stoichiometry was well analysed, a variation of the tRNA concentration
around the stoichiometric value (2:1) was the main cause of polymorphism
(95). In some cases, however, excess of DNA over protein concentration was
used, as for the crystallization of a X represser fragment (1-92) with a 20-mer
operator, and the correct stoichiometry, one DNA duplex per protein dimer,
was found in the crystals (96).
Purity is of general concern (Chapter 2). Since we are dealing with two
molecules the problem is more crucial here. An illustration of the importance
of the nucleic acid purity is given with crystallization of the operator binding
domain of the X represser with the X operator site. Much better crystals were
obtained when the synthetic operator was further purified with HPLC (96).
226
 8: Nucleic acids and their complexes
For the crystallization of yeast aspartyl-tRNA synthetase with tRNAAsp,
improvement of the protein purification protocol produced a new and better-
diffracting crystal form (97).

4.1.4 Crystallizing agent and pH

High salt conditions were long believed to be unfavourable to the stability of
nucleic acid protein complexes. This was the main reason for the success of
crystallization attempts with alcohols, and among them MPD is the most
popular (in the range 15-25%). However, many successful attempts were
realized with (NH4)2SO4 at high concentration (in the range of 2 M). This can
be explained by a screening effect of ammonium and/or sulfate ions which
hamper non-specific contact and prevent aggregation. It is an experimental
fact that high salt concentrations are disruptive of complexes. However high
concentrations of (NH4)2SO4 or ammonium citrate do not have such dis-
ruptive effects and thus can be used in crystallization attempts. Despite their
disruptive effects, salts such as NaCl can also sustain crystallization, as for
phage A. Cro represser complexed with its operator (98). We believe that the
salts have stronger disruptive effects on non-specific than specific complexes.
Thus choice of adequate amounts of salt may favour formation of homo-
geneous samples of specific complexes and hence their crystallization. Along
these lines, mixtures of salts and PEGs are also of particular interest. Slightly
acidic or neutral pH seem the best bet although attempts at slightly basic pH
(7.5-8.0) are not uncommon.

4.1.5 Additives
For this part we enter in more specific problems linked to the nature of the
systems investigated. MgCl2 and CaCl2 are the most common additives used.
Phosphate salts have to be avoided for two reasons: they often lead to
insoluble compounds and act as competitors for nucleic acid binding sites.
When existing, cofactors or small substrates (like ATP, GTP, L-tryptophan)
should be used as an important variable in crystallization screenings.

4.2 Complexes of synthetic oligodeoxynucleotides and

proteins
Table 3 illustrates the diversity of complexes that were crystallized and the
diversity in crystallization conditions employed. Crystals have been obtained
with salts, alcohols, or PEGs as crystallizing agents. A tendency that emerges
from a survey of recent crystallizations of DNA:protein complexes is the
usefulness of PEGs of rather low molecular weight and of mixtures of
medium sized PEGs with salts. Of frequent use is the addition of protein
stabilizing agents, like glycerol or ethylene glycol (111). Interestingly, and in
contrast to what observed for RNA:protein complexes, spermine is used as an
additive in many cases.
227
Table 3. Sampling of crystallization conditions for DNA:protein complexes

Protein and DNA Temp Concentration Crystals Ref.

Precipitanta (°C) Proteinb DNA Buffer Additives
Phage 434 repressor (fragment 1-69) + operator (14-mer, symmetric, blunt ends)
(NH4)2SO4 4 0.5 mM 0.25 mM Na phosphate I422 99
1.3 M (2:1) 5mM 3.2-4.5 A
pH 4.7
Phage 434 repressor (fragment 1-69) + operator (20-mer, asymmetric and complementary overhangs of 1 nt)
PEG 3000 4 2mM 1 mM NaCl 100 mM, P212121 100
12-14% MgCI2120 mM, 2.5 A
spermine 2 mM
Phage X repressor (fragment 1-92) + operator (20-mer, asymmetric, complementary overhangs of 1 nt)
PEG 400 20 0.91 mM 0.91 mM BTP 15 mM NaN31 mM P21 96
10 vs 20% pH 7.0 2.5 A
Phage X Cro repressor + operator (17-mer, asymmetric, blunt ends)
NaCI 0.1M - 2.5 mg/ml 2.5 mg/ml Nacacodylate P62(P64) 98
vs 3.5-4.0 M or 20 mM 3.7 A
slow evaporation pH 6.9
E.colitrp repressor + Trp + operator (18-mer + overhanging 5'-T)
MPD 20 0.4 mM 0.6-0.8 mM Na cacodylate L-Trp 2 mM P2, 101
20 vs 40% 10mM CaCI211mM 2.5A
pH 7.2
E. coli CAP protein + cAMP + DNA binding site (30-mer + overhanging 5'-G)
PEG 3350 - 4-6 mg/ml 1.5-fold MES NaCI 0.2 M, C2221 102
5-10% molar 50 mM CaCI20.1 M, 3.0 A
excess pH 5.0-6.0 cAMP 2 mM,
spermine 2 mM,
0.02% NaN3, DTT
2 mM, 0.3% n-octyl-
glucoside
E. coliKlenow fragment of Pol I + DNA substrate (8 bp + 3 bases, single-stranded 5' overhang)
(NH4)2S04 38% - 2-fold molar EDTA 1 mM 90
excess
E. coli restriction endonuclease EcoRI + DNA substrate (13-mer with overhanging 5'-T)
PEG 400 4 2.7 mg/ml 2.8 mg/ml BTP 40 mM NH4Ac 0.5M, P321 92
8 vs 16% pH 7.4 dioxane 15% 2.6 A
Bovine pancreatic DNase I + DNA substrate (8-mer, nicked)
PEG 600 4 EDTA 15 mM C2221 93
2.0 A
GCN4 (leucine zipper protein) + DNA substrate (20-mer pseudopalindrome with complementary overhangs of 1 nt)
PEG 400 12% 22 0.95 mM 0.57 mM MES 25 mM MgCI2 30mM, P212121 103
pH 5.8 spermine 1 mM, 2.9 A
NaCI 0.15 M
GLI (Zn finger protein) + DNA substrate (21-mer with complementary overhangs of 1 nt)
PEG 400 22 0.5 mM 0.6 mM BTP CoCI2 1 mM, P212121 104
20-25% pH 7.0 MgCI2 60-100 mM 2.6A
HNF-3/fork head + DNA substrate (13 bp with blunt end)
NH4 acetate 4 Complex K acetate KCl 100 mM, P31 105
550 mM 1 mM 20 mM MgCI2 2mM, 2.5 A
salting-in pH 5.5 DTT 20 mM
78 Resolvase + DNA substrate (34 bp)
PEG 3350 Room 0.36 mM 0.72 mM MES 50 mM EDTA 0.5 mM, P212121 106
15 vs 30% + pH 6.0 + Tris (NH4)2SO4 0.2 M 3.0 A
ethylene glycol 10 mM pH 8.0
2.5 vs 5%
Oestrogen receptor DNA binding domain + DNA substrate (17 bp with overhanging 1 nt)
MPD 10% 20 Complex MES 20 mM Spermine 1.8 mM, P212121 107
70 uM pH 6.0 ZnCI2 2uM, 2.4 A
CaCI2 2-8 mM,
NaCI 30-80 mM
Tables. Continued

Protein and DNA Temp Concentration Crystals Ref.

Precipitant3 (°C) Protein" DNA Buffer Additives
Yeast TATA binding protein + DNA substrate (29 nt:12 bp stem and 5 nt loop)
PEG 8000 22 0.25 mM 0.5 mM BTP20mM NaCI or KCI P43 108
15 vs 30% pH7.5 500 mM, glycerol 2.5 A
3.75%, ethylene
glycol 1%
Heterodimeric transcription factor c-Fos-c-Jun + DNA substrate (20 nt with complementary overhangs of 1 nt)
PEG 400 5.5-7.5 20 0.5 mM 0.6-0.8 mM bisTris 50 mM NH4acetate P21212 109
vs11-15% pH6.7 100-200mM, 3.0A
NaCI 150mM,
MgCI227.5mM,
spermine 1.0 mM,
DTT5-10 mM
Ternary complex TBP:TFIIB:DNA substrate (16 bp blunt end)
Salting-infrom 4 Complex Tris-HCI KCI 40 mM, P212121 110
NH4acetate 0.3-0.4 mM 40 mM MgCI25mM, 2.7 A
300 mM pH 8.5 CaCI2 5 mM,
DTTIOmM,
Zn acetate 10 uM,
glycerol 10%,
ethylene glycol 2%

a The concentration in the reservoir is given, or initial concentrations in the form: C(drop) versus C(reservoir).
bThe molar ratio of protein monomers versus DNA duplex is given in brackets.
c
When several DNA duplexes were tried, the conditions indicated are those producing the best crystals.
 8: Nucleic acids and their complexes
The main problem with binary complexes is the choice of the best DNA
sequence and of its optimal length. Two major constraints have to be taken
into account: the biological relevance of the sequence and the stability of the
duplex. An effect of the number of base pairs (which should have been a
multiple of seven) was thought to be important after the co-crystallization of
the DNA binding domain of phage 434 represser and its operator (99). Later
examples were no longer in this line. Some co-crystallizations were made with
blunt-ended oligomers: e.g. phage A Cro represser with a 17-mer operator
(98) or phage 434 repressor with a 14-mer operator (99). Others underline the
importance of overhanging nucleotides. These could reinforce the end-to-end
stacking of DNA duplexes which seem to be a common mode of packing.
Clearly, there is no generally applicable rationale that specifies the optimal
length and terminal structure of the oligonucleotides to be used in
crystallizing protein:DNA complexes. The principal limitation of the choice
seems to be the production of the oligonucleotides, especially if the required
sequence is large. This problem has been nicely overcome in the crystalliza-
tion of the CAP protein complexed with DNA (102). Ten oligonucleotides, up
to 20 nt in length, were synthesized. These are able to self-hybridize and were
mixed to generate 19 different double-stranded segments (of 28-36 bp) with
symmetric overhangs of zero, one, or two bases. Crystallization conditions
were examined with 26 different DNA segments, 28 or more bp in length, that
explored a variety of sequences (symmetric or not), length, and extended 5'-
or 3'-termini. Crystals of variable quality were produced, one of them
diffracting to 3.0 A resolution.

4.3 Complexes of RNAs and proteins

Complexes between tRNAs and their cognate aminoacyl-tRNA synthetases
were the first examples of co-crystallization of proteins and RNAs. Their
stability range is not very high (Kd values within 10'6 to 10-9 M). Well
characterized crystals, which led to high resolution structure determinations,
were first obtained in the E. coli glutamine (39) and the yeast aspartate (95,
97) systems. More recently, several other tRNA:synthetase complexes have
been crystallized: e.g. the phenylalanine complex from T. thermophilus with a
tetrameric synthetase (115), a serine complex with a long variable loop tRNA
(116), the E. coli glutamine complex with unmodified tRNAGln (113), the
heterologous aspartate complex between the E. coli synthetase and yeast
tRNA (112), and the lysine complex, either homologous with T. thermophilus
partners but unmodified tRNA or heterologous with the thermophilic
synthetase and E. coli tRNALys (mnm5S2UUU) (114), as well as other RNA:
protein complexes, e.g. the RNA binding domain of the U1A spliceosomal
protein with an RNA hairpin (118) or EF-Tu with tRNAphe (117).
An initial limitation in the field has been the poor understanding of the
conditions leading to complex formation in the presence of crystallizing agents.
231
Table 4. Sampling of crystallization conditions for RNA:protein complexes

Stoichiometry Concentration Temp pH Buffer Precipitant Additives Ref.

RNA and protein (°C)

Aminoacyl-tRNA synthetase:tRNA complexes

AspRS:tRNAAsp (yeast)
1:2 [protein] 10 mg/ml 4 7.5 Tris- (NH4)2SO4 MgCI25 mM 97
(80 uM) maleate 1.0 vs 2.4 M
[tRNA] 4.8 mg/ml 40 mM
(190 uM)
AspRS (E. coli):tRNAAsp (yeast) heterologous complex
1:2 [protein] 2 mg/ml 17 6.7 BTP75- (NH4)2S04 MgCI21 mM, 112
[tRNA) 1 mg/ml 100 mM 1.5 vs 1.9 M AMP-PCP 0.5 mM,
seeding required aspartic acid 1 mM
GlnRS:tRNAGln (E.coli)
1:1 [complex] 10 mg/ml 17 6.8- Pipes Na citrate MgCI220mM, 39
7.0 80 mM 44-64% ATP 4 mM,
2-mercaptoethanol
20 mM
or
7.0- Pipes (NH4)2SO4 MgSO4 20 mM,
7.5 80 mM 1.8-2.0 M ATP 8 mM,
NaN30.02%
GlnRS:tRNAGln(E coli), but with unmodified tRNA transcript
1:1 10-15 mg/ml 17 6.5- Pipes (NH4)2SO4 MgSO4 20 mM, 113
7.5 80 mM 1.6-2.0 M ATP 8 mM,
NaN3 0.02%
LysRS (T. themophilus):tRNALys(E.coli) + lysyl-adenylate analogue
1:2 [protein] 4 mg/ml - 7.6 Tris- (NH4)2SO4 MgCI2 10 mM, 114
[tRNA] 2.5 mg/ml maleate 24-26% NaN31 mM,
50 mM Lys-AMS 325 uM
PheRS:tRNAphe (T. thermophilus)
1:1 [protein] 5-7 mg/ml 15 7.2 Imidazole (NH4)2SO4 MgCI21mM 115
20 mM 15 vs 25-30%
SerRS:tRNAser (T. thermophilus)
1:2 [protein] 5.6 mg/ml 20 7.2 Tris- (NH4)2S04 NaN31mM, 116
[tRNA] 2.6 mg/ml maleate 20 vs 32% MgCI2 2.5 mM
25 mM

Other protein:RNA complexes

EF-Tu (T. aqoaticus):GDPNP:Phe-tRNAphe (yeast)
1:1 [complex] 4 6.7- Tris 20 mM (NH4)2SO4 NaN3 0.5 mM, 117
15 mg/ml 7.0 7.6 35 vs 47-49% MgCI27mM,
Mes 3 mM DTT 0.5 mM,
2.7 GDPNP 0.4 mM
RNA binding domain of U1A spliceosomal protein + RNA hairpin (21-mer)
1:1 [protein] 5.6 mg/ml 20 7.0 Tris-HCI (NH4)2S04 Spermine 5 mM 118
[RNA] 2.6mg/ml 40 mM 1.8 M
 A.-C. Dock-Bregeon et al.
Because complex formation between proteins and nucleic acids involves
electrostatic interactions (119), crystallizations were for long not tempted in
the presence of salts. Only when it was realized that tRNA:synthetase com-
plexes are stable and even active in the presence of (NH4)2SO4 (120), crystals
could be obtained in the presence of this salt. Another limitation has been the
poor supply of biological material with reliable physico-chemical integrity.
This prevented a good survey of crystallization conditions. In the case of the
yeast aspartate system, the first attempts of large scale purifications of the
synthetase and the tRNA were set up from wild-type yeast cells and
commercial bulk tRNA (95). Preparation of the enzyme necessitated three
weeks of work with rather poor yield of intact enzyme due to proteolysis. An
improvement of the purification procedure reduced the time scale to three
days with concomitant increase of the yield. Together with revised conditions
of crystallization, this improvement resulted in much better crystals diffract-
ing to 2.7 A resolution (97). In the case of the E. coli glutamine system, the
problem of sample quantities was overcome by cloning and construction of
over-producing strains for both the enzyme and the tRNA (39). At present,
all crystallized complexes are obtained from over-produced synthetases and
tRNAs.
For crystallization, most of the problems with tRNA:synthetase systems are
similar to those described in the general part. Table 4 shows conditions
leading to crystals of such complexes. Except in one case, crystallizations were
realized in high (NH4)2SO4 conditions. Spermine, which is an important
additive for obtaining good diffracting crystals of free tRNA, is not necessary
to obtain co-crystals of complexes. Interestingly, spermine is often used for
co-crystallization of DNA:protein complexes (Table 3). On the other hand,
Mg2+ is systematically included when crystallizing with RNA, which is not the
case with DNA. This difference should be related to the different nature of
the nucleic acids, the 2'-OH of riboses introducing the possibility of making
new contacts with protein as well as with neighbouring RNA molecules.
4.4 Ribosomes and their subunits
Protein biosynthesis takes place on the large ribonucleoprotein particles
called ribosomes. These organelles are made of two subunits which associate
upon initiation of protein synthesis to form a full particle. Although early
observations of crystalline material were made in vivo as part of a mechanism
of hibernation in a variety of lizards, the only real successful attempts to grow
large 3D crystals of ribosomes was achieved with bacterial particles. In
bacteria the smallest subunit (30S) has a molecular weight of 700 kDa and
contains ~ 20 proteins and one RNA chain (16S). The large subunit (50S) of
1600 kDa consists of ~ 35 different proteins and two RNA chains (23S and
5S). The large size of ribosomes, comparable to that of viruses, and the lack of
internal symmetry combined with conformational heterogeneity transforms
the problem in a formidable challenge.
234
 8: Nucleic acids and their compleplexes
Crystallization properties and biological activities of the particles are
strongly correlated, i.e. inactive particles do not crystallize. Conversely, re-
dissolved crystals are active. In the case of 70S ribosomes from T. thermo-
philus, the best crystals are grown from material obtained after dissolution of
previously formed egg-like crystals (121). Interestingly, addition of two
molecules of charged tRNA (phenylalanyl-tRNAphe) and of a piece of mRNA
(35-mer poly U) improves crystal quality with diffraction resolution improved
from 20 to 12 A (122). The best 3D crystals with highest resolution (~ 3 A),
however, are obtained with the isolated 50S subunit from Haloarculum
marismortui (122, 123); promising diffraction limits of 7.3 A and 8.7 A were
also reported for crystals of the 30S and 70S particles of T. thermophilus (122).
Since high salt conditions are disruptive for most ribosomes, crystallization
conditions were searched mostly with volatile organic solvents. Initially, the
crystallization droplets contain no precipitant or a very small quantity of it. To
reduce the rate of crystal growth and to avoid technical difficulties linked to
the use of volatile solvents, crystallization assays are sometimes realized
directly in X-rays capillaries (38). Crystallizing agents are often PEGs and
MPD (38, 121-123). In contrast, halophilic ribosomes and their isolated
subunits are stable at high salt concentrations and the high diffracting crystals
of the 50S subunit of H. marismortui were indeed grown from (NH4)2SO4
solutions (122). These growth conditions mimic to some extent the natural
salt-rich environment within the halobacteria that contains KC1, NH4C1, and
MgCl2. Altogether, crystal quality depends on the procedure used for the
preparation of the ribosomal material, the strain of a given bacterial species,
and on the fine-tuning of conditions such as the balance between Mg2+ and
monovalent ions.
A major problem of ribosome crystals is their poor limit of resolution and
their extreme sensitivity to X-ray damage. Data collection at very low
temperatures (-150 °C) increases crystal lifetime but does not allow improved
resolution. Nevertheless, data sets-up to 10 A could be collected and
preliminary phasing at 7.9 A could be achieved for crystals of the halophilic
50S particles (122). Further improvements led to a 9 A resolution map (123).
These promising crystallographic results became possible because of progress
in crystallization methods, in data collection strategies, and in production
of isomorphous crystals containing appropriate clusters of heavy atoms
(122,123).
4.5 Viruses
In viruses, proteins form the protecting shell which encapsidates the genetic
material (RNA or DNA). The quaternary structure of viruses is dominated by
the nature of protein-protein interactions within the external capsid. There-
fore crystallization of viruses resembles that of proteins. Two main shapes are
observed: helical rods as in tobacco mosaic virus and filamentous bacterio-
phages, and isometric capsids in spherical viruses. Viruses were among the
235
Table 5. Crystallization conditions of some viruses

Virusa Concentration Temp (°C) Buffer pH Precipitant Additives Ref.

RNA viruses
Animal viruses (Picornaviruses)
BEV 5 mg/ml 20 Na phosphate NaCI 30% NaN3 125
0.1 MpH 7.6
Coxsackie B110 mg/ml 20 Na acetate (NH4)H2P04 126
10 mM pH 5.0 0.1 M
HRV 14 5 mg/ml 20 Tris-HCl 10 mM PEG 8000 CaCI220 mM 127
pH 7.2 0.25-0.5%
Mengo 5 mg/ml Room Na phosphate PEG 8000 128
0.1 MpH 7.4 2.8%

Plant viruses
CCMV 20-50 mg/ml Room Succinate PEG 8000 NaN31 mM, 129
0.3 M pH 3.3 3.7-4.0% EDTA 1 mM
CpMV 35 mg/ml 20 K phosphate PEG 8000 (NH 4 )S0 4 0.4M 130
50 mM pH 7.0 2%
STMV 20 mg/ml 23 Cacodylate, (NH4)SO4 NaCI or 131
Na phosphate, 10-18% NaC2H4O4
or Tris 40 mM
pH 6,6.5, or 7
STNV 10-12 or 7-8 Na phosphate PEG6000 Mg 2+ 1 mM 132
mg/ml 50 mM pH 6.5 0.4%
TBSV 30 mg/ml 4 None (NH4)SO4 133
0.5 M
TYMV 25 MES (NH4)H2P04 134
100mMpH3.7 1.11-1.15M
Insect and bacterial viruses
BBV 8 mg/ml 20 Na phosphate (NH4)S04 130
50 mM 13.5%
pH 6.9-7.2
FHV 18 mg/ml Room bisTris PEG 8000 CaCI2 20 mM 135
10 mM pH 6.0 2.8%
MS2 1% 37 Na phosphate PEG 6000 NaN30.02% 136
0.4 M pH 7.4 1.5%

DNA viruses
CPV 10 mg/ml Room Tris PEG 8000 CaCI2 6mM 137
10 mM pH 7.5 0.75%
$X174 8 mg/ml 20,4 bisTris methane PEG 8000 138
90-93 mM 1.5-2.0%
pH 6.8
a
BBV, black beetle virus; BEV, bovine enterovirus; CCMV, cucumber chlorotic mottle virus; CpMV, cowpea mosaic virus; CPV,
canine parvovirus; FHV, flock house virus; HRV, human rhinovirus; STMV, satellite tobacco mosaic virus; STNV, satellite tobacco
necrosis virus; TBSV, tomato bushy stunt virus; TYMV, turnip yellow mosaic virus; MS2 and 0 X 174, two bacteriophages.
 A.-C. Dock-Bregeon et al.
first crystallized biological materials (Chapter 1). Now many viruses have
been crystallized and more than 20 structures of spherical viruses are deter-
mined (124). A list of typical representatives that yielded highly ordered
crystals is given in Table 5. For the best, diffraction limit often exceeds 3.0 A
resolution, probably as a consequence of their symmetric and isometric
structure.
The importance of the external capsid and hence the non-effect of RNA or
DNA on crystal formation is nicely demonstrated with cowpea mosaic virus
(CpMV). The genome of this virus consists of two RNA molecules, RNA1
(5.9 kb) and RNA2 (3.5 kb), which are encapsidated in separate particles.
Empty capsids are also formed in vivo. All three components are of the same
size and appear to have identical surfaces. Isomorphous crystals were
obtained with each of the isolated components or with a mixture of the three
components, and the same ratio of components was found in the crystals and
in the crystallizing solution (130).
As for other macromolecular systems, a wide diversity of conditions led to
crystal formation. Details on crystallization conditions can be found in Table 5
and in ref. 124. PEGs (2-3%), alone or mixed with (NH4)2SO4 in the 0.5 M
range, are the most currently used crystallizing agents. Interestingly, their
concentration range is low when compared to other systems. Crystallizations
are usually done at room temperature (20 °C). The pH range is larger than for
nucleic acids and reaches the acidic domain (i.e. 3.3-7.5). It is only limited by
the stability of the capsid. Thus, turnip yellow mosaic virus, an RNA spherical
virus, was crystallized at pH 3.7 (134). Finally, attempts leading to pre-
cipitation should not be discarded since crystals of viruses can also grow from
heavy precipitates by Ostwald ripening mechanisms as exemplified for tomato
bushy stunt virus (TBSV) (139). However, under such circumstances duration
of crystal growth can be long (several weeks and more).

References
1. Gilliland, G. L. and Ladner, J. E. (1996). Curr. Opin. Struct. Biol., 6,595.
2. Lietzke, S. E., Barnes, C. L., and Kundrot, C. E. (1995). Curr. Opin. Struct. Biol.,
5,645.
3. Dock, A.-C., Lorber, B., Moras, D., Pixa, G., Thierry, J.-C., and Giege, R. (1984).
Biochimie, 66,179.
4. Abdel-Meguid, S. S., Moore, P. B., and Steitz, T. A. (1983). J. Mol. Biol., 171, 207.
5. Cate, J. H., Gooding, A. R., Podell, E., Zhou, K., Golden, B. L., Kundrot, C. E., et
al. (1996). Science, 273, 1678.
6. Timsit, Y. and Moras, D. (1992). In Methods in enzymology (ed. D. M. J. Lilley
and J. E. Dahlberg), Vol. 211, pp. 409-29. Academic Press, London.
7. Wing, R., Drew, H., Takano, T., Broka, C., Tanaka, S., Itakura, K., et al. (1980).
Nature, 287, 755.
8. Timsit, Y., Westhof, E., Fuchs, R .P. P., and Moras, D. (1989). Nature, 341, 459.
9. Westhof, E., Auffinger, P., and Gaspin, C. (1996). In DNA and protein sequence

238
 8: Nucleic acids and their complexes
analysis: a practical approach (ed. M. Bishop and C. Rawlings), pp. 255-78. IRL
Press, Oxford.
10. Sigman, D. S., Mazumder, A., and Perrin, D. M. (1993). Chem. Rev., 93, 2295.
11. Ehresmann, C., Baudin, F., Mougel, M., Romby, P., Ebel, J.-P., and Ehresmann,
B. (1987). Nucleic Acids Res., 15, 9109.
12. Kolchanov, N. A., Titov, I. I., Vlassova, I. E., and Vlassov, V. V. (1996). Prog.
Nucleic Acid Res. Mol. Biol., 53, 131.
13. Westhof, E. and Michel, F. (1995). In RNA-protein interactions (ed. K. Nagai and
I. W. Mattaj), pp. 25-51. IRL Press, Oxford.
14. Gold, L., Polisky, B., Uhlenbeck, O., and Yarus, M. (1995). Annu. Rev. Biochem.,
64, 763.
15. Leonard, G. A., McAuley, H. K., Ebel, S., Lough, D. M., Brown, T., and Hunter,
W. N. (1994). Structure, 2, 483.
16. Pley, H. W., Flaherty, K. M., and McKay, D. M. (1994). Nature, 372, 68.
17. Schindelin, H., Zhang, M., Bald, R., Furste, J.-P., Erdmann, V. A., and
Heinemann, U. (1995). J. Mol. Biol., 249, 595.
18. Baeyens, K. J., De Bondt, H. L., Pardi, A., and Holbrook, S. R. (1996). Proc. Natl.
Acad. Sci. USA, 93,12851.
19. Ott, G., Dorfler, S., Sprinzl, M., Muller, U., and Heinemann, U. (1996). Acta
Cryst.,D52,871.
20. Ferre-D'Amare, A. R., Zhou, K., and Doudna, J. A. (1998). J. Mol. Biol., 279,621.
21. Oubridge, C., Ito, T., Teo, C. H., Fearnley, L, and Nagai, K. (1995). J. Mol. Biol.,
249, 409.
22. Pingoud, A., Fliess, A., and Pingoud, A. (1989). In HPLC of macromolecules: a
practical approach (ed. R. W. A. Oliver), pp. 183-208. IRL Press, Oxford.
23. Davanloo, P., Rosenberg, A. H., Dunn, J. J., and Studier, W. (1984). Biochemistry,
81, 2035.
24. Milligan, J. F., Groebe, D. R., Witherell, G., and Uhlenbeck, O. C. (1987). Nucleic
Acids Res., 15, 8783.
25. Frugier, M., Florentz, C., Hosseini, M. W., Lehn, J.-M., and Giege, R. (1994).
Nucleic Acids Res., 22, 2784.
26. Yin, Y. and Carter, C. W., Jr (1996). Nucleic Acids Res., 24, 1279.
27. Lapham, J. and Crothers, D. M. (1996). RNA, 2, 289.
28. Giege, R., Sissler, M., and Florentz, C. (1998). Nucleic Acids Res., 26, 5017.
29. Price, S. R., Ito, N., Oubridge, C., Avis, J. M., and Nagai K. (1995). J. Mol. Biol.,
249, 398.
30. Perreault, J. P., Wu, T., Cousineau, B., Ogilvie, K. K., and Cedergren, R. (1990).
Nature, 344, 565.
31. Anderson, A. C., Scaringe, S. A., Earp, B. E., and Frederick, C. R. (1996). RNA, 2,
110.
32. Dock-Bregeon, A.-C., Chevrier, B., Podjarny, A., Johnson, J., de Bear, J. S.,
Gough, G. R., et al. (1989). J. Mol. Biol., 209, 459.
33. Cruse, W. B., Saludjian, P., Biala, E., Strazewski, P., Prange, T., and Kennard, O.
(1994). Proc. Natl. Acad. Sci. USA, 91, 4160.
34. Monier, R. (1971). In Procedure in nucleic acids research (ed. G. L. Cantoni and
D. R. Davies), Vol. 2, pp. 618-28. Harper and Row, New York.
35. Morikawa, K., Kawakami, M., and Takemura, S. (1982). FEBS Lett., 145, 194.
36. Giege, R. and Ebel, J.-P. (1968). Biochim. Biophys. Acta, 161, 125.
239
 A.-C. Dock-Bregeon et al.
37. Kowalak, J. A., Dalluge, J. J., McCloskey, J. A., and Stetter, K. O. (1994).
Biochemistry, 33, 7869.
38. Yonath, A., Frolow, F., Shoham, M., Mussig, J., Makowski, I., Glotz, C., et al.
(1988). J. Cryst. Growth, 90,231.
39. Perona, J. J., Swanson, R., Steitz, T. A., and Soll, D. (1988). J. Mol. Biol., 202,121.
40. Dirheimer, G. and Ebel, J.-P. (1967). Bull. Soc. Chim. Biol., 49,1679.
41. Nishimura, S. (1971). In Procedure in nucleic acids research (ed. G. L. Cantoni and
D. R. Davies), Vol. 2, pp. 542-64. Harper and Row, New York.
42. Spencer, M., Neave, E. J., and Webb, N. L. (1978). J. Chromatogr., 166,447.
43. Gillam, I., Millward, S., Blew, D., von Tigerstrom, M., Wimmer, E., and Tener, G.
M. (1967). Biochemistry, 6,3043.
44. Singhal, R. P., Griffin, G. D., and Novelli, G. D. (1976). Biochemistry, 15, 5083.
45. Nishimura, S., Shindo-Okada, N., and Crain, P. F. (1987). In Methods in
enzymology (ed. R. Wu), Vol. 155, pp. 373-8. Academic Press, London.
46. Holmes, W. M., Hurd, R. E., Reid, B. R., Rimerman, R. A., and Hatfield, G. W.
(1975). Proc. Natl. Acad. Sci. USA, 72, 1068.
47. Dudock, B. S. (1987). In Molecular biology of RNA. New perspectives (ed. M.
Inouye and B. S. Dudock), pp. 321-9. Academic Press Inc., New York and
London.
48. Brown, D. M. (1974). In Basic principles in nucleic acids chemistry (ed. P. O. P.
Ts'O), Vol. 2, pp. 1-90. Academic Press Inc., New York and London.
49. Werner, W., Krebs, W., Keith, G., and Dirheimer, G. (1976). Biochim. Biophys.
Acta, 432, 161.
50. Gueron, M. and Leroy, J.-L. (1978). Anal. Biochem., 91, 691.
51. Kennard, O. and Hunter, W. N. (1989). Q. Rev. Biophys., 22, 327.
52. Neidle, S. (1996). In Methods in molecular biology (ed. C. Jones, B. Mulloy, and
M. Sanderson), Vol. 56, pp. 267-92. Humana Press Inc., Totowa, NJ.
53. Wahl, M. C., Ramakrishnan, B., Ban, C., Chen, X., and Sundaralingam, M. (1996).
Acta Cryst., D52, 668.
54. Wang, A. H. J., Fujii, S., van Boom, J. H., and Rich, A. (1982). Proc. Natl. Acad.
Sci. USA, 79, 3968.
55. Hunter, W. N., Langlois D'Estaintot, B., and Kennard, O. (1989). Biochemistry,
28, 2444.
56. Langlois d'Estaintot, B., Dautant, A., Courseille, C., and Precigoux, G. (1993).
Eur. J. Biochem., 213, 673.
57. Yoon, C., Prive, G. G., Goodsell, D. S., and Dickerson, R. (1988). Proc. Natl.
Acad. Sci. USA, 85, 6332.
58. Prive, G. G., Heinemann, U., Chandrasegaran, S., Kan, L. S., Kopka, M., and
Dickerson, R. (1987). Science, 238, 498.
59. Heinemann, U. and Alings, C. (1991). EMBO J., 10, 35.
60. Wang, A. H. J., Quigley, G. J., Kolpak, F. J., Crawford, J. L., van Boom, J. H., van
der Marel, G., et al. (1979). Nature, 282, 680.
61. Wang, A. H. J., Hakoshima, T., van der Marel, G., van Boom, J. H., and Rich, A.
(1984). Cell, 37, 321.
62. Chevrier, B., Dock, A.-C., Hartmann, B., Leng, M., Moras, D., Thuong, M. Y., et
al. (1986). J. Mol. Biol., 188, 707.
63. Zhou, G. and Ho, P. S. (1990). Biochemistry, 29, 7229.
64. Chen, L., Cai, L., Zhang, X., and Rich, A. (1994). Biochemistry, 33, 13540.
240
 8: Nucleic acids and their complexes
65. Wang, A. H. J., Fujii, S., van Boom, J. H., van der Marel, G., van Boeckel, S. A.
A., and Rich, A. (1982). Nature, 299,601.
66. Gao, Q., Williams, L. D., Egli, L., Rabinowitch, D., Chen, S.-L., Quigley, G. J., et
al. (1991). Proc. Natl. Acad. Sci. USA, 88, 2422.
67. Brown, D. G., Sanderson, M. R., Skelly, J. V., Jenkins, T. C, Brown, T., Garman,
E., et al. (1990). EMBO J., 9,1329.
68. Kim, S. H., Quigley, G., Suddath, F. L., McPherson, A., Sneden, D., Kim, J. J., et
al. (1973).J. Mol Biol, 75, 421.
69. Johnson, C. D., Adolph, K., Rosa, J. J., Hall, M. D., and Sigler, P. B. (1970).
Nature, 226,1246.
70. Giege, R., Moras, D., and Thierry, J.-C. (1977). J. Mol. Biol., 115, 91.
71. Baeyens, K. J., Jancarik, J., and Holbrook, S. R. (1994). Acta Cryst., D50,764.
72. Mikol, V., Rodeau, J.-L., and Giege, R. (1989). J. Appl. Cryst., 22, 155.
73. Sakai, T. T. and Cohen, S. S. (1976). Prog. Nucleic Acid Res. Mol. Biol., 17,15.
74. Young, J. D., Bock, R. M., Nishimura, S., Ishikura, H., Yamada, Y., Rajbandhary,
U. L., et al. (1969). Science, 166,1527.
75. Sauter, C., Ng, J. D., Lorber, B., Keith, G., Brion, P., Hosseini, M.-W., et al.
(1999). /. Cryst. Growth, 196, 365.
76. Ladner, J. E., Finch, J. T., Klug, A., and Clark, B. F. C. (1972). J. Mol. Biol, 72, 99.
77. Bancroft, D., Williams, L. D., Rich, A., and Egli, M. (1994). Biochemistry, 33,
1073.
78. Holbrook, S. R., Sussman, J. L., Warrant, W. R., and Kim, S. H. (1978). J. Mol.
Biol., 123,631.
79. Gessner, R. V., Quigley, G. J., Wang, A. H. J., van der Marel, G., van Boom, J. H.,
and Rich, A. (1985). Biochemistry, 24, 237.
80. Brennan, R. G., Westhof, E., and Sundaralingam, M. (1986). J. Biomol. Struct.
Dyn.,3,649.
81. Einspahr, H., Cook, W. J., and Bugg, C. E. (1981). Biochemistry, 20, 5788.
82. Hingerty, B. E., Brown, R. S., and Klug, A. (1982). Biochim. Biophys. Acta, 697,
78.
83. Doudna, J. A., Grosshans, C., Gooding, A., and Kundrot, C. E. (1993). Proc. Natl.
Acad. Sci. USA, 90,7829.
84. Scott, W. G., Finch, J. T., Grenfell, R., Fogg, J., Smith, T., Gait, M. J., et al. (1995).
J. Mol. Biol., 250, 327.
85. Dickerson, R. E., Goodsell, D. S., Kopka, M. L., and Pjura, P. E. (1987). J.
Biomol. Struct. Dyn., 5, 557.
86. Moras, D. and Bergdoll, M. (1988). J. Cryst. Growth, 90,283.
87. Anderson, A. C., Earp, B. E., and Frederick, C. A. (1996). J. Mol. Biol., 259,696.
88. Dock-Bregeon, A.-C. and Moras, D. (1987). Cold Spring Harbor Symp. Quant.
Biol., 52,113.
89. Golden, B, Gooding, A. R., Podell, E. R., and Cech, T. R. (1996). RNA, 2,1295.
90. Freemont, P. S., Friedman, J. M., Beese, L., Sanderson, M. R., and Steitz, T. A.
(1988). Proc. Natt. Acad. Sci. USA, 85, 8924.
91. Brown, D. G. and Freemont, P. S. (1996). In Methods in molecular biology (ed. C.
Jones, B. Mulloy, and M. Sanderson), Vol. 56, pp. 293-318. Humana Press Inc.,
Totowa, NJ.
92. Grable, J., Frederick, C. A., Samudzi, C., Jen-Jacobson, L., Lesser, D., Greene, P.,
et al. (1984). /. Biomol. Struct. Dyn., 1,1149.
241
 A.-C. Dock-Bregeon et al.
93. Suk, D., Lahm, A., and Oefner, C. (1988). Nature, 332,464.
94. Luger, K., Mader, A. W., Richmond, R. K., Sargent, D. F., and Richmond, T. J.
(1997). Nature, 389, 251.
95. Lorber, B., Giege, R., Ebel, J.-P., Berthet, C., Thierry, J.-C, and Moras, D.
(1983). J. Biol. Chem., 258, 8429.
96. Jordan, S. R., Whitcombe, T. V., Berg, J. M., and Pabo, C. O. (1985). Science,
230, 1383.
97. Ruff, M., Cavarelli, J., Mikol, V., Lorber, B., Mitschler, A., Giege, R., et al.
(1988).J. Mol. Biol, 201,235.
98. Brennan, R. G., Takeda, Y., Kim, J., Anderson, W. F., and Matthews, B. W.
(1986). J. Mol. Biol, 188, 115.
99. Anderson, J., Ptashne, M., and Harrison, S. C. (1984). Proc. Natl. Acad. Sci.
USA, 81, 1307.
100. Aggarwal, A. K., Rodgers, D. W., Drottar, M., Ptashne, M., and Harrison, S.
(1988). Science, 242, 899.
101. Joachimiak, A., Marmorstein, R. Q., Schevitz, R. W., Mandecki, W., Fox, J. L.,
and Sigler, P. B. (1987). J. Biol. Chem., 262, 4917.
102. Schultz, S. C., Shields, G. C., and Steitz, T. A. (1990). J. Mol. Biol., 213,159.
103. Ellenberger, T. E., Brandel, C. J., Struhl, K, and Harrison, S. C. (1992). Cell, 71,
1223.
104. Pavletich, N. P. and Pabo, C. O. (1993). Science, 261, 1701.
105. Clark, K. L., Halay, E. D., Lai, E., and Burley, S. K. (1993). Nature, 364, 412.
106. Yang, W. and Steitz, T. A. (1995). Cell, 82,193.
107. Schwabe, J. W. R., Chapman, L., Finch, T., and Rhodes, D. (1993). Cell, 75,
567.
108. Kim, Y., Geiger, J. H., Hahn, S., and Sigler, P. B. (1993). Nature, 365,512.
109. Glover, J. N. M. and Harrison, S. C. (1995). Nature, 373, 257.
110. Nikolov, D. B., Chen, H., Halay, E. D., Usheva, A. A., Hisatake, K., Lee, D. K.,
et al. (1995). Nature, 377,119.
111. Sousa, R. (1995). Acta Cryst., D51,271.
112. Boeglin, M., Dock-Bregeon, A.-C., Eriani, G., Gangloff, J., Ruff, M., Poterzman,
A., et al. (1996). Acta Cryst., D52,211.
113. Arnez, J. and Steitz, T. A. (1994). Biochemistry, 33, 7567.
114. Cusack, S., Yaremchuk, A., and Tukalo, M. (1996). EMBO J., 15, 6321.
115. Reshetnikova, L., Khodyreva, S., Lavrik, O., Ankilova, V., Frolow, F., and Safro,
M. (1993). /. Mol Biol, 231, 927.
116. Yaremchuk, A. D., Tukalo, L, Krikliviy, N., Malchenko, N., Biou, V., Berthet-
Colominas, C., et al. (1992). FEBS Lett., 310,157.
117. Nissen, P., Reshetnikova, L., Siboska, G., Polekhina, G., Thirup, S., Kjeldgaard,
M., et al. (1994). FEBS Lett, 356,165.
118. Oubridge, C., Ito, N., Evans, P. R., Teo, C. H., and Nagai, K. (1994). Nature, 372,
432.
119. Record, M. J. J., Anderson, C. F., and Lohman, T. M. (1978). Rev. Biophys., 11,
103.
120. Giege, R., Lorber, B., Ebel, J.-P., Moras, D., Thierry, J.-C., Jacrot, B., et al.
(1982). Biochimie, 64, 357.
121. Trakhanov, S., Yusupov, M., Shirokov, V., Garber, M., Mitschler, A., Ruff, M., et
al (1989). J. Mol Biol, 209,327.
242
 8: Nucleic acids and their complexes
122. Thygesen, J., Krumholz, S., Levin, I., Zaytzev-Bashan, A., Harms, J., Bartels, H.,
et al. (1996). /. Cryst. Growth, 168, 308.
123. Ban, N., Freeborn, B., Nissen, P., Penczek, P., Grassucci, R. A., Sweet, R., et al.
(1998). Cell, 93,1105.
124. Fry, E., Logan, D., and Stuart, D. (1996). In Methods in molecular biology (ed. C.
Jones, B. Mulloy, and M. Sanderson), Vol. 56, pp. 319-63. Humana Press Inc.,
Totowa, NJ.
125. Smyth, M., Fry, E., Stuart, D., Lyons, C, Hoey, E., and Martin, S. J. (1993). J.
Mol. Biol., 231, 930.
126. Li, T., Zhang, A., luzuka, N., Nomoto, A., and Arnold, E. (1992). J. Mol. Biol.,
223,1171.
127. Arnold, E., Erickson, J. W., Fout, S. G., Frankenberger, E. A., Hecht, H. J., Luo,
M., et al. (1984). J. Mol. Biol., 177, 417.
128. Luo, M., Vriend, G., Kamer, G., Minor, I., Arnold, E., Rossman, M. G., et al.
(1987). Science, 235,182.
129. Speir, J. A., Munshi, S., Baker, T. S., and Johnson, J. E. (1993). Virology, 193,
234.
130. Sehnke, P. C., Harrington, M., Hosur, M. V., Li, Y. R., Usha, R., Tucker, R. C., et
al (1988). /. Cryst. Growth, 90, 222.
131. Koszelak, S., Dodds, J. A., and McPherson, A. (1989). J. Mol. Biol., 209,323.
132. Jones, T. A. and Liljas, L. (1984). J. Mol. Biol., 177, 735.
133. Harrison, S. C. and Jack, A. (1975). J. Mol. Biol., 97,173.
134. Canady, M. A., Larson, S. B., Day, J., and McPherson, A. (1996). Nature Struct.
Biol., 3,771.
135. Fisher, A. J., McKinney, B. R., Wery, J. P., and Johnson, J. E. (1992). Acta Cryst.,
48,515.
136. Valegard, K., Unge, T., Montelius, I., Strandberg, B., and Fiers, W. (1986). J.
Mol. Biol., 190, 587.
137. Luo, M., Tsao, J., Rossmann, M. G., Basak, S., and Compans, R. W. (1988). J.
Mol. Biol., 200, 209.
138. Willingmann, P., Krishnaswamy, S., McKenna, R., Smith, T. J., Olson, N. H.,
Rossmann, M. G., et al. (1990). J. Mol. Biol., 212,345.
139. Ng, J. D., Lorber, B., Witz, J., Theobald-Dietrich, A., Kern, D., and Giege, R.
(1996). J. Cryst. Growth, 168, 50.

243
This page intentionally left blank
 9

Crystallization of membrane
proteins
F. REISS-HUSSON and D. PICOT

1. Introduction
Crystallization of membrane proteins is one of the most recent developments
in protein crystal growth; in 1980, for the first time, two membrane proteins
were successfully crystallized, bacteriorhodopsin (1) and porin (2). Since then,
a number of membrane proteins (about 30) yielded three-dimensional
crystals. In several cases, the quality of the crystals was sufficient for X-ray
diffraction studies. The first atomic structure of a membrane protein, a
photosynthetic bacterial reaction centre, was described in 1985 (3), followed
by the structure of about ten other membrane protein families. Crystallization
of membrane proteins is now an actively growing field, and has been discussed
in several recent reviews (4-8).
The major difficulty in the study of membrane proteins, which for years
hampered their crystallization, comes from their peculiar solubility properties.
These originate from their tight association with other membrane compo-
nents, particularly lipids. Indeed integral membrane proteins contain hydro-
phobic surface regions buried in the lipid bilayer core, as well as hydrophilic
regions with charged or polar residues more or less exposed at the external
faces of the membrane. Disruption of the bilayer for isolating a membrane
protein can be done in various ways: extraction with organic solvents, use of
chaotropic agents, or solubilization by a detergent. The last method is the
most frequently used, since it maintains the biological activity of the protein if
a suitable detergent is found. This chapter will be restricted to specific aspects
of three-dimensional crystallizations done in micellar solutions of detergent.
In some cases, it is possible to separate soluble domains from the membrane
protein either by limited proteolysis or by genetic engineering. Such protein
fragments can then be treated as soluble proteins and so will not be discussed
further in this chapter. We refer to Chapter 12 and the review by Kuhlbrandt
(9) for the methodology of two-dimensional crystallization used for electron
diffraction.
 F. Reiss-Husson and D. Picot

2. Crystallization principles
The general principles discussed in this book for the crystallization of soluble
biological macromolecules apply for membrane proteins; the protein solution
must be brought to supersaturation by modifying its physical parameters
(concentrations of constituents, ionic strength, and so on), so that nucleation
may occur. The main differences from the behaviour of soluble proteins stem
from the following two points:
(a) The entity which is going to crystallize is the protein-detergent complex,
not the protein alone. Yet, most of the detergent found in the crystal is
disordered. This has been demonstrated for three detergents (C10DAO,
C12DAO, and C8G) associated with two bacterial reaction centres (10,11)
and OmpF porin (12). Usually, only a few ordered detergent molecules
are seen in the electron density maps. But the amount of disordered
detergent in the crystals is fairly high; about 200 molecules of detergent
are associated with one reaction centre protein, and form a ring around
the hydrophobic transmembrane a helices. The detergent ring is inter-
connected with its neighbours by bridges. Thus ribbon-like detergent
structures run throughout the crystal. These findings explain why the
characteristics of the detergent molecules (such as their length) are so
crucial in the crystallization process. Indeed they should fit around the
hydrophobic regions of the protein without hindering the interprotein
contacts.
(b) The solubility of the protein-detergent complex is governed not only by
the protein properties, but also (and mainly) by those of the detergent
micellar solution. Generally, as will be discussed below, this detergent is
non-ionic; its micellar solution therefore only exists in a limited range of
concentration and temperature, defined in a phase diagram (Figure 1).
Outside of this range, the micellar solution may spontaneously break
apart into two immiscible aqueous phases; one being enriched in deter-
gent, the other one remaining essentially depleted in detergent. The
temperatures and concentrations at which phase separation is observed
define a curve, called the consolution boundary. Depending on the
detergent and the crystallizing agent used, this boundary may be reached
starting from the micellar solution either by increasing or by decreasing
the temperature as shown in Figure 1. When phase separation takes place,
the solubilized membrane protein generally partitions into the detergent-
rich phase (but exceptions are known for glycoproteins) (13). Phase
separation is a function of all constituents of the solution such as deter-
gent, protein, nature and concentration of salt, concentration of a crystal-
lizing agent like PEG. Phase separation seems to play a major role in the
crystallization because it is quite often observed that crystallization takes
place right before phase separation occurs. Choosing crystallization con-
246
 9: Crystallization of membrane proteins
ditions close to the consolution boundary thus appears equivalent to
bringing the protein-detergent complexes into a supersaturated state.
In conclusion, the chosen detergent plays a crucial role in membrane pro-
tein crystallization and it is therefore important to gain knowledge of its
properties before embarking on crystallization experiments.

3. Detergents for crystallization

The detergents which have already led to crystallization of membrane pro-
teins belong to several chemical classes (Table 1). Some have been specifically

Figure 1. Schematic phase diagrams commonly observed for non-ionic detergent-water

mixtures. Depending on the detergent, the region occupied by two immiscible liquid
phases may be situated either above (see upper diagram) or below (see lower diagram)
the consolution boundary.

247
 F. Reiss-Husson and D. Picot

Table 1, Structure of detergents

synthesized for this purpose. Most of them, however, were already used by
biochemists for solubilizing and purifying membrane proteins. As such they
have been reviewed in a recent volume of this series (14). With a few except-
ions, they are either non-ionic, with uncharged polar groups, or zwitterionic at
the pH used. Furthermore they all are short aliphatic compounds, the length
of their hydrophobic part not exceeding that of a normal C12 hydrocarbon
chain. This last feature results in moderate or high critical micellar
concentration (CMC) values (the CMC being the concentration limit between
molecular and micellar solutions, compare Figure 1). In a micellar solution,
micelles are in dynamic equilibrium with the monomeric detergent still pre-
sent in concentration equal to the CMC. Thus, the higher the CMC, the larger
this exchange. Such a mobile character might play a role in the crystallization
of the detergent-protein complexes.
One further requirement of these detergents is purity and chemical
homogeneity, which are important for:
(a) Reproducibility of experiments. When a detergent is heterogeneous (e.g.
it contains a mixture of various hydrocarbon chains) or impure (traces of
248
 9: Crystallization of membrane proteins
fatty alcohol, and so on), its composition is badly defined and may vary
from batch to batch. This may lead to the intolerable situation that, with a
new detergent batch, crystals may no longer be obtained or still grow but
with lower quality.
(b) Quality of the crystals. For example trace impurities in C8G have been
reported to interfere with crystal growth of bacteriorhodopsin.
Thus it is advisable to pay attention to purity of the selected detergent and
be able to check it. Thin-layer chromatography (TLC) is one of the useful
tools, simple and fast to perform (Protocol 1). Trace impurities may however
escape detection by TLC.

Protocol 1. Detergent TLC of n-alkyl-|3-glucosides and

-maltosidesa

Equipment and reagents

• TLC glass tank • Detergent of choice
• Silica gel plate (e.g. Merck 60, 0.25 mm • 5 (ul microsyringe (e.g. Hamilton)
thickness) . iodine: put I2 crystals in a closed glass
• Ethyl acetate:methanol (4:1, v/v) desiccator until brown fumes develop

Method
1. Equilibrate a silica gel plate in a TLC tank with enough ethyl acetate:
methanol (4:1, v/v) to wet the bottom of the plate over 5 mm.
2. Dissolve 1 mg detergent in 100 ul ethanol in a microcentrifuge tube.
3. Using a microsyringe, spot 10 ul of this solution at the bottom of the
plate, 1 cm from the edge, and let it dry.
4. Put the plate back in the tank and wait until the solvent front reaches
within 1 cm of the top.
5. Take the plate out and let it dry.
6. Put the plate in the desiccator containing iodine vapours and let it
stain.b Only one spot should be present.

"TLC of n-alkyl-oligoethylene glycol-monoethers can be performed with water saturated

methylethylketone as the solvent.
"Alternatively, spray under a hood with a 2 M H2S04 solution in water. Char in a 90°C oven or
on a hot plate.

Another parameter which is very sensitive to detergent composition is the

CMC. Comparing the experimental value to those in the literature gives a
clue to the purity of the sample. The CMC can be measured by changes in
several properties of the solution (surface tension, drop size) (15), or by
spectral changes in absorption (16) or fluorescence (17) of solubilized dyes
(Protocol 2).
249
 F. Reiss-Husson and D. Picot

Protocol 2. Determination of the CMC of a detergent using ANS

fluorescence

Equipment and reagents

• 400 (uMANS (8-anilino-l-naphthalene sul- • Detergent of choice
fonic acid, Mg2+ salt) in water • Fluorimeter (excitation at 370 nm)

Method
1. Prepare a working solution of 10 uM ANS in water by dilution of a
stock 400 (AM ANS solution.
2. Prepare 500 ul of the stock detergent solution in 10 (uM ANS at a
detergent concentration about 100 times the expected CMC. Mix
thoroughly. Fluorescence of this solution is taken as the 100% control.
3. Read the fluorescence emission at 490 nm with excitation at 370 nm
while titrating a 2 ml sample of 10 (uMANS with small aliquots of stock
detergent solution (up to 50 ul). Use the fluorescence of 10 uM ANS as
the blank.
4. Plot the relative fluorescence versus detergent concentration. A steep
increase in fluorescence indicates the onset of micellization. Draw a
straight line through the points in the steep increase region; its
intersection with the x axis is the CMC of the detergent.

Table 2 summarizes useful properties of the most often used detergents

(18-20). Specific points will be discussed below.

3.1 n-Alkyl-B-glucosides (CnG)

The C8 compound (C8G) has been used in most crystallizations reported so
far. It is also very useful in membrane protein biochemistry and generally
behaves like a mild detergent (but not always; some membrane proteins are
inactivated by C8G) (21). Its CMC is high and it forms small micelles in water.
Therefore the dialysis of C8G is fast (half-time of a few hours). Because of the
possible hydrolysis of the ether linkage C8G (and other glucosides) must be
stored dry and frozen, and freshly prepared solutions should be used. This
instability may explain why various commercial brands have been found to
contain unidentified trace impurities which could be eliminated by chromato-
graphy on a mixed-bed strong ion exchanger (22) (Protocol 3). This purification
primarily removes ionic contaminants, but might also remove organic impuri-
ties. In our hands, C8G from Bachem was satisfactory without repurification.
Members of the CnG class with C7 and C6 chain length are commercially
available, but their CMC is much too high. The C9 compound has been used
once for crystallization (4). Compounds with longer chain are of limited use
because of their poor solubility, not exceeding the CMC.
250
 9: Crystallization of membrane proteins

Table 2. Properties of some detergents used for crystallization

Detergent Molecular CMC (mM) Ref. Monomers Suppliersa

weight per micelle
C8G 292.4 23 18 78 Various
C9G 306.4 6.5 18 C,F
C10M 482.6 2.2 17 C,F
C11M 496.6 0.59 A
C12M 510.6 0.16 17 130 Various
Hecameg 335.4 20 57 C, V
C10DAO 201.4 10.4 4 F
C11DAO 215.4 O
C12DAO 229.4 2.0 20 73 Various
C8E4 306 8.5 58 82 B, K
C8E5 350 9.2 58 B, K
C10E8 515.1 0.10 58 C
C12E8 518 0.071 58 120 C,K
C12E9 583 0.071 58 C,F
C8HESO 206 29.9 27 B,O
MEGA-10 349.5 5 27 O

a A, Anatrace; B, Bachem; C, Calbiochem; F, Fluka; K, Kohyo; O, Oxyl; V, Vegatec.

Protocol 3. Purification of C8Ga

Equipment and reagents

• Chromatography column (10 x 150 mm) Strong mixed-bed ion exchanger (e.g.
• C8G Rexin I-300 or Bio-Rad AG501X8) in the H-
• Ethanol GH form

Method
1. Pour the column with the ion exchanger resin.
2. Wash the column with 120 ml ethanol then with 600 ml water. Stop the
flow when water is draining the gel surface.
3. Dissolve 5 g C8G in 50 ml water and put the solution on the gel. Elute
at a flow rate of 0.2 ml/min. Then wash with water at the same rate.
4. Collect the first 100 ml of eluate. Lyophilize and store at -20°C.

* This protocol may be scaled down for smaller detergent quantities.

3.2 n-Alkyl-thioglucosides
Although these detergents have not been used very often in crystallizations
(23, 24), they could be assayed instead of the glucosides. The C6, C7, C8, and
C10compounds are commercially available.
251
Table 3. Crystallization conditions for some membrane proteins

Proteina Organism Detergentb Precipitantc Additived Methode Ref.

Reaction centre Rhodopseudomonas C12DAO AS HT VD 59
viridis
Reaction centre Rhodobacter C8G PEG/NaCI VD 60
sphaeroides R26
Reaction centre Rhodobacter C12DAO PEG/NaCI HT VD 61
sphaeroides R26
Reaction centre Rhodobacter C8G PEG/NaCI HT VD 62
sphaeroides 241
Reaction centre Rhodobacter C8G PEG/NaCI MD 63
sphaeroidesY
Reaction centre Rhodobacter C8G PEG HT/BZ SD 64
sphaeroides 241
Reaction centre Rhodobacter C12DAO KPi HT/1,4-dioxane VD 65
sphaeroides 241
Reaction centre Chromatium tepidum C8G PEG/NaCI HD 66
Reaction centre Chloroflexus C10E8 PEG GAPA SD 67
aurantiacus
LH B800-850 Rhodopseudomonas C8G P04 BZ VD 68
acidophila
LH B800-850 Rhodospirillum C11DAO AS HT SD 69
molischianum
Porin OmpF Escherichia coli C8HESO PEG/MgCI2 MD 70
C8POE
Porin OmpF Escherichia coli C8G PEG/NaCI VD, MD 27
C8POE
Porin Rhodobacter C8E4 PEG/LiCI 35
capsulatus
Porin Rhodopseudomonas C8E4 PEG/LiCI 71
blastica
Porin PhoE Escherichia coli C8G PEG/NaCI 72
C8E4
Porin LamB Escherichia coli C10M PEG/MgCI2 MD 73
C12E9
Porin ScrY Staphylococcus C8G PEG/LiCI/ SD 41
typhimurium C6DAO MgSO4
Porin Paracoccus C8G PEG/KCI SD 74
denitrificans
PSI Synechocystis C12M MgS04 MD 42
elongatus
Cytochrome oxidase Bovine heart C10M PEG 75
Cytochrome Paracoccus C12M PEG-ME/ 37
oxidase/antibody denitrificans NH4 acetate
complex
Cytochrome bc1 Bovine MEGA-10 PEG/KCI Glycerol 50
(or SPC)
Cytochrome bc1 Bovine HECAMEG PEG 49
Prostaglandin H Sheep C8G PEG/NaCI HD 6
synthase l
Prostaglandin H Human C8E5 PEG/NaCI 76
synthase II
Prostaglandin H Murine C8G PEG-ME VD 46
synthase II
a-Haemolysin Staphylococcus C8G AS 77
aureus PEG-ME
Phospholipase Plda Escherichia coli C8G MPD/CaCI2 HD 43
a
LH, light harvesting; PS, photosystem.
b
For detailed properties see Table 7.
cAS,ammonium sulfate; PEG-ME, polyethylene glycol monomethylether; MPD, 2-methyl-2,4-pentane diol.
dHT, heptanetriol; BZ, benzamidine; GAPA, N-N-bis (gluconamidopropyl)-amine.
e
VD, vapour diffusion; SD, sitting drop; HD, hanging drop; MD, microdialysis.
 F. Reiss-Husson and D. Picot

3.3 n-Alkyl-maltosides (CnM)

The C12, C11, and C10 compounds have been used in crystallizations. C12M was
also used for purification of membrane proteins and is considered to be
superior to C8G. Its CMC is fairly low, and the size of the micelles in pure
water is larger than for C8G. Dialysis of the detergent is an extremely slow
process. It is likely to hydrolyse and should be stored frozen. Contamination
by dodecanol decreases its solubility and may be detected by appearance of a
white precipitate in solutions kept at 5°C (18). Contamination with the a
isomer influences the crystallization process (P. Fromme, personal com-
munication) but can be checked by reverse-phase HPLC of detergent batches.

3.4 n-Alkyl-dimethylamineoxides (CnDAO)

No hydrolysis is observed for this class of detergents; solutions are stable
when kept at 5°C. The C11 and C12 compounds have been used successfully
for the crystallization of bacterial reaction centres and antenna (see Table 3).
They are zwitterionic at pH > 3 and cationic at pH < 3. The micellar size of
C12DAO is similar to C8G, but its CMC is lower, resulting in a slower dialysis
rate. It is available from various sources, either as the pure C12 species, or as a
cheaper mixture containing primarily the C12 species but also other chain
lengths. Traces of H2O2 (left over from synthesis) may contaminate some
batches. They can be removed by adding 10 (ug of catalase per ml of a 30%
stock solution of C12DAO in water. Shorter chain analogues (C6-C10) are
available. Some of them have been used for crystallization added to C12 one.

3.5 n-Alkyl-oligoethylene glycol-monoethers (CnEm)

Besides pure C12E8 which has been used for biochemistry, pure compounds of
shorter hydrocarbon chain (C5-C8) are commercially available with a defined
number of ethylene glycol units ranging from one to five. The C8E4, C8E5, and
C12E9 detergents have been used in crystallizations, either pure or mixed with
C8G. A cheaper polydisperse mixture, C8Em, has also been used mixed with
C12M in some cases.
Aqueous solutions of these detergents are not stable; peroxides and
aldehydes are formed on storage under air, particularly in the light. The
purification protocol (Protocol 4} involves treatment with a reducing agent
(SnCl2 or Na2SO3) followed by solvent extraction (25).
Besides these five main classes, other detergents have been used for
crystallization such as the n-alkyl-glucamides (MEGA-9, MEGA-10) (26), n-
heptylcarbamoyl-methyl-a-D-glucoside (HECAMEG), n-methyl-n-decanoyl-
maltosylamine, but the relatively poor solubility of the MEGA and
HECAMEG detergents may however cause some problems (beware: they
crystallize easily).
254
 9: Crystallization of membrane proteins

Protocol 4. Purification of n-alkyl-oligoethylene glycol detergents

Equipment and reagents

• 10% (w/v) solution of the detergent in • Dichloromethane
distilled water . 1% NaOH, 10% NaCI
* SnCI2 • Anhydrous Na2SO4
• 10% NaCI • Flash evaporator (e.g. Buchi)

Method
1. Stir for 2 h a 10% (w/v) solution of the detergent in distilled water with
SnCI2 (0.5% (w/v) final concentration).
2. Add NaCI to 10% (w/v) final concentration, then add an equal volume
of dichloromethane and mix thoroughly. Let stand until the two layers
separate.
3. Discard the upper water layer and recover the lower organic layer
which contains the detergent.
4. Extract the organic phase with an equal volume of 1% NaOH, 10%
NaCI, then three times with 10% NaCI (the pH of the final NaCI layer
must be 7). Each time discard the water layer.
5. Dry the organic phase for 24 h over anhydrous Na2SO4, then in a flash
evaporator at 40 °C.
6. Store the purified detergent at -20°C.

New detergents related to those described appear steadily on the market.

Personal investigation of chemical catalogues may be fruitful.

4. Purification of membrane proteins before

crystallization
General methods for solubilization and purification of membrane proteins
have been given (14) and will not be detailed. This section only focuses on
several specific points: purity requirement, procedures for detergent exchange
and for sample concentration.

4.1 Purity requirements

As for the crystallization of soluble proteins, the starting protein solution
should be as pure and homogeneous as possible and the same precautions
should be taken to avoid denaturation, proteolysis, and microheterogeneities
(see Chapter 2).
Protein purity is usually checked by SDS-PAGE. However one should be
aware that contaminants may escape detection; e.g. lipopolysaccharides and
lipoprotein contaminants of porin preparations are not stained by Coomassie
Blue but only by silver staining (27).
255
 F. Reiss-Husson and D. Picot
Residual lipids represent another common source of impurity for mem-
brane protein preparations. Indeed, they may withstand the solubilization by
a mild detergent and remain associated with the detergent-protein com-
plexes. Their non-specific, random binding prevented crystallization in several
cases. Lipid content should therefore be checked by TLC (28) of organic
solvent extracts (29) (see Protocol 5). Alternatively, phosphorus content of
these extracts can be measured (30) as most residual lipids are phospholipids.
If present, lipids should be eliminated as much as possible; the ease of
removal depends on the protein-detergent couple and there is no general
recipe. Chromatographic techniques may be adequate even if they were not
devised for this purpose. Ion exchange chromatography has been reported in
several cases to lower the lipid content, probably because of the extensive
detergent washes of the adsorbed protein. Another chromatographic step
which has been used for purification of membrane proteins prior to crystal-
lization is chromatofocusing in the presence of a detergent (31). Besides
providing a homogeneous preparation of defined isoelectric point, it may also
result in lowering the phospholipid content. Purity of the detergent may play
a role in elution by IEF, as reported in the case of bacteriorhodopsin which
seemed heterogeneous when impure C8G was present (22).

Protocol 5. Analysis of lipids in membrane protein preparations

Equipment and reagents

• Protein sample (~ 1 mg/ml) Reagents for TLC analysis (as described in
• Hexane:isopropanol (3:2, v/v) Protocol 7 except use chloroform:methanol:
• Nitrogen gas (or flash evaporator) water (65:25:4, by vol.) as the solvent)
• Chloroform

Method
1. Mix the protein solution (~ 1 mg/ml) with 20 vol. of hexane:
isopropanol (3:2, v/v).
2. Shake well and then centrifuge at low speed (5000 g) for 20 min.
3. Recover the supernatant. Repeat steps 1 and 2 on the pellet.
4. Combine the supernatants and dry this under a stream of N2 or with a
flash evaporator.
5. Dissolve the residue in the minimal volume of chloroform.
6. Carry out TLC analysis of the lipid extract as described in Protocol 1,
except use chloroform:methanol:water (65:25:4, by vol.) as the solvent.
The neutral lipids run near the solvent front and the other lipids are
fractionated into various classes. They are identified with reference to
known standards and published Rf values, in addition to the use of
specific stains (28) instead of iodine staining or H2SO4 charring.

256
 9: Crystallization of membrane proteins
Homogeneity of the preparation requires also its monodispersity, i.e. all the
protein-detergent complexes should have the same composition. This is
verified by gel filtration experiments, e.g. on FPLC columns (Superose gels,
Pharmacia) or HPLC ones (such as the TSK-SW or TSK-PW gels, Toso
Haas). From these experiments one can estimate the size of the whole com-
plex, including detergent. On the other hand, the amount of bound detergent
may be determined by several techniques (see ref. 32 for a review). Com-
bining these results allows the aggregation state of the protein-detergent
complex to be determined and controlled.

4.2 Detergent exchange

Purification of a membrane protein is often done in the presence of a deter-
gent and crystallization performed with a different one. This may be because
of the cost of one detergent, or because solubilization and purification require
a particular detergent, or the variation of the detergent nature during crystal-
lization trials. In all cases, exchange of detergent hereafter called detergent 1
and detergent 2 has to be performed; several methods may be used.

4.2.1 Dialysis
This is the simplest procedure but not applicable in all cases. The meaningful
parameter of a dialysis membrane is its cut-off value. It must be low enough
for retention of the protein-detergent complex. Thus for large complexes,
highly permeable membranes can be used; e.g. for a complex of 100 kDa, a
Spectrapor 7 membrane with a cut-off value of 50 kDa may be used. With
such a pore size, exchange of detergents with CMC values higher than 1 mM
is relatively rapid (a few days). On the other hand, for a small complex (e.g.
15 kDa) a Spectrapor 1 membrane (cut-off value 6 kDa) should be chosen and
only detergents of high CMC (10 mM or so) will exchange at an acceptable
rate. Since the diffusion rate between two detergents may differ by several
orders of magnitude, care should be taken in order to avoid detergent depletion
leading to irreversible aggregation or increase of detergent concentration
causing inactivation of the protein.
In favourable cases dialysis may be performed in two steps; e.g. it is possible
with the bacterial reaction centre, to exchange 0.1% C12DAO for 0.8% C8G
as follows:
(a) First dialyse for 48 h against a detergent-free buffer, with several changes
of reservoir (Spectrapor semi-microtubing, cut-off 12 kDa). Removal of
C12DAO results in increased turbidity of the sample.
(b) Transfer the bag in a C8G containing buffer for another 24 h. Loss of
turbidity indicates redissolution of the protein.
From a practical point of view, dialysis may be performed either in the
familiar closed bags, or for small volumes (< 500 ul) in microdialysis
257
 F. Reiss-Husson and D. Picot
cells, either home-built (see Chapter 5) or commercially available (Pierce,
Amicon). Some dialysis membranes, particularly those stored wet (e.g.
Spectrapor type 7, Amicon) are specially prone to fungi contamination.
Before use, a good precaution is to boil membranes for 1 min in 1% (w/v)
NaHCO3, then to soak them three times in highly pure water (Milli Q grade),
and to use them immediately afterwards.

4.2.2 Chromatography
The sample which contains detergent 1 is chromatographed on a column
equilibrated with detergent 2, and eluted with detergent 2. This method is
feasible with any type of detergent, with various chromatographic supports.
(a) Gel filtration is gentle and can be used with all detergents. However, it
usually dilutes the sample appreciably.
(b) Ion exchange chromatography (with DEAE, CM exchangers, or some-
times hydroxyapatite) is restricted to non-ionic detergents; it has the
advantage of concentrating the sample when elution is done by a steep
salt increase, but one must check that the stability of the protein is not
affected by the shift of the CMC, which is induced by the higher salt
concentration.
If the presence of salt in the final sample is not wanted, a mixed column
consisting of ion exchanger superposed on gel filtration matrix (Sephadex
G25) will exchange the detergent and desalt the sample altogether (33).
Microcolumns built from Pasteur pipettes are useful for such ion exchange
procedure.

4.2.3 Precipitation
The protein in the presence of detergent 1 is first precipitated with cold ethanol,
which is a solvent of detergent 1; the precipitate is washed to eliminate deter-
gent 1, and redissolved in detergent 2. This method has been used only for
porins as it requires a very sturdy protein. Salt or PEG precipitation may also
be used in some cases but care should be taken to avoid phase separation.
Furthermore, since precipitated protein still bound a large amount of
detergent, several precipitation cycles are needed.
The efficiency of these procedures can be judged from the absence of
detergent 1 in the final sample. Unfortunately, very few detergents exist in a
labelled form and those are expensive. Colorimetric determination is possible
for glucosides and maltosides with reagents specific for reducing sugars (34).
In other cases, TLC of extracts is the only method.

4.3 Sample concentration

A concentrated stock solution (often 10 mg protein/ml or more) is required to
prepare the samples for crystallization trials. As already mentioned (see
258
 9: Crystallization of membrane proteins
Section 4.2), concentration may be achieved during the detergent exchange by
ion exchange chromatography. Another very useful method is ultrafiltration.
Here again a large number of devices is available. For final volumes of 1 ml
and up, stirred cells equipped with Amicon YM or XM membranes (with a
wide choice of cut-off values) are convenient, and may be operated under
nitrogen pressure. The major drawback is foaming during stirring. For smaller
volumes, a number of devices allow concentration down to 100 ul or so with
low speed centrifugation (Millipore, Amicon). Whatever the device, the cut-
off value should be chosen as high as possible, taking into account the size of
the protein-detergent complex; this avoids concentrating the detergent in
solution in the same time as the protein-detergent complexes.

5. Crystallization protocols
For conducting crystallization trials with a membrane protein, similar
strategies, like the incomplete factorial design (see Chapters 4 and 5), to those
developed for the soluble proteins can be used (7). Several parameters have
to be chosen (see Chapter 1, Table 7) to which should be added the nature
and concentration of the detergent. Some of these parameters have been
discussed earlier with reference to soluble proteins (see Chapter 5). One of
these parameters, the purification of the protein, will influence the crystal-
lization condition to such an extent that it may be as important to modify the
purification as the crystallization protocol. This has been the case for porin
(35) and prostaglandin H synthase (31), for which the change of the detergent
used for solubilization was critical to obtain good crystals, even if an other
detergent was then used for the crystallization.
Protein-protein, protein-detergent, and detergent-detergent interactions
can be observed in membrane protein crystals. Depending on the nature of
the protein and the detergent, the crystallization process will be more in-
fluenced by one or another type of interaction. Protein-protein interactions
are more specific than detergent-detergent interactions and should thus yield
better crystals. This may explain why crystals of the reaction centre from Rps.
viridis diffract better than those from Rb. sphaeroides since the former is
crystallized with an additional soluble subunit. Therefore, larger proteins, but
with a larger polar domain, may be easier to crystallize than their more hydro-
phobic counterparts. An increase of the hydrophilic surface of cytochrome c
oxidase of Paracoccus denitrificans has been obtained by forming a complex
with a conformation specific engineered Fv fragment (36). This has allowed
well diffracting crystals to be grown using the detergent C12M, that is able to
maintain the activity. Thus, the Fv fragment counterbalances the disadvantage
of C12M, i.e. to form large micelles (37). We will stress now a few specific
points based on published crystallization protocols (Table 3).

259
 F. Reiss-Husson and D. Picot

5.1 Detergent
The choice of detergent is still empirical. The first criterion is to maintain the
functional and structural integrity of the protein. Not only the type but also
the detergent concentration are important. Furthermore, the behaviour and
stability of the protein will be very different below and above the CMC as well
as above and below the consolution boundaries. The optimal stability of a
membrane protein is often observed around the CMC, which may therefore
be a good starting detergent concentration for a crystallization experiment. A
number of membrane proteins crystallize with a wide variety of detergents. In
one case, a systematic search has been done over 23 detergents with Omp F,
an E. coli porin (23). Among them, 16 non-ionic detergents (from classes
described in Table 7) could be used successfully. Interestingly, OmpF
crystallized also in micellar solutions of short chains lecithins (diC6- or
diC7-glycerophosphatidylcholines) or lysolecithins (monoC14- or monoC16-
glycerophosphatidylcholines) used as detergents. However, crystals could not
be obtained with ionic detergents, nor with detergents derived from bile salts
(cholate, CHAPS, or CHAPSO). A non-ionic, non-steroid polar group and a
short alkyl chain seemed thus to be the only requirement, without narrow
specificity.
The reverse situation may prevail, and strict requirements may exist for
chain length or detergent type. For example, a light-harvesting chloroplast
protein, LHCII, crystallizes reliably with C9G and poorly with C8G (4).
Therefore, for an unknown protein, screening should be done using at least
two homologues of each detergent class. However, before beginning an ex-
tensive screening, one could try the most popular C8G and C12DAO deter-
gents. Indeed they allowed a number of successful crystallizations such as
bacterial reaction centres and light-harvesting complexes; E. coli porins gave
crystals with one of them or both. From a certain point of view, these two
detergents may be considered as equivalent: in two bacterial reaction centre
crystals, the regions they occupied respectively around the hydrophobic a
helices could be nearly superimposed (10, 11).
C12DAO should be tried alone and also in the presence of an additive (see
Section 7) which was required in some cases. C8G has been used either alone,
or mixed with low amounts of other short chain detergents; these were
however not essential for crystallization but improved crystal growth. Thus
first trials may be done with pure C8G only.
The initial concentration of detergent in the sample should be chosen only
slightly higher than the CMC (see Table 2). Under this condition, the deter-
gent is present either as monomers or as part of protein-detergent complexes,
with very few pure detergent micelles. For porin crystallization, the optimal
range of C8G concentration is narrow: 8-9 mg/ml. Below the CMC (less than
7 mg/ml) or well above it (more than 10 mg/ml) growth rate and nucleation
are excessive (38). The same optimal range was found for C8G with bacterial
260
 9: Crystallization of membrane proteins
reaction centres, and it does not seem to be very sensitive to the protein
concentration.

5.2 Additives
Small molecules with amphiphilic character have been sometimes added to
the crystallization media (see Table 3). Most often used is heptane-l,2,3-triol
(high melting point isomer); hexane-l,6-diol, benzamidine, glycerol, and
triethylamine phosphate were also used. Their effects are various.
(a) They may be absolutely required: best example is heptane-l,2,3-triol,
essential for crystallizing Rps. viridis reaction centre with C12DAO.
(b) They improve crystal growth and quality, but crystallization still takes
place in their absence; this is the case of heptane-l,2,3-triol for Rb.
sphaeroides reaction centre with C12DAO.
(c) In other cases their presence has no effect whatsoever; it is the case of
E. coli porin with C8G.
These 'additives' have been usually used at quite high molarities, when
compared to those of detergent; in the cases cited above heptane-l,2,3-triol
was present at about 0.1 M. Their mode of action is still poorly understood.
One hypothesis is that their small size and amphiphilic character allow them
to localize between neighbouring protein molecules, in regions inaccessible to
detergent, filling thus voids in the lattice (39). Another explanation, which has
some experimental support, is they modify the micellar structure of the deter-
gent by partitioning into the micelles (40); thus they change the consolution
boundaries and could bring them in a favourable temperature range.
Detergents with short aliphatic chains and a CMC too high to be used alone
may also be used as additive (41). The addition of another detergent at low
concentration (C8E4) may also have a similar effect to the other type of
additive (31).
Whatever the case, such additives may be tried when all previous trials
done in their absence have failed. It is better to test them beforehand on the
protein in solution; indeed they may have a denaturing effect.

5.3 Crystallizing agent

In Table 3, one may notice that the conditions used so far to achieve super-
saturation are not very diverse: either the presence of PEG at a suitable con-
centration, or 'salting-out' at high ionic strength. There are a few exceptions:
the PSI complex has been crystallized under 'salting-in' conditions (42) and
OMPLA from E. coli uses the organic solvent 2,4-methyl-pentane diol (43).
5.3.1 PEG
PEG, which is a classical crystallizing agent for soluble proteins, probably acts
in the same way on detergent solutions by competing with the micelle polar
261
 F. Reiss-Husson and D. Picot
groups for water molecules and by modifying the structure of the solvent. This
destabilizes the micelles and the protein-detergent complexes. Thus, when
PEG is continuously added to a membrane protein in detergent solution, the
micellar solution is perturbed and one of two situations may occur: protein
and detergent will precipitate, or the solution spontaneously separates in two
immiscible liquid phases, with the protein and most of the detergent in one
phase and PEG in the other (see Figure 1). At a given temperature, the
solubility limit of the protein depends on all constituents of the solution
(PEG, detergent, protein, and salts). Crystals will eventually form when the
system is slowly approaching this limit by increasing PEG concentration.
PEG and the more recent PEG-ME (monomethyl ether) are available in a
variety of polymeric ranges. The optimal range of PEG concentration for
crystallization of a given protein depends on PEG molecular weight. It may
be very narrow (about 1.5%) as observed for porin (38) and a bacterial re-
action centre (44). Repurification is recommended before use (45) and is
described in Chapters 2 and 5. Low molecular weight PEG and PEG-ME are
also suitable with membrane protein (35, 46) and may be suitable for cryo-
crystallography (46).
Crystallization of soluble proteins with PEG is usually performed at low
ionic strength. On the contrary, for membrane proteins salt is generally
required along with PEG (see Table 3). Again, the optimal range of salt
concentration may be narrow. Thus the two meaningful parameters are the
concentrations of PEG and of salt, for a given protein with a given detergent
(pH and temperature being fixed). For C8G at room temperature, these
conditions are quite similar for unrelated proteins like E. coli OmpF porin,
Rb. sphaeroides reaction centre, prostaglandin H synthase, and cytochrome
bc1 (47) (see Table 3), as if they were little influenced by the protein but
mainly by the detergent. This observation, if it is generalized by further
experiments, would greatly simplify the search for crystallization conditions
with the system C8G-PEG-salt.
No systematic comparison of the influence of various salts on the crystal-
lization conditions of membrane proteins has been published. We have
observed that in the case of crystallization of a bacterial reaction centre NaCl
could be replaced by a number of other monovalent salts; their optimal con-
centrations were not identical and had to be optimized. However there was no
significant influence of the salt nature on crystal growth or characteristics
(Reiss-Husson, unpublished experiments).
5.3.2 'Salting-out'
Surprisingly, a few membrane proteins have been crystallized by high salt
concentrations; ammonium sulfate and phosphate have been used in the
presence of C12DAO, C11DAO, C8G, or C9G (Table 3). These salts decrease
the solubility of a membrane protein-detergent complex by a mechanism
probably more complicated than for the salting-out of soluble proteins (see
262
 9: Crystallization of membrane proteins
Chapter 9). Their presence may modify the interactions between water and
the hydrophilic regions of the protein. More importantly, they induce a
salting-out of the detergent itself, e.g. monovalent salts modify the upper
consolution boundary of C8E5; the solubility shift is mainly determined by the
anions and follows the Hofmeister series (48).
How to choose between these various precipitating agents? The choice may
be restricted by considering the stability of the protein in their presence. For
example LHCII, a light-harvesting chloroplast membrane protein, is de-
natured by PEG and by ammonium sulfate; it was therefore crystallized in the
presence of high phosphate concentrations, which do not affect it. The bio-
logical activity of the protein should be checked in the presence of increasing
amounts of these various precipitants before any crystallization trial. Then the
useful range of precipitant concentrations is determined by measuring the
lowest precipitant concentration leading to phase separation or precipitation,
at given protein and detergent concentrations.

5.4 Optimization
Once crystals (most often microcrystals) have been observed in trial experi-
ments, crystallization conditions have to be improved for crystal size and
quality. The strategy is based on the same principles as for soluble proteins
(see Chapter 4). Excessive nucleation, leading to a 'shower' of microcrystals,
should be avoided; at the same time, growth rate should be kept low enough,
as crystal defects are frequently observed when the rate is too high (hollow
crystals may even be obtained). Practically, this implies repeating the trials
with the different parameters (pH, concentrations, and so on) slightly modi-
fied around their initially positive values, over a fine grid. At this stage, use of
an additive or of a small amount of a second detergent (compare Section 5.2)
may be included as a further variation.
The crystal form of several membrane proteins has been shown to depend
on several parameters: type of detergent, pH, nature of the buffer, and the
ionic strength when PEG together with salt are present (38). By varying these
parameters, it has been possible to select a form which grows better, or is
more suitable for structure determination because of its symmetry or unit cell
dimensions.

6. Experimental techniques
Crystallization of membrane proteins may be performed with all the experi-
mental set-ups described in Chapter 5. Vapour diffusion and microdialysis
have been more frequently used than batch crystallization and free liquid-
liquid interface (see Table 3). Crystallization of cyt bc1 in gel of agarose has
also been described (26).
Because of the wetting properties of detergents, their drops tend to spread
263
 F. Reiss-Husson and D. Picot
when formed on a planar glass slide, and to fall out when the slide is inverted.
Therefore, vapour diffusion with hanging drops is restricted to drop volumes
less than 10 ul. With sitting drops formed on depression slides, there is no
restriction in volume.
Microdialysis is performed in capillaries or microtubes closed by dialysis
membrane with sample volumes less than 150 ul, equilibrated against a
reservoir. Choice of the cut-off value of the membrane should take into
account the molecular weights of the components of the sample (see Section
4.2.1). Depending on this cut-off value, and also on the thickness of the
membrane, dialysis rate, and diffusable species may be controlled; e.g. PEG
4000 and 6000 diffuse (but slowly) through a membrane of cut-off value 25 000
daltons, together with water and salts, but not if a cut-off value of 2000 daltons
is used. We have observed that with these two membrane types, all other
conditions being the same, crystallization of a bacterial reaction centre with
PEG 4000 does not occur similarly (unpublished experiments).
Choosing between microdialysis and vapour diffusion is often a matter of
personal preference. Cost of microdialysis is higher when the detergent is
expensive, as detergent must be present in the microdialysis reservoir but may
be omitted from the vapour diffusion reservoir. One of the advantages of the
microdialysis method over vapour diffusion for screening experiments is the
possibility of changing individual constituents of the mixture; furthermore
the detergent concentration may be kept constant throughout the crystalliza-
tion, by putting it at the same concentration in the sample and in the reservoir.
Changing the dialysis reservoir is also very easy.
The main disadvantage of microdialysis is the rapid equilibration between
sample and reservoir (much faster than through vapour diffusion), which may
be troublesome if growth rate has to be slowed down. In that case, double
dialysis (see Chapter 5) is recommended.
Finally, when crystals are obtained, it important to realize that they are
extremely fragile and that their stabilization and manipulation are often diffi-
cult. For example, crystals of prostaglandin H synthase are stable in artificial
mother liquor only with the detergent at its CMC; this value is critical enough
that changes of the CMC due to the addition of salt or sucrose have to be
taken into account (6). Recently, it has been shown that the cryo-crystallo-
graphic techniques used for soluble protein (see Chapter 13) may successfully
be applied to membrane protein, provided that suitable stabilization conditions
are found (46, 49, 50, 76).

7. Conclusion
The structures of several membrane proteins have been solved during the past
few years, some of them to high resolution (51, 52). This has shown that the
methodology originally developed for porin and bacteriorhodopsin has a
more general validity. Furthermore, crystallization conditions worked out for
264
 9: Crystallization of membrane proteins
one protein have been successfully used with other proteins, opening the way
to the design of more systematic strategies. Various detergents are suitable.
Their role is important: their properties and their phase diagrams influence
the crystallization conditions; they are still associated with the protein in the
crystal lattice. The methodology requires (as for soluble proteins) a systematic
search over the different parameters, including the nature of the detergent.
This adds one more factor to this empirical analysis. However, good quality
crystals are still not easy to obtain; the difficulties encountered with bacterio-
rhodopsin provide the more vivid example. This has stimulated the search for
alternative approaches; one of them takes advantage of the bicontinuous
cubic phases of lipids where the lipid molecules are arranged in curved three-
dimensional bilayers. Such a phase could incorporate bacteriorhodopsin and
was used as a matrix for its crystallization (53). This allowed well-ordered
crystals to be grown with an improved quality as compared to those pre-
viously grown in detergent solutions (54). It is to be hoped that this method
could be of general use for other membrane proteins. On the other hand, the
problem of maintaining a pure and active protein in solution has been
recently tackled with the use of polymeric amphiphiles (55). Nevertheless,
finding proper expression system to overexpress these types of protein is still a
difficult task (56) that will have to be overcome before membrane protein
structures could flood the Protein Data Bank.

References
1. Michel, H. and Oesterhelt, D. (1980). Proc. Natl. Acad. Sci. USA, 77, 1283.
2. Garavito, R. M. and Rosenbusch, J. P. (1980). J. Cell Biol., 86, 327.
3. Deisenhofer, J., Epp, O., Miki, K., Huber, R., and Michel, H. (1985). Nature, 318,
618.
4. Kuhlbrandt, W. (1988). Q. Rev. Biophys., 21, 429.
5. Michel, H. (ed.) (1991). Crystallization of membrane proteins. CRC Press, Boca
Raton.
6. Garavito, R. M., Picot, D., and Loll, P. J. (1996). J. Bioeng. Biomemb., 28, 13.
7. Song, L. and Gouaux, J. E. (1997). In Methods in enzymology (ed. C. W. Carter
and R. M. Sweet), Academic Press, London. Vol. 276, p. 60.
8. Ostermeier, C. and Michel, H. (1997). Curr. Opin. Struct. Biol., 7, 697.
9. Kuhlbrandt, W. (1992). Q. Rev. Biophys., 25, 1.
10. Roth, M., Lewit-Bentley, A., Michel, H., Deisenhofer, J., Huber, R., and
Oesterhelt, D. (1989). Nature, 340, 659.
11. Roth, M., Arnoux, B., Ducruix, A., and Reiss-Husson, F. (1991). Biochemistry, 30,
9403.
12. Pebay-Peyroula, E., Garavito, R. M., Rosenbusch, J. P., Zulauf, M., and Timmins,
P. A. (1995). Structure, 3, 1051.
13. Bordier, C. (1981). J. Biol. Chem., 256, 1604.
14. Findlay, J. B. C. (1990). In Protein purification applications: a practical approach
(ed. E. L. V. Harris and S. Angal), pp. 59-82. IRL Press, Oxford.
15. Zulauf, M., Furstenberger, U., Grabo, M., Jaggi, P. M. R., and Rosenbusch, J. P.
265
 F. Reiss-Husson and D. Picot
(1989). In Methods in enzymology, (ed. S. Fleischer and B. Fleischer), Academic
Press, London. Vol. 172, p. 528.
16. Rosenthal, K. S. and Koussale, F. (1983). Anal. Chem., 55, 1115.
17. De Vendittis, E., Palumbo, G., Parlato, G., and Bocchini, V. (1981). Anal.
Biochem.,115, 278.
18. Van Aken, T., Foscall-Van Aken, S., Castleman, S., and Ferguson-Miller, J.
(1986). In Methods in enzymology, ibid ref. 15. Vol. 125, p. 27.
19. Herrmann, K. W. (1966). J. Coll. Int. Sci., 22, 352.
20. De Grip, W. J. and Bovee-Geurts, R. (1979). Chem. Phys. Lipids, 23, 321.
21. Lund, S., Orlowski, S., de Foresta, B., Champeil, P., le Maire, M., and Moller, J. V.
(1989).J. Biol. Chem., 264, 4907.
22. Lorber, B., Bishop, J. B., and DeLucas, L. J. (1990). Biochim. Biophys. Acta, 1023,
254.
23. Eisele, J.-L. and Rosenbusch, J. P. (1989). J. Mol. Biol., 206, 209.
24. Soulimane, T., Gohlke, U., Huber, H., and Buse, G. (1995). FEBS Lett., 368, 132.
25. Ashani, Y. and Catravas, G. N. (1980). Anal. Biochem., 109, 55.
26. Yu, C.-A., Xia, D., Deisenhofer, J., and Yu, L. (1994). J. Mol. Biol., 243, 802.
27. Garavito, R. M., Markovic-Housley, Z., and Jenkins, J. (1986). J. Cryst. Growth,
76, 701.
28. Kates, M. (1972). In Techniques in lipidology (ed. T. S. Work and E. Work), Vol.
3, Part II. North Holland, Amsterdam.
29. Radin, N. S. (1981). In Methods in enzymology, (ed. J. M. Lowenstein), Academic
Press, London. Vol. 72, p. 5.
30. Bartlett, G. R. (1959). J. Biol. Chem., 234, 466.
31. Garavito, R. M. and Picot, D. (1990). Methods: a companion to methods in
enzymology, 1, 57.
32. M011er, J., Le Maire, M., and Andersen, J. P. (1986). In Progress in protein lipid
interactions (ed. A. Watts and J. J. De Pont), Vol. 2, pp. 147-96. Elsevier,
Amsterdam.
33. Rivas, E., Pasdeloup, N., and Le Maire, M. (1982). Anal. Biochem., 123, 194.
34. Rao, P. and Pattabiraman, T. N. (1989). Anal. Biochem., 181, 18.
35. Kreusch, A., Weiss, M. S., Welte, W., Weckesser, J., and Schulz, G. E. (1991).
J. Mol. Biol., 217, 9.
36. Kleyman, G., Ostermeier, C., Ludwig, B., Skerra, A., and Michel, H. (1995).
Biotechnology, 13, 155.
37. Ostermeier, C., Iwata, S., Ludwig, B., and Michel, H. (1995). Nature Struct. Biol.,
2, 842.
38. Garavito, R. M. and Rosenbusch, J. P. (1986). In Methods in enzymology, ibid ref.
15. Vol. 125, p. 309.
39. Michel, H. (1983). Trends Biochem. Sci., 8, 56.
40. Timmins, P. A., Hauk, J., Wacker, T., and Welte, W. (1991). FEBS Lett., 280, 115.
41. Forst, D., Schulein, K., Wacker, T., Diederichs, K., Kreutz, W., Benz, R., et al.
(1993). J. Mol. Biol., 229, 258.
42. Krauss, N., Schubert, W.-D., Klukas, O., Fromme, P., Witt, H. T., and Saenger, W.
(1996). Nature Struct. Biol., 3, 965.
43. Blaauw, M., Dekker, N., Verheij, H. M., Kalk, K. H., and Dijkstra, B. W. (1995).
FEBS Lett., 373, 10.
44. Ducruix, A., Arnoux, B., and Reiss-Husson, F. (1988). In The photosynthetic
266
 9: Crystallization of membrane proteins
bacterial reaction center: structure and dynamics (ed. J. Breton and A. Vermeglio),
Vol. 149, pp. 21-5. Plenum, New York.
45. Ray, W. J. and Puvathingal, J. M. (1985). Anal. Biochem., 146, 307.
46. Kurumbail, R., Stevens, A. M., Gierse, J. K., McDonald, J. J., Stegeman, R. A.,
Pak, J. Y., et al. (1996). Nature, 384, 644.
47. Berry, E. A., Shulmeister, V. M., Huang, L.-S., and Kim, S.-H. (1995). Acta Cryst.,
D51, 235.
48. Weckstrom, K. and Zulauf, M. (1985). J. Chem. Soc. Faraday Trans. 1, 81,
2947.
49. Lee, J. W., Chan, M., Law, T. V., Kwon, H. J., and Jap, B. K. (1995). J. Mol. Biol.,
252, 15.
50. Yu, C.-A., Xia, J.-Z., Kachurin, A. M., Yu, L., Xia, D., Kim, H., et al. (1996).
Biochim. Biophys. Acta, 1275, 47.
51. Weiss, M. S. and Schulz, G. E. (1992). J. Mol. Biol., 221, 493.
52. Song, L., Hobaugh, M. R., Shustak, C., Cheley, S., Bayley, H., and Gouaux, J. E.
(1996). Science, 274, 1859.
53. Landau, E. M. and Rosenbusch, J. P. (1996). Proc. Natl. Acad. Sci. USA, 93,
14532.
54. Pebey-Peyroula, E., Rummel, G., Rosenbusch, J. P., and Landau, E. M. (1997).
Science, 277, 1676.
55. Tribet, C., Audebert, R., and Popot, J.-L. (1996). Proc. Natl. Acad. Sci.USA, 93,
15047.
56. Grisshammer, R. and Tate, C. G. (1995). Q. Rev. Biophys., 28, 315.
57. Frindi, M., Michels, B., and Zana, R. (1992). J. Phys. Chem., 96, 8137.
58. Degiorgio, V. (1985). In Physics of amphiphiles: micelles, vesicles and microemulsions
(ed. V. Degiorgio and M. Cord), pp. 303-35. North Holland, Amsterdam.
59. Michel, H. (1982). J. Mol. Biol., 158, 567.
60. Chang, C.-H., Schiffer, M., Tiede, D., Smith, U., and Norris, J. (1985). J. Mol.
Biol., 186, 201.
61. Allen, J. P. and Feher, G. (1984). Proc. Natl. Acad. Sci. USA, 81, 4795.
62. Yeates, T. O., Komiya, H., Rees, D. C., Allen, J. P., and Feher, G. (1987). Proc.
Natl. Acad. Sci. USA, 84, 6438.
63. Arnoux, B., Ducruix, A., Reiss-Husson, F., Lutz, M., Norris, J., Schiffer, M., et al.
(1989). FEBS Lett., 258, 47.
64. Allen, J. P. (1994). Proteins, 20, 283.
65. Buchanan, S. K., Fritzsch, G., Ermler, U., and Michel, H. (1993). J. Mol. Biol., 230,
1311.
66. Katayama, N., Kobayashi, M., Motojima, F., Inaka, K., Nozawa, T., and Miki, K.
(1994). FEBS Lett., 348, 158.
67. Feiek, R., Ertlmaier, A., and Ermler, U. (1996). FEBS Lett., 396, 161.
68. Papiz, M. Z., Hawthornthwaite, A. M., Cogdell, R. J., Woolley, K. J., Wightman,
P. A., Ferguson, L. A., et al. (1989). J. Mol. Biol., 209, 833.
69. Koepke, J., Hu, X., Muenke, C., Schulten, K., and Michel, H. (1996). Structure, 4,
581.
70. Pauptit, R. A., Zhang, H., Rummel, G., Schirmer, T., Jansonius, J. N., and
Rosenbusch, J. P. (1991). J. Mol. Biol., 218, 505.
71. Kreusch, A., Neubuser, A., Schiltz, E., Weckesser, J., and Schultz, G. E. (1994).
Protein Sci., 3, 58.
267
 F. Reiss-Husson and D. Picot
72. Tucker, A. D., Jackman, S., Parker, M. W., and Tsernoglou, D. (1991). J. Mol.
Biol., 222, 881.
73. Stauffer, K. A., Page, M. G. P., Hardmeyer, A., Keller, T. A., and Pauptit, R. A.
(1990). J. Mol. Biol., 211, 297.
74. Hirsch, A., Wacker, T., Weckesser, J., Diederichs, K., and Welte, W. (1995).
Proteins, 23, 282.
75. Tsukihara, T., Aoyama, H., Yamashita, E., Tomizaki, T., Yamaguchi, H.,
Shinzawa-Itoh, K., et al. (1995). Science, 269, 1069.
76. Luong, C., Miller, A., Barnett, J., Chow, J., Ramesha, C., and Browner, M. F.
(1996). Nature Struct. Biol., 3, 927.
77. Gouaux, J. E., Braha, O., Hobaugh, M. R., Song, L., Cheley, S., Shustak, C., et al.
(1994). Proc. Natl. Acad. Sci. USA, 91, 12828.

268
 10
From solution to crystals with a
physico-chemical aspect
M. RIES-KAUTT and A. DUCRUIX

1. Introduction
Biological macromolecules follow the same thermodynamic rules as inorganic
or organic small molecules concerning supersaturation, nucleation, and crystal
growth (1). Nevertheless macromolecules present particularities, because the
intramolecular interactions responsible of their tertiary structure, the inter-
molecular interactions involved in the crystal contacts, and the interactions
necessary to solubilize them in a solvent are similar. Therefore these different
interactions may become competitive with each other. In addition, the bio-
logical properties of biological macromolecules may be conserved although
the physico-chemical properties, such as the net charge, may change depend-
ing on the crystallization conditions (pH, ionic strength, etc.). A charged
biological macromolecule requires counterions to maintain the electro-
neutrality of the solution; therefore it should be considered as a protein (or
nucleic acid) salt with its own physico-chemical properties, depending on the
nature of the counterions.
To crystallize a biological macromolecule, its solution must have reached
supersaturation which is the driving force for crystal growth. The understand-
ing of the influence of the crystallization parameters on protein solubility of
model proteins is necessary to guide the preparation of crystals of new
proteins and their manipulation. Only the practical issues are developed in
this chapter, and the reader should refer to recent reviews (2-4) for a de-
scription of the fundamental physical chemistry underlying crystallogenesis.

2. The concept of solubility and methods for solubility

diagram determination
The solubilization of a solute (e.g. a biological macromolecule) in an efficient
solvent requires solvent-solute interactions, which must be similar to the
solvent-solvent interactions and to the solute-solute interactions of the
compound to be dissolved. All of the compounds of a protein solution
 M. Ries-Kautt and A. Ducruix
(protein, water, buffer, crystallizing agents, and others) interact with each
other via various, often weak, types of interactions: monopole-monopole,
monopole-dipole, dipole-dipole, Van der Waals hydrophobic interactions,
and hydrogen bonds.

2.1 Solubility
Solubility is defined as the amount of solute dissolved in a solution in equilib-
rium with its crystal form at a given temperature. For example, crystalline
ammonium sulfate dissolves at 25°C until its concentration reaches 4.1 moles
per litre of water, the excess remaining non-dissolved. More salt can be
dissolved when raising the temperature, but if the temperature is brought
back to 25°C, the solution becomes supersaturated, and the excess of salt
crystallizes until its concentration reaches again its solubility value at 25°C
(4.1 moles per litre of water).
In the case of biological macromolecules, the solubility is additionally
defined by the characteristics of the solvent. Proteins are mostly solubilized in
water which acts through hydrogen bonds. In some cases another protic
solvent (an alcohol) or an aprotic solvent (e.g. acetone, DMSO, dioxane,...)
is added at low concentration. In addition, the solvent solutions contain at
least the ubiquitous buffer used to fix the pH of the solution and therefore the
net charge of the protein. Salts are added not only to ensure an ionic strength
but most often to reach supersaturation.
Throughout this chapter, protein solubility is defined as the concentration
of soluble protein in equilibrium with the crystalline form at given tempera-
ture and pH values, and in the presence of a given concentration of solvent
compounds others than the protein (i.e. water, buffer, crystallizing agents,
stabilizers, additives). The solubility values depend on the physico-chemical
characteristics of the protein itself (hydrophilicity, net charge, type of solvent
exposed residues) and of the solvent (pH, dielectric constant, ionic strength,
concentration, and nature of the additives).
Figure 1 illustrates the variability of protein solubilities, depending on the
protein itself or on the protein salt (e.g. different lysozyme salts). The
solubility of the three proteins: bovine pancreatic trypsin inhibitor (BPTI) (5)
in ammonium sulfate, collagenase from Hypoderma lineatum (Hl) in
ammonium sulfate (6), and hen egg white (HEW) lysozyme in NaCl (7), cover
a very large range of both ionic strength and solubility values, in their
standard crystallization conditions. Furthermore the solubility of a same
protein, HEW lysozyme, can be changed drastically when changing the nature
of the crystallizing salt, as shown by the solubility curves of lysozyme/KSCN,
lysozyme/NaCl, and lysozyme/NH4OAc.
In the literature other conventions of defining solubility are encountered; it
may be the protein concentration measured before the actual equilibrium is
reached, or it can be evaluated in the presence of precipitate instead of crystals
(8). Their applications are discussed at the end of this section.
270
 10: From solution to crystals with a physico-chemical aspect

Figure 1. Solubility curves at 18°C of BPTI (5), collagenase from Hypoderma lineatum (6),
and HEW lysozyme (7, 10). The crystallizing agents are AS (ammonium sulfate), NH4OAc
(ammonium acetate), NaCI (sodium chloride), and KSCN (potassium thiocyanate).

The zone of the solubility diagram where crystals appear (nucleation zone)
depends on the supersaturation, which is the ratio, Cp/Cs, of the protein
concentration over the solubility value, but also on the kinetics to reach these
conditions.
The protein purity (see Chapter 2) must be checked before doing any
screening experiments. In a mixture of proteins, the first crystals contain the
most supersaturated protein and this may not be the most concentrated one.
Furthermore the solubility of the major protein may be affected differently
from the contaminant when changing a parameter.

2.2 Measurements of the solubility

Solubility measurements are necessary to understand the effect of crystal-
lization parameters on the solubility of model proteins, which can then be
transposed for the crystallization of a new protein.
2.2.1 Conventional methods
The solubility can be determined either by crystallization of a supersaturated
solution (see Protocol 7) or by dissolution of crystals in an undersaturated
solution. In both cases the protein concentration in the supernatant converges
271
 M. Ries-Kautt and A. Ducruix
toward the same asymptotic value at equilibrium. For double checking of
solubility values, both crystallization and dissolution methods can be run
in parallel. For both crystallization and crystal dissolution, crystallization
conditions must be previously defined.
During the course of the experiments, all parameters, except the one under
investigation, (e.g. pH, temperature, salt and buffer concentrations, nature of
salt and buffer) must be carefully kept constant and the stability of the
biological macromolecule versus time and proteases must be checked.

Protocol 1. Solubility measurements by crystallization

Equipment and reagents

• ACA boxes • Incubator
* 10 ml buffer • 10 ml salt stock solution

Method
1. Define one parameter to vary (e.g. salt concentration), keeping all
others strictly constant (e.g. pH, temperature, nature of salt and
buffer).
2. Choose at least four values of the variable (different salt concentra-
tions over a large range), because solubility curves usually do not fit
with linear curves.
3. Set up the batch experiments (> 10 ul) in duplicate at two or three
different initial protein concentrations for a given parameter value.
4. Follow the decrease of the protein concentration of the supernatant,
by withdrawing a crystal-free aliquot of the duplicate set-up, each
week for optical density (OD) measurements. If microcrystals are
present, filter or centrifuge the aliquot before the dilution for the OD
measurement.
Once crystallization has started, the protein concentration in the
supernatant will converge to a constant value, solubility, with time.
This value is identical for the different initial protein concentration at a
same ionic strength.
5. When the protein concentrations remain constant for at least two
weeks and are identical for the different initial protein concentration at
the same ionic strength, confirm the measurement by testing the
original undisturbed set of experiments.

Solubility measurements are often performed by batch methods (9, 10).

Hanging, sitting, or sandwich drops systems can be used as long as the salt
272
 10: From solution to crystals with a physico-chemical aspect
concentration is identical in both the drop and the reservoir all over the
process. In this case the ratio of the salt concentration initially in the drop and
in the reservoir is 1:1, whereas in a classical vapour diffusion experiment it is
1:2. The role of the reservoir here is only to keep vapour pressure constant
during the experiment. The vapour diffusion technique is not less suited,
because:
(a) Only the initial conditions at the beginning of the experiment are well
known. The accurate ionic strength, once the drop/reservoir equilibration
is achieved, is difficult to verify when working with small drop volumes.
(b) All components in the drop will concentrate: the crystallizing agent and
the protein as expected, but also the buffer and additives (and impuri-
ties!). As a consequence more than one parameter may change during the
experiment.

2.2.2 Alternative methods

More sophisticated methods for the solubility dependence with temperature
have been described in the literature. They yield not only solubility values
versus temperature, but also the crystallization enthalpies and sometimes the
crystallization induction times.

i. Column method
A solution of either supersaturated or undersaturated protein solution is
poured in a microcolumn filled with crystals. An aliquot of the solution is
periodically withdrawn from the bottom of the column to follow the change of
the protein concentration by optical density measurements. This method (11)
is based on the maximization of the exchange between the available crystal-
line surface area and minimal free solution volume to reach equilibrium. It
overcomes the problem of prolonged equilibration time, as equilibrium, i.e.
the solubility value, is reached within one to five days.
Two microcolumns are run in parallel: one for crystallization (super-
saturated state), one for dissolution (undersaturated state). This method is
mostly appropriate for the study of the influence of temperature. To deter-
mine protein solubility at different concentrations of a given salt, crystals can
be prepared from a same batch, but must then be equilibrated carefully at
respective salt concentrations. When solubility is checked in different salts, a
batch of crystals is prepared in each appropriate salt.

ii. Microscopic observation and OD measurements

About 0.5 ml crystallization solution containing crystals are placed in a glass
vessel inserted in a thermoregulated cell. The temperature is monitored and
controlled by a Peltier element (± 0.1 °C). The whole set-up is placed under a
microscope. Dissolution and growth are followed by microscopic observation
in parallel with OD measurements of the crystallization solution (5, 12).
273
 M. Ries-Kautt and A. Ducruix
Equilibration is achieved within about 50 days but several cells can be run in
parallel in the Peltier element.

iii. Scintillation
A thermoregulated cell (50-100 ul) is filled with the solution to crystallize.
The bath temperature is changed until crystallites occur inducing scintillation
which is detected by a photodiode signal. The temperature is changed back-
wards and forwards to define the solubility limit, defined by the appearance
and disappearance of the crystallite detected by the scintillation signal. One
solubility value is obtained within approximately 12-24 hours. In addition this
technique (13) allows crystallization induction times to be measured which
were shown to follow supersaturation.

iv. Temperature controlled static light scattering method

This method (14) is similar to the scintillation method described above, but
using light scattering to follow the occurrence or dissolution of crystallites
when changing the temperature. The crystallization solution (about 1 ml) is
stirred by a Teflon coated magnet to maintain the particles in suspension. The
solution is illuminated by a 1 mm2 cross-section laser beam. The light
scattered, normal to the incident beam, is focused on a photodiode whose
signal is amplified and analysed by a phase-sensitive detector. The tempera-
ture is changed until faceted crystals nucleate and the scattered intensity
reaches a plateau. Then the temperature is changed backwards until the
crystals dissolve.

v. Michelson interferometry
A Michelson interferometer is used for the observation of concentration
gradients around a crystal to determine whether the crystal is growing or
dissolving when changing the temperature (15). The volume of the cell is
about 70 ul. The equilibrium temperature is obtained within two hours.

vi. Calorimetry and OD measurements

The heat signal from a 1 ml crystallizing solution is recorded every two
minutes over a period of two to three days, using a differential scanning
calorimeter to follow the heat of crystallization. The final protein concentra-
tion, i.e. solubility, is obtained by removing an aliquot of crystal-free solution
from the cell and measuring the absorbance at 280 nm (16).

2.2.3 Estimating the residual protein concentration

It is obvious that accurate solubility measurements are necessary in funda-
mental research to understand protein crystal growth. However when dealing
with the crystallization of a new protein, one is probably not willing to invest
much protein and time to define the solubility values in various conditions.
Nevertheless it is very helpful to have at least an order of magnitude of the
274
 10: From solution to crystals with a physico-chemical aspect
residual protein concentration, in contact with the crystals, to guide further
experiments and handle the crystals (Protocol 2). The estimation of the
residual protein concentration is easy to perform even with a small drop size.

Protocol 2. Measurement of the residual protein concentration in

hanging dropsa

Equipment and reagents

• Buffer solution • U.V. spectrophotometer

Method
1. Open the coverslip with the drop in which crystals have grown for at
least two weeks.
2. Withdraw 1 (preferably 2) ul of clear supernatant under the binocular.
If too many microcrystals are present, centrifuge the drop and take the
aliquot from the clear supernatant to avoid diluting crystals.
3. Dilute to the minimal volume required for an OD measurement at
280 nm. Eventually measure the protein concentration by the Bradford
method using a coloured dye to increase the absorbance of the
protein-dye complex.

aWhen using the dialysis technique, an aliquot can be withdrawn from the dialysis cell with a
Hamilton syringe. The dialysis cell can of course no longer be immersed in the reservoir, but it can
be rescued in a vessel with some reservoir solution around in order to avoid the solution drying.

Measuring the residual protein concentration is helpful for:

(a) Having an estimate of the solubility. Even though the drop has not
reached equilibrium, this measurement gives an order of magnitude
whether the solubility is low, medium, or high. It will be explained later in
this chapter that low solubility conditions (< 1 mg/ml) are difficult for the
optimization of growing few and large crystals.
(b) Knowing the starting conditions, the ratio of initial over final protein
concentrations tells an order of magnitude of the supersaturation where
these crystals were obtained. This helps to choose the range of super-
saturation, and therefore initial protein concentration, for further
experiments.
(c) Defining the amount of protein which is available to grow crystals. For
HEW lysozyme, 1, 8.5, 29, 68 ug of protein are necessary to grow
respectively crystal of 0.1, 0.2, 0.3, 0.4 mm3.
In Figure 2 are illustrated two quite different conditions, A and B, for
275
 M. Ries-Kautt and A. Ducruix

Figure 2. Schematic phase diagram showing the crystallization conditions at a

supersaturation, B, of 10, and the solubility curve. Conditions A and B illustrate high and
low solubility conditions, respectively.

which respectively 55 or 0.9 ug/ul of protein are in supersaturation, when

starting from a tenfold supersaturated protein solution. A 0.3 mm3 crystal
can be grown from a drop of = 0.5 ul in A but from 32 ul in B.
(d) Estimating the slope of the solubility curve, by measuring the residual
protein concentration at three different values of the variable which can
also be ionic strength, temperature, pH, etc. This helps for extrapolating
the nucleation zone to lower or higher values of the variable depending
on whether the solubility variation is steep or smooth.
(e) Guiding the preparation of seeding experiments. A classical vapour
diffusion experiment, with a 1:2 salt concentration ratio between drop
and reservoir, may evolve in two ways as shown on Figure 3.
(i) The concentrations in the drop equilibrate from I (initial) to F (final),
then nucleation occurs and the protein concentration drops from F to
P2, the residual protein concentration, while the salt concentration
remains constant in the drop and equals the one of the reservoir.
(ii) The drop concentrations increase from I to an intermediate value B,
at which nucleation starts already without ever reaching F. Then both
protein and salt concentrations continue to change to reach P2.
276
 10: From solution to crystals with a physico-chemical aspect

Figure 3. Residual protein concentration in a drop after a vapour diffusion process, and
selection of the seeding conditions.

Once the residual protein concentration is measured, news drops are pre-
pared at protein concentrations ranging from P1 to P5, as indicated on
Figure 3, directly at the salt concentration of the reservoir (batch). The
concentrations P3 to P5 aim at covering the searched metastable zone.
The protein concentration P1 (< solubility) is added to have a better
estimate of the lower limit of the metastable zone. Crystals dissolve in
undersaturated drops (< P1) and remain unchanged in saturated drops
(between P1 and P2). They grow in slightly supersaturated drops (P2 to
P4, where you should seed) whereas new nucleation occurs in more
supersaturated conditions (> P4-P5).
(f) Mounting the crystals. When a crystal is recovered from a drop, it is in a
solution of a given protein and crystallizing agent concentrations as
shown in Figure 2. Very often reservoir solution is used to transfer the
crystal. In fact, this can be done safely only if the remaining protein
concentration in the drop is lower than = 0.5 mg/ml (e.g. B in Figure 2). If
the solubility value is higher (e.g. A in Figure 2), then the crystal would
start to dissolve as the reservoir contains no protein. Knowing the
residual protein concentration in the mother liquor gives the amount of
protein to introduce in additional mounting solutions.
Similarly protein should be added when soaking crystals in cryo-
protectant solutions for cryo-crystallography when crystals dissolve. Cryo-
277
 M. Ries-Kautt and A. Ducruix
protectants often change the protein solubility. Although the knowledge of
the residual protein concentration of the crystallization drop is not repre-
sentative of the solubility of the protein in the cryo-protectant solution,
crystals should be soaked in a series of cryo-protectant solutions containing
a protein concentration higher than the residual protein concentration.

2.3 Phase diagram

As the solubility of a biological macromolecule depends on various para-
meters (see Chapter 1, Table 1), a phase diagram is a useful representation of
its solubility (mg/ml or mM biological macromolecule in solution) as a
function of one parameter, all other parameters being kept constant.
The diagram, represented in Figure 4, comprises the following zones:
(a) The solubility curve delimits the under- and supersaturated zones. In an
experiment where crystallizing agent and biological macromolecule con-
centrations correspond to solubility conditions, the saturated macro-
molecule solution is in equilibrium with the crystallized macromolecule.
This corresponds to the situation at the end of the process of crystal
growth: additional crystalline macromolecule does not dissolve, but adding

Figure 4. Schematic description of a two-dimensional solubility diagram showing the

different zones of the supersaturation domain. Note that the metastable zone covers a
larger range of supersaturation, when solubility is low (i.e. at high salt concentration)
than when it is high (i.e. at low salt concentration). Conversely, the nucleation zone is
larger for high solubility than for a low one.

278
 10: From solution to crystals with a physico-chemical aspect
reservoir solution without the macromolecule leads to the dissolution of
the macromolecule crystals.
(b) Below the solubility curve the solution is undersaturated, the system is
thermodynamically stable, and the biological macromolecule will never
crystallize.
(c) Above the solubility curve, the concentration of the biological macro-
molecule is higher than the concentration at equilibrium. This corresponds
to the supersaturation zone. A supersaturated macromolecule solution
contains an excess of macromolecule which will appear as a solid phase
until the macromolecule concentration reaches the solubility value in the
solution (supernatant). In some cases the excess of macromolecule may
concentrate in oily drops in a liquid-liquid separation. The rate of
supersaturation is defined as the ratio of the biological macromolecule
concentration over the solubility value. The higher the supersaturation
rate, the faster this solid phase appears.
It is often difficult to understand how supersaturated macromolecule
solutions are achievable. In terms of molarity, it must be remembered that
macromolecule solubilities are very low (uM to mM) compared to small
molecules or inorganic molecules (mM). This also corresponds to very low
volume fractions of solute which allow macromolecule solutions to be
prepared at supersaturations as high as 10 to 20 times the solubility. For small
molecules supersaturation of only 1.1 to about 1.5 are achievable. Recently
we observed crystallization of HEW lysozyme at supersaturations around 1.5
when working with 400 mg/ml (28 mM) protein which correspond to 30%
volume fraction (17).
However the higher the supersaturation, the faster the solid phase appears
in the solution, as described below. Contrary to macromolecule purification
which implies precipitation, crystallization requires an accurate control of the
level of supersaturation. This allows nucleation of crystals, i.e. a solid phase
with a three-dimensional periodicity, by controlling the nucleation rate to
yield few single crystals.
2.3.1 Precipitation zone
Precipitation occurs at very high supersaturation (= 30 to 100 times the
solubility value for HEW lysozyme). Insoluble macromolecules rapidly sep-
arate from the solution in an amorphous state. If the solution is centrifuged,
the supernatant is in fact still supersaturated and crystallization may occur. To
differentiate amorphous precipitate from microcrystals, fresh drops can be
seeded (see Chapter 7) with this material; amorphous precipitate dissolves
whereas microcrystals grow.
2.3.2 Nucleation zone
At a sufficient supersaturation, nucleation spontaneously occurs, once critical
activation free energy is overcome. This is called homogeneous nucleation.
279
 M. Ries-Kautt and A. Ducruix
Crystallization often occurs at lower saturation when it is induced by vibra-
tions or the presence of particles (dust, precipitate, irregularities of crystal-
lization cell); it is then called heterogeneous nucleation. The latter is usually
characterized by non-reproducibility, therefore it is recommended to filter all
solutions and blow the coverslips with an air stream before setting up the
hanging drops.
Nucleation requires a lower supersaturation than precipitation. To give an
order of magnitude, the nucleation range for HEW lysozyme is = 5 times the
solubility for dialysis and batch crystallizations, and about 10 times the
solubility for vapour diffusion. Crystals appear faster and in larger numbers
with increasing supersaturation. High supersaturation may be useful to find
the nucleation zone, but growing crystals for X-ray diffraction may benefit
from a search of the optimal supersaturation where few but large crystals are
grown.
The nucleation rate, defined as the number of nuclei formed per unit
volume and unit time, is linked (1, 18) to:
(a) Supersaturation, as illustrated by the curve A of Figure 5. For super-
saturations higher than B*, the critical supersaturation, nucleation occurs.
When increasing the supersaturation, the number of crystals increases.

Figure 5. Nucleation rate versus supersaturation for A, high solubility, and B, low
solubility. The curves A and B delimit the metastable zone from the nucleation one.

280
 10: From solution to crystals with a physico-chemical aspect
(b) The number of molecules per unit volume. When the solute is sparingly
soluble, the solution remains in a metastable state over long periods.
Nucleation requires much higher supersaturation to occur (Figure 4).
Once B* is reached, the nucleation rates becomes drastic, as illustrated by
the curve B in Figure 5. The different curves A and B of Figure 5
correspond respectively to the situations A and B of Figure 2.
Crystallization conditions for which the solubility is very low should be
avoided. The precipitation curve is then very close to the solubility curve, the
domain of crystallization becomes very narrow, which brings difficulties in
defining the right conditions for growing large crystals. This is the case for the
crystallization of HEW lysozyme/KSCN, where the crystallization zone is
limited to a range of 100 mM KSCN. On the counterpart, HEW lysozyme/
Nad crystallizes over a broad range of 1400 mM. To enlarge the nucleation
zone, solubility must be increased. This can be done by:
• decreasing the ionic strength while increasing the macromolecule concen-
tration, if the slope of the solubility curve is smooth enough
• using another salt in which the solubility is higher (e.g. from HEW
lysozyme/KSCN to HEW lysozyme/NaCl shown in Figure 1)
• changing the pH to increase the protein net charge
• changing the temperature.

2.3.3 Metastable zone

In the metastable zone, the critical supersaturation is not yet reached.
Spontaneous nucleation does not occur, unless it is induced by vibrations or
introduction of a particle which will promote heterogeneous nucleation.
As shown in Figure 5, the metastable zone is much larger when the solubil-
ity is very low. It becomes then extremely difficult to reduce the nucleation
rate. One can use the metastable zone to seed crystals which will grow, fed by
the amount of protein in supersaturation. When a low solubility system
cannot be brought to higher solubility for technical reasons, seeding remains
nearly the only way to grow large single crystals (Chapter 7).

2.4 Kinetic aspects

Dealing with a supersaturated protein solution implies a system which is
thermodynamically out of equilibrium. Therefore nucleation and growth
depend on various kinetics. In other words, the solubility is unchanged as long
as the pH, the temperature, the ionic strength, and the nature of solvent
constituents are constant, but the nucleation zone, as well as the precipitation
zone, may be shifted depending on the crystallization technique, the geometry
of the device, etc.
281
 M. Ries-Kautt and A. Ducruix

2.4.1 Time lag for nucleation

When the protein solution is directly prepared at a given supersaturation
(using the batch method) or if the protein solution has completely equilib-
rated with its reservoir before crystals have nucleated, the crystals do not
appear immediately, but after a time lag. This time lag is related with the
supersaturation as illustrated for HEW lysozyme (18); the higher the super-
saturation the faster crystals appear. It must be remembered that the number
of crystals also increases with supersaturation.
2.4.2 Protein solution/reservoir equilibration
When using vapour diffusion or dialysis (see Chapter 5), the first kinetics to
control are the equilibration of the protein solution with the reservoir.
These kinetics have been shown (19, and refs therein) to increase with the
drop/reservoir distance, with the initial gradient of crystallizing agent between
drop and reservoir, and when using PEGs instead of salts. These kinetics are
obviously not relevant for the batch method because the protein solution is
directly prepared at the protein and crystallizing agent concentrations to be
tested.
Figure 6 shows the equilibration of a set of vapour diffusion experiments

Figure 6. Variations of the protein and salt concentration in four drops A to D during
equilibration with their reservoirs for a classical 1:2 ratio. The bold arrows start at the
initial conditions in the drops, and end at the expected final conditions, if no
crystallization occurs during the drop/reservoir equilibration.

282
 10: From solution to crystals with a physico-chemical aspect
with a typical 1:2 ratio for the salt concentration initially in the drop and in the
reservoir. In condition A the salt gradient between the initial drop conditions
and the reservoir is 0.5 M whereas it is 0.8 M in D. This implies a faster
equilibration for A than for D. As a consequence the drop/reservoir equilib-
ration may become much faster than the time lag for nucleation. Therefore
nucleation occurs at a higher supersaturation, implying a higher nucleation
rate. To help the system to crystallize before the drop/reservoir equilibration
is achieved, the experiment may be run in two steps of lower gradient, or with
a larger drop/reservoir distance, or in ACA instead of Linbro plates. Similarly
for dialysis, capillaries may be preferred to Cambridge buttons, because the
protein solution equilibrates more slowly with the reservoir solution, allowing
nucleation to occur before equilibration is achieved.
2.4.3 Equilibration to reach the solubility value
Once crystallization has started, the protein concentration in the solution
decreases until it reaches the solubility value. These kinetics depend on the
growth kinetics of a given crystal form of a given biological macromolecule,
the number of crystals growing, the amount of protein to crystallize, and
stirring (or not) the solution.
For the photochemical reaction centre of Rhodobacter sphaeroides (20), the
solubility value was reached within 12 days even though the unstirred batch
method was used. In this case the crystals grew very quickly.
For tetragonal HEW lysozyme crystals, equilibration requires up to nine
months if using unstirred batch methods. The delay can be reduced to two
months by stirring (17) the crystallization vials. However this was not
successful for very high protein concentrations which are very viscous. These
experiments were kept for one month at low temperature to accelerate the
crystallization before letting them equilibrate at 18°C.

3. Proteins as polyions
Throughout the process of crystallization and of structure determination, a
protein must be considered as a polyion of a given net charge, surrounded by
counterions. The number of counterions is at least equal to the net charge to
ensure the electrostatic compensation. Even though the biochemical activity
may not be altered, the physico-chemical properties of a protein, and hence its
behaviour in crystallization, may be changed significantly by variations of its
net charge or by adsorption of small molecules or ions onto its surface. It is
useful to begin a new crystallization project by first calculating the probable
net charge of the protein as a function of pH, although it is an approximation.
Additives are well defined solvent constituents (chemical nature and con-
centration); otherwise they are impurities. To improve the reproducibility of
crystallization experiments and the reliability during a structural investigation
involving different protein batches, impurities should be eliminated whenever
283
 M. Ries-Kautt and A. Ducruix
possible (see Chapter 2). This is usually done for biological contaminants, but
seldom for small molecules or salts. After a routine purification step, aqueous
protein solutions contain various additives, including at least a buffer, various
salts from elution gradient, and additives to prevent oxidation of free SH
groups, EDTA, NaN3, stabilizers, etc. Some compounds or ions may be bound
by the protein. Proteins from commercial sources often contain organic or in-
organic compounds whose nature depends on the source, and the amount on
the batch. Up to 14% (w/w) salt were observed for commercial HEW lysozyme
(21), although the purity was otherwise excellent in terms of biological activity.
All the uncontrolled organic and inorganic solutes should be eliminated or
replaced by co-ions and counterions of known concentration and nature.

3.1 Estimation of the net charge

The net charge of a protein, Zp, is the difference between the number of
positive and negative charges arising from the deprotonation of acidic groups
and the protonation of basic groups. On average, the pKa values of solvent-
accessible charged residues are respectively: 3.5 (COOH-terminus), 4.5 (Asp,
Glu), 6.2 (His), 7.6 (NH2-terminus), 9.5 (Tyr), 9.3 (Cys), 10.4 (Lys), and 12.0
(Arg).
The estimated protein net charge, Zp, for different pH values can be
calculated with programs such as Excell, Kaleidagraph, according to:

where [H+] = l0 pH ; Ka = 10-pKa for the individual charged groups; ni the

number of each type of charged residue, and assuming that:
(a) Identical groups have the same pKa value. Of course actual local pH
values may be different depending on their environment, but this is rarely
known as long as the protein structure is not determined.
(b) All potentially charged groups are accessible, i.e. neither buried inside the
protein, nor involved either in a salt bridge with another charged residue,
or in ion complexation. However some ionizable side chains buried inside
the protein may exist in their uncharged forms. Similarly, charged groups
may complex ions (e.g. Zn, Fe, Mg, Ca, etc.). The contribution of such
groups can be adjusted, a priori if they are known or, a posteriori, when
comparing the estimated isoelectric points with the experimental ones or
with titration data.
(c) The protein is monomeric. This requirement is important unless it is
known which charged residues are accessible or buried in the interface of
the di- or multimeric macromolecule.
284
Table 1. Calculation of protein charges versus pHa

a-COOH Asp + Glu Tyr Cys-SH a-NH2 His Lys Arg

pK = 3.5 4.5 9.5 9.3 7.6 6.2 10.4 12
n n
ni na-COOH = n
Asp + Glu = Tyr = n
Cys-SH = n
a-NH2 = nHis = n
Lys = Arg =
pH z zx n z zx n Z ZX n z zx n z zx n z zx n z zx n z zx n Zp
2.00 -0.03 -0.003 -0.00 -0.00 + 1.00 + 1.00 + 1.00 + 1.00
2.50 -0.09 -0.01 -0.00 -0.00 + 1.00 + 1.00 + 1.00 + 1.00
3.00 -0.24 -0.03 -0.00 -0.00 + 1.00 + 1.00 + 1.00 + 1.00
3.50 -0.50 -0.09 -0.00 -0.00 + 1.00 + 1.00 + 1.00 + 1.00
4.00 -0.76 -0.24 -0.00 -0.00 + 1.00 + 1.00 + 1.00 + 1.00
4.50 -0.91 -0.50 -0.00 -0.00 + 1.00 +0.99 + 1.00 + 1.00
5.00 -0.97 -0.76 -0.00 -0.00 + 1.00 +0.97 + 1.00 + 1.00
5.50 -0.99 -0.91 -0.00 -0.00 +0.99 +0.91 + 1.00 + 1.00
6.00 -1.00 -0.97 -0.00 -0.00 +0.97 +0.76 + 1.00 + 1.00
6.50 -1.00 -0.99 -0.00 -0.00 +0.91 +0.50 + 1.00 + 1.00
7.00 -1.00 -1.00 -0.00 -0.00 +0.76 +0.24 + 1.00 + 1.00
7.50 -1.00 -1.00 -0.01 -0.02 +0.50 +0.09 + 1.00 + 1.00
8.00 -1.00 -1.00 -0.03 -0.05 +0.24 + 0.03 + 1.00 + 1.00
8.50 -1.00 -1.00 -0.09 -0.14 +0.09 +0.01 +0.99 + 1.00
9.00 -1.00 -1.00 -0.24 -0.33 +0.03 +0.00 +0.97 + 1.00
9.50 -1.00 -1.00 -0.50 -0.61 + 0.01 +0.00 +0.91 + 1.00
10.0 -1.00 -1.00 -0.76 -0.83 +0.00 +0.00 +0.76 +0.99
10.5 -1.00 -1.00 -0.91 -0.94 +0.00 +0.00 +0.50 +0.97
11.0 -1.00 -1.00 -0.97 -0.98 +0.00 +0.00 +0.24 +0.91
11.5 -1.00 -1.00 -0.99 -0.99 +0.00 +0.00 +0.09 +0.76
12.0 -1.00 -1.00 -1.00 -1.00 +0.00 +0.00 +0.03 +0.50
12.5 -1.00 -1.00 -1.00 -1.00 +0.00 +0.00 +0.01 +0.24
13.0 -1.00 -1.00 -1.00 -1.00 +0.00 +0.00 +0.00 +0.09

a The charge contribution z of the charged residue is given in each column for the indicated pH value. For a given protein. note: ni, the number of the amino acid
in each corresponding column. Multiply zi by ni of each column. Sum all charges for a given pH (row) to obtain the net charge at this pH.
 M. Ries-Kautt and A. Ducruix
In Table 1 the contribution zi of a given type of charged residue at a given
pH value appears in each column. For a given protein, each column has to be
multiplied by n, the number of a given type of amino acid. Summing all
charges for a given pH (row) gives the net charge at this pH.
The estimation of a protein net charge is also accessible on the web at:
• http://www-biol.univ-mrs.fr/d_abim/compo-p.html
• http://www.expasy.ch/sprot/protparam.html
• http://www.infobiogen.fr/service/deambulum
The calculated pI (pH for a net charge of zero) should be supplemented by
the electrophoretic measurement of the experimental pI. If a difference is
observed between the estimated pI and the experimental one, it means that
one of the conditions detailed for the estimation is not met or that some
additives of the experimental conditions interact with the protein.
Water soluble proteins can be classified broadly, according to their pI
(experimental or estimated from the content of charged residues), as:
(a) Acidic proteins, having a higher content of Asp and Glu, than His, Lys,
and Arg. Their pI is lower than 6. This arbitrary value is linked to the pKa
value of the basic group His.
(b) Basic proteins, with a higher content of His, Lys, and Arg, than Asp and
Glu. Their pI is above 7.5-8.
(c) Neutral proteins containing roughly equal numbers of acidic and basic
residues, and therefore presenting a pI near the neutrality.
In the pH range between pH 6-8, acidic proteins bear a negative net charge,
basic proteins a positive one, and neutral proteins a net charge of about zero.
This classification does not include membrane proteins, which naturally
occur in hydrophobic, lipid environments and which are known to be poorly
soluble in water (see Chapter 9). Their solubilization requires detergents, and
the physical chemistry of protein-detergent systems is very different from the
discussion of this chapter. Nevertheless, the preceding discussion does apply
to their water soluble surfaces, and should therefore also be considered in
those studies.

3.2 Desalting of proteins

3.2.1 Dialysis against water
Dialysis removes most of the solvent compounds, except those tightly bound
by the protein and the counterions necessary for electrostatic compensation
(21). For example HEW lysozyme bears a net charge of about 10 at pH 5,
thus the dialysed solution contains at least ten counterions leading to an
anion concentration of 35 mM for a 50 mg/ml (3.5 mM) HEW lysozyme
solution.
286
 10: From solution to crystals with a physico-chemical aspect

3.2.2 Mixed-bed resins

A more efficient procedure consists in passing the protein solution through
strong cation and anion exchange resins in H+ and OH~ form, where all
cations and anions of the protein solutions are exchanged for H+ and OH~,
except Li and F (4). The resins used for this purpose have a pore size of about
103 Da preventing the protein molecules from adsorbing in the exchange sites.
The eluted protein solutions contain only the protein, water, H+, and OH~,
and are isoionic by definition. Isoionic protein is perhaps the simplest possible
system, free of the modulating effects of ligands and bound ions. So far we
have tested the desalting procedure with six proteins having either an acidic
or a basic pI, among which are HEW lysozyme, BPTI, and Hl collagenase.
Two other proteins, which are very hydrophobic and show tendency to
aggregate, were irreversibly adsorbed on the resins. Thus, it is recommended
to first use a small amount in order to verify its stability in isoionic conditions.
When using resins with dyes to indicate the saturation of the resins, we have
observed by NMR the contamination of the eluted proteins with this dye.
Desalting can be performed in:
(a) One step by passing a previously dialysed protein solution through a
mixed-bed resin. This is quicker than the two-step procedure, but the
resins cannot be regenerated as they are mixed. It is recommended for
small amounts of proteins, thus small quantities of resins, or when work-
ing with different proteins, as contamination is limited when changing the
resins for each type of protein.
(b) Two steps by exchanging first the co-ions then the counterions (Protocol
3). Thus a solution of a protein presenting a basic pI is successively passed
through a cation exchange resin and then a anion exchange resin, and in
reverse sequence for proteins having an acidic pI. The advantage of a
two-step desalting is the possibility to regenerate separately the cation
exchange resins with HC1 (1 M) and the anion exchange resins with
NaOH (1 M).

Protocol 3. Preparation of up to 100 mg of isoionic basic protein

Equipment and reagents

• Two 5 ml syringes • Bio-Rad AG 1-X8 20-50 mesh, OH- form
. Bio-Rad AG 50W-X8 20-50 mesh, H+ form (No. 140-1422)
(No. 142-1421)

Method
1. Fill two 5 ml syringes with 1.5-3 ml of respectively Bio-Rad AG 50W-X8
20-50 mesh, H+ form for the cation exchange, and Bio-Rad AG 1-X8
20-50 mesh, OH" form for the anion exchange. Rinse five times with

287
 M. Ries-Kautt and A. Ducruix
Protocol 3. Continued
1 ml of pure water. Minimize the dead volume of water to avoid
dilution of the protein sample.
2. Aspirate the dialysed protein solution (< 1 ml) in the syringe
containing the cation exchange resin. Shake the syringe for = 5 min,
then remove the solution from the syringe through a 0.22 um filter.
The pH of the solution becomes more acidic (pH = 3-4), depending on
how extensive the dialysis was.
3. Aspirate the acidic protein solution in the syringe containing the anion
exchange resin. Shake the syringe for = 5 min, then remove the
solution from the syringe through a 0.22 um filter.
4. Aspirate 1 ml of pure water into the first syringe and shake for a few
minutes to recover protein remaining in the dead volume. Remove this
solution and rinse the second syringe. This step is repeated twice for a
better recovery of the protein.
5. Isoionic protein solutions can be rapidly deep-frozen in liquid nitrogen
and freeze-dried for storage. The freeze-dried protein is stored at
-80°C.
6. To prepare protein solutions, solubilize the isoionic protein powder in
pure water, centrifuge the solution, and filter it to remove insoluble
protein. Adjust to the required pH and add desired additives.

Alternatively the resin may be prepared in a small column, or on a 0.45 um

filter system equipped either with a vacuum system at the bottom, or using a
nitrogen pressure on the top to accelerate the recovery of the solutions. This
was necessary in the case of HEW lysozyme where a slight precipitation is
observed, which may indicate some denaturation.
3.2.3 Exchange dialysis
Because the counterions of a charged protein cannot be eliminated by a
dialysis step against water, and if the protein does not resist a treatment over a
mixed resin, then the initial unknown counterions may be exchanged against
the desired ions and buffer. This can be achieved by repeated dialysis steps or
by repeated washing over concentration devices (e.g. Centricon or Microcon
from Amicon, or Ultrafree with tangential flux from Millipore). The con-
centration must be higher than the counterions concentration (i.e. the protein
concentration in mM multiplied by the net charge of the protein) and possibly
follow the efficiency of ions described in Section 4.4.1.

3.3 Net charge and crystallization conditions

In June 1998, 3258 crystal forms of 2297 biological macromolecules were
listed in the Biological Macromolecule Crystallization Database (22)
288
 10: From solution to crystals with a physico-chemical aspect

Table 2. Composition of three proteins having a basic pl

Protein No. of amino acids Experimental pl % of neutral % of polar %of

charged
BPTI 58 10-10.5 35 41 24
Erabutoxin 62 9.45 23 55 22
HEWIysozyme 129 11 35 44 21

(http://ibm4 .carb.nist.gov:4400/bmcd.html). The data gathered

in this database are limited to the information given by the authors of the
published crystallization conditions, and very often the description is
incomplete. The main question remains however of how to transpose the
information of the previously crystallized protein to a new one.
Although the intrinsic solubility of a monomeric protein cannot yet be
predicted, it is qualitatively related to the content of hydrophobic, hydro-
philic, and charged amino acids. For the same molecular weight, a protein
containing the highest content of charged amino acids will be more soluble.
For example, the highly soluble cytochrome c contains 35% charged amino
acids, whereas the poorly soluble elastase contains only 13%, and HEW
lysozyme 21%. This can also be illustrated by three basic proteins, HEW
lysozyme, BPTI, and erabutoxin (a snake venom) described in Table 2.
The information of the protein's pI3 and the net charge at the pH of
crystallization is missing in the database, therefore render the search of
'related' proteins difficult. HEW lysozyme, BPTI, and erabutoxin are not
biologically related, but have all three a basic pI. The estimated net charge
for these proteins (Figure 7) at pH 4.5 is higher for BPTI than for the two
others.
As to the nature and the concentration of the salt, the crystallization
conditions of BPTI and erabutoxin at pH 4.5 could be found by starting from
the knowledge of the crystallization conditions of HEW lysozyme at this pH.
In all three cases, NaCl was much less effective for crystallizing the protein
than KSCN (Figure 8).
Concerning the protein concentration, crystallization at pH 4.5 and 100-200
mM KSCN, requires a higher protein concentration for BPTI (= 12 mM),
than for erabutoxin (= 6 mM), and for HEW lysozyme (= 1.5 mM) (Figure 8).
This may be linked to a difference of solubility due to their respective net
charge. However for erabutoxin and HEW lysozyme, which bear approxi-
mately the same net charge at pH 4.5 (Figure 7), a difference of their
respective content of polar and neutral amino acids may also been involved.
An estimation of the solubility change, and crystallization conditions, is also
worth doing if the net charge of a protein is modified when preparing mutants
or complexes of your protein.
289
 M. Ries-Kautt and A. Ducruix

Figure 7. Example of the change of the protein net charge versus pH for three basic
proteins. The pH for which the net charge is zero is the pI. At pH 4.5, the net charge of
BPTI is higher than for the two other proteins.

Figure 8. Crystallization conditions (bold segments) for three basic proteins, BPTI,
erabutoxin, and HEW lysozyme, at pH 4.5 and 18°C.

290
 10: From solution to crystals with a physico-chemical aspect

4. Influence of physico-chemical parameter changes

Crystallogenesis depends on the chemical properties of the protein as well as
on its interactions in a given solvent. Thus a physico-chemical characteriz-
ation of crystallization conditions for a protein requires an exhaustive list of
the chemical constituents of the solution, and the knowledge of how they
change with the temperature and the pH of the solution.
The crystallization parameters which affect protein solubility are:
• temperature
• protein net charge, i.e. the pH and the pKa of the buffer
• the ionic strength, i.e. the concentration of the salts, the buffer, the
counterions, and co-ions of the protein
• the nature of the crystallizing agent
• the dielectric constant, i.e. addition of organic solvents or heavy water.
These parameters affect the protein solubility but may also act on
nucleation, protein crystal growth, and/or the crystallization kinetics.
Although exhaustive solubility diagrams are most often not achievable for a
new protein available in small amounts, we describe here what general rules
have been drawn from model proteins and how to adapt these to a particular
case.

4.1 Interactions in a protein solution

A biological macromolecule is a polymer of amino acids or nucleotides, which
is folded in a tertiary structure mainly by dipole-dipole interactions (H-
bonds, e.g. C=O...H—N—, and Van der Waals interactions), by some covalent
bonds (S-S bridges), and occasionally by salt bridges (e.g. -COO-...+H3N-)
between charged residues.
'Water soluble' proteins contain mostly the hydrophobic side chains in the
core, exposing the hydrophilic side chains on their surface. They are thus
considered as polyions able to dissolve in water (protic polar solvent). The
special case of membrane proteins (see Chapter 8) should be mentioned as
they bear, at least on the surface embedded in the membrane, hydrophobic
residues which interact in their natural medium with lipidic (apolar) com-
pounds. In practice, detergents are added to the water solutions in order to
induce hydrophobic interactions between the hydrophobic residues of the
protein and the hydrophobic tail of the detergent. The hydrophilic head of the
detergent allows then interactions with the solvent.
The balance of interactions controlling the solubility and/or the conform-
ation of a macromolecule can be modified (23, 24) as summarized in Table 3.
The stability of biological macromolecules in solution relies on the com-
petition of solvent-solute interactions with the intramolecular interactions
which are necessary to maintain the tertiary structure.
291
Table 3. Effects of crystallizing parameters on the solvent and/or on the biological macromolecule

Parameter Effect on solvent Effect on the macromolecule Examples

Temperature Disorder of solvent Formation of conformations of higher
increase molecules total free energy
PH H+ and OH- Protonation or deprotonation of Arg, Lys, His, Asp, Glu,
concentrations charged groups C-and N-terminus
Salts Ionic strength Chemical activity coefficient
Shielding of macromolecular electrostatic
interactions
Monopole-monopole interactions with Anions with lysine or arginine side chains, or
accessible charged residues cations with glutamic or aspartic side chains
Monopole-dipole interactions with Peptide bonds, amino, hydroxyl, orcarboxyl
dipolar groups of the macromolecule groups, or amides
Non-polar interactions between solvent Carboxylates, sulfonates, or ammonium salt
exposed hydrophobic residues and the Solubilization of solvent exposed hydrophobic
hydrophobic part of organic salts residues by the hydrophobic tail of an ionic
detergent
Association with binding sites Protein-ion interaction with a specific part, of well
defined geometry, of the biological macromolecule
H-bond Competition at high concentration (> 4 M) Formamide, urea, guanidinium salts
competitors with H-bonds of water and the structural
intramolecular H-bonds of the protein
Hydrophobic Alteration of the Interaction with hydrophobic parts Non-ionic detergents
additives solvent structure of the protein
Organic Modification of the Interaction with hydrophobic or polar Alcohols, DMSO, MPD
solvents dielectric constant parts of the protein
 10: From solution to crystals with a physico-chemical aspect
Two levels of protein-protein interactions can be distinguished:
(a) Long-range interactions (a few nm) are essentially governed by non-
specific electrostatic interactions according to the Debye-Huckel theory.
The individual macromolecules are then considered as spheres having a
given net charge with randomly distributed charges. The Derjaguin-
Landau-Verwey-Overbeck (DLVO) theory (25, 26) describes the inter-
action between two molecules as the net interaction resulting from:
• an electrostatic repulsion and
• a Van der Waals attraction (4).
Considering a macromolecule as a particle with a given net charge Zp,
different from zero, the long-range electrostatic protein-protein inter-
actions are repulsive at low ionic strength or in pure water. For electro-
static compensation, the solution provides at least Z counterions. Once
the net charge is sufficiently screened, protein-protein interactions become
less repulsive and finally attractive. The charges can be screened by
increasing the salt concentration of the solution, because an increasing
ionic strength lowers the repulsive contribution of the protein-protein
interaction. Additionally, specific ions may be bound by the protein and
therefore change its net charge. Thus binding of ions can also lead to less
repulsive (or more attractive) protein-protein interactions. Attractive
interactions are necessary, but not sufficient for protein crystallization.
They may lead to amorphous precipitate, as well as to crystals.
(b) Specific short-range interactions, occurring at the intermolecular level,
promote specific and periodic protein-protein contacts to build the
crystal. The chemical equilibria become determinant when local charges
on the protein surface interact specifically with another protein molecule,
water, or a solvent component. Protein contacts in the crystal are due to
hydrogen bonds, hydrophobic interactions/Van der Waals, and salt
bridges (27). However, the balance of interactions finally leading to
protein contacts in the crystal are difficult to predict. Moreover, the
situation is complicated by the fact that both static and dynamic aspects
should be considered.

4.2 pH
A change of pH implies a change of the protein net charge:
(a) Near the pKa of most numerous charged residues, solubility varies very
rapidly.
(b) Outside the range of the pKa values of charged residues, the solubility
changes smoothly.
(c) Solubility is minimal at the pI of the protein as shown in the case of insulin
(28) (Figure 9), egg albumin (29), haemoglobin (30), and B-lactoglobulin
(31). Conversely solubility is higher when the net charge increases (17).
293
 M. Ries-Kautt and A. Ducruix

Figure 9. Variation of the solubility of insulin versus pH (redrawn from ref. 28). Note that
decreasing the pH by 1 unit (i.e. from pH 5 to pH 4), the net charge changes from +1 to
+5 and the solubility from = 0.4 to = 1.2 mg/ml. When increasing the pH by the same
increment (+1 unit, but from pH 5 to pH 6), the net charge changes from +1 to only-0.5
and the solubility from ~ 0.4 to only = 0.6 mg/ml.

It is of practical interest to remember that the net charge is independent of

the ionic strength at the pI (32). Consequently solubility is also constant at the
pI whatever the ionic strength (17). A change of the protein net charge may
induce polymorphism. The effect of pH on protein solubility is amplified at
low ionic strength (17, 33).
The pH of a protein solution is set by the buffer, whose importance is often
neglected. The buffering capacity of a weak acid, or base, is limited to a pH
range from its pKa ± 1 pH unit. Some additives and crystallizing agents (phos-
phate, citrate, acetate) are themselves weak acids or bases. Their solutions
have to be adjusted to the desired pH of crystallization. A polyacid changes its
own charge depending on the pH; its efficiency can be different depending
whether it is mono-, di-, or trivalent.
By definition buffers are soft acids or bases and may present preferential
binding, even of low affinity, with the biological material. Even at exactly the
same pH, the protein solubility can be different depending on the nature, and
of course the concentration, of the buffer. Therefore two, otherwise identical,
crystallization experiments may behave differently at a same pH when using
different buffers.
294
 10: From solution to crystals with a physico-chemical aspect
To make use of the effect of pH on protein solubility, it is recommended to
try crystallization conditions at the pI and on both sides of it (Figure 9).
However the range of pH in which the protein is stable must first be checked.
A change of pH is achievable with volatile acids or bases by vapour diffusion,
otherwise by dialysis or batch method.

4.3 Ionic strength

The ionic strength, I, of a salt solution is due to the concentration Ci of this
salt, but also to the valency Zi of each ion. It is easily calculated by:

Therefore the ionic strength of a 0.1 M salt solution of:

• a [1:1] electrolyte like NaCl is: 1/2 (([Na+] X 12) + ([Cl-] X 12)) = 0.1 M
• a [1:2] electrolyte like (NH4)2SO4 is: 1/2 ((2 X [NH4+] X 12) + ([SO42-] X
22)) = 0.3 M
• a [2:2] electrolyte like MgSO4 is: 1/2 (([Mg2+] X 22) + ([SO42-] X 22)) = 0.4 M
For salt concentrations above 0.2 M, the concentration should be corrected
by the chemical activity coefficient which can be found in most handbooks. In
a phase diagram, it is more convenient to express the salt concentration as
ionic strength rather than molarity, especially for comparing the solubility in
mono-, di-, or polyvalent ions.
If a salt of a weak acid or base is used, the actual concentration must be
calculated depending on the pH of the solution in respect with the pKa,
according to the same rules as detailed in Section 3.1 and Table 1. For a 0.1 M
sodium acetate (pKa = 4.76) solution:
• at pH = 8 only acetate is present, I = 1/2 (([Na+] X 12) + ([AcO-] X 12)) =
0.1 M
• at pH = pKa, 50% of the acetate is protonated and 50% is charged,
I = 1/2 ((0.5[Na+] X 12) + (0.5[AcO-] X 12)) = 0.05 M.
This shows the importance of how the buffer is prepared, if the pKa is
reached by adding:
• acetic acid to a 0.1 M sodium acetate solution, I = 0.1 M
• sodium hydroxide to a 0.1 M acetic acid, I = 0.05 M.
As for cationic species, an increasing pH may also promote the formation
of hydroxides, the cation then no longer acts as Mn+, but as M(OH)(n-1)+, or
more generally as M(OH)i (n-i)+ .
The effects of salts on protein solubility are complex and rely on a balance
between protein-water, protein-salt, and salt-water interactions. In addition,
295
 M. Ries-Kautt and A. Ducruix
the variation of protein solubility over the whole range of salt concentration
reflects the resultant effect of both electrostatic and hydrophobic interactions,
the first being predominant at low salt concentration and the second at high
salt concentration. The change of protein solubility at increasing salt con-
centrations was studied in term of salting-in and salting-out (34-36). However
it was shown more recently that salting-in is not systematic at low ionic
strength, but seems to be also correlated with the protein net charge (17).

4.3.1 Salting-in
Solubility data of carboxyhaemoglobin (Figure 10a) showed that protein
solubility first increased (salting-in) and then decreased (salting-out) with
increasing ionic strength (34). This phenomenon is explained by the decrease
of the chemical activity of the protein when the ionic strength of its environ-
ment increases (36). It is worth emphasizing that the solubility variation of
carboxyhaemoglobin at 25°C and pH 6.6 corresponds to its minimal
solubility, i.e. near the pI.
As for HEW lysozyme (Figure 10b) bearing a net charge different from
zero, no salting-in could be evidenced (17). Here, the screening of the salt on
the electrostatic protein-protein interactions seems to dominate the effect of
the protein chemical activity. Furthermore, salting-in may be reduced or
emphasized depending whether co-ions or counterions bind to the proteins, as
will be discussed in Section 4.4.

4.3.2 Salting-out
Salting-out corresponds to a decrease of protein solubility at high ionic
strength, where the protein behaves as a neutral dipole and solubility is
mainly governed by hydrophobic effects. Theoretically a crystallizing agent
added to the protein-water (solute-solvent) system can either bind to the
protein (preferential binding) or be excluded (preferential exclusion) depend-
ing on preferential protein-additive or protein-water interactions (37). The
net interaction of salting-out is preferential exclusion, even though molecules
or additives may bind to the protein.
Protein solubility has been expressed (34) according to:

where S is the protein solubility (in mg/ml), m the molal salt concentration
(g salt/1000 g water), and 3 the intercept at m = 0. 3 is a constant at high salt
concentration and function of the net charge of the protein, thus strongly pH-
dependent. Therefore it is minimal at the isoelectric point. The magnitude of
3, as well as charge distribution, varies with temperature. Ks is the salting-out
constant. It is independent of pH and temperature, but depends on the nature
of the salt.
However, experimentally defined solubility curves rarely fit with a linear
296
 10: From solution to crystals with a physico-chemical aspect

Figure 10. Variation of the solubility versus ionic strength. (a) Salting-in for Carboxy-
haemoglobin near its pl and in the presence of various salts (redrawn from ref. 34). (b) No
salting-in for positively charged lysozyme in the presence of NaCl and at different pH
values (17).

297
 M. Ries-Kautt and A. Ducruix
function (Figure 1). This non-linearity may be essentially due to the following
reasons:
(a) The solubility tends to the solubility value of the protein in the buffer at
low concentrations of the crystallizing agent, the efficiency of the crystal-
lizing agent, and of the buffer becoming comparable. The curves of HEW
lysozyme solubility in the presence of various salts (10) converge at low
salt concentrations toward its solubility curve in the sodium acetate
buffer.
(b) Higher amounts of protein are required for crystallization at low salt
concentrations, so the solubility value can be affected by the presence of
higher amounts of protein related salts; they can either be counterions, or
salts which were not eliminated by a previous dialysis step.
(c) Protein binding of counterions can no more be neglected compared to
preferential exclusion. The solubility is then affected by the change of
both the net charge of the protein and the ionic strength.

4.4 Nature of salts

The way the nature of the salts acts on protein solubility is complex and not
yet clearly understood. They affect the ionic strength depending on the con-
centration and the valency of their ions. Even though two ions bear an
identical charge, their size and their polarizability are different, affecting
therefore their own hydration, the interaction with charged residues of the
protein, and potential binding sites of the protein.

4.4.1 Inversion of the Hofmeister series, depending on the protein net

charge
A longstanding and apparently general observation is that ions differ greatly
in their ability to salt-out protein solutions. In 1888, Hofmeister (38) ranked
various ions toward their precipitation ability by adding increasing amount of
salts to a mixture of hen egg white proteins. The Hofmeister series, sometimes
called the lyotropic series, have since been associated with many biological
phenomena and extensively reviewed (39, 40). It has been shown (24) that
ions act on the structures of biological macromolecule structures according to
the same series:
• cations: Li + > Na+ > K+ > NH4+ > Mg2+
• anions: sulfate2- > phosphate2- > acetate- > citrate3- > tartrate2- >
bicarbonate" > chromate2- > chloride- > nitrate- >> chlorate- >
thiocyanate-
Ions such as sulfate reinforce the structures of water and biological macro-
molecules (they are called lyotropic), whereas ions such as chlorate and
thiocyanate denature them (these are called chaotropic) (24).
298
 10: From solution to crystals with a physico-chemical aspect
i. Protein bearing a negative net charge
The solubility data of Hl collagenase (pI of 4.1) crystallized at pH 7.2 and at
18°C, and in the presence of various ammonium salts (6) showed that the
efficiency of anions to lower the solubility of Hl collagenase is consistent with
the precipitation observations of Hofmeister:
phosphate2-/phosphate- > sulfate2- > citrate3-/citrate2- >>chloride-
Hl collagenase is moderately soluble in ammonium sulfate and becomes
extremely soluble in ammonium chloride when it bears a negative net charge.
This was used to increase the solubility of Hl collagenase and suppress growth
of twinned crystals in ammonium sulfate by adding 200 mM NaCl.

ii. Protein bearing a positive net charge

Solubility measurements for basic proteins (10, 41) were undertaken at pH 4.5
(50 mM Na acetate) and 18°C in the presence of a large variety of salts with
HEW lysozyme (pI 11). The results of the solubility curves (some shown in
Figure 1) show an inversion of the Hofmeister anion series, which becomes:
thiocyanate- ~ para-toluene sulfonate (pTS~) > nitrate- > chloride- >
acetate" ~ phosphate- > citrate2-
whereas the efficiency of cations is weak and follows Hofmeister series,

iii. Inversion of Hofmeister series

It is of considerable potential relevance to protein crystal growth, that the
order of these effects on protein solubility depends on the net charge of the
protein. This was evidenced by systematic solubility measurements for these
two model proteins. It was since confirmed with a number of other proteins
for which the solubility, as well as the nucleation zones and the precipitation
zones are affected in the same order.
The anion series follows the order of Hofmeister in affecting the solubility
of Hl collagenase, amylase (42), parvalbumin (43), the nucleation zone of
Grb2 (44), and the precipitation of ovalbumin, the major protein in hen egg
white tested by Hofmeister. All these proteins have an acidic pI and were
tested at a higher pH where they bear a negative net charge.
The reverse order of the anion series is observed for the solubility of HEW
lysozyme, BPTI (5), and the nucleation zone of toxins (45) (erabutoxin,
fasciculin, and muscarinic toxin 2), and lysin from spermatozoa (46). These
proteins have a basic pI and were tested at a lower pH where they bear a
negative net charge.
Furthermore it was shown that the anion series is reversed depending
whether the precipitation of a same protein is achieved below or above the pI
in the case of the precipitation fibrinogen (47) and of insulin (28).
299
 M. Ries-Kautt and A. Ducruix
iv. Adsorption of anions by basic proteins
Thiocyanate, which is well known as a chaotropic agent at high concentration,
appears to be very effective at crystallizing HEW lysozyme at low concentra-
tion. This is also observed with other proteins having a high isoelectric point:
BPTI (48), toxins (45) (erabutoxin, fasciculin, and muscarinic toxin 2), and
lysin from spermatozoa (46). A similar efficiency was observed with organic
salts (sodium p-toluenesulfonate, benzenesulfonate, and benzoate) which
successfully crystallized HEW lysozyme at low concentrations (48), typically
0.1-0.2 M. It appeared that the efficiency of carboxylates to crystallize HEW
lysozyme was: pTS- ~ benzoate > propionate > acetate. The high efficiency
of these anionic species was interpreted by the occurrence of anion binding to
the protein, prior to the process of exclusion at high salt concentration. They
were thought to interact with positively charged residues of the protein (48).
The presence of one SCN ion could be demonstrated unequivocally in the
electron densities of erabutoxin b (49) and turkey egg white lysozyme (50).
Assuming a protein anion association constant of = 0.1 M, a protein salt
would be formed. This protein salt would have a lower net charge than the
protein itself if a counterion is bound, or a higher net charge if a co-ion is
bound. Therefore any solvent constituent, as well salts as buffers or other
additives, may play an important role in the crystallization, if they interact
with a protein.
4.4.2 Testing the Hofmeister series
The precipitation zone, the nucleation zone, and the solubility curve occur at
different supersaturation in the phase diagram. However the relative efficiency
of the salts can be detected on any of these zones. If no crystallization con-
ditions are known, the Hofmeister series can be tested by precipitation tests.
Precipitation is achievable with any crystallizing agent, when raising sufficiently
either the protein or the salt concentration. An example of a rapid test is
given in Protocol 4.

Protocol 4. Testing the Hofmeister series by dialysis

Equipment and reagents

• Five dialysis buttons • 2.0 M stock solution of KSCN, NaNO3, NaCI,
• Linbro box NaOAc, and NH4SO4

Method
1. Prepare five dialysis buttons with protein solution at a concentration
as high as possible.
2. Prepare 2 ml reservoir solutions at 0.05, 0.1, 0.2, and 0.5 M of KSCN,
and at 0.1, 0.5, 1.0, and 2.0 M ionic strength of NaNO3, NaCI, NaOAc,
and NH4S04 in the same buffer as the protein.
300
 10: From solution to crystals with a physico-chemical aspect
3. Place a dialysis button in the reservoir at the lowest concentration of
each salt. Close the well with a coverslip.
4. Transfer it after at least 2 h to the next concentration. Respect the
same delay for each change.
5. As soon as precipitation is observed, place the dialysis button in the
previous reservoir and prepare an intermediate concentration to refine
the value of the precipitation limit.

The relative position of phosphate and citrate versus sulfate in the series
may change, depending on the pH, since their ionic strength varies rapidly
around their pKa values with Zi2 according to Equation 1.

4.4.3 Peculiar behaviours

The salt effect of the Hofmeister series was shown to be general, depending
on the net charge of the protein. It was shown that formation of a protein salt
may occur when an ion weakly binds to the protein surface, probably by the
formation of an ion pair. This is different from binding to a specific site of the
protein, because the binding is then stronger. It becomes typical of a given
protein and no longer transposable to another one which has not this specific
site. Apart from known binding sites of cations, like Zn or Ca, which inter-
vene in the structure or the biological activity of a protein, binding may
incidentally occur in given crystallization experiments. The two following
examples illustrate such situations.
(a) HEW lysozyme crystallizes in similar concentrations when using Na ben-
zoate, NapTS, or KSCN (48) as crystallizing agent. In the case of BPTI
which also crystallizes at low concentration of KSCN, no crystallization
could be achieved with NapTS, even when adding solid NapTS in the
reservoir of a crystallization experiment where BPTI crystals had pre-
viously grown with NaCl. This led to the dissolution of the BPTI crystals
inside the dialysis cell, although pTS had reached saturation and started
to crystallize in the reservoir. This may possibly be due to an interaction
of the hydrophobic part of pTS with a small protein like BPTI, thus
acting like a solubilizing agent, instead of interacting through its sulfonic
group.
(b) In the Crystallization Database (22), ammonium sulfate is more frequently
cited than NaCl among the crystallizing agents (respectively about 800
citations versus 300). However HEW lysozyme is known to crystallize
more easily with NaCl, and to resist crystallization with ammonium sulfate.
Using the sulfate anion with diverse counterions (ammonium, sodium, or
lithium) no crystallization occurred at pH 4.5."Crystals of HEW lysozyme
sulfate could be grown at pH 8 instead of pH 4.5, from isoionic HEW
lysozyme in the presence of sulfate ions up to 100 mM (51).
301
 M. Ries-Kautt and A. Ducruix

4.5 Temperature
The variation of protein solubility with temperature may be either direct,
i.e. increasing with temperature, or retrograde. The behaviour of protein
solubility with temperature cannot been foreseen, and is not characteristic,
neither of the protein, nor of the crystallizing agent, but of the protein salt as
illustrated with BPTI. This protein has a retrograde solubility change with
temperature in the presence of ammonium sulfate (5) and sodium chloride
(52), but direct with potassium thiocyanate (5) (Figure 11),

Figure 11. Solubility behaviour of different BPTI salts versus temperature. It is retrograde
with ammonium sulfate (AS) and NaCl, but direct with KSCN. Redrawn from refs 5 and 52.

Protocol 5. Testing solubility changes versus temperature

Equipment and reagents

• Incubator at 4°C « Incubator at 37°C

Method
1. Take a series of screening experiments (vapour diffusion, batch, or
dialysis) covering the nucleation zone, i.e. from clear drops to ones
containing slight precipitation at 18-20°C.

302
 10: From solution to crystals with a physico-chemical aspect
2. Introduce the Linbro plate in a Styrofoam box and place these
experiments in an incubator at 4°C.
3. On the next day, check whether the nucleation or the precipitation has
increased or is reduced compared to the initial observations.
4. Bring the experiments again to 18-20°C for a day or two to check the
reversibility.
5. Place these experiments in an incubator at 37°C for another day or
two.
6. Check again whether the nucleation or the precipitation has increased
or is reduced.
7. Analyse the results:
(a) No change whatever the temperature. This may indicate that:
• The sampling of ionic strength or pH is too large (and the
nucleation zone very small). Repeat the experiments with drops
differing by smaller steps of ionic strength or pH.
• The ionic strength is too high. Repeat the experiments with
drops at lower ionic strength and higher protein concentration.
• The pH of crystallization is too close to the pl. Repeat the experi-
ments with drops at lower or higher protein net charge.
(b) The lower the temperature, the more intense the precipitation.
Solubility is directly related to temperature. Verify nevertheless
the reversibility of the precipitation by bringing the experiments to
a higher temperature, to avoid confusion between precipitation
which is reversible, with denaturation which may not be
reversible.
(c) The lower the temperature, the less intense the precipitation.
Solubility is retrograde with temperature, at least with this salt
combination. Repeat the experiments when changing the
crystallizing agent.

Likely to the effect of pH, the variation of protein solubility with tempera-
ture is amplified at low ionic strength (33, 53). To benefit the effect of
temperature this means working at rather low ionic strength, but conversely
carrying out experiments at higher ionic strength for the transport of crystals
with their crystallization solution to stabilize them (i.e. to avoid further
nucleation or dissolution of the crystals).

4.6 H2O versus D2O

The measurement of the solubility of lysozyme in H2O and D2O (7) has
shown that the solubility was decreased by about 30%. This has been
explained by the difference of the density of the solvent (14).
303
 M. Ries-Kautt and A. Ducruix

4.7 Combined effects of crystallization variables

When testing one variable at a time toward crystallization or solubility, the
other parameters are kept constant at a fixed value. The advantages of
varying all variables at the same time are treated in Chapter 4. We illustrate
here the combined effects of two variables on the solubility and on the
nucleation zone.

4.7.1 Low ionic strength amplifies pH and temperature effects

It has been shown with different proteins and in different crystallization con-
ditions that low ionic strength amplifies pH and temperature effects (33, 41,
53). The practical applications of this are the following:
(a) To make use of the pH or of the temperature effect, crystallization
experiments should be carried out at low ionic strength.
(b) To prevent changes of solubility when the temperature is less controlled
(e.g. transportation of the crystals), the samples should be gently soaked
in high ionic strength solutions.
(c) It must be reminded that the solubility is constant at the pI whatever the
ionic strength.

4.7.2 The nature of the salt and temperature

When combining the nature of the crystallizing salt and temperature on
the crystallization of HEW lysozyme at pH 4.5 (41), a similar solubility of
~ 0.8 mg/ml was observed in the presence of NaSCN at 40 °C as well as of
Nad at 0°C. In this case the temperature effect is direct for both protein salts.
The effect of the very efficient NaSCN salt can be compensated at high
temperature and become equivalent to a less efficient salt, NaCl, used at low
temperature.

4.8 Step wise replacement technique

When a known crystallization condition is unsatisfactory or if a crystallization
parameter (new salt, other pH or temperature values) needs to be explored,
the search to locate the nucleation zone in the new phase diagram can be done
by testing the shift of the initial nucleation zone while replacing stepwise the
parameters to be changed. This can be performed by hanging drop or dialysis
method, the dialysis offering the advantage to work on the same sample while
changing the nature of buffer or salt of the protein sample.
The stepwise replacement technique is very useful to qualitatively check
the relative effectiveness of ions on protein solubility, as illustrated in
Protocol 6.
304
 10: From solution to crystals with a physico-chemical aspect

Protocol 6. Stepwise replacement

Equipment and reagents

• Linbro box and silanized coverslips • 25 ml of 4 M sodium chloride stock solution
in
« 25 ml of 3 M ammonium sulfate stock buffer
solution in buffer • 100 ml buffer

Method
1. Choose a range of crystallizing agent concentration to cover the
nucleation zone, i.e. from clear drops to those containing slight pre-
cipitation, at a constant protein concentration. As an example, these
reference conditions may be 20 mg/ml protein and reservoirs from
0.8-1.3 M ammonium sulfate (with 0.1 M steps for the reservoirs).
2. Prepare a series of six drops in the reference conditions in the first row
of a Linbro plate (A1 to A6).
3. Set up the second row at the same protein concentration (B1 to B6),
but with reservoirs where 0.2 M ammonium sulfate is replaced by
0.3 M NaCI. It is recommended to take into account the ionic strength
rather than molarity (see Section 4.3) when replacing crystallizing
agents.
4. Observe whether crystallization occurs at higher or at lower ionic
strength in the second row compared to the first one:
(a) The nucleation zone is shifted by one column to lower ionic
strength. NaCI is slightly more efficient than ammonium sulfate.
Continue by setting up the third row while replacing 0.4 M
ammonium sulfate by 1.2 M NaCI.
(b) The nucleation zone starts already in B1. NaCI is much more
efficient than ammonium sulfate. Set up the third row to centre
again the nucleation zone by replacing 0.4 M ammonium sulfate
by only 0.4 M NaCI and starting at lower ionic strength (from
1.2-2.7 M total ionic strength instead of 2.4-3.9 M total ionic
strength).
5. Set up the last row at the same protein concentration (D1 to D6), but
with reservoirs containing only NaCI at concentrations chosen
depending on the previous results to centre the nucleation zone.

5. Crystallization
The aim of crystallization experiments is first to locate the nucleation zone,
then to optimize the physico-chemical parameters and the kinetics to grow
large single crystals. Testing a large number and combination of variables may
305
 M. Ries-Kautt and A. Ducruix
yield different crystal forms. It is worth optimizing different polymorphs
because the number of molecules in the asymmetric unit and the diffraction
quality may be very different.

5.1 Crystallization strategies

Before the crystallization protocol is defined, the batch of protein has to be
characterized as accurately as possible. This definitively helps the repro-
ducibility when using news batches. The analyses should be done by different
techniques: electrophoretic gel, IEF, mass spectrometry, UV spectrum, light
scattering, etc. (see Chapter 2). All information about the stability of the
protein should also be listed in order to select the parameters for which the
protein is known to be stable with time, and eliminate those for which the
protein denatures.
5.1.1 Solubilization in the buffer
When concentrating a protein solution, roughly three situations can be
encountered:
(a) The protein concentration in the buffer solution is low (< 5 mg/ml).
Nucleation will be difficult to control, and little protein will be available
to feed the crystals as described in Section 2.3.2. It is thus advisable to
search for conditions where solubility is expected to be higher, e.g. at a
different pH where the protein net charge is higher.
(b) The protein concentration in the buffer solution can reach at least 10-50
mg/ml. The screening tests can be performed.
(c) The protein concentration in the buffer solution is very high (> 100
mg/ml). Nucleation will be easier to control and will probably occur at
low supersaturation (1, 17). However if high protein concentration pre-
sents practical constraints, conditions of lower solubility can be sought,
e.g. by approaching the pI.
5.1.2 Screening
The aim of the screening step is to locate the nucleation zone for each
solubility diagram tested. This can be performed by the batch, the vapour
diffusion, or the dialysis technique (Protocol 7). Dialysis is the most recom-
mended as it includes exchange of the initial buffer and that of the unknown
counterions. When working with hanging drops, in which the constituents are
no longer exchanged (except water), a previous step of dialysis is recom-
mended to exchange the chemicals coming from the protein purification steps.
In agreement with Section 2.3.2, the tests should benefit from working at
high protein concentrations.
5.1.3 Refining the nucleation conditions
After the screening step, the list of crystallization parameters is examined
again. Some conditions may have shown to be incompatible with the stability
306
 10: From solution to crystals with a physico-chemical aspect
of the protein, others may be interesting for further investigations. For
example if the carboxylate salt has given interesting results, other carboxylate
salts with different cations may be worth testing. Additional parameters to
test the crystallizability of a given protein salt can be added: mixing inorganic
and organic crystallizing agents, effects of stabilizing agents, additives,
glycerol, the temperature, etc. Depending on the number of combinations to
test, it may be suitable to perform this step with an incomplete factorial design
(compare Chapter 4).

Protocol 7. Search of the nucleation zone by dialysis

Equipment and reagents

• Three Linbro plates with silanized cover- • 21-25 dialysis buttons
slips . 25 ml of stock solution of the six selected
• 100 ml buffer at the three selected pH crystallizing agents in appropriate buffers
values

Method
1. According to the estimation of the variation of net charge with pH (see
Section 3.1), select one pH close to the pI, and two on both sides of the
pI so that the net charge is about the same value, but of opposite sign.
When this is in conflict with the stability of the protein, select three
other pH values for which the variation of the protein net charge is as
large as possible. Taking the example of BPTI (Figure 7), a pH higher
than the pI would not be well suited. Therefore one would select pH
10.5, 9, and 4.5 for which the net charge of this protein is 0,
=>7 + , and = 14+ respectively.
2. Prepare three Linbro plates one for each pH. Fill the six reservoirs of
the first row (A) of each Linbro plate with 2 ml of buffer corresponding
to the pH of each plate.
3. Prepare three dialysis buttons filled with the stock protein solution.
Introduce one of them in A1 of each Linbro plate. Observe if the
protein solution remains clear for a few days. If precipitation occurs:
(a) It happens for the lower net charge. Prepare a dialysis button at
the lower protein concentration until the drop remains clear for a
week.
(b) It seems not to be linked to the net charge. Transfer the sample in
the original buffer to search for reversibility of the precipitation.
Replace the buffer solution twice or more to ensure a good
exchange of the buffer solution. If the precipitate remains, check
for possible denaturation and choose another pH for the following
steps.
4. Prepare 15 dialysis buttons filled with the stock protein solution.

307
 M. Ries-Kautt and A. Ducruix
Protocol 7. Continued
Introduce them in the remaining A2 to A6 reservoirs. Let them stand
for one day. Eventually change the reservoir when using large protein
volumes to ensure the buffer exchange.
5. Fill the six reservoirs of the row B with six different crystallizing agents
(2 ml), each in the appropriate buffer of a given Linbro plate. The
crystallizing agents should be chosen according to the protein net
charge (see Section 4.4.1). They should preferably be of different
chemical types; thiocyanate, halide (Cl-, Br-, l-, or F-), carboxylate
(acetate, citrate, tartrate), sulfate (or phosphate), PEG, divalent cation
(Mg2+ or Ca2+). As a rule of thumb, the concentration of the first
reservoir may be 0.1-0.5 M, or 5-10% PEG.
6. Transfer the dialysis buttons from row A to row B. After two to five
days, observe the protein solutions:
(a) Case 1: the solution is clear. Prepare the next reservoir C at twice
the concentration of the one in B.
(b) Case 2: the solution precipitates. Prepare the next reservoir C at
half the concentration of B. Transfer the dialysis button, first back
to row A to dissolve the precipitate, then to C.
(c) Case 3: the solution B is neither clear, nor precipitated. Wait for
another period of two to five days to decide whether the next
reservoir concentration should be increased or decreased by only
10%.
7. Continue until the limits between clear solutions and precipitation (i.e.
lower and upper limits of the nucleation zone) are defined.
8. Set up a new set of experiments to refine the concentration of each crys-
tallizing agent at each pH with a small step in between the nucleation
zone limits. At this stage, the vapour diffusion may be more suitable.

5.1.4 Optimization
Once the nucleation zone is defined, the optimal conditions to grow large
single crystals must be sought. At this step the protein/reservoir equilibration
kinetics should be included among the variables to be adjusted. The tools to
perform optimization can also be found in Chapter 4.

5.2 Polymorphism
Table 4 illustrates the variety of crystal forms observed for lysozyme when
changing the crystallization conditions (temperature, pH, nature of the
crystallizing agent) or a combination of them. Apart from the crystal form,
also the number of molecules in the asymmetric unit can change. Both can
present an advantage for the crystallographer.
308
Table 4. Polymorphism of HEW lysozyme

Lattice Space Parameters z Crystallizing agent pH T(°C) Ref

group a, b, c in A/ a, B, -y in ° (mol/au)
Tetragonal P432,2 a = b = 79.2 c = 38.0 1 NaCI (0.3-1. 5 M) 4.3-4.7" 18 10,
54-57
or KCI (0.5-1. 1M) 4.3-4.7" 18 10
or NH4CI (0.5-1. 1 M) 4.3-4.7" 18 10
or MgCI2 (0.4-1. 1 M) 4.3-4.7" 18 10
or NH4OAc (0.9-1.5 M) 4.3-4.7" 18 10
or Na2HP04(1. 1-1.2 M) 4.3-4.7a 18 10
or Na ptoluenesulfonate (0.08-0.25 M) 4.3-4.7a 18,40 41
or Na benzenesulfonate (0.21-0.24 M) 4.3-4.7" 18 48
or Na benzoate (0.1-0.2 M)/benzoic acid 5c 18 48
a = b = 78.1 c = 38.2 1 NaCI/HCI 4, 6, 8C 18 17
a = b = 78.8 c = 38.3 1 NH4citrate(0.5-1.2M) 4.7" 18 10
a = b = 78.9 c = 38.5 1 None/H2SO4 8 18 51
Orthorhombic P2,2121 a = 56.3 b = 65.2 c = 30.6 1 NaCI (0.88 M) 10 55-57
a = 56.5b = 73.9c = 30.5 1 NaCI (0.88-1. 37 M) 4.7a 37,40 41, 54
a = 58.6b = 68.4c = 54.3 2 Ethanol (55%) + NaCI 8.4* 18 58
Hexagonal P6,22 a = b = 87.0c = 70.4 1 Acetone (10%) + NaNO3 (saturated) 8.46 18 58
Monoclinic P2, a = 28.1 b = 63.1 c = 60.6 2 KSCN (0.075-0.2 M) 4.5" 18 10
p = 90.6 or NaSCN (0.1-1.0 M) 4.5" 18,40 41
orNaNO3(0.36M)/HN03 4.5c 18 57
or Nal (3%) 4.5C 18 57
a = 28.6 b = 63.0 c = 60.6 2 Na2S04(0.77M) + 4.5 18 57
p = 93.5 NaAcO (0.5 M) /H2S04
a = 27.9 b = 63.0 c = 66.3 2 KN03 (5%) 4.0 - 56
P =114.2
Triclinic P1 a = 27.5 b = 32.1 c = 34.4 1 NaN03(0.24M) 4.5" 18 57
a = 88.3B = 109.0y= 111.0
a = 41.7b = 58.4c = 58.9 3 NDsulfo-betaine195 + (NH4)2S04 4.5" 18 59
a = 1 16.5 p = 97.6 -y = 105.7
"The pH is adjusted by 50 mM NaAcO buffer.
'The pH is adjusted either by NaHC03 buffer.
'The pH is adjusted only by the indicated acid.
 M. Ries-Kautt and A. Ducruix

By definition a polymorph is a variant of the crystal form for an identical

molecule. Thus monoclinic HEW lysozyme thiocyanate is not a true poly-
morph of tetragonal HEW lysozyme chloride if we accept that the molecule
which crystallizes is the protein salt and not the protein by itself. In the case of
polymorphism of HEW lysozyme with pH, the importance of the initial
desalting of the protein has also been shown to be important (60). Different
crystal forms may appear during the process of crystallization, either because
the crystallization conditions are at the borderline between two crystal forms,
or because a parameter value has varied during the crystallization process
(pH or temperature shift).

References
1. Boistelle, R. and Astier, J.-P. (1988). J. Cryst. Growth, 90, 14.
2. Rosenberger, F., Vekilov, P. G., Muschol, M., and Thomas, B. R. (1996). J. Cryst.
Growth, 168,1.
3. Chernov, A. A. (1997). Phys. Rep., 288,61.
4. Ries-Kautt, M. and Ducuix, A. (1997). In Methods in enzymology (ed. C. Carter
and R. Sweet), Academic Press, London. Vol. 276, Part A, Chap 3, p. 23.
5. Lafont, S., Veesler, S., Astier, J.-P., and Boistelle, R. (1997). J. Cryst. Growth, 173,
132.
6. Carbonnaux, C., Ries-Kautt, M., and Ducruix, A. (1995). Protein Sd., 4, No. 10,
2123.
7. Broutin, I, Ries-Kautt, M., and Ducruix, A. (1995). J. Appl. Crystallogr., 28,
614.
8. Schein, C. H. (1990). Biotechnology, 8, 308.
9. Ataka, M. and Tanaka, S. (1986). Biopolymers, 25,337.
10. Ries-Kautt, M. and Ducruix, A. F. (1989). J Biol. Chem., 264, 745.
11. Cacioppo, E., Munson, S., and Pusey, M. L. (1991). J. Cryst. Growth, 110, 66.
12. Boistelle, R., Astier, J.-P., Marcis-Mouren, G., Desseaux, V., and Haser, R.
(1992). J. Cryst. Growth, 123,109.
13. Rosenberger, F., Howard, S. B., Sower, J. W., and Nyce, T. A. (1993). J. Cryst.
Growth, 129,1.
14. Gripon, C., Legrand, L., Rosenman, L, Vidal, O., Robert, M.-C., and. Boue, F.
(1997). J. Cryst. Growth, 177,238.
15. Sazaki, G., Kurihara, K., Nakada, T., Miyashita, S., and Komatsu, H. (1996).
J. Cryst. Growth, 169, 355.
16. Darcy, P. A. and Wiencek, J. M. (1998). Acta Cryst., D54, 1387.
17. Retailleau, P., Ries-Kautt, M., and Ducruix, A. (1997). Biophys. J., 73, 2156.
18. Feher, G. and Kam, Z. (1985). Methods in enzymology (ed. S. P. Colowick and N.
O. Kaplan), Vol. 114, pp. 77-112.
19. Luft, J. R. and DeTitta, G. T. (1997). Methods in enzymology, ibid ref. 4. Vol. 276,
p. 110.
20. Gaucher, J. F., Ries-Kautt, M., Reiss-Husson, F., and Ducruix, A. (1997). FEBS
Lett, 401,113.
21. Jolivalt, C., Ries-Kautt, M., Chevallier, P., and Ducruix, A. (1997). J. Synchrotron
Rad., 4,28.
310
 10: From solution to crystals with a physico-chemical aspect
22. Gilliland, G. L., Tung, M., Blakeslee, D. M., and Ladner, J. E. (1994). Acta Cyst.,
D50,408.
23. von Hippel, P. H. and Schleich, T. (1969). Ace. Chem. Res., 2,257.
24. von Hippel, P. H. and Schleich, T. (1969). In Structure and stability of biological
macromolecules (ed. S. N. Timasheff and G. D. Fashman), pp. 417. Marcel
Dekker, Inc., New York.
25. Israelachvili, J. (1992). In Intermolecular and surface forces. Academic Press,
London, pp. 246.
26. Israelachvili, J. (1992). Surface Sci. Rep., 14,113.
27. Salemme, F. R., Genieser, L., Finzel, B. C., Hilmer, R. M., and Wendoloski, J. J.
(1988). J. Cryst. Growth, 90, 273.
28. Fredericq, E. and Neurath, J. (1950). Am. Chem. Soc., 72,2684.
29. S0rensen, S. P. L. and H0yrup, M. (1915-17). Compt. Rend. Trav. Lab. Carlsberg,
12, 312.
30. Green, A. A. (1931). J. Biol. Chem., 93, 517.
31. Gronwall, A. (1942). Compt. Rend. Trav. Lab. Carlsberg, 24, 8.
32. Tanford, C. and Wagner, M. (1954). J. Am. Chem. Soc., 76, 3331.
33. Mikol, V. and Giege, R. (1989). J. Cryst. Growth, 97, 324.
34. Green, A. A. (1932). J. Biol. Chem., 95, 47.
35. Melander, W. and Horvath, C. (1977). Arch. Biochem. Biophys., 183, 200.
36. Arakawa, T. and Timasheff, S. N. (1985). In Methods in enzymology (ed. S. P.
Colowick and N. O. Kaplan), Academic Press, London. Vol. 114, p. 49.
37. Timasheff, S. N. and Arakawa, T. (1988). J. Cryst. Growth, 90, 39.
38. Hofmeister, F. (1888). Arch. Exp. Pathol. Pharmakol, (Leipzig) 24, 247.
39. Collins, K. D. and Washabaugh, M. W. (1985). Q. Rev. Biophys., 18, 323.
40. Cacace, M. G., Landau, E. M., and Ramsden, J. J. (1997). Q. Rev. Biophys., 30,
241.
41. Guilloteau, J.-P., Ries-Kautt, M., and Ducruix, A. (1992). J. Cryst. Growth, 122,
223.
42. Veesler, S., Lafont, S., Marcq, S., Astier, J. P., and Boistelle, R. (1996). J. Cryst.
Growth, 168, 124.
43. Lafont, S. (1996). These de I'Universite d'Aix-Marseille III.
44. Guilloteau, J. P., Fromage, N., Ries-Kautt, M., Reboul, S., Bocquet, D., Dubois,
H., etal. (1996). Proteins: Structure, Function, Genetics, 25,112.
45. Menez, R. and Ducruix, A. (1993). J Mol. Biol., 232,997.
46. Diller, T. C., Shaw, A., Stura, E. A., Vacquier, V. D., and Stout, C. D. (1994). Acta
Oysr.,D50,620.
47. Leavis, P. C. and Rothstein, F. (1974). Arch. Biochem. Biophys., 161, 671.
48. Ries-Kautt, M. and Ducruix, A. (1991). J. Cryst. Growth, 110, 20.
49. Saludjian, P., Prange, T., Navaza, J., Guilloteau, J.-P., Ries-Kautt, M., Menez, R.,
et al. (1992). Acta Cryst., B48,520.
50. Howell, P. L. (1995). Acta Cryst., D51, 654.
51. Ries-Kautt, M., Ducruix, A., and Van Dorsselaer, A. (1994). Acta Cryst., D50,
366.
52. Lafont, S., Veesler, S., Astier, J.-P., and Boistelle, R. (1994). J. Cryst. Growth, 143,
249.
53. Howard, S. B., Twigg, P. J., Baird, J. K., and Meehan, E. J. (1988). J. Cryst.
Growth, 90,94.
311
 M. Ries-Kautt and A. Ducruix
54. Alderton, G. and Fevold, H. L. (1946). J. Biol. Chem., 164,1.
55. Palmer, K. J. (1947). Struct. Rep., 11,729.
56. Steinrauf, L. K. (1959). Acta Cryst., 12, 77.
57. Jolles, P. and Berthou, J. (1972). FEBS Lett, 23,21.
58. Haas, D. J. (1967). Acta Cryst., 23,666.
59. Vuillard, L., Rabilloud, T., Leberman, R., Berther-Colominas, C., and Cusack, S.
(1994). FEBS Lett., 353,294.
60. Elgersma, A. V., Ataka, M., and Katsura, T. (1992). J. Cryst. Growth, 122,31.

312
 11
Diagnostic of pre-nucleation and
nucleation by spectroscopic
methods and background on the
physics of crystal growth
S. VEESLER and R. BOISTELLE

1. Introduction
Unlike the crystallization of small inorganic molecules, the problem of protein
crystallization was first approached by trial and error methods without any
theoretical background. A physico-chemical approach was chosen because
crystallographers and biochemists needed criteria to rationally select crystal-
lization conditions. In fact, the problem of the production of homogeneous
and structurally perfect protein crystals is set the same as the production of
high-quality crystals for opto-electronic applications, because, in both cases,
the crystal growth mechanisms are the same. Biological macromolecules and
small organic molecules follow the same rules concerning crystallization even
if each material exhibits specific characteristics.
This chapter introduces the fundamentals of crystallization: super-
saturation, nucleation, and crystal growth mechanisms. Phase diagrams are
presented in Chapter 10. Special attention will be paid to the behaviour of the
macromolecules in solution and to the techniques used for their analysis: light
scattering (LS), small angle X-ray scattering (SAXS), small angle neutron
scattering (SANS), and osmotic pressure (OP).

2. Concentration and supersaturation

Before obtaining any nucleation or growth, it is necessary to dissolve the
biological macromolecules under consideration in some good solvent. How-
ever, it may immediately be asked whether a good solvent is a solvent in
which the material is highly soluble, or in which nucleation is easily con-
trolled, or in which growth is fast, or solvent in which the crystals exhibit the
appropriate morphology. In practice, the choice of the solvent often depends
 S. Veesler and R. Boistelle
on the nature of the material to be dissolved, taking into account the well
known rule which says that 'like dissolves like'. This means that, for dis-
solution to occur, it is necessary that the solute and the solvent exchange
bonds: between an ion and a dipole, a dipole and another dipole, hydrogen
bonds, and/or Van der Waals bonds. Therefore, the nature of the bonds
depends on both the nature of the solute and the solvent which can be dipolar
protic, dipolar aprotic, or completely apolar.
Once the material has dissolved, the solution must be supersaturated in
order to observe nucleation or growth. The solution is supersaturated when
the solute concentration exceeds its solubility. There are several ways to
achieve supersaturation. The simplest is to partly evaporate the solvent, the
drawback being that all species in the solution (salts, impurities) concentrate
as well. This is the case for hanging drop, but, for a better control of growth, it
is more advisable to cool or heat the solution depending on whether the
solubility decreases with decreasing temperature or conversely. However this
method is not recommended when the temperature dependence of solubility
is too low. Besides, supersaturation can also be achieved by pH variation,
chemical reaction, addition of a poor solvent or a precipitant, and so on.
However, the evolution of the system is often more difficult to control with
these latter methods.
Supersaturation is the driving force for nucleation and growth. From a
thermodynamical point of view, it is the difference between the chemical
potential of the solute molecules in the supersaturated (u,) and saturated (us
states respectively. For one molecule which will crystallize one has:

where kB is the Boltzmann constant, T the absolute temperature, and

where C and Cs are the actual concentration and the saturation concentration,
i.e. the solubility, respectively. This ratio is dimensionless but its value
depends somewhat on the concentration units (g/litre, mol/litre, mol fraction,
activities, and so on). For protein crystallization, the concentrations are
mostly expressed as mg/ml, i.e. g/litre, which is the easiest way but probably
not the best for explaining crystallization kinetics. Since activities of proteins
cannot yet be calculated, molar fractions are the more appropriate units.
Unfortunately, due to the complexity of the protein solutions, they are seldom
used.
For the sake of simplicity, supersaturation is mostly defined as B, the ratio
defined in Equation 2, or as another dimensionless ratio a = B - 1:

As we will see in the sequel, several growth rate equations contain the term
InfJ included in Equation 1. Traditionally, for the growth of crystals made
314
 11: Diagnostic of prenucleation
from small molecules, ln(3 is approximated by a. This is permitted because
supersaturations are often low or very low. For a supersaturation of 5%, i.e. p
= 1.05, we haveCT= 0.05 and lnB ~ 0.049. However, proteins often nucleate
and grow from highly supersaturated solutions, e.g. at supersaturation (3 = 3,
we have ln(3 ~ 1.099, whereas a = 2. It is self-evident that in that case
replacing lnB by a should be avoided. In general, lnB is significantly different
from a as soon as B exceeds about 1.5 which is a very low value for protein
crystallization.
It is also worth noting that supersaturation is sometimes defined as the
difference C - Cs. In this case, its value drastically depends on the con-
centration units. The difference C - Cs = 100 mg/ml, for example, reduces to
about 1 X 10-2 if the concentrations are expressed as mol fractions, for a
molar weight of the solute Mw = 10000. In general, it is more suitable to use
B or cr to solve the nucleation or growth rate equations. However this may
conceal the specific influence of the concentration on crystallization. As an
example, let us consider the case for which the solubility of the protein
decreases when increasing the concentration of the crystallization agent, salt,
or poor solvent. Thereby, in the case of BPTI in NaCl solutions (1), a
supersaturation of twice the solubility, P = 2, can be achieved in the area of
the solubility diagram where solubility is large (44 mg/ml in 1.4 M NaCl
solutions at 25 °C) or low (3 mg/ml in 2.3 M NaCl solutions at 25 °C). In the
former case the mass of solute which will be deposited is 44 mg/ml whereas in
the latter case it is only 3 mg/ml. Despite the same B value, nucleation and
growth will be favoured in the former case.

3. Nucleation
When a solution is supersaturated, the solid phase forms more or less rapidly
depending on the conditions: concentration of solute, crystallization agent,
pH, supersaturation, temperature, nature and concentration of impurities,
stirring, presence of solid particles. Primary nucleation occurs in a solution
that is clear, without crystals. It is called homogeneous nucleation if the nuclei
form in the bulk of the solution. On the other hand, it is called heterogeneous
if the nuclei preferentially form on substrates such as the wall of the
crystallizer, the stirrer, or solid particles (dust particles, and so on). Secondary
nucleation which is induced by the presence of already existing crystals is less
frequent during protein crystallization because the crystallizers are rarely
equipped with stirrers which generate attrition, shear at the crystal surface.

3.1 Nucleation rate

From a theoretical point of view (2-4) nucleation is considered as an addition
of monomers to clusters made of a few molecules called i-mers. When the
system is in a steady state, the rate of formation of an i-mer is equal to its rate
315
 S. Veesler and R. Boistelle
of disappearance. Ifks/i-1_iis the rate constant for subtracting a monomer from a
(i + l)-mer and ka/i-1is the rate constant for adding a monomer to a (i - l)-mer,
one has:

where N1 and Ni are the concentrations of monomers and i-mer, respectively.

The flux of clusters going from a lower class into an upper class is:

This is the steady state rate of nucleation, J being a number of clusters per
unit time and unit volume of solution. It is further assumed that, at
equilibrium between the size classes, the rates at which a monomer leaves or
sticks on a cluster are equal, so that:

As it will be seen hereafter the small clusters turn into stable nuclei only if
they contain a critical number of monomers. Accordingly, the nucleation rate
J is mainly dependent on the class sizes around i*. It is the product of the
nuclei concentration times the frequency at which they exceed the critical
number i* by addition of a monomer. It expresses as:

where AG* is the activation free energy for forming a nucleus of critical size
and Z the so-called Zeldovich factor:

In solution Z ~ 1 X 10 2 and the pre-exponential term typically ranges from

102 to 1020 nuclei per cubic centimetre and second. Usually Equation 7 is
rewritten as:

where d is the frequency at which the critical size becomes supercritical

allowing the nucleus to grow and turn into a crystal.

3.2 Activation free energy for homogeneous nucleation

In order to solve Equation 9, it is necessary to precisely calculate or estimate
the activation free energy AG*. Since the solute concentration is the same in
the whole bulk, nucleation occurs if there are energy fluctuations, somewhere
in the solution, around the mean value imposed by the supersaturation. To
create a nucleus it is necessary to create a volume and a surface. Assuming
316
 11: Diagnostic of prenucleation
that the nucleus is limited by only one type of face, the activation free energy
for homogeneous nucleation is:

where i is the number of molecules in the nucleus, Aj the area of the nucleus,
and 7J its interfacial free energy with respect to the solution. The first term
represents the energy to create the volume whereas the second term is the
excess energy to create the surface. To simplify the demonstration we can also
suppose that the nucleus is a sphere so that:

At equilibrium, when dAG/jr = 0, the nucleus has the critical radius r*, as
shown in Figure 1,

where V is the volume of a molecule.

Inserting Equation 12 into Equation 11 yields:

which can also be written as:

Figure 1. Variation of the activation free energy for three-dimensional nucleation versus
nucleus size.
317
 S. Veesler and R. Boistelle
The critical activation free energy for creating the nucleus with critical radius
r* is one-third of the energy required for creating its surface. As shown in
Figure 7, at the critical size r*, the nucleus is in a very labile equilibrium. If it
gains one molecule so that r > r* it grows. But if it loses one molecules so that
r < r*, then it spontaneously dissolves. In both cases there is a gain in energy.
Inserting Equation 13 or Equation 14 into Equation 9 allows for the
calculation of the nucleation rate if nucleation is homogeneous.

3.3 Activation free energy for heterogeneous nucleation

Heterogeneous nucleation often occurs prior to homogeneous nucleation
especially when superstauration is low. However, this implies that the solute
molecules have some affinity for the substrate onto which they stick. Here
also, it is convenient to consider that the nucleus is a sphere, actually cap-
shaped, making the contact angle a with the substrate (Figure 2). Three
surface free energies are involved in heterogeneous nucleation: -yj between
the nucleus and the solution, -ya between the nucleus and the substrate, and y0
between the substrate and the solution. They are related by Young's
equation:

If we name St the area of the nucleus and Sa the area of the interface between
the nucleus and the substrate, the activation free energy for heterogeneous
nucleation is:

Taking Equation 15 into account, AGj,et becomes:

Figure 2. Cap-shaped nucleus forming by heterogeneous nucleation on a substrate.

318
 11: Diagnostic of prenucleation
At equilibrium, when 8AGhet /8r = 0, the radius of the critical nucleus is:

The critical radius of the nucleus formed by heterogeneous nucleation is the

same as for homogeneous nucleation. However, the cap-shaped nucleus
contains fewer molecules than does the full sphere. Inserting Equation 18 into
Equation J 7 yields:

which is the product of AG* for homogeneous nucleation times a term depend-
ing on the contact angle. It is worth noting that for a = 180°, AG het = AG*.
The substrate does not have any effect on nucleation. For a = 90°, AGhhet =
AG*/2. If a tends toward zero, then AG*het tends also toward 0. That means
that the substrate induces nucleation even at very low supersaturation since
less and less energy is required to form the nucleus. The nucleation rate
Equation 9 drastically increases when the contact angle a decreases and
subsequently the activation free energy for nucleation.

3.4 Examples
Let us first imagine a system for which nucleation of small molecules is
homogeneous. To solve Equation 9 we suppose that, in Equation 13, each
molecule occupies a volume V = (5 X 5 x 5) X 10-24 cm3. If the solubility is
rather high (typically 10-50 g/litre) then the interfacial free energy, -ft, is
rather low, e.g. 10 erg cm"2 (10 mJ m-2). Typically in Equation 9 one has uN1 =
1020 cm-3 s-1. Inserting T = 293 K and kB = 1.38 X 10-16 erg/K yields:

In order to have a nucleation rate J = 1 nucleus cm 3 s1, the super-

saturation 3 has to be ~ 1.34, what is a rather low value. But, if (3 were only 1.2,
then the nucleation rate would be catastrophically low (1.8 X 10~32 cm"3 S"1).
On the other hand, if (3 were 2.0, then the nucleation rate would be drastically
high (2.6 X 1016 cm3 s'1). This demonstrates that nucleation is highly super-
saturation-dependent. In the low supersaturation range the solution remains
metastable over a long period of time, whereas in the high supersaturation
range nucleation occurs spontaneously. That means, for protein crystal-
lization, that a condition which leads to a large number of crystals can be
improved by decreasing supersaturation.
The second parameter which greatly influences the nucleation rate is the
surface free energy -yt of the nucleus. If the value of yl is increased up to 20 or
50 erg cnr2, all other parameters being unchanged in the previous example,
then J = 1 nucleus cnr3 s-1 for (3 ~ 2.29 or 26.5, respectively. The metastable
319
 S. Veesler and R. Boistelle
zone, where no nucleation occurs after a reasonable time lag, drastically
widens out with increasing the interfacial free energy. In that case, the only
way to avoid it is to change the solvent or the solution composition. The rule
is 'the higher the solubility, the better the affinity of the solvent for the
nucleus and the lower the surface free energy'. All other things being
unchanged, a better solvent gives rise to higher solubilities and smaller
surface free energies, both these facts contributing to increase the nucleation
rate.
Considering again Equations 9 and 13, it can be seen that increasing the
volume V of the molecule has the same effect on the nucleation rate as
increasing the surface free energy. Accordingly, macromolecules should
nucleate with a lower rate than small molecules if all other parameters are
unchanged. Assuming that the volume of the macromolecule is (50 X 50 X
50) X 10-24 cm3, instead of (5 X 5 X 5) X 10-24 cm3, then Equation 20
becomes:

In that case a very large supersaturation is needed to obtain 1 nucleus cm-3 s-1.
In fact B equals 7718 which is completely unrealistic! However it is a good
illustration of the difficulty often encountered for nucleating proteins. Further-
more, it is difficult to consider that the kinetic coefficient, taken as 1020 cm-3 s-1
in the previous example, would be greater for macromolecules than for small
molecules. Hence, the only way to obtain reasonable values of J and of (3 is to
assume that the surface free energy of a protein crystal is significantly lower
than the energies usually encountered for crystals of small molecules. If to
calculate Equation 21, yi = 1 erg cm-2 is used instead of 10 erg cm-2 then the
J values are given by Equation 20 and subsequently the same low p values
are obtained. Consequently lower surface free energies can compensate
higher molecular volumes in the case of protein crystallization. Values of
•yj = 0.5-0.7 erg cm"2 for thaumatin (Mw = 22000 Da) were observed by
Malkin et al. (5).
As a concluding remark, it should be emphasized that there is no special
reason that, a priori, all proteins should have a low or very low surface free
energy. If the general rule holds for proteins, it might be, then for sparingly
soluble proteins the surface free energies are relatively high. But, these
proteins also nucleate, sometimes even after rather short induction periods.
The only explanation would be that nucleation is heterogeneous. As a matter
of fact, it cannot be homogeneous because the required supersaturation
would be much too high. The existence of heterogeneous nucleation can be
checked with a very well known model protein, i.e. with hen egg white
lysozyme (HEWL). With p values of ~ 3, the solution deposits several tens of
crystallites per cubic centimetre of solution within a few hours if the solution
is not carefully filtered. On the other hand, it deposits sometimes only one or
320
 11: Diagnostic of prenucleation
two crystals after one or two days, if most of the solid particles are removed
by filtration.
Finally, in order to estimate the nucleation rate and the interfacial free
energy -yj, it is possible to measure the time lag for nucleation, or induction
time tj, as a function of different supersaturations. Assuming that after the
time tj J = 1 cm'3 s"1 on has:

so that:

Plotting Intj versus l/ln2p should give a straight line, the slope of which is
proportional to yl which is the only unknown in the term on the right side of
Equation 23. This method was often used for determining the •/! values of
crystals of small molecules. Due to the uncertainties of the measurements of t;
it only gives a good order of magnitude.

4. Pre-nucleation — investigation of the solution

The main questions addressed in this section are:
• what is the behaviour of the solution before nucleation occurs?
• why do some solutions lead to precipitate whereas others yield crystals?
From the classical nucleation theory, it is known that crystals form prefer-
entially instead of precipitates when the first clusters (built up of several
molecules) exhibit some crystalline arrangement. Therefore, the very first
stage of nucleation, called pre-nucleation, is very important. Accordingly, the
systematic investigation of the solutions, in both under- and supersaturated
state, is essential to understand and control the crystallization process of
proteins.

4.1 Methods
Depending on the techniques used, different information can be obtained on
the solutions. Since we only give a brief survey of these techniques, the reader
can refer to the different monographs where they are widely described. The
references are given hereafter. In this section, solution scattering and osmotic
pressure techniques will be presented. These techniques aim at obtaining
information on molecules in solution: molecular weight, size, aggregation
states, polydispersity, and interactions. As will be seen in Section 4.3 probing
the protein interactions is very important in the field of protein crystallization.
In dilute solution interactions include excluded volume term, repulsive
321
 S. Veesler and R. Boistelle
electrostatic term, and attractive Van der Waals term, the last two terms are
described in the so-called DLVO theory of colloidal stability (6).
4.1.1 Solution scattering techniques
For all scattering experiments in solution the principle is the same: a mono-
chromatic beam of visible light, X-rays, or neutrons impinges on the protein
molecules and induces an oscillating polarization of their electrons. The
molecules then serve as a secondary source which is radiated and scattered.
For neutrons the interaction with the matter is different, because neutrons are
scattered by the atomic nuclei, the scattered intensity depending on the
scattering length density.
i. Light scattering (7, 8)
Depending on the way the data are analysed, two types of experiments are
possible: elastic or static light scattering (SLS) and quasi-elastic scattering or
dynamic light scattering (DLS). The experimental set-up is shown in Figure 3.
Static light scattering (SLS)
The experiment consists in measuring the photons intensity scattered by the
solution at different angles and for different concentrations. Let us consider
two cases depending on the size of the protein with respect to the wavelength
of the light.
(a) The light is scattered by spherical particles which are small compared to
the wavelength of the light, d < X/10, where d is the characteristic size of
the protein and the wavelength of the light (400 < X < 650 nm). In that
case the particle is assumed to be a punctual source of light, and the
intensity scattered is independent of the angle. The experiments are
usually carried out at 90°, then:

where K is a constant, dn/dC the increment of refractive index with pro-

tein concentration C, Iref the intensity scattered by a reference at 90°, and

Figure 3. A schematic representation of the LS experiment.

322
 11: Diagnostic of prenucleation
AI the difference between the intensity scattered by the solvent and the
solution at 90°.
Information obtained: Mw (the molecular weight of the particle) and A2
(the second virial coefficient), also noted B22 in the literature. The sign of
the second virial coefficient is indicative of the type of interactions: it is
negative when the interactions between molecules are attractive and
positive if the interactions are repulsive.
(b) The light is scattered by larger particles, A/10 < d < X; in that case the
particle is assumed to be a multiple source of the light so that there is a
phase difference between the light scattered by different portions of the
particle at any time, the intensity scattered is angular dependent, and then:

where AI(6) is the difference between the intensity scattered by the

solvent and the solution at angle 0.
Information obtained: Mw, A2, and RG (the radius of gyration of the
particle).
Dynamic light scattering (DLS)
The experiments consist in measuring the fluctuation of the scattered intensity
at constant scattering angle but at different concentrations. These fluctuations
are correlated to the diffusion (translational and rotational) of the particles in
solution. In the following, it is assumed that the molecules are spherical and
d < A/10 so that rotational diffusion and internal motion of flexible macro-
molecules can be neglected.
Two cases must be considered:
(a) In the absence of interaction, there is no concentration dependence of the
measured coefficient diffusion.
Information obtained: D0 (the free particle diffusion coefficient), Rh (the
hydrodynamic radius of the particle), and v (the quality factor or the
polydispersity index which expresses the dispersion of the sizes of the
particles in solution).
DO is related to the hydrodynamic radius by means of the
Stokes-Einstein equation:

where kB is the Boltzmann constant, T the absolute temperature, and n0

the solution viscosity.
(b) In the presence of interactions, there is a concentration dependence of
the measured diffusion coefficient.
Information obtained: Deff (the effective diffusion coefficient of the particle)
and KD (the interaction parameter with Deff = D0(l + KD C), C being the
323
 S. Veesler and R. Boistelle
protein concentration). KD is negative when the interactions are attractive
and positive when the interactions are repulsive.
ii. Small angle X-ray and neutrons scattering (9-12)
The experiments consist in obtaining the angular distribution of the X-ray or
neutrons intensity scattered by the solution. There are also two cases depend-
ing on the interactions:
(a) In the absence of interactions, the intensity scattered is the sum of the
scattering of the individual particles, namely, the form factor, and is given
by the Guinier's law.
Information obtained: Mw, RG, and the form factor.
(b) In the presence of interactions, the intensity scattered is the product of
the form factor and the structure factor, the interference term related to
particle distribution.
Information obtained: A2 and the structure factor.
4.1.2 Osmotic pressure techniques (13, 14)
In an osmotic pressure experiment, the solvent and the protein solution are
separated by a semi-permeable membrane. The excess pressure due to the
difference in the chemical potentials of the two solutions creates a flux of
solvent through the membrane. The osmotic pressure (IT) is therefore pro-
portional to the number of particles in solution and is often expanded as a
series of virial coefficients.

where C is the protein concentration, R the molar gas constant, 8.31 107 erg
mor1 K-1, and T the absolute temperature.
Information obtained: Mw and A2-

4.2 Practical recommendations

4.2.1 Generals
All the above methods (listed in Table 1) often require investigation of
several solutions at different concentrations in order to precisely deduce the

Table 1. Comparison of the different methods

Method Information Domain-range Wavelength

SLS M W ,A 2 d < 50 nm A = 500 nm
SLS Mw, A2, and RG 50 d< 500 nm A = 500 nm
DLS Deff, D0, Rh, v, and KD 1 <d < 1000 nm A = 500 nm
SAXS Mw, A2, and RG 1 < d< 100 nm 0.1 <A.< 0.5 nm
SANS Mw, A2, and RG 1 <d< 100 nm 0.2<X<2nm
OP Mw ,A 2 5000<M w <10 6 Da

324
 11: Diagnostic of prenucleation
concentration dependence of the measured parameter. Furthermore, the
experimenter has to check whether the particles are all the same in the whole
concentration range under consideration.
4.2.2 LS
The volume of solution required for a measurement is about 100-300 ul. A
typical experimental protocol is described in Protocol 1. Prior to the experi-
ment it is necessary to control several points:
(a) Absence of fluctuation in the laser intensity.
(b) Absence of parasitic light due to reflections or refractions. These con-
ditions are difficult to obtain in experiments carried out at angles below
30° with a commercial set-up.
(c) Absence of dust, air bubbles, glass particles, and other foreign tiny
materials in solution.
(d) Measurement achieved at the proper angle.
(e) Good transparency of the solution is required in order to avoid the
multiple diffusion, if needed a dilution must be done.
Therefore, before any experiment it is necessary to check the laser quality,
the optical trajectory, and to filter and/or centrifuge the solution. To treat the
signal it is also essential to know the refractive index and viscosity of the
crystallization medium (buffer + crystallization agent). In addition, for SLS
experiments, it is essential to know the increment of the refractive index of
the solution as a function of the protein cpncentration.

4.2.3 SAXS-SANS
These experiments can only be carried out using a synchrotron radiation or a
nuclear reactor. Runs are always allocated by a program committee to which
the application must be submitted. Since the number of runs is limited, it is
recommended to test the sample quality before using one of these techniques.
For instance DLS is a good tool for checking whether the molecules are
aggregated or not. If the polydispersity is high the SAXS and SANS will fail.
(a) SAXS: the volume of solution required for one measurement is ~ 100 ul.
The electron density of the crystallization agent should be as low as
possible, otherwise the signal due to the particles which is under investi-
gation disappears in the background. Thus electron-rich buffer at high
concentration (e.g. ammonium sulfate) should be avoided.
(b) SANS: the volume of solution required for one measurement is about
150 ul. In these experiments it is often necessary to dissolve the protein
in D2O solutions. Moreover, Broutin et al. (15) and Gripon et al. (16)
recently showed a shift of the solubility of lysozyme when H2O is replaced
by D2O. Furthermore, D2O affects the interactions between the particles
325
 S. Veesler and R. Boistelle
in solution, and care must be taken when working with materials which
have a tendency to aggregate. Accordingly, the experimenter has to check
the aggregation behaviour of the protein solution before any experiment.

4.2.4 OP
The volume of solution required for a measurement is about 120 ul. Actually,
three measurements performed with 40 ul are necessary.
(a) The main point of this experiment concerns the equilibration on both
sides of the membrane. The higher the salt concentration is, the more
difficult the equilibration is. Practically the upper concentration limit is
about 500 mM.
(b) High viscosities solutions can generate very long equilibrium time.
(c) It is important to check the temperature and the pH stability.

4.3 Examples
DLS is the most widely used method for the characterization of protein
solutions and was first proposed as a diagnostic tool for protein crystallization
by Kam et al. (17). Zulauf et al. (18) studied 15 proteins, in dilute solutions and
in the absence of crystallization agent, and suggested that the detection of
aggregates indicates that crystallization will not be successful. Ferre-D'Amare
et al. (19) have determined the crystallizability of three different RNAs by
DLS with this criterion. In addition, more recent studies showed the absence
of large molecular aggregates in supersaturated solutions for different proteins
(1, 20, 21), contrary to Georgalis et al. (22, 23) who observed the formation of

Figure 4. Polydispersities (%) measured by concentrating or diluting a-amylase solutions

with respect to the solubility curves of the A and B polymorphs (15).

326
 11: Diagnostic of prenucleation
praggs (precipitating aggregates) or craggs (crystallizing aggregates) depend-
ing on the conditions.
Measurements of A2 (or B22) by SLS for different proteins in different sol-
vents have shown that conditions which promote crystallization are grouped
within a narrow range of A2 values (24, 25). Moreover different studies have
shown a systematic relationship between solubility and second virial co-
efficient (25). Recently Bonnete et al. (26) showed the complementary nature
of DLS, SAXS, and OP in order to probe interaction in protein solution.
Rather than reviewing all the outputs for crystallogenesis of the above
methods, we discuss here an example of the application of DLS with the
associated experimental protocol, and another example dealing with an
application of SAXS.
4.3.1 DLS study applied to porcine pancreatic a-amylase (20)
In this example special attention was paid to the polydispersity of under- and
supersaturated solutions of a-amylase. The results are presented in Figure 4,
and can be summarized as follows: polydispersity is very high (v > 10%) when
the protein concentration is much lower than solubility whereas it is very low
(v < 10%) when the protein concentration is nearly equal or even slightly
higher than solubility. Even more important, monodispersity is a prerequisite
for obtaining good crystals. Some polydispersity seems to be acceptable if
there are no large aggregates in solution.

Protocol 1. DLS experiment

Equipment and reagents

• Protein in solution • A high performance light scattering
• A solution containing the buffer plus the apparatus
crystallization agent

Method
1. Prepare a solution of 300 ul of the protein at the desired concentration.
2. Filter three times the solution with the same LCR13 (Millipore) filter
and pour the solution in a glass cell. Dust can also be removed by
centrifugation (20000-30000 g, for 1-2 h).
3. Put the glass cell in the sample holder.
4. Switch on the laser and check the optical trajectory. Avoid continual
illumination of the solution for several hours because of potential
protein denaturation.
5. Run the analysis for 1-3 min; the correlator receives the signal from
the detector (photomultiplier).
6. Analyse the data: the cumulant method (27) directly gives the diffusion
coefficient and the polydispersity.

327
 S. Veesler and R. Boistelle

Figure 5. X-ray scattering curves recorded at 100 mg/ml lysozyme concentration in acetate
buffer plus added salt. Anion series as indicated in the figure with Na+ as counterion,
ionic strength = 0.150 (17).

4.3.2 Lysozyme interactions as seen by SAXS (28)

In this example SAXS is used to characterize the influence of various salts on
the protein-protein interactions in undersaturated lysozyme solutions at
constant pH and temperature. Attractive and repulsive interactions respect-
ively result in an increase or decrease of the structure factors at low scatter-
ing angles. Practically, as shown in Figure 5, when the scattering intensity at
low angle is increased, the protein-protein interactions are moved toward
attraction. Hence, the addition of different salts to lysozyme solutions results
in changes from repulsion to attraction. Interestingly anions can be ordered
according to their effectiveness to create attractive interactions and follow the
reverse order of the Hofmeister series (see Chapter 10).

4.4 Practical considerations

The practical application of the studies on pre-nucleation is to diagnose
whether a protein solution will deposit crystals or not. It is noteworthy that
this ability should be studied under crystallization conditions, in both under-
and supersaturated states. This implies that the phase diagram is known. As a
general rule it seems that the solution monodispersity is a prerequisite to
crystallization. Large molecular aggregates hinder growth especially if they do
not dissociate into small aggregates or monomers once growth proceeds.
Growth is also favoured by the occurrence of attractive interactions in the
solution due to the addition of a crystallization agent. Finally there is a clear
correlation between the occurrence of attractive protein interactions and the
decrease of protein solubility.
328
 11: Diagnostic of prenucleation
From a practical point of view polydispersity of the protein solution can
only be studied by DLS whereas the molecular interactions can be observed
by LS, SAXS, SANS, and OP. Accordingly, LS is a powerful and interesting
tool for knowing some important solution characteristics, and can be easily
carried out at the laboratory scale.

5. Crystal growth
When a nucleus grows and transforms into a crystal, the different faces of the
growing crystal exhibit growth mechanisms and rates that depend on external
factors (supersaturation, impurities, temperature, and so on) and internal
factors (structure, bonds, defects, and so on). According to the periodic bond
chain (PBC) theory (29-32) there are three types of crystal faces (Figure 6).
• F (flat) faces: they contain at least two PBCs in the slice of thickness dhkl,
where dhkl is the interplanar distance of the face (hkl).
• S (stepped) faces: they contain only one PBC in the slice dhkl.
• K (kinked) faces: they do not contain any PBC in the slice dhkl.
Let us just recall that a PBC is an interrupted chain of strong bonds running
along a crystallographic direction in the crystal. Since all sites on the K faces
are growth sites, more commonly called kinks, the K faces grow by direct
incorporation of the growth units which hit them. The growth rate is high and,
normally, these faces do not occur on the crystal morphology, because the
growth form of the crystal is made up only of the faces which have the slowest
growth rate.
Conversely, the F faces are poor in kinks. They grow by lateral spreading of

Figure 6. Schematic representation of a crystal exhibiting flat (F), stepped (S), and kinked
(K) faces. The front face exhibits a polygonized growth spiral, whereas the top face
exhibits a two-dimensional nucleus.

329
 S. Veesler and R. Boistelle
the growth layers. For being integrated into the crystal, the solute molecules
must first adsorb on the surface, and later on diffuse toward the step of a
growth layer along which they migrate toward a kink. Such faces grow either
by a two-dimensional mechanism or a spiral growth mechanism (Figure 6).
Since the number of kinks is low, the growth rates are low too.
At last, the S faces are in an intermediate situation. Their growth rate is
lower than that of the K faces but higher than that of the F faces. Normally,
the S faces do not appear on the crystal morphology, except when their
growth rate is slowed down by adsorption of an impurity for example.
The growth mechanisms have been discussed in detail elsewhere (33).
Hereafter we summarize the general trends.

5.1 Growth controlled by surface processes

5.1.1 Growth by two-dimensional nucleation
This growth mechanism occurs when the crystal face is perfect, without any
defect. The molecules which adsorb, randomly diffuse on the surface,
encounter, coalesce into a two-dimensional nucleus which spreads across the
crystal face if its size exceeds a critical size. For a square-shaped 2D nucleus
the critical size (number of molecules n* in the nucleus) is:

where X is the so-called edge free energy expressed here per molecule in the
edge. In the mononuclear model, there is only one 2D nucleus which spreads
across the surface so that the growth rate R of this face is:

where d is the height of the growth layer, S the area of the face, and B2 the 2D
nucleation rate, i.e. the number of nuclei forming per unit time and unit area
(cm-2 s-1). B2 can be written:

where n1 is the number of growth units adsorbed per unit area of the face and
6 the frequency at which the 2D nucleus of critical size become supercritical
and grows. AG2* is the activation free energy for 2D nucleation:

Inserting Equation 31 into Equations 30 and 29 shows that the growth rate is
an exponential function of supersaturation. As in the case of three-
dimensional nucleation, there is a critical supersaturation below which the
growth rate is zero or nearly zero. A dead zone is observed at low super-
330
 11: Diagnostic of prenucleation
saturation when measuring R as a function of p. Once this critical super-
saturation is exceeded, the growth rate drastically increases with increasing
supersaturation. Growth is difficult to control.
If several nuclei spread at the same time across the crystal face, growth is
determined by a 2D polynuclear mechanism. In that case the expression for
the growth rate is somewhat more complicated (33-35).
5.1.2 Growth by a spiral mechanism
When a screw dislocation emerges on a crystal face, it generates a growth
spiral (Figures 6 and 7). Since the growth spiral is made of a parallel sequence
of steps, growth can take place even at low supersaturation since the growth
units which adsorb onto the crystal face easily find growth sites where they are
incorporated into the crystal. It can be seen (Figure 7} that the growth rate of
the face can schematically be written as:

where v is the lateral velocity of the steps, d their height, and y their
equidistance. If the spiral is circular one has:

where A. is also the edge free energy of the steps, a being the distance between
two molecules in the step. Considering Figure 7 and Equation 33, we see that
an F face which exhibits a growth spiral is really flat only if the super-
saturation is low (y large). Conversely, it takes a conical outline when the
supersaturation is high (y small) due to the high step density.
The theories of the spiral growth mechanism were extensively discussed
elsewhere (36-39). Here, we give only a few possibilities which can be derived
from the general growth rate equation that is not commented here. Depend-
ing on the influence of the different parameters, growth depends on surface
diffusion, kink integration kinetics, and so on. As an example, let us suppose

Figure 7. Profile of a face growing by a spiral growth mechanism.

331
 S. Veesler and R. Boistelle
that growth is controlled by surface diffusion of the growth units toward the
steps. Then we have two possibilities.
At low supersaturation:

At high supersaturation, due to the high step density:

In both equations, Ds is the surface diffusion coefficient and xs the mean free
path of diffusion; n1 is again the number of adsorbed growth units per unit
area, and V the volume of a molecule in the crystal.
Inserting Equation 33 into Equation 34 shows that R is proportional to a X
InB. With the approximation lnB = a the so-called primary quadratic growth
rate law is obtained at low supersaturation:

whereas at high supersaturation the primary linear growth rate law is obtained:

It is worth noting that Equation 36 is valid only if the supersaturation is very

low as discussed in Section 1. For protein crystals it is normally not allowed to
replace R = klalnB by R = k1a2 because of the high values of the super-
saturation.
Among the different processes predicted by the spiral growth mechanism,
there is one which directly depends on the kink integration kinetics. All
volume diffusion and surface diffusion processes are supposed to be fast with
respect to the kinetics at which the growth units enter into the kinks. The
growth units reach the growth sites but need some time to find the proper
conformation before being incorporated. This time is called relaxation time Tk
for entering into the kinks. It depends on the activation free energy for
entering into the kinks:

where v is a frequency often taken, in the vapour phase, as kBT/h ~ 6 X 1012 s -1

where h is Planck's constant. In solution v should have a lower value. If this
mechanism is rate determining, one has:

where a is the length of the elementary jump of the growth units which diffuse
toward the kinks. Once more, if we insert Equation 33 into Equation 39, and
332
 11: Diagnostic of prenucleation
with the assumption that lnB = a a secondary quadratic growth rate law is
obtained:

Such a mechanism is likely to occur with large molecules such as proteins

which have to reorient for being trapped into the growth sites. But the
approximation lnB = cr should be avoided.
When the diffusion of the growth units towards the crystal is slow with
respect to the surface processes, the theories (36-39) predict that growth is
controlled by volume diffusion. This happens especially in stagnant systems
where there is no solution flow. The basic equations describing the growth
rates are rather cumbersome, but after some assumptions, some of them take
a rather simple form. As an example, one has:

where n0 is the number of growth units per unit solution volume, Dv the
volume diffusion coefficient, and 8 the thickness of the boundary layer. This
equation is similar to that derived from Fick's law (33, 40):

In the latter equation C - Cs is expressed as a number of mol cm -3, and V as

cm3 mol-4.
For theoretical reasons, it is sometimes interesting to know whether growth
depends on volume diffusion or on surface kinetics. For doing this, the crystal
is placed in a flow system and the growth rates of the faces are measured as a
function of the flow velocity U of the solution. At least for crystals of small
molecules, it is always observed that the growth rates first increase with
increasing flow velocity up to a final value, a plateau, where they become
independent of the flow velocity. Thus in the former case R is controlled by
volume diffusion, whereas in the latter case it is controlled by surface
processes. With growth rates, respectively as follows:

where C, Q, and Cs refer to the solute concentrations in the bulk of the

solution, at the crystal-solution interface and at saturation, respectively. Kv
and Ks are kinetic coefficients which depend on temperature, solvent,
solubility, and so on. When both rates are equal:

333
 S. Veesler and R. Boistelle
Plotting R/U1/2 versus R1/n for all curves obtained at different supersaturations,
provides the highest possible growth rates (for U = oo) by extrapolating the
straight lines thus obtained to R/U1/2 = 0.
As concerns proteins crystallization, it is practically always carried out in
stagnant systems. The reason for this is the missing of instrumentation, but
perhaps also the fragility of proteins. Moreover, Dv of proteins are two order
of magnitude smaller than the ones of small molecules. Accordingly, the
growth rates are mainly controlled by volume diffusion (Equations 39, 42) or
by the kink integration kinetics (Equation 39).

5.2 Kinetic measurements

Before leaving growth kinetics, we must emphasize that growth rate curves
must be determined at constant temperature if one wants to speculate on the
growth rate laws and growth processes. This means that supersaturation must
be varied by changing the solute concentration. As a matter of fact the growth
rate equations show that there are many parameters which depend on
activation energies, i.e. on temperature (volume and surface diffusion,
integration into the kinks, desolvation of the surface of the growth units, and
so on). It is therefore not surprising that, at constant supersaturation, growth
can drastically increase with increasing temperature. It is often not really
important if there is a variation of ± 1 °C around the mean crystallization
temperature, but large gaps are prohibited. The second reason for which it is
necessary to work at constant temperature is that misleading results and
interpretations of the growth rate curves can be given. Changing the
temperature is easier than changing the solution concentration in order to
change supersaturation. Unfortunately, there are conflicting effects between
increasing supersaturation and decreasing temperature. If, for instance,
solubility decreases with decreasing temperature, cooling the solution results
in an increase of supersaturation, but at the same time in a decrease of the
diffusion coefficients, of the mean free paths for diffusion and in an increase
of all relaxation times involved in the growth processes. This is the reason why
the growth rate first increases with cooling down the solution, but later on
passes through a maximum before drastically decreasing. A conclusion drawn
from such a curve are therefore subject to criticism. The drastic effect of the
temperature on the growth rate on protein crystal was pointed out earlier (41)
for porcine pancreatic a-amylase crystallization. In Figure 8 it clearly appears
that the growth rate of the porcine pancreatic a-amylase crystals is extremely
temperature-dependent above 18 °C.
The most current observation on protein crystallization and growth
measurements are made by optical microscopy. For instance, Boistelle et al.
(41) measured the growth rate of the porcine pancreatic a-amylase as a
function of supersaturation at 18 °C. The principle of the experiment was the
following. Once the solution was at the right temperature and supersaturation
334
 11: Diagnostic of prenucleation

Figure 8. Growth rate of polymorphs A1 and BII of porcine pancreatic a-amylase, at

constant supersaturation, versus crystallization temperature (37).

a few seed crystals were introduced in the crystallization cell and the displace-
ment of the faces was recorded as a function of time. A linear dependence
with supersaturation was obtained and growth was interpreted as a process
controlled by volume diffusion. On a more microscopic scale, direct measure-
ments of the step velocity are possible using laser Michelson interferometry
(42) or in situ atomic force microscopy (5). Such measurements are especially
interesting if the step equidistance can be related to supersaturation in order
to deduce the edge free energy of the step of the growth layer (Equation 33).
Knowing the step velocity, also allows a better estimation of the parameters
involved in the growth rate equations (mean free path for diffusion, surface
diffusion coefficient, relaxation time for entering into the kinks, and so on).

6. Crystallization in the presence of impurities and

additives
6.1 General trends
In the vocabulary of crystallization, the words impurities and additives play an
important role. The latter word concerns foreign substances voluntarily added
to the solution in order to obtain a special effect (inhibition of nucleation or
growth, habit change of the crystal, and so on). On the contrary, the former
335
 S. Veesler and R. Boistelle
word concerns foreign substances that exist in the solution, their source being
either the material which is going to crystallize, the solvent, the crystallization
agents, and so on. If impurities cannot always be avoided, it may be asked why
additives are put into the solutions. Actually, there are several reasons. Some
are biological in essence, for instance when adding a ligand to a protein, or
any compound that will interact specifically with the protein. Other reasons,
are of physico-chemical nature (Chapter 10). An additive properly chosen
allows a better control of nucleation and growth. It can also induce a modi-
fication of the crystal morphology, more commonly called habit change. In
some special cases, it induces phase transitions that are commonly observed
with protein crystals. However, from a general standpoint, additives should
also be considered as impurities as all other constituents of the solution.
The first impurities for the crystal are its own components. For small
molecules this is easy to understand; calcium carbonates for instance can be
grown in the presence of an excess of calcium or of carbonate. Nucleation,
growth, and habits will be different from those obtained in a strictly stoichio-
metric solution. As concerns the crystallization of biological macromolecules,
similar effects could be observed since the stoichiometry of the salts used as
crystallization agents can be widely changed. Since these salts are inside the
crystal structure, they can also be considered as constituents of the crystals.
The solvent is also an impurity. Sodium chloride grows as cubes from aqueous
solutions, but as octahedra from formamide solutions. Since organic solvents
are often used to induce the crystallization of proteins by decreasing their
solubility, they also influence the growth kinetics and the crystal habit. The
pH variations affect both the interfaces between crystal and solution, and the
nature and activities of the impurities. As an example glycine mainly exists as
zwitterions H3+N-CH2-COO~ in the pH range 3.5-8.5, but as positively
NH3+-CH2-COOH or negatively NH2-CH2-COO~ charged ions outside this
range. Adsorption of glycine on crystal faces, or interactions with other solute
species will obviously be affected by the charges of the molecules. Many
examples and interpretations concerning the different effects of impurities in
the crystallization of small molecules are found in the literature (43-45).
The matter has also been discussed in the macromolecule field (Chapter 2).
Here, the situation is even more complex, since salts (the most important
additives used in protein crystallization) interact with the solvation shell of
the macromolecules. In addition macromolecules can present conformational
and sequence heterogeneities which will affect the crystallization process (46).
6.2 Additives, phases, and polymorphs
The polymorphs of a same chemical compound have all the same composition
but different crystal structures, whereas different phases of a compound have
both different compositions and crystal structures. Stricto sensu, calcium
oxalates trihydrate, dihydrate, and monohydrate are three solid phases of
calcium oxalate. They should not be called polymorphs. On the other hand,
336
 11: Diagnostic of prenucleation
calcite, aragonite, and vaterite which are the rhombohedric, orthorhombic,
and hexagonal varieties of calcium carbonate are real polymorphs.
From these definitions, the crystallization of a protein in different solutions,
in the presence of different crystallization agents, gives rise not to real poly-
morphs but to different crystalline phases of the same protein. The crystalliza-
tion agents, salts in general, belong to the crystal structure so that the phases
of the same protein have different compositions. Despite these different com-
positions and for the sake of simplicity, it is accepted in the world of protein
crystallization to call polymorphs crystal varieties which should be called
phases of the same compound.
When the additive does not enter into the crystal structure, their main role
is to stabilize metastable phases or polymorphs. Metastable phases form due
to kinetic reasons and are favoured by high supersaturations. When several
phases are possible in the same solution, each of them has its own solubility so
that the solution can be supersaturated with respect to several phases at the
same time. According to the Ostwald's rule of stages, the phase which first
forms is not the most stable one, i.e. the less soluble one, but the phase lying
nearest to the original state in free energy. In other words, nature prefers to
follow a sequence of nucleations, growths, and phase transitions rather than
using a high energy level to directly nucleate the most stable phase. The meta-
stable phase later undergoes a phase transition as soon as nuclei of a more
stable phase, i.e. a less soluble phase, occur. Several phases or polymorphs
may temporarily coexist but all except one are subject to transformation. In
most cases, the phase transition occurs by dissolution of the metastable phase
and recrystallization into the stable one. It is called a solution-mediated phase
transformation. If some impurities, or additives, strongly adsorbs on the
crystals, the phase transition can be inhibited for a very long time. This
explains the rather long metastability of some protein crystals.

6.3 Crystallization kinetics, impurities, and additives

Impurities adsorb on the terraces between the growth steps, along the steps or
in the kinks, i.e. the growth sites. Depending on the energy of the bonds
between impurity and adsorption sites, adsorption is more or less reversible.
When growth proceeds, there is a competition between the kinetics of
molecule incorporation and the kinetics of impurity adsorption and
desorption. Accordingly, impurities hinder the crystallization processes so
that nucleation and growth rates are sometimes drastically slowed down.
Vekilov (47) has measured HEWL growth rates five to six times lower in the
presence of impurity than in pure media at low supersaturation. When the
impurity adsorption selectively takes place on a crystal face, the growth rate
of this face is selectively reduced and its relative development rapidly
increases at the expense of the development of the other faces. This induces
the habit changes or influences the crystal quality. Lorber et al. (48) observed
337
 S. Veesler and R. Boistelle
that the addition of ovalbumin or bovine serum albumin to pure HEWL is
correlated with an increase of the proportion of twinned crystals. When the
impurity adsorption takes place on all crystal faces, and is irreversible, i.e.
without exchange with the surrounding solution, then growth is completely
inhibited. Then the so-called growth cessation that often occurs with protein
crystal is observed. The only way to overcome this difficulty is to drastically
increase supersaturation; in some cases, new surface nuclei form and growth
starts again. However, if the crystal surface is too energetically poisoned,
three-dimensional nucleation becomes easier than growth. The solution
deposits fresh crystals.

6.4 Impurity incorporation

For a long time, it was believed that the impurities which induce habit changes
of crystals were incorporated into the crystals and that this incorporation was
liable for the habit changes. Actually, impurity absorption can accompany the
habit change but it is not its cause. This was demonstrated by growing
negative crystals, i.e. holes, in solutions poisoned by impurities. A hole is first
bored in a crystal. Then, all parts of the crystal but the hole, are coated by glue
to prevent it from dissolution. The crystal is then immersed in a solution
slightly undersaturated so that, by dissolution, the size of the hole increases
while, at the same time, the hole becomes faceted. When the impurity pro-
vokes the occurrence of facets different from those which occur in pure
solution, it is the proof that the habit change results from impurity adsorption
and not from impurity incorporation. Impurities cannot be trapped in a
crystal which is dissolving.
In the macromolecules field, impurity incorporation takes place especially
when the molecule of the impurity resembles the molecule of the crystal. It
was first observed in the case of small molecules (e.g. glutamic acid incorpor-
ated into asparagine monohydrate crystals) and later on in the case of biological
macromolecules, contamination of turkey egg white lysozyme crystallizing
solutions by HEWL (49). Pure materials are difficult to grow when impurity
and crystal molecules are homologues.

References
1. Lafont, S., Veesler, S., Astier , J.-P., and Boistelle, R. (1994). J. Cryst. Growth,
143,249.
2. Abraham, F. F. (1974). Homogeneous nucleation theory. Academic Press, New
York.
3. Zettlemoyer, A. C. (1969). Nucleation. Marcel Dekker, New York.
4. Katz, J. L. (1982). In Interfacial aspects of phase transformations (ed. B.
Mutaftschiev), p. 261. D. Reidel, Dordrecht, The Netherlands.
5. Malkin, A. J., Kuznetsov Yu, G., Glantz, W., and McPherson, A. (1996). J. Phys.
Chem., 100, 11736.
338
 11: Diagnostic of prenucleation
6. Israelachvili, J. N. (1991). Intermolecular and surface forces. Academic Press,
London.
7. Tanford, C. (1961). In Light scattering in physical chemistry of macromolecules,
p. 275. John Wiley & Sons, Inc., New York, London.
8. Pecora, R. (1985). Dynamic light sacttering: application of photon correlation
spectroscopy. Plenum Press, New York and London.
9. Guinier, A. and Fournet, G. (1955). Small-angle scattering of X-rays. John Wiley,
New York.
10. Tardieu, A. (1994). In Neutron and synchrotron radiation for condenses matter
studies (ed. J. Baruchel, J. L. Hodeau, M. S. Lehmann, J. R. Regnard, and C.
Schlenker), Vol. III, p. 145. Les Editions de Physique, Springer-Verlag.
11. Lindner, P. and Zemb, T. (1991). Neutron X-ray and light scattering: introduction
to an investigative tool for colloidal and polymeric systems. North-Holland,
Amsterdam.
12. Jacrot, B. and Zaccai, G. (1981). Biopolymers, 20, 2413.
13. Einsenberg, H. (1976). Biological macromolecules and polyelectrolytes in solution.
Clarendon Press, Oxford.
14. Parsegian, V. A., Rand, R. P., Fuller, N. L., and Rau, D. C. (1986). In Methods in
enzymology, (ed. L. Packer), Academic Press, London. Vol. 127, p. 400.
15. Broutin, L, Ries-Kautt, M., and Ducruix, A. (1995). J. Appl. Cryst., 28, 614.
16. Gripon, C., Legrand, L., Rosenman, I., Vidal, O., Robert , M.-C., and Boue, F.
(1996). C. R. Acad. Sci. Paris, 323,215.
17. Kam, Z., Shore, H. B., and Feher, G. (1978). J. Mol. Biol., 123, 539.
18. Zulauf, M. and D'Arcy, A. (1992). J. Cryst. Growth, 122, 102.
19. Ferre-d'Amare, A. R., Zhou, K. Z., and Doudna, J. A. (1998). J. Mol. Biol., 279,
621.
20. Veesler, S., Marcq, S., Lafont, S., Astier, J.-P., and Boistelle, R. (1994). Acta
Cryst., D50, 355.
21. Muschol, M. and Rosenberger, F. (1996). J. Cryst. Growth, 167, 738.
22. Georgalis, Y., Zouni, A., and Saenger, W. (1992). J. Cryst. Growth, 118, 360.
23. Georgalis, Y., Zouni, A., Eberstein, W., and Saenger, W. (1993). J. Cryst. Growth,
126, 245.
24. George, A. and Wilson, W. W. (1994). Acta Cryst., D50, 361.
25. George, A., Chiang, Y., Guo, B., Arabshahi, A., Cai, Z., and Wilson, W. W.
(1997). In Methods in enzymology, (ed. C. W. Carter and R. M. Sweet), Academic
Press, London. Vol. 276, p. 100.
26. Bonnete, F., Malfois, M., Finet, J., Tardieu, A., Lafont, S., and Veesler, S. (1997).
Acta Cry'St., D53, 438.
27. Koppel, D. E. (1972). J. Chem. Phys., 57, 4814.
28. Ducruix, A., Guilloteau, J.-P., Ries-kautt, M., and Tardieu, A. (1996). J. Cryst.
Growth, 168, 28.
29. Hartman, P. and Perdok, W. G. (1955). Acta Cryst., 8, 49.
30. Hartman, P. (1973). In Crystal growth: an introduction (ed. P. Hartman), p. 367.
North-Holland, Amsterdam.
31. Hartman, P. (1982). Geol. Mijnbouw, 61, 313.
32. Hartman, P. and Bennema, P. (1980). J. Cryst. Growth, 49,145.
33. Ohara, M. and Reid, R. C. (1973). Modeling crystal growth rates from solution.
Prentice Hall, Englewood Cliffs, NJ.
339
 S. Veesler and R. Boistelle
34. Hillig,W. B. (1966). ActaMetalL, 14,1868.
35. Madsen, L. and Boistelle, R. (1979). J. Cryst. Growth, 46, 681.
36. Burton, W. K., Cabrera, N., and Frank, F. C. (1951). Phil. Trans. Roy. Soc., 243,
299.
37. Chernov, A. A. (1961). Sov. Phys. Usp., 4, 116.
38. Gilmer, G. H., Ghez, R., and Cabrera, N. (1971). J. Cryst. Growth, 8, 79.
39. Bennema, P. and Gilmer, G. H. (1973). In Kinetics of crystal growth, an
introduction (ed. P. Hartman), p. 263. North-Holland, Amsterdam.
40. Nielsen, A. E. (1964). Kinetics of precipitation. Pergamon, Oxford.
41. Boistelle, R., Astier, J.-P., Marchis-Mouren, G., Desseaux, V., and Haser, R.
(1992). J. Cryst. Growth, 123, 109.
42. Vekilov, P. G., Ataka, M., and Katsura, T. (1993). J. Cryst. Growth, 130,317.
43. Boistelle, R. (1982). In Interfacial aspects of phase transformation, p. 621. Erice,
Sicily.
44. Parker, R. L. (1970). In Solid state physics, Vol. 25, p. 151. Academic Press, New
York.
45. Kern, R. (1968). Bull. Soc. Fr. Mineral. Cristallogr., 91, 247.
46. Giege, R., Dock, A.-C, Kern, D., Lorber, B., Thierry, J.-C, and Moras, D. (1986).
J. Cryst. Growth, 76, 554.
47. Vekilov, P. G. (1993). Prog. Cryst. Growth, 26, 25.
48. Lorber, B., Skouri, M., Munch, J.-P., and Giege, R. (1993). J. Cryst. Growth, 128,
1203.
49. Abergel, C., Nesa, P. M., and Fontecilla-Camps, J. C. (1991). J. Cryst. Growth,
110, 11.

340
 12

Two-dimensional crystallization of
soluble proteins on planar lipid
films
A. BRISSON, O. LAMBERT, and W. BERGSMA-SCHUTTER

1. Introduction
Electron crystallography of protein two-dimensional (2D) crystals constitutes
a fast-expanding method for determining the structure of macromolecules at
near-atomic resolution (1, 2). The main limitation in the application and
generalization of this approach remains in obtaining highly ordered 2D
crystals, as is the case of 3D crystals in X-ray crystallography.
Several methods of 2D crystallization are available which can be classified
into two families, depending on the type of proteins under investigation,
either membrane proteins (3, 4) or soluble proteins (5, 6). In both cases, 2D
crystallization is a self-organization process which spontaneously occurs
between macromolecules which are restricted to diffusing by translation and
rotation in a 2D space, with a fixed orientation along the normal to this plane.
The scope of this chapter is restricted to the 2D crystallization of soluble
proteins on planar lipid films, by the so-called 'lipid monlayer crystallization
method' (5). Our aim is to present a step-by-step description of the experi-
mental procedures involved in the application of this method.

2. Two-dimensional crystallization of soluble proteins

on planar lipid films
The method of protein 2D crystallization on planar lipid films was introduced
about 15 years ago (5) and has since been successfully applied to about 30
proteins (Table 1). Its principle is based on the specific interaction between
soluble proteins and lipid ligands inserted in a lipid monolayer, at an air-water
interface (Figure 1). In practice, a lipid monolayer is formed by spreading
lipids dissolved in an organic solvent on a water surface. Proteins present in
the aqueous subphase bind to their ligand of lipidic nature and spontaneously
 A. Brisson et al.

Table 1. List of macromolecules crystallized on planar lipid layers

Protein Ligand lipid Ref.

(a) Natural lipids
Cholera toxin GM1 ganglioside 1
Tetanus toxin GT1 ganglioside 2
Botulinum toxin GT1b ganglioside 3
Staphylococcus a-toxin platelets lipids 4
AnnexinVI 14:0-PE/ 18:1-PS 5
Annexin V 18:1-PS/brain extract 6
Coagulation factor Va 18:1-PS 7
Coagulation factor IX 18:1-PS 8
Protein kinase C 18:1-PS 9

(b) Synthetic lipids made of a protein ligand coupled to a lipid molecule

anti-DNP IgG DNP-PE 10
Ribonucleotide reductase dATP-PE 11
Streptavidin biotin-PE 12
DNAgyraseB novobiocin-PA 13
C-reactive protein DS8PE 14
(His6)-HIV-1 reverse transcriptase Ni-NTA-18:1-PE 15
(His6)-peptide-MHC Ni-NTA-18:1-PE 16
(His6)-HupR Ni-NTA-DOGA 17
(His6)-(M-MuLV) Nter-capsid protein Ni-DHGN 18
RNA polymerase 1 Ni-NTA-DOGA 19
Streptavidin Cu-DIODA 20

(c) Charged lipids

ferritin eicosic-N+(CH3)3 21
RNA polymerase (E. coli) octadecylamine 22
RNA polymerase II, I (yeast) octadecylamine / cetyl-N+(CH3)3 23
a-actinin DDMA 24
Mitochondrial creatine kinase cardiolipin 25
50Sribosome 18:1-PS 26
Brush border myosin I 18:1-PS 27
Oestrogen receptor-LBD 14:0-PC 28
chaperonin 29
I) a) Ludwig et al. (1986) P.N.A.S., 83, 8585; b) Mosser et al. (1992) J. Mol. Biol., 226, 23; 2) Robinson et
al. (1998) J. Mol. Biol., 200, 367; 3) Schmid et al.. Nature (1993) 364, 827; 4) Olofsson et al. (1990) J.
Mol. Biol., 214, 299; 5) a) Newman et al. (1989) J. Mol. Biol., 206, 213; b) Benz et al. (1996) J. Mol. Biol.,
260, 638; 6) Mosser et al. (1991) J. Mol. Biol., 217, 241; 7) Stoylova ef al. (1994) FEBS Lett., 351, 330;
8) Stoylova et al., FEBS Lett. (1998) 1383, 175; 9) Owens et al. (1998) J. Struct. Biol., 121, 61; 10) a)
Uzgiris, Kornberg (1983) Nature, 301, 125; b) Uzgiris (1986) Biochim. Biophys. Res. Comun., 134, 819;
II) Ribi et al. (1987) Biochemistry 26, 7974; 12) a) Blankenburg et al. (1989) Biochemistry 28, 8214; b)
Kubalek et al. (1991) Ultramicroscopy, 35, 295; 13) Celia et al. (1994) J. Mol. Biol., 236, 618; 14) Sui ef
al. FEBS Lett. (1996) 388, 103; 15) Kubalek et al., J. Struct. Biol. (1994) 113, 117; 16) Celia et al.
(submitted); 17) Venien-Bryan et al., J. Mol. Biol. (1997) 274, 687; 18) Barklis et al., EMBO J. (1997).16,
1199; 19) Bischler et al. (1998) Biophys. J., 74, 1522; 20) Frey et al. (1998) Biophys. J., 74, 2674; 21)
Fromhertz (1971) Nature, 231, 267; 22) Darst et al. (1989) Nature, 340, 730; 23) a) Darst et al. (1991) Cell,
66, 1; b) Schultz et al. (1993) EMBO J., 12, 2601; 24) Taylor and Taylor (1993) J. Mol. Biol., 230,196; 25)
Schnyder et al., J. Struct. Biol. (1994) 112, 136; 26) Avila-Sakar ef al. J. Mol. Biol. (1994) 239, 689; 27)
Celia ef al., J. Struct. Biol. (1996) 117, 236; 28) 29) Ellis et al. (submitted).
Abbreviations: DDMA, didodecyldimethylammonium; DNP, dinitrophenyl; DHGN, dihexadecylglycaro-;
DOGA, dioleoylglyceroxyacetylamino; DS8PE, dioctadecanyl N-[/V-(aminoethyl phosphatoethyl)
succinamido-/V-yl]-aspartate inner salt; Ni-NTA, Ni-/V-nitrilotriacetic acid-chelated nickel; PA, phos-
phatitic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PS, phosphatidylserine; 12:0,
dilauroyl; 14:0, dimyristoyl; 18:1, dioleoyl.

342
 12; Two dimensional crystallization of soluble proteins

Figure 1. Scheme of the lipid-layer crystallization method. Lipids are deposited at an

air-water interface and spread into a monomolecular layer. The monolayer presented
here is made of two types of lipids, ligand lipids and dilution lipids. Proteins present in
the subphase interact by molecular recognition with their ligand at the level of the planar
layer (1). Protein-lipid complexes diffuse freely and concentrate in the plane of the
monolayer (2). These complexes interact with each other and spontaneously assemble
into 20 domains and 2D crystals, provided that favourable interactions are stabilized (3).

form 2D domains and, in favourable cases, 2D crystals. The process of 2D

crystal formation relies on three successive steps:
(a) Molecular recognition between a protein and its ligand.
(b) Diffusion and concentration of the protein-lipid complexes in the plane
of the lipid film.
(c) Self-organization of the proteins into 2D crystals.
As indicated in Table 1, three different types of systems can be
distinguished, depending on the nature of the lipid ligand:
• natural lipids
• synthetic lipids made of a protein ligand coupled to a lipid molecule
• charged lipids.
343
 A. Brisson et al.
Although most applications up to now have concerned proteins, this method
has also been used to crystallize other types of macromolecules, as the 50S
ribosomes on negatively charged lipids (7).
The main advantages of this method are:
(a) It is universal, as indicated by the wide diversity of proteins already
crystallized.
(b) It is easy to apply and gives fast results (either positive or negative).
(c) Only small amounts of protein are required—about 1 1 ug per incubation.
(d) Crystallization conditions are gentle, as proteins can be maintained in a
physiological buffer.
(e) 2D crystals can be ordered up to high resolution, around 3 A (8-10), and
thus amenable to high resolution structure determination by electron
crystallography (11).
Several review articles have already been published on the lipid-layer crystal-
lization method (12, 13). This chapter will focus on the experimental aspects
involved in the application of this method, and will cover the following steps:
(a) Setting up a crystallization experiment.
(b) Transfer of protein-lipid films to an electron microscopy (EM) grid.
(c) Characterization of protein-lipid domains/crystals by EM. Only the first
steps of the characterization will be described here, namely the prepa-
ration of negatively stained specimens and the analysis of electron
micrographs by optical diffraction.
Examples taken from studies on annexin V and streptavidin will be used as
illustrations.

3. Setting up a crystallization experiment

3.1 Lipid solutions
3.1.1 General considerations
Interfacial lipid films used for growing protein 2D crystals must fulfil both
requirements of stability, as some experiments last for several days, and
fluidity, allowing for the necessary diffusion of the protein-lipid complexes.
Lipid molecules spread in monolayers at an air-water interface interact more
or less strongly, depending on the chemical nature of the lipids, their con-
centration, surface pressure, temperature, or composition of the subphase.
Three states are distinguished: gas, liquid expanded and condensed, or solid
state, which reflect the strength of the intermolecular interactions (14, 15).
Most, if not all, successful crystallization experiments reported until now have
been performed with lipid layers in a liquid expanded or fluid state. This is the
state of most unsaturated phospholipids, such as dioleoylphosphatidylcholine
344
 12: Two dimensional crystallization of soluble proteins
(DOPC) or egg phosphatidylcholine (PC), at a temperature close to 20 °C.
This is the reason why lipid solutions contain, in addition to the lipid ligands, a
second lipid, often referred to as dilution lipid. DOPC or a mixture of un-
saturated PC molecules are often used as dilution lipids. Mixtures with ligand
lipid:dilution lipid molar ratios ranging from 1:3 to 1:6 are frequently used. It
is highly recommended to characterize the physico-chemical properties of the
monolayers formed by each new lipid mixture under study, by measuring
surface pressure/area isotherms with a Langmuir balance or a Wilhelmy plate
(15).
An illustrative example of the stabilizing role of dilution lipids is provided
by the cholera toxin system (16). The monosialoganglioside GM1, which is the
natural lipid ligand of cholera toxin, has a tendency to form micelles in
aqueous solutions and does not form stable monolayers. On the other hand,
lipid films formed with GM1/DOPC mixtures are stable and support the 2D
crystallization of cholera toxin, over a wide range of GM1/DOPC mixtures.
Most 2D crystallization experiments are performed in Teflon wells of small
size (see Section 3.3). To saturate the non-specific lipid binding sites present
at the Teflon surface, an excess of lipids is commonly used with respect to the
amount needed to form one monolayer. Although there are no strict rules, a
tenfold excess is often used in practice. As a first approximation, we can con-
sider that lipid films formed with a lipid excess consist of a monolayer to
which a reservoir of lipids is associated, probably in the form of small 3D
aggregates adsorbed at Teflon edges. This hypothesis is supported by ellipso-
metry experiments which show no increase in film thickness after a monolayer
is formed (17). Although it might be more correct to refer to planar lipid films
instead of lipid monolayers, this is a rather semantic distinction.

3.1.2 Practical considerations

Particular attention must be given to the preparation and storage of lipid
solutions. Most lipids can be dissolved in either chloroform or chloroform:
methanol (2:1, v/v). Lipid mixtures used for spreading interfacial films often
contain hexane in addition to these solvents, in a 1:1 (v/v) ratio. Lipid
solutions must be stored in hermetic glass containers, such as Wheaton vials
Z10 equipped with Teflon-faced rubber caps. Lipid solutions are stored at
-20 °C. Pipettes or syringes made of either glass or metal must be used with
lipid solutions. It is recommended to frequently control the level of the lipid
solutions, as it is difficult to maintain tight seals and to prevent evaporation of
organic solvents.

3.2 Protein solutions

The preparation of protein solutions is comparatively simple, as proteins can
be used in their physiological buffer, as for other biochemical experiments.
345
 A. Brisson et al.
Protein concentrations ranging from 50-200 ug/ml are used in practice.
Concentrations as low as 10 ug/ml have been used with success (16).
Due to the known property of proteins to denature at air-water interfaces,
it is often questioned whether it is better to inject proteins under a preformed
lipid layer or to spread lipids over a protein solution. The few comparative
studies performed until now have reported no significant differences between
both procedures. For convenience, proteins are deposited first in the crystal-
lization wells before spreading the lipids.
The affinity between proteins and lipid ligands is obviously a critical para-
meter for the success of crystallization. For several systems—streptavidin,
cholera toxin, annexin V—apparent dissociation constants of the complexes
are available, which in all three cases correspond to a very tight binding, with
Kd ranging from 10~15 M to 10-9 M. However, high affinity is not sufficient to
induce crystallization, as shown by the fact that only a few members of the
annexin family form 2D crystals, while other members that bind negatively
charged lipids with similar affinity (Kd ~ 10-9 M) only form disordered 2D
protein domains (Bergsma-Schutter and Brisson, unpublished results).

3.3 Preparation and cleaning of Teflon supports

Teflon supports are made of ordinary Teflon material. Discs of 60 mm
diameter and 8 mm height are convenient in practice. Wells of 4 mm diameter
and 1 mm depth are drilled at regular intervals in these discs; discs with 3X3
or 4 X 4 wells allow several crystallization experiments to be performed
simultaneously. The diameter of each well—4 mm—is adapted to receive, for
the transfer step, an EM grid with a standard diameter of 3 mm (see Section 4).
Care must be given to the wells' edges, which must be devoid of irregularities,
as they might constitute traps for lipids or affect the overall homogeneity of
the planar lipid layer. To minimize possible problems due to uneven edges, it
is recommended to select wells with sharp edges with a magnifying glass.
The state of hydrophobicity of the surface of Teflon wells plays an
important role in crystallization experiments. The main goal of the cleaning
step is to regenerate a highly hydrophobic Teflon surface, in order to prevent
lipids overflowing over the Teflon surface.

Protocol 1. Cleaning of Teflon supports

Method
1. Handle Teflon discs with gloves.
2. Place a disc in a bath of 1% Hellmanex II for 2 h.
3. Rinse the disc extensively with warm tap-water for 1 h.
4. Rinse with several baths of deionized water.

346
 12: Two dimensional crystallization of soluble proteins
5. Smash the disc against a piece of tissue paper lying on a table, to get
rid of the water.3 The disc is ready for use.d
6. For prolonged storage, keep the discs either in water or dry, in a box.
Perform the treatment (steps 1-5) before use.
aA good indication of the state of hydrophobicity of a Teflon support is obtained by smashing
it on a piece of tissue paper; the wells must look completely dry after two smashes. If some
water remains at the bottom of the wells or at their edges, washing of the disc must be
repeated. When discs have been used repeatedly, a second washing step might not be
sufficient to obtain a hydrophobic surface. It is then recommended to brush the wells in order
to eliminate lipids stuck onto their surface.
'Teflon surfaces tend to become electrostatic, which can lead to several surprising effects; a
microscope grid may 'jump' upon deposition on the lipid-coated wells, or a lipid droplet
approached from the Teflon surface may become deformed or even 'explode'. Although it is
difficult to judge the influence of such a behaviour on crystallization, the best remedy is to
discharge the surface with an antistatic device.

Protocol 2. Setting up a crystallization experiment, e.g. annexin V

(see Figure 2)

Equipment and reagents

• A Hamilton syringe of 10 p.1, with a 90° • A Petri dish of 9 cm diameter, in which the
bevel, used for depositing lipids—rinse Teflon disc is deposited, serves as a humid
thoroughly with chloroform before and chamber—a small hole is made in the
after use plastic lid to allow a gentle opening
• Protein solution: 100 ug/ml annexin V in • Lipid solution: 150 H.M DOPS (dioleoyl-
150 mM NaCI, 2 mM CaCI2, 25 mM Hepes phosphatidylserine), 450 n-M DOPC, in
pH 7.5 chlorofornrhexane (1:1, v/v)

Method
1. Place a freshly conditioned Teflon disc in a Petri dish.
2. Poor water around the disc up to about mid-height.
3. Deposit in each well 17 uJ of the protein solution.
4. Rinse the 10 ul syringe three times with chloroform.
5. Deposit 0.6 ul of the lipid solutiona on top of each protein droplet.
6. Install the lid on the Petri dish and close the hole with a piece of tape,
in order to limit evaporation.
7. Incubate.b-c

a Lipids must not flow over the edges. If this occurs, irreproducible results can be expected and
it is wise to use another well or another Teflon disc.
"The incubation time required for crystal growth is variable and depends on the protein and
lipid system. For example, one hour or less is sufficient to get 20 crystals of streptavidin or
cholera toxin; in the case of annexin V, one hour is also sufficient to get one type of crystal,
with p6 symmetry, while several days are required to get highly ordered p3 crystals (18).
c
Most studies reported until now have been performed at ambient temperature, around 20°C.
It is of course important to use conditions in which the protein is stable.

347
 A. Brisfion et al.

Figure 2. Step-by-step scheme of the 2D crystallization of proteins on lipid monoiayers.

348
 12: Two dimensional crystallization of soluble proteins

4. Transfer of protein-lipid films to an EM grid

4.1 General considerations
The characterization of 2D crystals of proteins, in situ at the air-water inter-
face, presents technical difficulties due to the very nature of the specimen.
Optical techniques, such as fluorescence microscopy (19), Brewster angle
microscopy (20), or ellipsometry (17), as well as mechanical techniques (17),
can provide valuable information on the formation of 2D interfacial domains.
However, this information is limited to macroscopical features, due to the
diffraction limit of light. EM remains the most commonly used method for
characterizing and investigating the structure of 2D protein crystals formed by
the lipid-layer method. A recent study has demonstrated the potential of
atomic force microscopy to follow, in situ and in real time, the formation of
2D protein crystals on supported lipid bilayers (21). This study opens new
possibilities in various areas and will certainly generate interest for
investigating the process of crystal growth in 'live' conditions.
The transfer of protein-lipid films from the air-water interface to an EM
grid is certainly the most important step in the whole procedure, as far as the
influence of the experimenter is concerned. Most of the practical aspects
presented below (Protocols 3-7} concern this step of transfer, from which
depends the ultimate quality of the structural results.

4.2 Preparation of EM grids

Specimens observed by transmission electron microscopy are in general de-
posited onto grids coated with a support material. Carbon is the most popular
support material because thin carbon films—several nm thick—are easy to
prepare, transparent to electrons, mechanically stable, and conducting (22).
Two main types of carbon films can be used for transferring protein-lipid
films (Figure 3):
(a) Continuous carbon films (Protocol 3).
(b) Perforated carbon films, presenting holes and commonly named holey
carbon films (Protocols 4 and 5).
Many methods of preparation of EM grids coated with carbon films have
been reported (22). We will only describe here some standard methods which
have given satisfactory results for the transfer of protein-lipid interfacial
films.

4.2.1 Preparation of EM grids coated with a continuous carbon film

Three successive steps can be distinguished:
(a) Formation of a plastic film onto which grids are deposited (Protocol 3,
steps 1-8).
349
Figure 3. Overall aspect of different types of carbon film supports, (a) Grid coated with a
continuous carbon film (adapted from ref. 23), This image was recorded by reflected light
microscopy, which constitutes an appropriate method for visualizing the planarity of
carbon films. The carbon film is mostly flat, except near some corners, where wrinkles
are observed (arrowheads). Each square is 37 un wide, (b) <c) EM images of holey films
prepared by the methods described in Protocol 4 and 5, respectively. The films presented
here have been used to transfer interfacial films of streptavidin (b) and annexin V Ic).
Most of the holes are covered with a continuous layer exhibiting a homogeneous
greyness, indicating that the transferred material is of uniform thickness. Several holes
present breaks. Note the almost complete absence of vesicles and multilayered domains,
in comparison with Figure 4. Scale bars: (b) 1 um; (c) 2 um.

350
 12: Two dimensional crystallization of soluble proteins
(b) Evaporation of carbon on the plastic film (Protocol 3, step 9).
(c) Dissolution of the plastic film (Protocol 3, steps 10 and 11).

Protocol 3. Preparation of EM grids coated with a continuous

carbon film

Equipment and reagents

• EM grids (300-400 mesh): clean before use • Glass slides: clean with alcohol and let
by washing them either in acetone, or them dry by placing them vertically on a
successively in an 0.1 M H2SO4 solution, filter paper
then water, and then acetone (30 sec each) • 0.3% (w/v) nitrocellulose (collodion)
• A beaker (10 cm diameter) filled with about solution in amyl acetate
5 cm water

Method
1. Deposit four droplets of the nitrocellulose solution on a clean glass
slide, with a Pasteur pipette.
2. Form a continuous and homogeneous liquid film by tilting the slide.
3. Eliminate most of the liquid by holding the slide vertically against a
piece of filter paper.
4. Let dry for 10 min.
5. Float off the plastic film on a water surface, by slowly inserting the
slide at glancing angle into the water-bath.
6. Deposit EM grids on top of the plastic film.
7. Deposit a piece of absorbent paper or a piece of Parafilm on top of the
plastic film covered with the grids, while maintaining the paper/
Parafilm by one edge. Wait until a good contact is formed with the
underlying plastic film.
8. Lift up the paper/Parafilm and deposit it on a clean surface. Allow for
a complete drying, under a lamp.
9. Evaporate a thin film of carbon on the nitrocellulose side, using
standard EM procedures (22).a
10. Place the grids, carbon side up, on top of several pieces of filter paper
soaked with amyl acetate, in a glass Petri dish. Close the dish and
leave for several hours to overnight.
11. Transfer the grids on a dry filter paper.b

a Grids prepared up to here can be used for negative staining without removal of the plastic
film. On the other hand, for cryo-microscopy experiments, it is mandatory to remove the
plastic film in order to avoid artefacts.
b Carbon films prepared by this method are in general flat (Figure 3a) (23). The flatness of
carbon films is certainly an important parameter for allowing a good contact with the lipid
chains and thus achieving an efficient transfer.

351
 A. Brisson et al.

4.2.2 Preparation of EM grids coated with a holey carbon film

The fabrication of holey films has received considerable attention from
electron microscopists. Several methods have been developed and more or
less modified in many laboratories for particular needs. We describe here two
methods (Protocols 4 and 5) used routinely in our laboratory, which give two
types of films differing by the nature, the size, and the distribution of holes.

Protocol 4. Preparation of EM grids coated with a holey carbon

filma

Reagents
• 0.3% (w/v) formvar solution in dichloroethane

A. Formation of a plastic film with 'pseudo'-holes

1. Bring to the boil a 500 ml beaker filled up with water.
2. Dip a glass slide into a 50 ml beaker filled with the formvar solution for
about 10 sec.
3. Pull it gently out of the solution.
4. Keep it vertically against a filter paper to drain most of the liquid.
5. When the slide is still wet, place it horizontally in the stream of water
vapour.
6. Take the slide away from the stream when the film at its surface
becomes milky. Let it dry.
7. Check the aspect of the network with a light microscope, and mark the
good areasb,c with a needle.

B. Formation of holes in the plastic film

1. Scratch with a needle the borders of the good areas.
2. Float off the plastic film on water, deposit EM grids, and pick up the
whole as in Protocol 3, steps 5-9.
3. Place a filter paper in a Petri dish and soak it with acetone.
4. Place few (approx. five) grids on a support made of a thin wire mesh,
standing 5 mm above the bottom of the Petri dish.
5. Leave the grids in the acetone vapour for 60 sec.
6. Check the aspect of the lacy network with a light microscope, and
repeat part B, step 5 by varying the length of the etching time until a
convenient aspect is observed.d
7. Repeat the procedure for the other grids.
8. Evaporate a thick layer of carbon (as in Protocol 3, step 9).e

352
 12: Two dimensional crystallization of soluble proteins
C. Dissolution of the plastic film
1. Dissolve the formvar film by depositing the grids on top of several
pieces of filter paper soaked with dichloroethane, in a glass Petri dish.
2. Let stand overnight.
3. Transfer the grids on a dry paper.f

•This method is adapted from Sjostrand (24).

"The selection step is crucial. The network present on the glass slide must be of the 'good'
size, as an open network might be too fragile while a dense network might result in too small
holes.
c
As this stage, there is still a thin layer of plastic covering the 'pseudo'-holes.
d
The time of etching has to be optimized. It is recommended to start with a time of 1 min,
which often gives good results, and with a small number (approx. five) of grids.
e
Due to the lacy structure of the holey film, it is recommended to evaporate a thick layer of
carbon, to improve the mechanical stability of the film and to ensure good conductivity.
'It is recommended to check grids before use for EM.

Protocol 5. Preparation of EM grids coated with a holey carbon

filma

Equipment and reagents

• A metal block pre-cooled in a freezer at • 0.25% (w/v) cellulose acetate butyrate
-20°C solution in ethyl acetate (triafol)

A. Formation of a plastic film with holes

1. Clean glass slides as in Protocol 3.
2. Plunge the slides in a bath of 0.1% (w/v) Tween 20.
3. Let them dry vertically against a filter paper.
4. Take the metal block out of the freezer and deposit a glass slide on it.
5. When moisture appears at the surface of the slide—this takes few
seconds—pour a few droplets of the cellulose acetate butyrate
solution on the slide.b
6. Eliminate the excess of solution by blotting with a filter paper.
7. Let the slide dry vertically.
8. Check the quality of the holey film with an optical microscope.
9. Float off the thin holey plastic film on a water-bath, deposit EM grids
onto it, and evaporate a thin carbon layer, following the procedure
described in Protocol 3, steps 5-10.

B. Dissolution of the plastic film

1. Remove the plastic film by placing the grids over a filter paper soaked
with chloroform or ethyl acetate, in a closed glass chamber saturated
with vapour.

353
 A. Brisson et al.
Protocol 5. Continued
2. Check each grid with an optical microscope for homogeneity and
integrity of the holey film,c,d
a This method has been adapted from Fukami and Adachi (25) by Chretien et al. (26) (adapted
from ref. 26 with permission).
b It is possible to adjust the size of the holes at this step. A long exposition on the metal block
will produce larger holes. Several trials are necessary to obtain holes with the desired size.
c With holey films prepared by this method, about half of the surface is covered with holes and
the other half with carbon (Figure 3c).
d 'The main advantage of this method, as compared with most other methods of fabrication of
holey films, is that floating of the plastic film is easy and reproducible.

4.2.3 Transfer of interfacial films to EM grids

Interfacial films are transferred to EM grids by depositing an EM grid coated
with a carbon film on top of a crystallization well. EM grids are deposited with
the carbon film facing the lipid tails. This method of transfer is often referred
to as the Schaeffer method (15) (Protocol 6).
Another method of transfer has recently been proposed, which makes use
of a wire loop to pick up the protein-lipid film and to transfer it to an EM
grid, of either type mentioned before (27). This method has only been applied
to a limited number of specimens and will not be further described here.

Protocol 6. Transfer of a protein-lipid interfacial film to an EM

grid: 'fishing' step
Method
1. Hold a carbon coated grid with tweezers, and deposit it horizontally,
carbon side facing down, on top of the lipid film.
a 2. Wait 2-5 min.
3. Lift up the grid with the tweezers. A thin layer of water covering the
carbon surfacebmust be visible.
4. The grid is now ready for the specimen preparation step, which must
be performed immediately (see Protocol 7).
a Observe carefully the deposition of the grid on the film. When 2D crystals are present at the
interface, the grid 'sticks' to the film upon deposition. When no crystals are present at the
interface, as for example with pure fluid lipid layers, the grid moves and rotates freely.
b
As a large excess of lipid is present at the interface, a new lipid film will spontaneously cover
the water surface after the first film has been picked up. The well is in principle ready for a
second cycle of crystallization. It must be noted that results obtained with a 'second' or even a
'multiple' fishing are more variable, which may be due to the formation of heterogeneous lipid
layers or to some overflowing of lipids on the Teflon surface.

354
 12: Two dimensional crystallization of soluble proteins

5. Characterization of protein 2D crystals by EM

The observation of biological specimens by EM requires the use of specific
methods of preparation, the role of which is to protect these specimens
against dehydration and electron radiation. In the case of 2D crystals of
proteins, as in general for observations at the molecular level, two preparation
methods are best adapted; the classical negative staining method (Protocol 7)
and the more complex cryo-methods, applied to unstained specimens. For each
new study, the initial steps are performed by negative staining, which provides
an efficient way for screening various conditions of crystallization. The resolu-
tion achievable by negative staining is however limited to ~ 10-20 A and the
structural information is restricted to the molecular envelope. Therefore,
further steps of the analysis, abutting ideally to the determination of the
molecular structure at high resolution (28), must be performed by cryo-EM
on unstained specimens. This latter part is outside the scope of this chapter.

Protocol 7. Negative staining of protein 2D crystals

Equipment and reagents
• 1% (w/v) uranyl acetate solution in water, pH ~ 3.5

Method
1. Immediately after the fishing step (Protocol 6, step 4), add a 5 ul
droplet of the uranyl acetate solution to the liquid film present on the
grid.a
2. Wait for 30 sec.
3. Remove the excess liquid by touching a grid border with a filter paper.
4. Allow for complete drying.
5. In the case of holey films, evaporate a thin layer of carbon on the side
of the protein-lipid film.b
6. The grid is ready for observation in the microscope.
a
At the beginning of each new study, it is recommended to compare the results obtained with
several negative stains. Sodium phosphotungstate (2% (w/v) aqueous solution, pM 7) is
another commonly used negative stain.
b
The deposition of a carbon layer enhances the mechanical stability and improves the
conductivity of self-supported interfacial films.

EM of negatively stained specimens provides two types of 'low resolution'

structural information:
1. At low magnification (X 2000), the overall aspect of the material trans-
ferred on the EM grid is visible. The nature and the aspect of the transferred
material depend highly on the type of carbon films used (Figures 4-6). As a
355
 A. Brisaon et al.

Figure 4. Aspect of interfacial films transferred with continuous carbon films, (a) and |b)
correspond to annexin V and streptavidin, respectively, (a) Domains exhibiting a uniform
greyness and vesicles cover the carbon film support. The vesicles surround the domain
areas and most probably form during the transfer step. Extensive carbon film areas are
devoid of domains and vesicles (*). (b) Domains showing multilayered structures are
characteristic of films transferred with continuous carbon supports. On these low
magnification images (scale bar: 1 um), the crystalline nature of the domains is not visible.

356
 12: Two dimensional crystallization of soluble proteins
rule that undoubtedly has exceptions, interfacial films picked up with a con-
tinuous carbon film present domains easily distinguishable from the carbon
background, together with vesicular material (Figure 4). These domains are
often folded or overlap in multilayered structures, and present morphologies
characteristic of each protein-lipid system. On the other hand, interfacial films
transferred with a holey carbon film appear as homogeneous layers of uni-
form greyness and thus uniform thickness, mostly devoid of domains or large
vesicles (Figures 3b, 3c, 5, and 6). It is now commonly accepted that the inter-
facial films are transferred without, or with minor, reorganization when holey
films are used (8, 18, 29), while they are submitted to profound reorganization
upon transfer/drying with continuous carbon films (13, 18, 30). The annexin V
system constitutes an extreme case in this context, as p6 crystals are obtained
with holey films, while p3 crystals are observed with continuous carbon films
(18). Most strikingly, these p3 crystals do not pre-exist at the air-water
interface and their formation is induced by the transfer step. Cholera toxin
constitutes another interesting case as highly ordered 2D crystals are obtained
after transfer with continuous carbon films while close-packed 2D domains
are obtained with holey films, and thus pre-exist at the air-water interface
(18). The coherent picture which emerges from these studies is that upon
specific binding to ligands incorporated into lipid monolayers at the air-water
interface, some proteins form 2D crystals, while many others self-organize in
close-packed assemblies. Upon transfer with a continuous carbon film, these
close-packed assemblies are 'stressed' and may reorganize into more compact
and better ordered 2D crystals.
2. At high magnification (X 50 000), the crystalline nature of the transferred
material can be visualized. However, it is in general not possible to get a
quantitative evaluation of the crystalline order by a mere 'eye' observation.
Even when strongly contrasted stain striations are observed, this information
is of low resolution as it represents most often the accumulation of stain
between molecules (see for example Figure 5). The most objective way to
evaluate the crystalline quality of protein-lipid interfacial films is by optical
diffraction or Fourier transform calculation (see Section 6).
The characterization of interfacial films by EM is one of the most time-
consuming steps in the whole procedure. The main reason is the huge number
of areas of potential interest on each grid and the variability of aspect existing
between grids and also within different areas of one given grid. It is important
to consider that:
(a) In the case of holey films, holes covered by either 2D crystals or close-
packed assemblies present the same aspect, and there are of the order of
104 to 106 holes per grid.
(b) With continuous carbon films, the number of domains of potential
interest present on a grid is even larger.
357
 A. Brinnon et al.

Figure 5. Interfacial film of streptavidin transferred with a holey carbon film. The
streptavidin film consists of a mosaic of crystalline domains. Two large crystals, one in
each hole, are indicated by two arrows aligned along the main directions of stain
striations. Their frontiers with adjacent 2D crystals are delineated with dashed lines. Next
to the carbon threads, the streptavidin film is often disordered (*), suggesting that the
crystalline structure is disorganized when the carbon film touches the interfaciai film or
during the 'fishing step'. Scale bar: 1000 A.

358
 12: Two dimensional crystallization of soluble proteins

Figure 6. Interfacial film of annexin V transferred with a holey carbon film. Single crystal-
line domains of annexin V cover the holes. The main orientation of the lattice (arrows) is
almost conserved between adjacent holes, suggesting that the interracial film was a
single monocrystal before transfer. The crystal is built up of trimers of annexin V (circles)
assembled with p6 symmetry. Scale bar: 1000 A.

359
 A. Brisson et al.
(c) Crystalline domains of 1 um2 are large enough to provide high resolution
information and the surface of a grid is equivalent to about 106 such
domains. This explains why screening EM grids for the presence of crystals
and optimizing crystallization conditions are extremely time-consuming.

6. Characterization of the protein—lipid crystals by

optical diffraction
As mentioned above, the most standard method for characterizing the crystal-
Unity of a specimen is by optical diffraction (Figure 7). An optical diffraction
set-up consists merely of a laser source, a convergent lens, and an observation
screen placed in the diffraction plane of the lens. When an EM negative is
placed along the optical system, either before or after the lens, its diffraction
image, or diffraction pattern, is displayed on the screen. An optical diffraction
bench provides a simple and fast way of calculating on-line Fourier trans-
forms. Optical diffraction gives access to:
(a) The resolution of the crystalline order, from the position of the peaks
farthest away from the direct unscattered beam.
(b) The overall crystalline quality, from the sharpness of the diffraction peaks
and the completeness of the information.
(c) The optical conditions, like focusing and astigmatism, from the position
and shape of the rings of the contrast transfer function.

7. Conclusion
The lipid-layer crystallization method is a rational and general method for
growing highly ordered 2D crystals of macromolecules. Until now, it has been
Figure 7. Optical diffraction, (a) Scheme of an optical diffraction set-up. The principal
components of an optical diffraction set-up are: a laser source, a convergent lens, and a
screen placed in the diffraction plane of the lens. In the set-up presented here, the lens,
placed at a distance p from the laser source, is illuminated by a non-parallel beam. The
rays emerging from the lens converge at the diffraction plane, located at a distance p'
from the lens, such as: 1/p + 1/p' = 1/f (f: focal length of the lens). The advantage of using
a non-parallel beam illumination is that the size of the diffraction pattern from the EM
negative can be easily adjusted by changing the distances between the lens and the
source and/or between the EM negative and the lens. Sub-areas of the negatives are
evaluated for their crystalline quality and selected for further processing. According to
the diffraction theory, a periodic grating of period d, illuminated by a coherent beam of
wavelength \, gives rise to two diffracted beams forming an angle 9 with the direct beam,
such as: d sin O = y. On the screen placed at a distance L from the negative, two
diffraction peaks will be observed, located at a distance D from the centre, such as: D = L
tg O. As 9 is small, dD = L X = cst. (b) Example of a diffraction pattern of a negatively
stained 2D crystalline domain of annexin V (adapted from ref. 35). The diffraction peaks
are arranged onto a hexagonal lattice. The (0,6) and (6,2) reflections, at 1/13.4 and 1/11.2
A-1 respectively, are circled. Scale: 1 cm = 0.028 A-1.

360
 12: Two dimensional crystallization of soluble proteins

applied almost exclusively to soluble proteins. Its application to the field of

membrane proteins is confronted to the presence of detergents required to
maintain membrane proteins in a soluble form. These detergents are likely to
intercalate and possibly alter the integrity of the planar lipid films. However,
one can expect that solutions to this problem will be found, as for example by
combining a rapid elimination of the detergent molecules and/or the use of
lipids in a solid state during the initial step of binding.
An extension of the lipid-layer crystallization method has recently been
developed in our group with the helical crystallization of proteins by specific
361
 A. Brisson et al.
interaction with lipids forming tubular aggregates (31, 32). In this work,
biotinylated lipids were synthesized which self-assembled as tubules in
aqueous solutions. Binding of streptavidin molecules to the biotin head
groups was followed by their self-organization into helical arrays at the
tubular surface. This approach is of particular interest in electron crystall-
ography as the complete 3D structure of macromolecules arranged with
helical symmetry can be retrieved by image analysis of individual images of
tubular crystals. A variant of this novel approach has been proposed with
the incorporation of Ni+-chelating lipids into lipid tubules and the helical
crystallization of polyhistidine-containing proteins (33). The use of 2D crys-
tals formed on planar lipid layers as seeds for epitaxial growth of 3D crystals
constitutes another potential application of the lipid-layer crystallization
method (34).
The main problem associated with the application of the lipid-layer crystal-
lization method is the difficulty in obtaining reproducible results. This problem
is also encountered with other crystallization methods, and is due to a lack of
understanding and control of some of the parameters involved in crystal-
lization. We believe that the main source of irreproducibility in the case of the
lipid-layer crystallization method originates at the transfer step, particularly
when continuous carbon films are used for transfer: the presence and number
of crystalline domains, their size, or the extent of vesicular material are almost
unpredictable. This is certainly related to the lack of planarity of EM grids
and/or carbon films; the lipid layer is almost atomically flat, while the surface
of the carbon film is far from ideally flat (23). It is therefore highly recom-
mended to use holey films, at least as a control, with which more reproducible
results are obtained. On the other hand, the examples of annexin V and
cholera toxin demonstrate that the formation of highly ordered crystals can be
induced during the transfer step with continuous carbon films (18). Therefore,
the variability of the results must not hide the main interest of this method of
crystallization: it works!

References
1. Kimura, Y., Vassylyev, D. G., Miyazawa, A., Kidera, A., Matsushima, M.,
Mitsuoka, K., et al. (1997). Nature, 389, 206.
2. Nogales, E., Wolf, S. G., and Downing, K. H. (1998). Nature, 391, 199.
3. Kiihlbrandt, W. (1992). Q. Rev. Biophys., 25, 1.
4. Jap, B. K., Zulauf, M., Scheybani, T., Hefti, A., Baumeister, W., Aebi, U., et al.
(1992). Ultramicroscopy, 46, 45.
5. Uzgiris, E. E. and Kornberg, R. D. (1983). Nature, 301, 125.
6. Harris, J. R. (1992). Microsc. Anal., 13.
7. Avila-Sakar, A. J., Guan, T. L., Arad, T., Schmid, M. F., Loke, T. W., Yonath, A.,
et al. (1994). J. Mol. Biol., 239, 689.
8. Kubalek, E. W., Kornberg, R. D., and Darst, S. A. (1991). Ultramicroscopy, 35,
295.

362
 12: Two dimensional crystallization of soluble proteins
9. Mosser, G., Mallouh, V., and Brisson, A. (1992). J. Mol. Biol., 226, 23.
10. Celia, H., Hoermann, L., Schultz, P., Lebeau, L., Mallouh, V., Wigley, D. B., et al.
(1994). J. Mol. Biol., 236, 618.
11. Avila-Sakar, A. J. and Chiu, W. (1996). Biophys. J., 70, 57.
12. Kornberg, R. D. and Darst, S. A. (1991). Curr. Opin. Struct. Biol, 1, 642.
13. Brisson, A., Olofsson, A., Ringler, P., Schmutz, M., and Stoylova, S. (1994). Biol.
Cell, 80, 221.
14. Gaines, G. L., Jr. (1966). Insoluble monolayers at liquid-gas interphases. Wiley,
New York.
15. Roberts, G. (ed.) (1990). Langmuir-Blodgett films. Plenum Press, New York.
16. Mosser, G. and Brisson, A. (1991). J. Struct. Biol, 106, 191.
17. Venien-Bryan, C., Lenne, P.-F., Zakri, C., Renault, A., Brisson, A., Legrand, J.-F.,
et al. (1998). Biophys. J., 74, 2649.
18. Brisson, A., Bergsma-Schutter, W., Oling, F., Lambert, O., and Reviakine, I.
(1999). J. Cryst. Growth, 196, 456.
19. Blankenburg, R., Meller, P., Ringsdorf, H., and Salesse, C. (1989). Biochemistry,
28, 8214.
20. Frey, W., Schief, W. R., and Vogel, V. (1996). Langmuir, 12, 1312.
21. Reviakine, L, Bergsma-Schutter, W., and Brisson, A. (1998) J. Struct. Biol, 121,
356.
22. Baumeister, W. and Hahn, M. (1978). In Principles and techniques of electron
microscopy: biological applications (ed. M. A. Hayat), Vol. 8, p. 1. Van Nostrand
Reinhold Co., New York.
23. Schmutz, M., Lang, J., Graff, S., and Brisson, A. (1994). J. Struct. Biol, 112, 252.
24. Sjostrand, F. S. (1956). In Stockholm Conf. Electron Microscopy, Proc. 20.
25. Fukami, A. and Adachi, K. (1965). J. Electron Microsc., 14, 112.
26. Chretien, D., Fuller, S. D., and Karsenti, E. (1995). J. Cell Biol, 129, 1311.
27. Asturias, F. J. and Kornberg, R. D. (1995). J. Struct. Biol, 114, 60.
28. Henderson, R., Baldwin, J. M., Ceska, T. A., Zemlin, F., Beckmann, E., and
Downing, K. H. (1990). J. Mol Biol, 213, 899.
29. Voges, D., Berendes, R., Burger, A., Demange, P., Baumeister, W., and Huber, R.
(1994). /. Mol Biol, 238, 199.
30. Darst, S. A., Ahlers, M., Meller, P. H., Kubalek, E. W., Blankenburg, R., Ribi, H.
O., et al. (1991). Biophys. J., 59, 387.
31. Ringler, P., Mttller, W., Ringsdorf, H., and Brisson, A. (1997). Chem. Eur. J., 3,
620.
32. Huetz, P., van Neuren, S., Ringler, P., Kremer, F., van Breemen, J. F. L.,
Wagenaar, A., et al (1997). Chem. Phys. Lipids, 89, 15.
33. Wilson-Kubalek, E. M., Brown, R. E., Celia, H., and Millagan, R. A. (1998). Proc.
Natl Acad. Sci. USA, 95, 8040.
34. Edwards, A. M., Darst, S. A., Hemming, S. A., Li, Y., and Kornberg, R. D. (1994).
Nature Struct. Biol., 1, 195.
35. Olofsson, A., Mallouh, V., and Brisson, A. (1994). J. Struct. Biol., 113, 199.

363
This page intentionally left blank
 13

Soaking techniques
E. A. STURA and T. GLEICHMANN

1. Introduction
Once crystals of a macromolecule are obtained there are many circumstances
where it is necessary to change the environment in which the macromolecule
is bathed. Such changes include the addition of inhibitors, activators, sub-
strates, products, cryo-protectants, and heavy atoms to the bathing solution to
achieve their binding to the macromolecule, which may have sufficient free-
dom to undergo some conformational changes in response to these effectors.
In fact, macromolecular crystals have typically a high solvent content which
ranges from 27-95% (1, 2). Although, part of this solvent, 'bound solvent'
(typically 10%) is tightly associated with the protein matrix consisting of both
water molecules and other ions that occupy well defined positions in refined
crystal structure it can be replaced in soaking experiments, at a slower rate
compared to the 'free solvent'.
In this chapter we will consider the relative merits of various methods for
modifying crystals, the restraints that the lattice may impose on the macro-
molecule, and the relative merits of soaking compared to co-crystallization.

1.1 The crystal lattice

The size and configuration of the channels within the lattice of macro-
molecular crystals will determine the maximum size of the solute molecules
that may diffuse in. The solvent channels are sufficiently large to allow for the
diffusion of most small molecules to any part of the surface of the macro-
molecule accessible in solution except for the regions involved in crystal
contacts, although in some cases lattice forces may hinder conformational
changes or rearrangements of the macromolecule in crystal. In other cases,
the forces that drive the conformational changes can be sufficient to overcome
the constraints imposed by the crystalline lattice leading to the disruption of
intermolecular and crystal contacts resulting in the cracking and dissolution of
the crystals. Some lattices may be more flexible and capable of accommo-
dating conformational changes, and while crystals may crack initially, they
may subsequently anneal into a new rearrangement and occasionally improve
their crystallinity.
 E. A. Stura and T. Gleichmann
In general small changes are easily accommodated and many macro-
molecules maintain their activity in the crystalline state. This is exploited in
time-resolved crystallography to obtain structural information of transition
states of enzymes.
1.2 Reasons for soaking
The major use of soaking is for the introduction of heavy atom substances into
crystals for the determination of phases in the techniques of single (SIR) and
multiple isomorphous replacement (MIR) which is needed for the deter-
mination of macromolecular crystal structures that lack a model suitable for
molecular replacement (MR). In SIR and MIR, phase information is obtained
by analysing the changes in the intensity of reflections as a result of derivatiz-
ation with heavy atom containing reactants. The magnitude of the changes
depends on the number of electrons in the 'derivative' relative to the 'native'
protein. Changes in intensity may also be the result of changes in the unit cell
parameters of the crystal. These latter changes, which are referred to as non-
isomorphous changes, are undesirable and decrease the resolution to which
the intensity differences can be used in phase determination. In soaking we
try to minimize such changes while maximizing the incorporation of the
heavy-atom into the crystal lattice.
Another method for the determination of phases is multiple anomalous dis-
persion (MAD). This method obviates the need for heavy-atom derivatization
by incorporating anomalous scattering atoms in the crystal and by collecting
data sets at three or four different wavelengths to make the best use of the
anomalous dispersion. Because data are collected from crystals grown under
identical conditions, in some cases the same crystal, this method does not
suffer from lack of isomorphism. However, because the anomalous differences
are considerably smaller, more accurate data are needed. This is accom-
plished using longer exposures, and to avoid differences from one crystal and
another, data are collected from the same crystal at a synchrotron source.
Such experiments are carried out with crystals flash-frozen at close to liquid
nitrogen temperatures so that the diffraction is preserved for the longer periods
of time needed to collect complete data sets at each of the wavelengths on and
on either side of the absorption edge. To avoid loss of crystallinity the solvent
used in the crystallization must be exchanged for a cryo-solvent before the
crystals are flash-frozen. This is done by soaking or dipping them in an appro-
priate solution. Because crystals maintain their diffraction for longer periods
of time at cryogenic temperatures, this method is extensively utilized to
collect high resolution data from crystals of proteins irrespective of the
method used for the phasing of their structure, and is becoming common for
data collection at synchrotron facilities.
Soaking may also be the method of choice for the determination of ligand
binding sites in proteins although co-crystallization is a better alternative
when conformational changes occur as a result of ligand binding.
366
 13: Soaking techniques

1.3 Soaking of crystals versus co-crystallization

Complexes of macromolecules can be obtained either by co-crystallizing
directly from solution or by soaking preformed crystals of the macromolecule
in a ligand or reactant solution. Both methods have their own advantages and
disadvantages. Soaking of crystals of a macromolecule whose structure has
been determined in that crystal form, reduces the complexity of the crystallo-
graphic problem to that of determining the positions of the newly introduced
atoms by difference Fourier methods. However it may not be possible to
ensure that 'true' binding occurs as rearrangements in the macromolecule
may be inhibited by the crystal lattice. In co-crystallization, the formation of
the complex does not have to contend with lattice forces, but the solubility
and the conformation of the complex may be sufficiently different from that
of the native molecule that new crystallization conditions may need to be
determined. Co-crystallization of complexes can however yield crystals which
are totally isomorphous with the uncomplexed protein crystals. This can be
encouraged through the use of seeding (Chapter 7). However, the soaking of
relatively hydrophobic ligands in aqueous solution may present problems.
Such ligands have poor solubility in aqueous solutions. In these cases, a large
volume of soaking solution is used so that a one- to five-fold stoichiometric
ratio is achieved when the number of molecules in the solution are integrated
over the entire volume. Long soak times are commonly used. To increase the
water solubility of such ligands, and achieve higher ligand concentrations,
organic compounds which are miscible with water are used (Protocol 1).

Protocol 1. Preparation of solution for soaking hydrophobic

ligands

A crystal suitable for data collection will contain from 5-100 ug of protein,
a proportionate amount of ligand must be present in the soaking solution
so that a stoichiometric or higher concentration is achieved in an
appropriate solution compatible with crystal stability. The exact molarity
or stoichiometric ratio required will depend on the affinity of the
compound and is likely to vary from case to case.

Equipment and reagents

• Microbalance • Crystallization tray
• Ligand • Spot plate or capillary
. Various solvents • Microscope

Method
1. Measure 1-10 mg of ligand. Make a saturated solution of the ligand in
a suitable solvent (see Table 7) by adding solvent to the ligand until
fully dissolved. Calculate the molarity of the solution obtained.
367
 E. A. Stura and T. Gleichmann
Protocol 1. Continued
2. Test the solubility by adding 1 ul of the saturated solution to 1 ml of
the same buffer solution used for the protein. Continue adding until
the solution becomes opalescent. The molarity of the resulting
solution can be calculated.
3. If the solubility is in the millimolar range, soaking can be done directly
in the drops or capillaries (Protocol 5). If the solubility is less than
0.2 mM a soaking volume of 300 ul or greater will be needed and spot
plates or vials should be used (Figure 1).
4. Mix the appropriate volume of ligand saturated solution with the
precipitant used in the crystallization experiment so that the desired
molarity or so that ligand-protein stoichiometry will be achieved.
5. Test that the precipitant-solvent mixture to be used for soaking is
compatible with the crystal. For volatile solvents, just replace the
reservoir solution with the precipitant-solvent mixture and allow the
crystal to equilibrate with the reservoir by vapour diffusion. Check for
cracks and if possible test that crystals equilibrated in such a manner
still diffract. Later exchange the mother liquor in the drop with this
solution and repeat checks.

An alternative to soaking of hydrophobic compounds is co-crystallization.

This presents similar problems. In general it is advisable to mix the ligand
with the protein at low protein, low ligand concentrations and subsequently
concentrate and purify the complex.
Other situations in which co-crystallization may be needed is to achieve
binding of large multi-metal clusters used in the phasing of large macro-
molecular assemblies (> 106 Da) as such clusters may be too large to diffuse
through the crystal lattice. Since the channels between such assemblies are
likely to be proportionately large, soaking experiments should also be
attempted. See Thygesen et al. (3) for a review of the usage of multi-metal
clusters in phasing large assemblies.

1.4 Soaking techniques

While in some cases it is possible to transfer crystals directly from the mother
liquor from which they are grown to a fresh soak solution, more gradual
changes in the crystal soaking, ending with the desired soak solution, can be
effective in slowly annealing the crystal into the new conditions. The number
of steps to the final soaking conditions varies from crystal to crystal. Typically,
in the first stage the free solvent is replaced without major disruption, and in
later stages, pH is changed and ligands are soaked-in or exchanged. The time
needed for diffusion of ligands will vary from crystal to crystal. Intuitively,
crystals with large channels and high solvent content will equilibrate faster.
368
 13: Soaking techniques

Table 1. Organic solvents, additives, and cryo-protectantsa

Additive Concentration Usage

Ethanol 5-20% Solubilization (steroids) additive for crystallization
Methanol 5-15% Solubilization (phospholipids combined with MPD)
Hexafluoropropanol 1-5% Solubilization (very versatile, peptides and
mimetics, steroids, etc.)
2-Propanol 5-20% Solubilization (steroids), (cryo-protectant at > 70%
best in combination with others), additive in
crystallization
Glycerol 15-45% Cryo-protectant and additive for crystallization
DMSO 2-20% Solubilization of ligands and cryo-protectant
Ethylene glycol 15-45% Cryo-protectant
PEG 200-600 35-50% Cryo-protectant, precipitant
Sucrose > 50% (w/v) Cryo-protectant, best in combination with others
MPD 0.5-55% Solubilization (phospholipids combined with
methanol)
Additive for crystallization, cryo-protectant,
precipitant
Erythritol 5-35% Cryo-protectant, best in combination with others
Xylitol 5-35% Cryo-protectant, best in combination with others
Inositol 5-35% Cryo-protectant, best in combination with others
Raffinose 5-35% Cryo-protectant, best in combination with others
Trehalose 5-35% Cryo-protectant, best in combination with others
Glucose 5-35% Cryo-protectant, best in combination with others
L-2,3-Butanediol 15-45% Cryo-protectant (levo isomer, racemic mixture also
useful)
Propylene glycol 15-45% Similar to ethylene glycol

a Typical concentrations and usage for organic solvents and additives. As it is suggested throughout
the table, combinations of these compounds can be more effective to solubilize ligands and less likely
to be incompatible with the crystals. Several organic compounds are suitable both as additives to
crystallization set-ups as for use as cryo-solvents. For example the effect of slow equilibration of MPD
onto crystals of the multisubstrate adduct complex of glycinarnide ribonucleotide transformylase,
which under room temperature conditions diffracted to only 2.0 A was to extend the resolution to 1.96
A at cryogenic temperature collected with a conventional X-ray source (35). The improvement in
resolution may have been due to MPD rather than cryo-cooling.

Experimentally, by soaking the crystals in suitable dyes (for example, the

mercury containing dye merbromin and rose Bengal containing iodine, as
used by the authors) the time needed for soaking can be determined by
observing the crystals to becoming coloured. In practice, 20 minutes to one
hour between steps is a good starting point. Longer intervals, and repeated
soaks should be used when exchanging one ligand for another. For fragile
crystals, it is preferable to add further ligand to the soaking solution than to
transfer crystals between solutions. By adding volatile organic solvents or
further salt to the reservoir such solvent can be introduced into the crystals or
the salt concentration increased to stabilize crystals before soaking. This latter
step is necessary if it is known from crystallization trials that the protein-
ligand complex has a higher solubility than the protein alone. Flow cells,
369
 E. A, Stum and T. Gleichmann

which have been used to change the mother liquor in which the crystals are
bathed, such as for the introduction of a substrate, are well described
elsewhere (4, 5), pressure cells are used far the incorporation of krypton and
xenon into crystals.
Buffers are exchanged in order to change the pH, to analyse pH-induced
changes, to favour heavy-atom or drug binding, or to avoid conditions of
370
 13: Soaking techniques
Figure 1. Schematic drawing of the transfer of crystals from either a sitting drop set-up,
or a hanging drop set-up, to a well for soaking. (A) A crystal is dislodged from the drop
using a probe. If the crystal adheres firmly to the glass (or plastic) use a sacrificial crystal
placed between the probe and the crystal to push against. The crystal is then floated to
the surface to easily be picked up as in Figure 3 or in (B). (B) Crystals can be transferred
directly from a sitting drop to a capillary because the vapour from the reservoir solution
(not shown) protects the drop containing the crystal from dehydration. For a hanging
drop vapour diffusion experiment the coverslip is placed at the centre of a plastic Petri
dish within a ring of filter paper soaked in water. Evaporation of water from the filter
paper will ensure that the drop does not dry out while the crystal is picked up into the
capillary connected to the syringe. (C) The crystal is transferred into a small vial or a well
in a spot plate containing the soak solution. (D) After soaking the crystal is removed from
the soak solution for mounting. The walls of the vial used for soaking the crystal should
not be high, as this will restrict the angle at which the crystal can be picked up from the
soak solution into the capillary, as further restrictions are also imposed by the dissecting
microscope, which also limits the working angle. For soak volumes below 1 ml a spot
plate may be preferable.

incompatibility between certain heavy-atoms and the crystallization buffer.

For example, ammonium sulfate is a poor mother liquor for heavy-atoms
binding at pHs above 6 because of the production of NH3 which acts as a
nucleophile. Ammonium sulfate can be replaced by sodium and potassium
phosphate, except when uranium and rare earth compounds are used as
heavy-atoms as these form insoluble phosphates. Soaking crystals in a cryo-
protectant is necessary to favour the formation of vitreous rather than crystal-
line ice when crystals are flash-frozen. Since the solvent within the crystals
does contribute to the diffraction at low resolution it is advisable to collect a
new 'native' data set for these crystals after the buffer change to differentiate
between those changes induced by the buffer, and those caused either by
heavy-atoms, cryo-temperature, or other modifications to the crystals.

2. Soaking of substrates, activators, and inhibitors

Many enzymes remain catalytically active in the crystal (e.g. see refs 6 and 7)
and the soaking of substrates may yield a mixture of substrates and products.
In fact, with the use of synchrotron radiation, the conversion of heptenitol to
heptulose-2-phosphate in the presence of inorganic phosphate, was followed
crystallographically (8). Soaking is the method of choice for the determin-
ation of the binding sites for analogues, inhibitors, activators, substrates, and
products, since it is often difficult to determine the binding of substrates,
activators, and inhibitors to crystals when reliable phases for the native
crystals have been obtained. Co-crystallization is advisable until then, as a
means of confirming the result obtained by soaking, and when soaking fails.
When soaking crystals, the development of hair-line cracks can be a good
indication of ligand binding. Control experiments in which the native crystals
are handled in the same fashion with the same buffer should be performed.
371
 E. A. Stura and T. Gleichmann

2.1 Soaking techniques for crystals in drops

When sitting drops are used for crystallization (see Chapter 5 and ref. 9),
soaking experiments can follow Protocol 2 with the use of a syringe or
Protocol 4 using a loop. The protocol for sitting drops is easier than that used
for hanging drops since the evaporation from the reservoir solution is able to
slow down sufficiently the evaporation from the drop containing the crystals
during the procedure.

Protocol 2. Soaking of crystals grown in sitting drops

(see Figure 7)

This work is carried out under a dissecting microscope using a

magnification of x 10 to x 100.

Equipment and reagents

• Dissecting microscope • Tweezers
• Glass syringe . Whisker
• C-f lex tubing • Spot plate or vial
• Glass capillary • Modelling clay

Method
1. Connect a glass or quartz capillary tube to a 1 ml glass syringe with a
short piece of rubber tubing such as c-flex (Fisher, 14-169-5c) which
gives an excellent seal.
2. Snap open the end of the capillary with tweezers or scissors. The glass
capillary may be siliconized if the experimental situation can benefit
from a diminished adhesion of the solution to the glass wall such as
when viscous solutions are handled. After siliconizing it should be
extensively washed.
3. To increase the volume available for handling crystals, mother liquor
(20-50 ul) can be added to the drop (some crystals require the mother
liquor to contain protein for stability). If the crystals adhere to the well,
withdraw liquid from the drop and gently eject it onto a chosen crystal.
Check that the desired crystal moves in the flow.
4. If flushing with liquid fails to dislodge the crystals the probes for streak
seeding (Chapter 5) are used for this purpose. Select a thick whisker
with a sharp point or cut a new point if needed. Run the point around
the contour of the crystal, this will detach the crystal from precipitated
or denatured protein in the depression. Now push gently on the crystal
with the wide side of the whisker. Slowly apply pressure and watch for
movement. Should the crystal show signs of breaking up or cracking,
select a smaller crystal that can be sacrificed and utilize it as shown in
372
 13: Soaking techniques
Figure 1. Using a loop to dislodge the crystal is another option
(Protocol 4). Unfortunately, some crystals have severe adhesion
problems and cannot be dislodged without breaking them. For such
problematic crystals, glass pots with depressions, or microbridges,
coated with a thin film of Corning vacuum silicone grease should be
used in the original crystallization set-up.
5. Pick up crystals into the capillary by pulling back the plunger of the
syringe.
6. Transfer crystals from the syringe directly into a soak solution from
which they are later picked up and mounted for X-ray studies.

Crystals grown by hanging drop (see Chapter 5) can be flushed with mother
liquor from the coverslip into a larger container and then picked up as
described in Protocol 2 for sitting drops. Since it can be difficult to find small
crystals in a large container, the method described in Protocol 3 may be
preferable.

Protocol 3. Soaking crystals grown by hanging drops

Equipment and reagents

• Petri dish • Microscope
• Filter paper • Syringe with capillary
• Distilled water • Mother liquor

Method
1. Cut a circular piece of filter paper to fit a Petri dish 4-5 cm inside
diameter.
2. Cut out a small circle from the centre of the filter paper such that the
coverglass from the hanging drop can fit inside this without touching
the paper.
3. Soak the filter paper with distilled water.
4. Place the coverglass with the hanging drop in the centre. Mother
liquor is added to the drop (20-50 ul) and the crystals for soaking can
be picked up and soaked as described in Protocol 2. steps 2-6.

This set-up has been used for the stable transportation of crystals to
synchrotron facilities, by soaking the filter paper with mother liquor instead of
water.
Instead of a capillary and a syringe a loop can be used for the handling of
crystals as described in Protocol 4. This method is widely used for soaking
crystals in cryo-solvents.
373
 E. A. Stum and T. Gleichmann

Protocol 4. Handling of crystals using loops (see Figure 3)

Equipment and reagents

• Cryo-loop • Tweezers to open
• Microscope • Crystallization setup

Method
1. Follow Protocol 2, steps 1 and 2 if it is necessary to increase the
volume of the drop. Select a loop with a diameter about 1.5 times the
maximum size of the crystal. Loops can be made with individual fibres
from plain dental floss or can be purchased pre-made from Hampton
Research.
2. Clean the loop in methanol and wash with water. If the crystal is stuck
it can be dislodged as in Protocol 2, steps 3 and 4 or by gently pushing
with the loop. Tease the crystal to the top of the drop. When close to
the surface of the drop, place the loop under the crystal and lift it out
of the drop. Keep the loop only slightly above the drop, to avoid it
drying out and focus the microscope on the loop to ensure that the
crystal is in the loop.
3. Transfer the crystal to the soaking well as rapidly as possible. Drying
out of the solvent around the crystal is the main disadvantage of the
method.

2.2 Soaking of crystals in capillaries

Once the crystals have been introduced into the thin glass capillary, using
either of the two above described procedures, the mother liquor can be
removed by allowing the crystal to adhere to the capillary wall and pushing
the mother liquor out of the capillary onto a piece of absorbent paper, while
the crystal remains in situ because of surface tension. Crystals that do not
adhere to the capillary wall can be stopped from flowing with the mother
liquor by wedging a hair against the crystal while the solution is removed
(Figure 2).
Alternatively the solution can be removed with a thin glass capillary tube
(0.1-0.05 mm outside diameter) or a thin strip of filter paper. Once the
mother liquor is removed, with the syringe still connected to the capillary
tube, soaking is performed following Protocol 5.
This technique is particularly important for soaking compounds which are
available only in limited quantities. After data collection, capillaries may be
opened and a soak solution added to the crystal with a Hamilton syringe and
the capillary sealed for the duration of the soak with paraffin oil. The oil and

374
 13: Soaking techniques

Figure 2, Schematic representation of the various stages involved in capillary soaks. (Al
After the crystal is picked up into the capillary (Figure 1A) the mother liquor is removed
from around the crystal by pushing the liquid out, while holding the crystal in position
with a hair (in many cases the surface tension between the crystal and the capillary is
sufficient to hold the crystal in place). (B) With a thin strip of filter paper, taking care not
to touch the crystal the excess liquid is removed. (C) The soak solution is drawn into the
capillary to bathe the crystal. (D) Paraffin oil or buffer is added to the open end and the
capillary is sealed with molten wax. (E) After soaking the capillary can be snapped open
with thin-nosed forceps and the soak solution is removed using a thin piece of filter
paper. (F) The capillary can now be sealed at both ends and the crystal used for X-ray
diffraction studies. (G) Crystals that have been used for X-ray work can be soaked by
snapping of one end of the sealed capillary with forceps and opening the other end with
a hot needle. A piece of wet filter paper is placed over the crystal to prevent the crystal
from warming up during this procedure. A solution is then introduced at the broken end
of the capillary so that it bathes the crystal. Petroleum jelly is used to seal the experiment
as it is easy to remove prior to resealing the thin walled capillary tube with wax. (H) The
soaked crystal can be used for X-ray diffraction analysis.

375
 E. A. Stum and T. Gleichmann
the soak solution are then removed and the crystal used for further X-ray
studies such as for collecting an inhibitor complex data set after the native
protein data have been measured, if the crystal has survived the damage from
the first irradiation.

Protocol 5. Soaking crystals in capillaries (Figure 2)

Equipment and reagents
• Syringe with capillary . Wax
• Forceps • Filter paper
• Paraffin oil • Syringe with needle
• Microscope

Method
1. Suck the new solution into the capillary fully immersing the crystal.
2. Add paraffin oil to the open end of the capillary, leaving an air gap
between the oil and the soak solution, for the duration of the soak.
3. After the soak period has elapsed the oil and the solution are removed.
4. Remove the excess solution around the crystal with filter paper.
5. Add either soak solution and or oil to the open end of the capillary to
maintain a moist environment for the crystal.
6. Seal with wax while still attached to the syringe. A wet strip of filter
paper (5 mm wide) can be placed on the outside of the capillary to
keep the crystal cool while the ends are sealed. The crystal is now
mounted for X-ray diffraction work. Other techniques for mounting
crystals can be found in Chapter 14 and elsewhere (10).

2.3 Soaking of crystals in dilute ligand solutions

There are many situations in which it is desirable to soak crystals in diluted
solution. High concentration of heavy-atom substances can cause non-
isomorphous changes, while prolonged soaking at low concentrations is well
tolerated, some ligands are poorly in aqueous solutions or may need to be
introduced slowly into the crystals for the crystals to be able to anneal to the
changes. Protocol 6 gives some suggestions to increase the success with this
method.

Protocol 6. Soaking of crystals in dilute ligand solution

Equipment and reagents

• Microscope • Syringe with capillary for crystal transfer
• Vacuum grease • Glass cover to fit depression well or vial
• Depression plate or vial

376
 13: Soaking techniques
Method
1. Follow Protocol 2 using a large depression plate (Figure 1) or large
volume in a soak vial.
2. Soak several crystals for 20 min to several days. Harvest one crystal at
a time, mount in a capillary, and test to determine whether the ligand
or heavy-atom has bound to the crystal.
3. Remove old soak solution leaving the remaining crystals in the vial or
depression plate and add fresh solution. Continue testing the crystals
and replacing or adding more ligand or heavy-atom solution.

2.4 Cross-linking of crystals

The use of glutaraldehyde to stabilize crystals is well known from the early
work on carboxypeptidase A (11), where it was shown that cross-linked
crystals were resilient to changes in mother liquor. The cell dimensions of the
cross-linked crystals were shown to remain relatively constant, under a variety
of low and high salt conditions as well as extremes of pH, from 5-11. Such
cross-linked crystals also retained catalytic activity. Further examples, such as
the cross-linking of phosphorylase a, with 0.03% glutaraldehyde for 1 h, also
indicated that the reagent produces little change in the diffraction pattern of
the cross-linked crystals to a resolution of 5.5 A, while maintaining crystal
integrity even after major conformational changes (7).
Most cross-linking reagents link between the e-amino groups of lysine
residues (12, 13). Bifunctional diimidates of variable length provide a means
of restricting the length of the cross-links from 3.7 A for dimethyl malonic
diimidate to 14.5 A for dimethyl dodecanoic diimidate as the reacting groups
must be within the maximal distance of the reactive groups (14). Dimethyl
malonic diimidate was used to study complex crystals of glycogen phos-
phorylase b with the inhibitor glucose-6-phosphate (15). In the above study, a
2 mg/ml solution of dimethyl malonic diimidate in 0.1 M triethanolamine-
HC1, 10 mM magnesium acetate pH 7.8, was used to cross-link crystals for 2 h
before the reaction was stopped by lowering the pH to 7.1.

3. Soaking application
In this section we will consider some of the more typical applications of
soaking: heavy-atom derivatization and the soaking for cryo-crystallography.

3.1 Heavy-atom soaking and isomorphous replacement

The method of isomorphous replacement has been central to X-ray analysis
of protein crystals from the initial work on haemoglobin (16). In this pro-
cedure a single or a limited number of heavy-atoms per macromolecule are
377
 E. A. Stum and T. Gleichmann
introduced, as an addition or replacement of an endogenous atom, without
disrupting or significantly altering the crystal lattice. This addition of electrons
in the structure causes significant changes in X-ray recorded intensities which
can be used to obtain an estimate of the 'phase' for each reflection, which are
then used for the calculation of the electron density and the solution of the
structure. Such modifications to crystals of macromolecules are normally
carried out by soaking the native crystals in the mother liquor containing the
heavy-atom compound. Although obtaining a good heavy-atom derivative is a
trial and error process, there are general considerations which give the best
chance of success.
It is clear that the preparation of isomorphous derivative crystals will
depend on the pH, composition of the mother liquor, and temperature. Many
successful pH values for heavy-atom soaking are 6-8. If the pH value is below
6, most reactive groups which could bind the metal ion will be protonated
and blocked. Since many heavy-atom compounds are alkaline labile, at high
pH they may form insoluble hydroxides. Except at low pH (i.e. below 6),
ammonium sulfate is a poor mother liquor for heavy-atom binding due to the
production of the good nucleophile NH3 (17). If possible, the crystals should
be transferred to Mg or Na sulfate, or Na and K phosphate. However, an
excess of phosphate is undesirable for the binding of uranium and rare earth
metals. The temperature can change the rate of reaction and sometimes the
degree of binding. In most cases the soaking temperature will be the same as
the crystallization temperature.
Other important considerations are the heavy-atom concentration and
soaking time. The necessary concentration will depend on the solubility of
heavy-metal compound. Typically, 1-2 mM is appropriate as a starting value.
The soaking time can vary from 20 minutes to months, but for the initial
screening, 4-18 hours is sufficient. The crystal is observed continuously for
the first ten minutes of the soak, and hourly for the first four hours, and again
at the end of the soak. If the crystal appearance is unchanged from the native
crystal, and small changes are found from the data reduction, it is advisable
to increase the concentration of the heavy-atom substance and soak time. On
the other hand, if the crystal cracks, indicating that the changes have caused
non-isomorphism, or if diffraction resolution dramatically decreases, a lower
concentration and shorter soaking time should be tried. Sometimes, back-
soaking is necessary in order to reduce the cell changes to acceptable values
(18). For back-soaking, after the initial soak the crystals are transferred to
solutions with a lower concentration of heavy-atoms or to the original
mother liquor in order to reduce binding to the lower affinity sites. Another
useful criteria for evaluating the soaking conditions is the relative tempera-
ture factor of the derivative data compared to the native. When the deriva-
tive data have significantly larger temperature factors than the native it is
likely that some disorder has been introduced into the derivative crystal. A
reasonable temperature factor can often be achieved by lowering the
378
 13: Soaking techniques
concentration of the heavy-atom compound. As the heavy-atom concentra-
tion is lowered the volume of the soak solution should increase in proportion
to ensure a good stoichiometric ratio between the heavy-atom compound
and the protein.
Since many heavy-atom compounds have a very vigorous photochemistry,
soaking should be carried out under low power illumination or in the dark. A
drawer is sufficient for this. Finally, freshly prepared soaking solutions should
be used whenever possible. It is frequently the case that many heavy-atoms
(10 to 50) need to be tried before a good isomorphous derivative is found.
There is no substitute for patience and hard work at this stage.

3.2 Selecting a heavy-atom compound

This chapter is not dedicated to the selection of a heavy-atom for structure
determination. The frequency with which the various compounds have been
used successfully in phasing protein structures are listed in Tables 2-12 and
the table legends provide some chemical guidelines. The frequency can be
used in a statistical manner to evaluate the likelihood that these compounds
may be useful for the phasing of new proteins. It is suggested that other
references (e.g. 10 and 19) should also be consulted. The use of xenon and
krypton at high pressure is becoming popular for the phasing of proteins, and
equipment for their use is now available at some synchrotron facilities such as
LURE and SSRL (20, 21). The number of sites can be varied with pressure,
typically from 0.4-2 MPa. Crystals are pressurized for 30 minutes before data
collection. Since xenon binding sites are generally different from those for
other heavy-atoms this technique can be used as a second resort when the
initial soaks in heavy-atoms are not successful.

3.3 Soaking for cryo-crystallography

Cryo-crystallography provides a means of increasing the lifetime of some
protein crystals by reducing radiation damage during data collection (22, 23)
and allowing a complete data set to be collected from one crystal, often to the
same or sometimes higher resolution than crystals analysed at room
temperature (24). It has also been suggested that cooling may also increase
the internal order of parts of the protein which are mobile at room tempera-
ture (25) and also provide ways to observe enzyme substrate complexes and
unstable intermediates (26) with the use of Laue X-ray photography. It is now
widely used for collecting data at synchrotron facilities (27).
The use of several cryo-protective solvents and combinations of such
solvents was pioneered by Petsko (25). The most commonly used cryo-
protectants are glycerol (28), ethylene glycol, 2-methyl-2,4-pentane diol
(MPD) (also commonly used in crystallization) sometimes in combination
with others such as low molecular weight polyethylene glycol (200-600),
379
 E. A. Stura and T. Gleichmann

80
Table 2. Hg—mercury compoundsa

Frequency of Name Abbreviation Supplierb

usage
24 HgCI2 Ac, AI, Af
21 Hg(CH3C02)2 HgAc Ac, AI, Af, Si
20 C2H6HgP04 EMP N
17 Ethyl mercury thiosalicylate EMTS Ac
15 CH3HgCI MMCI Af, St
13 Mersalyl Al, Si
12 CH3Hg(CH3C02)2 MMAc P
12 K2Hgl4 PMTI Af, M
9 p-Chloromercuribenzene sulfonate pCMBS Al, I
8 Tetrakis (acetoxymercury) methane TAMM St
8 p-Chloromercuribenzoate pCMB Al, I
4 C2H6HgCI EMCI Af
4 Baker's dimercurial Baker's An
4 p-Hydroxymercuribenzoate pHMB Al, I
4 2-Chloromercuri-4-nitrophenol CNP Ac
4 Hg2(CO2)2 (oxalate/malonate) DMMA P
4 3-Chloromercuri-2 methoxypropyl urea CMMPU
(chlormerodrin)
3 Hg-deoxyuridine triphosphate HgdUTP Si
3 K2Hg (CN)4
3 <CH3)2Hg DMHg Al, Af, St
2 2-Chloromercuriphenol CMP CS
2 Phenyl mercuriglyoxal PMG
2 HgBr2 A, St
2 Hgl2 St
1 Hg(N03)2 St
1 Hg(CN)2 Ac, Al, St
1 HgO St
1 Dimercuri acetate DMA An
Hg2(CH3C02)2 DMDA P
1 CH3HgOH MMOH Af, St
1 CH3HgBr MMBr Af
1 p-Hydroxymercuriphenyl sulfonate pHMPS Si
1 p-Chloromercuriphenyl sulfonate pCMPS Al, Fl

a Mercury compounds are targeted to sulfhydryl groups. Short soak times 1-3 h can produce useful
derivatives with concentrations as low as 0.01 mM. Mercury has also a strong tendency to bind to zinc
sites at a histidine nitrogen. The mercury compounds can be grouped in three classes: the ionic group,
the most commonly used are mercury chloride and mercury acetate; the alkyl chain mercury
compounds, ethyl mercury phosphate, ethyl mercury chloride, methyl mercury chloride, methyl
mercury acetate; and the aromatic group, the most popular being EMTS, mersalyl, pCMBS, and
pCMB. K2Hgl4 cannot be grouped with the ionic mercurials as it tends to give different results, but it is
definitely a compound worth trying. Baker's dimercurial consisting of two mercury atoms, and TAMM,
a heavy metal cluster of four mercury atoms which has been used in the phasing of large molecular
assemblies (36-39), have good solubility and are definitely worth trying.
"See Table 13.

380
 13: Soaking techniques

Table 3. 78Pt—platinum compoundsa

Frequency Name Supplierb

71 K2PtCI4 Af, FI, St
24 cis-Pt(NH3)2CI2 Af, St
17 K2PtCI6 Ac, Af, St
H K2Pt(NO2)4 Af, AI, St
14 K2Pt(CN)4 Af, St
12 Di-u,-iodobis(ethylenediamine) diplatinum nitrate (PIP) St
6 Pt(NH2CH2CH2NH2)CI2 Af, Al, St
5 (2,2':6',2")-terpyridinium platinum chloride Fl
4 Pt(NH3)2(NO2)2 Af, AI, St
4 K2PtBr4 St
3 K2Ptl6 Af
3 K2Pt(SCN)4 Al
3 Pt(NH2CH2CH2NH2)2CI2 Af, AI, St
2 frans-Pt(NH3)2CI4 Af, St
2 K2PtBr6 St
2 Pt(NH3)2CI2 Ac, Af, AI
1 K2Pt(CN)6 Af, AI, St
12  Pt(SCN)6 P
1 K2Pt(CN)2 Af, St
a
K2PtCI4, is the most widely used compound. Platinum compounds are good ligands for methionine,
histidine, and cysteine residues (see refs 5 and 10 for more details).
bSee Table 13.

Table 4. 79Au—gold compoundsa

Frequency Name Supplierb

19 KAu(CN)2 Af, Fl, St
12 KAuCL4 FI, St
5 NaAuCI4 Fl
2 AuCI3 Af
1 KAuBr4 P, St
1 HAuCI4 Af, Fi

aGold compounds have a propensity for binding to sulfhydryl

groups, and may provide a good alternative to ionic mercurials
with which they often share sites.
bSee Table 13.

381
 E. A. Stura and T. Gleichmann

82
Table 5. Pb—lead compoundsa

Frequency Name Supplierb

14 (CH 3 )3Pb(CH 3 CO 2)2 Af, An
9 Pb(CH3CO2)2 Ac, Af
5 Pb(NO3)2 Ac, Af
3 (CH3)3PbCI AI, Af
2 PbCI2 AI, Af
1 (C2H5)3PbCI Af
(C2H5)3Pb(CH3CO2)2 An

"Outstanding among the lead compounds is trimethyl lead acetate. It

is sparsely soluble and long soaks in large volumes may be
necessary. It is an excellent reagent for hydrophobic sites, in the a/B
TCR structure determination it was found to bind in close proximity
to EMTS (40) which binds to the only cysteine not involved in a
disulfide bond found in a hydrophobic pocket. Lead can bind at zinc
and other divalent metal sites, the acetate and nitrate salts are
commonly used.
bSee Table 13.

81
Table 6. TI—thallium compoundsa

Frequency Name Supplierb

3 TICI3 Af, Al
1 TICI Af, St
1 TI(CH3CO2)3 Af, St
a
Thallium is not a heavy-atom of choice due to its extreme toxicity. It
is wise to experiment with various mercury compounds first before
resorting to this option. The relatively good solubility of thallium
salts, particularly in phosphate buffers is its main attraction.
bSee Table 13.

Table 7. 77lr—iridium compoundsa

Frequency Name Supplierb

12 K3lrCI6 Af
3 lrCI3 Ac, Af
2 Na3lrCI6 Af
1 (NH4)3lrCI6 St
1 H2lrCI6 Af, Fl, I

a Iridium shares many properties with platinum, it gives stable

anionic (e.g. IrCle63-) and cationic complexes (e.g. lr(NH3)63+).
bSee Table 13.

382
 13: Soaking techniques
76
Table 8. Os—osmium compoundsa

Frequency Name Supplierb

10 K2OsO4 Af, Al, St
6 K2OsCI6 St
2 (NH4)2OsBr6 St
1 Na2OsCI6 Al, St
1 OsCI3 Ac, AI, St

"Osmium, OsO4 is extremely toxic. It is a good reactant for ribose

moieties and the 3' terminus of RNA (41, 42).
bSee Table 13.

Table 9. 92U—uranium compoundsa

Frequency Name Supplierb

21 K3UO2F5 Sp
16 UO2(CH3CO2)2 AI, FI
15 UO2(NO3)2 AI, FI, St
6 UO2SO4 P
4 UO2CI2 P, St
1 (NH4)2U2O7 P
a
Uranyl is the third most popular heavy-atom reagent after mercury and
platinum compounds. Uranyl nitrate and uranyl acetate give similar
results. In the structure determination of the erythropoietin EMP1
complex (43) the best results were obtained with uranyl nitrate in acetate
buffer. It should be noted that K2UO2F5 is the fifth most popular single
compound used for protein structure phasing.
bSee Table 13.

74
Table 10. W—tungsten compoundsa

Frequency Name Suppliera

1 Na2WO4 Ac, Al
1 (NH4)2WS4 Af, AI, St
a
The low frequency with which tungsten compounds have been used Is
somewhat surprising. Tungstate is a phosphate mimic and its complex
with human protein phosphatase 1 was used for multiple wavelength
anomalous dispersion experiments (44). Analysis of the changes
induced by tungstate on several enzymes able to bind to phosphate
moieties by the method of reverse screening (45), it has been noticed
that in general, the solubility of the tungstate complex improved. Indeed,
crystals of FIV dUTPase, dissolved when soaked in tungstate solution
(Stura and Prasad, unpublished results) and was not be used in its
structure determination (46). The tungstate complex with deoxyribose-5-
phosphate aldolase was obtained by co-crystallization. This resulted in a
crystal form of this enzyme different from that reported in ref. 47. Multi-
tungsten clusters have been used for the phasing of fumarase C (48) and
riboflavin synthetase (49), see review (3).
bSee Table 13.

383
 E. A. Stura and T. Gleichmann

Table 11. Lanthanide compoundsa

Frequency Name Supplierb

62
8 SmCI3 Al, St
67
3 La(NO3)3 Af, St
3 63EuCI3 Af, St
64
3 GdCI3 Af
62
3 Sm(CH3CO2)3 St
71
2 Lu(CH3CO2)3 Af, St
70
2 YbCI3 Af, St
66
1 DyCI3 Af, St
63
1 Eu(NO3)3 Af, St
62
1 Sm(NO3)3 St
71
1 LuCI3 Af, St
59
1 PrCI3 St
60
1 NdCI3 Af
67
1 HoCI3 Af
66
1 Dyl3 Af
a
The combined usage of lanthanide compounds compares with
that of the acetate and nitrate salts of uranyl. The overall low
frequency of each compound reflects the very similar behaviour
of each. SmCI3, Gd2(SO4)3,and Yb2(SO4)2 all occupy the same site
as the magnesium ion at the threefold axis of the FIV dUTPase
trimer otherwise occupied by magnesium (46).
bSee Table 13.

Table 12. Other compoundsa

Frequency of Name Supplierb

usage
73
1 Ta35Br5 Af, AI, CS
58
1 BaSO4 Al
58
1 BaCI2 Al
56
1 Ba38Sr41Nb4Ol2 Af, AI
48
1 CdCI2 Ac, Af, Al
3 I2 + Kl Al

" lodination is well described in refs 5 and 50. Certain iodide salts
of heavy metals may derivatize proteins due to I" rather than
the metal itself. Tantalum and niobium compounds are of interest
for their use in phasing large macromolecular assemblies (3).
Brominated and iodinated nucleotides are used in the phasing of
protein nucleotide complexes.
bSee Table 13.

384
 13: Soaking techniques

Table 13. Suppliers of compounds in Tables 2-12"

Ac: Acros Chemicals; 711 Forbes Avenue, Pittsburgh, PA 15219, USA. Tel: (800) 227-6701.
http://www.fishersci.com/catalogs
Af: Alfa; Johnson Matthey Catalog Company, Inc., PO Box 8247, Ward Hill, MA 01835-
0747, USA. Tel: (800) 343-0660.
Al: Aldrich Chemical Co.; 1001 West St Paul Avenue, Milwaukee, Wl 53233, USA. Tel:
(800) 558-9160.
http://www.sigma.com/SAWS.nsf/Pages/Aldrich?EditDocument
An: Anatrace Inc.; 434 West Dussel Drive, Maumee, OH 43537-1624, USA. Tel: (800) 252-
1280. http://www.anaTrace.com
CS: Chem Service; PO Box 3108, West Chester, PA 19381-3108, USA. Tel: (610) 692-3026.
Fi: Fisher Scientific; 711 Forbes Avenue, Pittsburgh, PA 15219, USA. Tel: (800) 766-7000.
http://www.fishersci.com/catalogs
FI: Fluka Chemie AG; Industriestrasse 25, CH-9471 Buchs, Switzerland.
http://www.sigma.aldrich.com/SAWS.nsf/Pages/Fluke?EditDocument
I: ICN Pharmaceutical Inc.; 3300 Hyland Avenue, Costa Mesa, CA 92626, USA. Tel: (714)
545-0100.
M: Mallinckodt; 470 Frontage Road, West Haven, CT 06516, USA. Tel: (203) 933-7064.
N: Noah Technologies; 1 Noah Park, San Antonio, TX 28249, USA. Tel: (210) 691-2000.
P: Pfaltz and Bauer; 172 E Aurora Street, Waterbury, CT, USA. Tel: (203) 574-0075.
Si: Sigma Chemical Company; PO Box 14508, St. Louis, MO 63178, USA.
http://www.sigma.aldrich.com/SAWS.nsf/Pages/Sigma7EditDocument
Sp: SPECS and BioSPECS bv; Koninginnegracht, 94-95, 2514 AK The Hague, The
Netherlands. PO Box 85586, 2508 CG The Hague, The Netherlands (mailing address)
Tel: 31-70-355-4473. Fax: 31-70-355-8527.
Brandon/SPECS Inc.; (North American sales company), PO Box 1244, Merrimack,
New Hampshire 03054, USA. Tel: 603-424-2035. Fax: 603-424-2035.
St: Strem Chemical; 7 Mulliken Way, Newburyport, MA 01950, USA. Tel: (508) 462-3191.

aData for Tables 2-12 has been compiled from structures reported in: Macromolecular structures
(1991-4), and Atomic structures of biological macromolecules (1990-3) (ed. W. A. Hendrickson and K.
Wuthrich). Current Biology Ltd., London.

ethanol, propanol, xylitol, erythriol, inositol, raffinose, trehalose, glucose,

L-2,3 butanediol. The methods for soaking of crystals for this application does
not vary substantially from that used for other applications. The loop
technique is used more often specially in the 'dip and shoot' method, where
the crystal is soaked in the cryo-solvent for a few seconds allowing for only the
liquid outside the crystal to be replaced. This can be compared to the oil
technique, where crystals are transferred to oil prior to freezing them (29).
Alternatively, the cryo-solvent is added to the mother liquor in small steps,
slowly increasing the summed concentration of all the cryo-solvents to about
35%. All such techniques are aimed at preserving crystal integrity by
obtaining a transition from water to vitreous ice, preventing crystallization of
the water in the mother liquor.
385
E. A, Stura and T. Gleichmann

386
 13: Soaking techniques
Figure 3. (A) Prepare a loop (cryo-loop) from a rayon fibre. This is best done by placing
both ends of the fibre into a needle and pulling the two ends until a loop of the correct
diameter is obtained. Place a spot of epoxy at the junction of the fibre. (b) Pick up the
crystal from the drop using the loop mounted on a wooden or glass rod. (C) Soak crystal
in spot plate (easier than in vials). For long soaks, place a ring of petroleum jelly or
vacuum grease around the well, and cover with a coverglass. Place the spot plate in a
drawer for platinum and other light-sensitive heavy-atom solutions. Warning: loops used
for heavy-atom work may retain some heavy-atom soaked in the fibre. (D) After the soak,
pick up the crystal from the soak solution. (E) If the drop attached to the loop is very large,
the excess liquid can be removed by touching the outside of the loop with a thin strip of
filter paper, being careful not to get close to the crystal. (F) The crystal is then plunged
into the nitrogen stream or the stream is blocked by a card or ruler and when the crystal
is in position the card is rapidly removed. In the design used at SSRL a thin bent brass
plate clips onto the nozzle and diverts the beam while the crystal is being positioned.
When the crystal is in place the brass plate is made to springs back allowing the stream
to flash-freeze the crystal.

Protocol 7. Soaking of crystals for cryo-crystallography

Equipment and reagents

• Cryo-solvents • Microscope
• Cryo-loops • Spot plate
• Vacuum grease • Cover glass

Method
1. Select crystals for freezing of roughly 0.4 mm in each dimension or
smaller. Larger crystals are more problematic as they may develop
cracks, and form crystalline ice due to slow heat transfer. Since the
radiation damage is small or negligible at cryo-temperature, the
strength of the X-ray source and exposure times will be able to
compensate for the smaller crystal size. Fast data collection on a
wiggler line at a synchrotron radiation source is preferable than long
exposure times with conventional sources.
2. Select a cryo-solvent. See Table 1 for suggestions. Garman and
Mitchell (28) give the minimum amount of glycerol to be added to 50
typical crystallization conditions. The cryo-buffer typically is a
combination of the crystallization reservoir and a cryo-solvent. For
crystallizations from small molecular weight PEG (200-600) or MPD no
cryo-solvent is required, although the precipitant concentration may
need to be increased. If the crystallization is from PEG 4000 or higher
molecular weight PEG, replace some of the PEG 4000 by PEG 200 or
just add PEG 200 at varying concentrations. The next most popular
choices are ethylene glycol or MPD. The alcohol sugars are generally
milder, and may be tolerated by those crystals that crack under the
other conditions.

387
 E. A. Stura and T. Gleichmann
Protocol 7. Continued
3. A small amount of this buffer is picked up in a loop (Figure 3) and
shock-frozen in the nitrogen stream. Plunge loop into nozzle to flash-
freeze or block stream with a paper card, then remove it quickly. If the
buffer stays transparent it has formed vitreous ice, whereas
opaqueness indicates formation of crystalline ice which will give a
powder diffraction pattern, 'ice-rings'. If there is no loss in resolution
or increased mosaicity (30) or anisotropy minimal ice-rings can be
tolerated as most integration programs can cope with this problem if
not too severe.
4. The next step is to optimize the cryo-solvent. Check whether the cryo-
protectant causes lattice damage, this normally results in cracks, very
fine hair-line cracks manifest themselves as a brown tinge when the
crystal is observed under the microscope. Take a few diffraction
images at various concentrations and with different cryo-protectants
at room temperature with crystals mounted in capillaries to select the
least damaging cryo-solvent and to maximize the resolution limit.
Some cryo-solvents may indeed enhance diffraction. Since this pro-
cess can take a long time, if the crystals are large enough to be
analysed on a conventional source, it is best to perform these tests in
advance of synchrotron data collection. Often the functional plot of
cryo-protectant versus resolution limit has a minimum (30).
5. Pull crystal through cryo-protectant in the loop to transfer the crystal
to the cryo-buffer (Figure 3). In a stepwise transfer using solutions with
increasing amounts of cryo-protectant is important in order to reduce
osmotic shock. It is also important to keep the number of operations
small to reduce damage and subsequent increase of mosaicity. For a
small crystal 20-30 sec between transfers is sufficient and the crystal
should be frozen immediately, else the mosaicity might increase.
Evaporation of buffer also requires speedy transfer to the cryo-stream.

4. Conclusions
Soaking is most commonly used to obtain heavy-atom derivatives, although
crystallization of previously modified proteins either chemically or biologically
have also been used (31, 32). Soaking and co-crystallization are two different
approaches to achieving complexes of macromolecules. The two procedures
are both alternative and complementary to each other. By soaking effectors
into preformed crystals it is possible to analyse the structure of complexes,
only if crystal lattice constraints permit. The problem of cracking of the
crystals which may occur both when binding effectors (15) and heavy-atoms
(33), can be often resolved by the use of cross-linking agents. One must how-
ever understand that complexes obtained by soaking may differ from com-
plexes obtained by co-crystallization. Flow cells in which a constant supply of
388
 13: Soaking techniques

substrate is supplied to the enzyme in the crystal and product is washed away
may answer the problem in cases where the rate of product formation is
significantly slower than the rate of diffusion through the crystal (19, 34).

Acknowledgements
We would like to thank Dr Ping Chen for her contribution to the first edition
of this chapter and Dr Ian A. Wilson for reading and support of that work
through his grants by the National Institutes of Health Grants AI-23498, GM-
38794, and GM-38419. T. G. was supported by BMBF grant 05 641BJA 4 (to
Rolf Hilgenfeld) and by the Australian Research Council grant AD 984283
(to B. Kobe). E. S. thanks the French Atomic Energy Commission (CEA) for
support during the revision work.

References
1. Matthews, B. W. (1968). /. Mol. Biol. 33, 419.
2. Cohen, C., Caspar, D. L. D., Parry, D. A. D., and Lucas, R. M. (1971). Cold Spring
Harbor Symp. Quant. Biol., 36, 205.
3. Thygesen, J., Weinstein S., Franceschi, F., and Yonath, A. (19%). Structure, 4, 513.
4. Wyckoff, H. W., Doscher, M. S., Tsernoglou, D., Inagami, T., Johnson, L. N.,
Hardman, K. D., et al. (1967). J. Mol. Biol., 127, 563.
5. Petsko, G. A. (1985). In Methods in enzymology (eds H. W. Wychoff, C. H. W.
Hirs, and S. N. Timasheff), Academic Press, London. Vol. 114, p. 141.
6. Remington, S., Wiegand, G., and Huber, R. (1982). J. Mol. Biol., 158, 111.
7. Kasvinsky, P. J. and Madsen, N. B. (1977). J. Biol. Chem., 251, 6852.
8. Hadju, J., Acharya, K. R., Stuart, D. L, McLaughlin, P. J., Barford, D.,
Oikonomakos, N. G., et al. (1987). EMBO J., 6, 539.
9. McPherson, A. (1982). Preparation and analysis of protein crystals. Wiley, New
York.
10. Blundell, T. L. and Johnson, L. N. (1976). Protein crystallography. Academic
Press, New York.
11. Quicho, F. A. and Richards, F. M. (1964). Proc. Natl. Acad. Sci. USA, 52, 833.
12. Hunter, M. J. and Ludwig, M. L. (1962). /. Am. Chem. Soc., 84, 3491.
13. Browne, D. T. and Kent, S. B. H. (1975). Biochem. Biophys. Res. Commun., 67,
126.
14. Dombradi, U., Hadju, J., Bot, G., and Friedrich, P. (1980). Biochemistry, 19, 2295.
15. Lorek, A., Wilson, K. S., Sansom, M. S. P., Stuart, D. L, Stura, E. A., Jenkins, J.
A., et al. (1984). Biochem. J., 218, 45.
16. Green, D. W., Ingram, V. M., and Perutz, M. F. (1954). Proc. Roy. Soc., A225, 287.
17. Sigler, P. B. and Blow, D. M. (1965). J. Mol. Biol., 14, 640.
18. Mondragon, A., Wolberg, C., and Harrison, S. C. (1989). J. Mol. Biol., 205, 179.
19. Petsko, G. A. (1985). In Methods in enzymology ibid ref. 5. Vol. 114, p. 147.
20. Schiltz, M., Prange, T., and Fourme, R. (1994). J. Appl. Cryst., 27, 950.
21. Stowell, M. H. B., Soltis, S. M., Kisker, C., Peters, J. W., Schindelin, H., Rees, D. C.,
et al. (1996). J. Appl. Cryst., 29, 608.
22. Rodgers, D. W. (1994). Structure, 2,1135.
389
 E. A. Stum and T. Gleichmann
23. Tsernoglou, D., Hill, E., and Banaszack, L. J. (1972). J. Mol. Biol, 69, 75.
24. Hope, H., Frolow, R, von Bohlen, K., Makowski, I., Kratky, C., Halfon, Y., et al.
(1989). Acta Cry St., B45,190.
25. Petsko, G. A. (1975). J. Mol. Biol., 96, 381.
26. Douzou, P., Hui Bon Hoa, G., and Petsko, G. A. (1975). J. Mol. Biol., 96, 367.
27. Hadju, J., Machin, P. A., Campbell, J. W., Greenhough, T. J., Clifton, I. J., Zurech,
S., et al. (1987). Nature, 329, 176.
28. Garman, E. F. and Mitchell, E. P. (1996). J. Appl. Cryst., 29, 584.
29. Hope, H. (1988). Acta Cryst., B44,22.
30. Mitchell, E. P. and Garman, E. F. (1994). J. Appl. Cryst., 27,1070.
31. Stura, E. A., Johnson, D. L., Inglese, J., Smith, J. M., Benkovic, S. J., and Wilson,
I. A. (1989). J. Biol. Chem., 264, 9703.
32. Yang, W., Hendrickson, W. A., Crouch, R. J., and Satow, Y. (1990). Science, 249,
1398.
33. Ringe, D., Petsko, G. A., Yakamura, F., Suzuki, K., and Ohmori, D. (1983). Proc.
Natl. Acad. Sci. USA, 80, 3879.
34. Farber, G. K., Glasfeld, A., Tiraby, G., Ringe, D., and Petsko, G. A. (1989).
Biochemistry, 28, 7289.
35. Klein, C., Chen, P., Arevalo, J. H., Stura, E. A., Marolewski, A., Warren, M. S., et
al. (1995). J. Mol. Biol., 249, 53.
36. Deisenhofer, J., Epp, O., Miki, K., Huber, R., and Michel, H. (1984). J. Mol. Biol.,
180, 385.
37. O'Holloran, T. V., Lippard, S. J., Richmond, T. J., and Klug, A. (1987). J. Mol.
Biol., 194, 705.
38. Bentley, G. A., Boulot, G., Riottot, M. M., and Poljak, R. J. (1990). Nature, 348, 254.
39. Reinemer, P., Dirr, H. W., Ladenstein, R., Schaeffer, J., Gallay, O., and Huber, R.
(1991). EMBO J., 10, 1997.
40. Garcia, K. C., Degano, M., Stanfield, R. L., Brunmark, A., Jackson, M. R.,
Peterson, P. A., et al. (1996). Science, 274,209.
41. Kim, S.-H., Shin, W.-C, and Warrant, R. W. (1985). In Methods in enzymology
ibid ref. 5. Vol. 114, p. 156.
42. Stout, C. D., Mizuno, H., Rao, S. T., Swaminathan, P., Rubin, J., Brennan, T., et al.
(1978). Acta Cryst., B34, 1529.
43. Livnah, O., Stura, E. A., Johnson, D. L., Mulcahy, L. S., Wrighton, N. C., Dower,
W. J., et al. (1996). Science, 273, 464.
44. Egloff, M. P., Cohen, P. T., Reinemer, P., and Barford, D. (1995). J. Mol. Biol,
254, 294.
45. Stura, E. A., Satterthwait, A. C., Calvo, J. C., Kaslow, D. C., and Wilson, I. A.
(1994). Acta Cryst., D50, 448.
46. Prasad, G. S., Stura, E. A., McRee, D. E., Laco, G. S., Hasselkus-Light, C., Elder,
J. H., et al. (1996). Protein Sci., 5, 2429.
47. Stura, E. A., Ghosh, S., Garcia-Junceda, E., Chen, L., Wong, C.-H., and Wilson, I.
A. (1995). Proteins, 22, 67.
48. Weaver, T. M., Levitt, D. G., Donnelly, M. L, Stevens, P. P., and Banaszak, L. J.
(1995). Nature Struct. Biol., 2, 654.
49. Ladenstein, R., Bacher, A., and Huber, R. (1987). J. Mol. Biol. 195, 751.
50. Ghosh, D., Enman, M., Sawicki, M., Lala, P., Weeks, D. R., Li, N. et al. (1999).
Acta Cryst. D55, 779.
390
 a4

X-ray analysis
L. SAWYER and M. A. TURNER

1. Introduction
This chapter covers the preliminary characterization of the crystals in order to
determine if they are suitable for a full structure determination. Probably
more frustrating than failure to produce crystals at all, is the growth of beauti-
ful crystals which do not diffract, which have very large unit cell dimensions,
or which decay very rapidly in the X-ray beam, though this last problem has
been largely overcome by freezing the sample.
It is impossible in one brief chapter to give more than a flavour of what the
X-ray crystallographic technique entails and it is assumed that the protein
chemist growing the crystals will have contact with a protein crystallographer,
who will carry out the actual structure determination and in whose laboratory
state-of-the-art facilities exist. However, preliminary characterization can
often be carried out with little more than the equipment which is widely avail-
able in Chemistry and Physics Departments and so the crystal grower remote
from a protein crystallography laboratory can monitor the success of their
experiments. The reader should refer to the first edition for protocols useful
for photographic characterization but such techniques are seldom used
nowadays. It must be remembered, in any case, that X-rays are dangerous and
the inexperienced should not try to X-ray protein crystals without help.

2. Background X-ray crystallography

It is necessary to provide an overview of X-ray crystallography, to put the
preliminary characterization in context. For a general description of the tech-
nique the reader should refer to Glusker et al. (1) or Stout and Jensen (2). For
protein crystallography in particular, the books by McRee (3) and Drenth (4)
describe many of the advances since the seminal work of Blundell and
Johnson (5). Amongst many excellent introductory articles, those by
Bragg (6), published years ago, and Glusker (7) are particularly recom-
mended.
 L. Sawyer and M. A, Turner

2.1 X-rays
2.1.1 Why use X-rays
The scattering or diffraction of X-rays is an interference phenomenon and the
interference between the X-rays scattered from the atoms in the structure
produces significant changes in the observed diffraction in different direc-
tions. This variation in intensity with direction arises because the path
differences taken by the scattered X-ray beams are of the same magnitude as
the separation of the atoms in the molecule. Put another way, to 'see' the
individual atoms in a structure, it is necessary to use radiation of a similar
wavelength to the interatomic distances, typically 0.15 nm or 1.5 A and
radiation of that wavelength lies in the X-ray region of the electromagnetic
spectrum. It is also important to realize that it is the electrons which scatter
the X-rays and so what is in fact observed is the electron density of the sample.
Because the electrons cluster round the atomic nuclei, regions of high
electron density correspond to the atomic positions.

2.1.2 X-ray sources

X-rays are produced in the laboratory by accelerating a beam of electrons into
an anode, the metal of which dictates what the wavelength of the resulting X-
rays will be. Monochromatization is carried out either by using a thin metal
foil which absorbs much of the unwanted radiation or, better, by using the
intense low order diffraction from a graphite crystal. To obtain a brighter
source, the anode, which is water cooled to prevent it melting, can be made to
revolve in what is known as a rotating anode generator. For most work with
proteins, the target is copper and the characteristic wavelength of the
radiation is 0.1542 nm (1.542 A).
An alternative source of X-radiation is obtained when a beam of electrons
is bent by a magnet. This is the principle behind the synchrotron radiation
sources which are capable of producing X-ray beams some thousand times
more intense than a rotating anode generator (3). A consequence of this high
intensity radiation source is that data collection times have been drastically
reduced, making kinetic crystallography feasible (8). A further advantage is
that the X-ray spectrum is continuous from around 0.05-0.3 nm, dependent
upon the particular machine, and this has distinct advantages for the crystallo-
grapher. The use of shorter wavelengths has usually been found to prolong
the room temperature lifetime of a crystal in the X-ray beam. The main draw-
back is that synchrotrons are centralized facilities and consequently access is
significantly less convenient, particularly for preliminary work.

2.2 What is a crystal?

A crystal is a regular, repeating array of atoms or molecules in three
dimensions. It is convenient to describe such an object with the aid of a lattice,
392
 14: X-ray analysis
which is a geometric construction defined by three axes and the three angles
between them. Along each axis direction; a point will repeat at a distance
referred to as the unit translation or unit cell repeat and labelled a, b, and c,
respectively. The angles between b and c, a and c, and a and b are a, 3, and -y,
respectively. The basic building block of a crystal, then, is a parallelepiped
described by the dimensions a, b, and c and a, (3, and -y and called the unit cell.
There are seven crystal systems which arise from the only possible com-
binations of these unit cell parameters. However, it is sometimes easier to
consider a larger unit cell but with a simpler shape, for example with mutually
perpendicular axes. This choice can be illustrated in the two-dimensional
example shown in Figure 1. The choice of the basic building block containing
a single 'molecule' can be made in a variety of ways because the lattice is no
more than a geometrical construction affording a convenient description of
the repeating figure. Crystallographers adopt the convention that the unit cell
which is chosen is the one with angles nearest to 90°. Such a cell with only one
copy of the molecular structure is called primitive but, as noted above, a more
convenient cell may have two or even four copies (see Figure 1, where the
non-primitive, centred cell is at the right). There are 14 so-called Bravais
lattices which can be constructed in three dimensions (there are five in 2D).
As an example of the limited number of lattices, construct a centred square
lattice and it is evident that a smaller, primitive square lattice is also present.
Although the basic building block of a crystal is the unit cell and the lattice
produced by its repetition has a characteristic symmetry (see Table 1), within
the unit cell there may be further symmetry. For example, the molecule itself
may have symmetry about an axis which is either a proper rotation of 360°,
180°, 120°, 90°, or 60° only, or an improper one which involves 'inversion'
through the point. Both of these can be illustrated with a molecule like
methane. A threefold rotation axis (120°) is evident when the molecule is
viewed along an H-C bond whereas a fourfold improper rotation axis bisects
an H-C-H angle in the plane of the other H-C-H so that a 90° rotation of one

Figure 1. Each of the unit cells shown in this two-dimensional example is a valid choice
for the lattice of points. The cell on the right is a centred cell and has twice the contents
of the others.

393
Table 1. The crystal systems and related data for a chiral molecule

System Necessary Bravais Classb Numbersc Available space groupsd Multiplicitye

cell parameters latticea
Triclinic a, b, C, a, B, y P 1 1 P1 1
Monoclinlc a, b, c, B P 2 3-4 P2, P21 2
(a = y = 90°) C 5 C2 4
Orthorhombic a, b, c P 222 16-19 P222, P2221, P212121, P21212 4
(a = B = y = 900) C 20-21 C222, C2221 8
F 22 F222 16
I 23-24 I222, 1212121 8
Tetragonal a(= b), c P 4 75-78 P4,P41mP42, P43 4
(a = p = 7 = 90°) I 79-80 14,141 8
P 422 89-96 P422, P42,2, P4,22, P4122, 8
P4222, P42212, P4322, P43212
I 97-98 1422, 14122 16
Trigonal a(= b), c, y = 1200 P 3 143-145 P3, P31, P32 3
(a = B = 90°) R 3 146 R3 3
P 312 149, 151, 153 P312, P3,12, P3212 3
P 321 150,152,154 P321, P3121, P3221 6
a (= b = c), R 32 155 R32 6
a = p = -y = 90°
Hexagonal a(=b), c, y = 120° P 6 168-173 P6, P61, P62, P63, P64, P65 6
(a = p = 90°) 622 177-182 P622, P6122, P6222, P6322, 12
P6422, P6522
Cubic a ( = b = c) P 23 195, 198 P23, P213 12
(a = P = y = 90°) F 196 F23 48
I 197 I23 24
P 432 207-8, 212-3 P432, P4232, P4332, P4,32 24
F 209-210 F432, F4,32 96
I 211,214 I432, 14132 48
a
b
The lattice types are: P, primitive; C, C-face centred; F, all faces centred; I, body centred. Alternative lattice types may occasionally be chosen.
The symbols under class refer to the rotational symmetry axes which are a characteristic of it.
cNumber
d
refers to the number in International tables.
The Herman-Mauguin nomenclature for space groups gives the lattice type first, then the symmetry elements in an order which depends upon the crystal
system. Refer to International tables for X-ray crystallography, Volume A for a fuller explanation of these symbols.
eMultiplicity gives the number of copies of the asymmetric unit in the unit cell.
 14: X-ray analysis
hydrogen about this axis brings it to a point on the opposite side of the C atom
to an adjacent H atom. Proteins are made up of L-amino acids and nucleic
acids have a chiral ribose unit which preclude centres or mirrors. The com-
bination of these symmetries and the crystal systems leads to the 32 point
groups or crystal classes, of which only 11 can accommodate protein molecules.
The rotations referred to above are the only ones allowed in the formation of
a crystal but of course other rotations about a point within the molecule are
possible as in the case of a spherical virus which has 532 point group
symmetry. Only the threefold and twofold axes can be exploited in building
up the crystal, leaving the fivefold axis as a non-crystallographic symmetry
element.
As well as the rotational symmetry possibly present in a unit cell, trans-
lational relationships between molecules also exist. The spatial repetition of a
crystal is such that a convenient packing involves axes which combine a
rotation with a translation. For example, if a rotation of 180° together with a
translation of half a unit cell along the axis of rotation is applied twice, it will
produce not the initial molecule (as with a pure rotation) but an equivalent
one in the next cell. Such an axis is a screw axis and several consistent types
exist.
It can be shown mathematically that there are only 230 combinations of
these symmetry elements possible in three dimensions. Thus, any crystal must
have a unit cell which conforms to one of these combinations, its space group.
Further, the presence of symmetry elements within a unit cell means that
there are at least two copies of the molecule which are related by an algebraic
relationship: if there is an atom at position x, y, z in a cell with a screw axis
parallel to the b axis, there must be an atom at -x,1/2+ y, -z. The effect of this,
is to reduce the crystallographer's problem to one of locating the atoms in the
asymmetric unit, rather than in the whole unit cell. Because all proteins and
nucleic acid crystals comprise only one optical isomer, there are only 65 space
groups available for such chiral molecules. Table 1 shows the available crystal
systems, classes and space groups for a protein.

2.3 How do X-rays interact with crystals?

The explanation of how X-rays are scattered by crystals is largely the result of
a beautiful simplification by Bragg, resulting in the law which bears his name.
Consider a crystal lattice, represented in Figure 2 by the rows of points A, B,
C. For X-rays X2 scattered from row 2 to enhance those scattered from row 1,
X1, there must be an integral number of wavelengths difference. The relation-
ship between the spacing of the rows, d, the wavelength, \, and the angle at
which the emergent ray is observed relative to the direction of the rows, 9, is:
n.\ = 2d.sin 0
Thus, as Bragg pointed out, X-ray diffraction can be regarded as the reflection
of the beam of X-rays from the planes of points in the crystal lattice. Provided
395
 L. Sawyer and M. A. Turner

Figure 2. X-rays (X1 X2, X3) reflected from lattice planes A, B, C. To observe a scattered
beam of X-rays in direction R, the thickened path must equal a whole number of
wavelengths. The ray from plane C travels twice as far as that from B, and so on.

there are a large number of planes contributing to the interference, the

position in space at which a given reflection is observed is highly defined.
These positions are defined by the crystal lattice and since very few, if any,
atoms actually lie on the lattice points, the scattered intensity is modulated by
the atomic arrangement within the unit cell. To repeat, the direction of a
diffracted ray is defined by the crystal lattice, the intensity of the ray depends
upon the atomic arrangement within the unit cell. One further point concerns
n, the order of diffraction, which is the number of wavelengths difference
between the scattering from adjacent planes; the higher the order, the larger
the angle of scattering. Alternatively, the scattering can be considered as
arising from planes which are closer together: e.g. using the equation above, it
can be seen that a reflection at 0 can be considered either as the nth order
from planes of spacing d, or the first order from planes of spacing d/n.
Crystallographers generally adopt the latter approach.
A diffraction pattern for a protein crystal contains many reflections which
must be appropriately indexed and the most convenient system is to use the
order of diffraction with respect to each of the unit cell axes. The Miller
indices as they are called, which were derived originally to label crystal faces
for mineralogical studies, are illustrated in Figure 3. Each index along the a, b,
and c axes, respectively, is derived by taking the reciprocal of the intercept
that the first plane of the set not passing through the origin, makes with each
axis in turn. Thus, the 100 planes are the set which have a spacing of a X 1/1
on the x axis, b X 1/oO on the y axis, and c X l/oo on the z axis. The 200 planes
have a spacing a X 1/2 on the x axis, and so on. Notice that the planes h00 are
all parallel to one another but the spacing decreases with increasing h. Hence
the angle of diffraction increases with increasing h, consistent with Bragg's
Law. The letters h, k, and l are used to refer to the indices in general terms.
Each of the many sets of planes defined by the lattice gives rise to one
reflection and Figure 4 shows the relationship in two dimensions of the planes
in the crystal (real space) to the points in diffraction space or reciprocal space.
396
 14: X-ray analysis

Figure 3. The set of planes 123 are shown as they cut a unit cell. The intercepts on b
occur every 1/2 and on c every 1/3.

Figure 4. A diagram illustrating the relationship between sets of planes in a crystal in real
or direct space and points representing a diffracted X-ray beam in reciprocal (or diffrac-
tion) space. Notice that the direction from the origin of reciprocal space (large point) to
any point, e.g. 130, is perpendicular to the planes in the crystal and that the length is
proportional to the reciprocal of the plane spacing.

The points can be seen to make up another lattice (reciprocal lattice) whose
axes and angles are derived from those of the crystal. This idea can be
extended to three dimensions. It is important to realize that each reflection
contains a contribution from every atom in the crystal and, conversely, each
atom in the crystal contributes to every reflection. Thus, as the crystal is
moved about in the X-ray beam, reflections flash out and can be recorded
when the geometrical arrangement of X-ray beam, crystal orientation, and
detector satisfies Bragg's Law.
To help understand diffraction from a crystal, there is a construction intro-
397
 L. Sawyer and M. A. Turner

Figure 5. The Ewald construction. For clarity, this is shown as a planar diagram but IXO is
the diameter of a sphere of radius 1/\.

duced by Ewald and shown in Figure 5. As we move the crystal, the reciprocal
lattice also moves about a fixed origin. With the crystal, X, as centre, a sphere
is drawn of radius 1/\. and the origin, O, of the reciprocal lattice is taken as the
point where the X-ray beam leaves the sphere after passing through the
crystal. As the crystal is rotated about the z axis (perpendicular to the page)
the reciprocal lattice rotates until the point P lies on the surface of the sphere.
The point P is the 410 reflection arising from the planes of spacing d410. The
angles at IX and XP, i.e. IXA and BXP are equal to 0 so that OXP = 20 and
OP is perpendicular to the crystal planes AXB. Now OP = 2 X XO X sin 0 =
2 X (1/X) X sin 0. However, OP = l/d410 and so l/d410 = (2/\) X sin 0 which is
Bragg's Law. Thus, the Ewald sphere gives a readily understandable way of
relating the orientation of the crystal to the diffraction pattern observed. In
order to collect a set of X-ray data, it is necessary to move the crystal (and in
some methods, the detector) in such a way that every reciprocal lattice point
passes through the sphere of reflection (Figure 6). There are various ways of
achieving this, some of which are described in Section 4.
The space group in which a molecule crystallizes may impose certain con-
ditions on the reflections which can be observed so that by looking at the
diffraction pattern of the crystal, it is often possible to determine the space
group unambiguously. Furthermore, the higher the symmetry of the crystal,
the less data is actually required to be collected. A diffraction pattern has a
centre of symmetry since reflections in opposite directions from the same
planes must have the same intensity (I(h k l) = l(h k l) is Friedel's Law) (see
Figure 3). Thus the diffraction symmetry shown in Table 2 has a centre
of symmetry even though the space groups do not. The effects of the lattice
type and symmetry elements upon the diffraction pattern are shown in Table 3
398
 14: X-ray analysis

Figure 6. The Ewald sphere intersected by several reciprocal lattice layers. As the crystal
is moved, the reciprocal lattice pivots about O but XA and OB remain parallel at all times.
As shown, a film placed perpendicular to XO will record a series of concentric circles. As
the crystal is rotated through a small angle about AX, the circles will become extended
into lunes as the neighbouring spots on each level pass through the sphere of reflection.

and the effect can be explained with reference to Figure 2. If the beam X3
scattered from row C is one wavelength behind X1 scattered from row A, then
X2 scattered from row B is exactly half a wavelength behind and it will cancel
out the reinforcing contributions from rows A and C. Thus, interposing planes
midway between the planes separated by the unit cell repeat as is the case for
a centred lattice, leads to a systematic absence of reflections. Further, if a
twofold screw axis is perpendicular to the planes, there will always be an
identical (but rotated) set of scatterers to row A, on row B. Only when the
index is even along the axial direction will constructive interference occur and
the reflection be observed. Notice that simple rotation axes do not generate
any systematic absences.

2.4 How is a protein crystal structure solved?

The formation of a magnified image by a light microscope involves collecting
all of the scattered light waves in the objective lens which recombines them in
the correct way to produce the magnified image. But what is 'the correct
way'? Associated with each wave is not only its amplitude but also its phase
relative to the unscattered light. The focusing by the objective lens uses both
amplitude and phase to produce the magnified image. In the case of X-rays,
the crystal produces a diffraction pattern which needs to be recombined in the
correct phase relationship but in this case, no lens exists which is able to
perform the task and in recording the pattern as one must, the vital phase
399
 L. Sawyer and M. A. Turner

Table 2. Equivalent data for the chiral point groups

System Class Laue group Equivalent reflectionsa

Triclinic 1 -1 l(hkl)
Monoclinic 2 2/m I(hkI), I(hkI)b
Orthorhombic 222 mmm l(hkl), I(hkl), I(hkI), l(hkl)
Tetragonal 4 4/m I(hkl), I(khl), I(hkl), I(khl)
422 4/mmm l(hkl), l(k h l
) , l(hkl), l(khl)
I(hkl), l(khl), l(hkl), l(khl)
Trigonalc 3 -3 I(hkl), I(kil), l(ihl)d
312 -3m1 I(hkl), I(kil), l(ihl), I(khl), l(ikl), I(hil)
321 -31m I(hkl), I(kil), l(ihl), I(khl), l(ikl), I(hil)
Rhombohedral 3 -3 I(hkl), l(klh), I(Ihk)
32 -3m I(hkl), I(klh), l(lhk), I(khl), l(khl), l(hlk)
Hexagonalc 6 6/m I(hkl), l(kil), I(ihl), I(hkl), I(kil), l(ihl)
622 6/mmm l(hkl), I(kil), I(ihl), I(hkl), I(kil), l(ihl),
I(khl), I(ikI), I(hil), l(hil), I(khl), l(ikl)
Cubic 23 m-3 I(hkl), I(hkl), I(hkl), l(hkl), I(klh), l(klh),
l(klh), I(klh), l(klh), l(Ihk), l(Ihk), l(lhk)
432 m-3m I(hkl), I(hkl), I(hkl), I(hk-l), I(klh), I(lhk),
I(klh), l(klh), I(lhk), l(Ihk), l(lhk), l(klh),
I(khl), l(hlk), l(lkh), l(hlk), l(lkh), l(khl),
l(hlk), I(lkh), I(khl), l(khl), l(lkh), l(hlk)

aThe reflections listed here are identical . If Friedel's Law holds then l(hkl) = l(hkl) and this generates
an equal number of equivalent reflections. In protein crystallography, anomalous scattering which
leads to a breakdown in Friedel's Law, is used to help with phasing the reflections and so the two sets,
equivalent to I(hkl) and I(hkl) must be kept separate.
bThe underlined reflections are those which are required to specify the Laue symmetry with the
others being generated by repeated application of the symmetry elements.
cThe axes in the trigonal and hexagonal systems referred to here are a = b, c, a = B = 90°, y = 120°
dWhen hexagonal axes are being used, i = hk.

information is lost. It must be calculated and this 'phase problem' is central to

crystallography. Ironically, if the positions of the atoms are known, then the
phase for each reflection can be calculated. Whilst this phase problem may
seem insuperable, if the positions of only a few heavy atoms are known,
whether these are added by soaking into crystals in the traditional way, or
introduced during protein biosynthesis with selenomethionine (see Chapter
3), their contribution can be calculated and this is generally sufficient to solve
the phase problem for a protein. The preparation of heavy metal derivatives
of proteins has been dealt with in Chapter 13. It should be pointed out that
molecular replacement (9) is applicable where a similar structure already
exists and this is increasingly found to be the case.
A phase must be calculated for each reflection to be included in the calcula-
tion of the electron density map. The more X-ray reflections that are phased
and included, the clearer the map will be and the better will be the resulting
model of the protein. Thus the resolution of the data is usually reported and
this refers to the minimum plane spacing included in the calculation; thus for a
400
Table 3. Conditions affecting possible reflections

Element Symbol Reflection observed for Notes

Primitive lattice
Lattice centred on the C face C hkl with h + k even The C face is contained by a and b
Face centred lattice hkl with h, k, and I all odd or all even
Body centred lattice I hkl with h + k + I even
Rhombohedral lattice -h + k + I = 3n '= 3n' means divisible by 3
Twofold screw axis II c 2, 001 with I even For an axis along a, the row is h00
Threefold screw axes II c 31,32 001 with I = 3n The two possible threefold axes have the same
pitch but opposite hands
Fourfold screw axes II c 001 with I = 4n
001 with I even cf. the twofold screw axis
Sixfold screw axes II c 61, 65 001 with I = 6n
62, 64 001 with I = 3n cf. the threefold screw axes
63 001 with I even cf. the twofold screw axis
 L. Sawyer and M. A. Turner
3.5 A map, all reflections with plane spacings greater than or equal to 3.5 A
will be included. The higher the resolution, the greater the amount of X-ray
data which must be measured. Disregarding the symmetry of the reflection
data, the total number of reflections is approximately 5V/d3 where V is the
unit cell volume and d is the resolution.

2.5 Importance of preliminary characterization

There are a number of reasons why the preliminary characterization of a
newly crystallized molecule is important. Most obviously, the first point to
establish is that the crystal does diffract X-rays. As part of the process of
checking that the crystal does diffract, some idea of the crystal lifetime in the
X-ray beam will be obtained together with the resolution which can be
achieved. Even when a crystal appears perfect, it should not be assumed that
it will be suitable for X-ray work. Occasionally, some or all of the crystals in a
batch give no discernible diffraction pattern. The reason for this is obscure
but possible avenues to explore before abandoning the particular
crystallization conditions used, are:
(a) Try crystals from different drops, tubes, or preparations.
(b) Search for crystals with a different morphology and X-ray them; some-
times different forms appear in the same tube.
(c) Cool the crystal before and during the X-ray exposure, possibly down to
liquid nitrogen temperatures (10). If the crystal has already been cooled,
try it at room temperature.
(d) Use synchrotron radiation with a short wavelength. It has been found to
extend the lifetime of sensitive crystals (11).
(e) Attempt to crosslink the molecules in the crystal with a bifunctional
reagent such as glutaraldehyde (12).
These ideas may also be worth trying if the crystals produce feeble or
rapidly fading diffraction, and (e), in particular, may allow successful handling
of crystals which are very fragile.
If the spots obtained on initial images are streaked rather than the well-
defined spots illustrated in Figure 7, then the crystal is likely to have a degree
of disorder which may render successful structure determination impossible.
The only recourse then is to re-examine the crystallization procedure.
The aim of the preliminary X-ray investigation should be to determine the
unit cell dimensions and the space group. Not only must these be known to
solve the crystal structure but also, with the crystal's X-ray lifetime, they
dictate the strategy for efficient data collection. The amount of data to be
collected is determined by the diffraction symmetry of the crystal and it is
often possible to reduce the number of exposures by ensuring that the crystal
402
 14: X-ray analysis

Figure 7. A 3° oscillation photograph of a hexagonal, cytochrorne c4 crystal taken with

synchrotron radiation. Lunes from several major zones ([100], [1T01, and (0101) can be
seen clearly. The rotation axis was horizontal.

is mounted in a particular way. For example, it is best to mount a hexagonal

crystal with the sixfold axis roughly parallel to the rotation axis of the
instrument.
It is normal practice to determine the volume occupied per unit molecular
weight (V m , often called the Matthews' coefficient) since this can be used to
determine the number of molecules in the asymmetric unit. V m has been
found to be around 2.4 A 3 /dalton for globular protein crystals, although this
value is subject to quite large fluctuations (13). It is oblained by dividing the
unit cell volume by the product of the protein molecular weight and the
number of equivalent positions (asymmetric units). Unfortunately, it is often
found with large unit cells that more than one value of Vm is reasonable and in
such a case some biological insight may help resolve any ambiguity. If it is
possible to determine the crystal density (see Chapter 2) and the weight loss
on drying, the protein molecular weight can be calculated, which when com-
pared with the known value, also gives the number of molecules per asym-
metric unit. The approximate solvent content can also be calculated from the
formula:
VJ7 0 ) = 100(l-1.23/V m ).
One final point about the preliminaries is that biological information may
emerge about the subunit structure. If it is found that the asymmetric unit
contains half of the expected molecular weight, the protein must consist of an
403
 L. Sawyer and M. A. Turner
even number of subunits and it is probable that the molecular twofold axis
coincides with a crystallographic one. This will be consistent with the space
group which must possess such a symmetry element. Conversely, if a crystal is
found to have three or four molecules in the asymmetric unit of a relatively
low symmetry space group, then one should be alerted to the possibility of
having missed a higher symmetry space group.

3. Mounting crystals
Mounting a protein crystal is a procedure which requires a reasonable degree
of manual dexterity. It is impossible to be dogmatic about the right and wrong
way, and each person develops their own technique, modifying it as required
from protein to protein depending on the size, strength, temperature behaviour,
need to exclude oxygen, or toxicity. Although early workers did dry their
crystals (14), drying out of mother liquor in the crystal generally disrupts it
such that no useful data can be collected. Mounting methods are therefore
designed to maintain the interstitial mother liquor as it is in the drop from
which the crystal grew. 'Flash-cooling' is a way of greatly reducing radiation
damage (15-17) but it can also help with the problem of fragile crystals by
preventing the loss of the interstitial water necessary to maintain crystal
integrity. Indeed, nowadays many laboratories routinely freeze their crystals.

3.1 Initial examination with a microscope

3.1.1 Observation
Well-formed protein crystals examined under the light microscope exhibit a
symmetric arrangement of edges and faces which are related to the packing of
the molecules. Thus, examination of crystal morphology may give a first
glimpse of the symmetry of the unit cell. A stereo-zoom dissecting micro-
scope, ideally fitted with a crossed polarizing attachment, with a magnification
in the range X 10 to X 40 is best for such examination since crystals which
cannot be readily seen with such an instrument are probably not going to
diffract sufficient X-rays, even with synchrotron radiation. It is important to
ensure that the illuminating light source does not heat the microscope stage
lest undue evaporation and denaturation occurs. The use of crossed polarizers
can indicate the direction of a principal axis. Rotating the crystal on the stage
in the dark field (polarizer and analyser at 90°), the crystal appears as a light
colour until an optic axis lies along the direction of the polarizer whereupon
extinction occurs, depending on the crystal system. During a full rotation,
extinction occurs every 90°. This effect will not be observed for cubic crystals,
for tetragonal, trigonal, and hexagonal crystals viewed along their unique axis
or for non-crystalline material. Note that the crystal should not be contained
in a plastic container (like a tissue culture plate) if polarized light is to be used
because these containers affect the polarization, usually producing splendid
404
 14: X-ray analysis
colours. Salt crystals are usually highly coloured, even if they are small and, if
a crystal is thought to be salt rather than protein, pressure with a fine probe
will produce an audible 'plink' as the tip slips off the hard salt crystal. A
protein crystal on the other hand will shatter with very little pressure at all. It
is well worth getting to know the crystal habit and its relationship to the axes
since this saves considerable time if alignment in the X-ray beam is required.

3.1.2 Selection of a crystal for mounting

Protein crystals with dimensions of 0.2-0.5 mm are most suitable for use in an
X-ray diffraction experiment. Use of smaller crystals is possible, however the
diffraction pattern tends to be weaker requiring longer exposure times and
possibly poorer resolution. On the other hand, crystals much larger than
0.5 mm may not be uniformly bathed in the X-ray beam (depending on the
size of beam collimator used) and generate their own problems associated
with absorption of X-rays. Larger crystals may also pose problems in a low
temperature experiment because of difficulties in freezing them uniformly.
Crystals for X-ray work should be single and should appear transparent
(containing no cracks) with well defined edges and faces. Birefringent single
crystals, when observed under a polarizer, extinguish light sharply when
rotated through 360°. Less obviously twinned or multiple crystals may some-
times be detected if different sections of the crystal extinguish light at differ-
ent rotations of the microscope stage. Crystals which have grown into one
another, or have grown as clumps may be carefully split using a fine probe or
a fresh scalpel blade. Gently touch the crystal at the point where the extra
piece joins the chosen crystal keeping the blade parallel to the direction
in which the crystals are to be separated. A gentle pressure is usually all that
is required since crystals will generally cleave readily along the axial
directions.

3.2 The basic techniques

The methods suggested take practice and a number of trials, preferably with a
batch of old or non-precious crystals, before a decision can be made as to
which steps are best suited to maintaining the crystal. The first involves draw-
ing the crystal up into the Lindemann tube by a pipette or syringe attachment,
allowing controlled movement of the crystal and buffer in the capillary. For
flash-freezing, several methods are possible but the one described in Section
3.2.4 has the advantage of being straightforward and as such is worth
practising until perfected.

3.2.1 Mounting for room temperature

This procedure is summarized by Protocol 1 and illustrated by Figure 8.
405
 L. Sawyer and M. A. Turner

Figure 8. A diagram of the steps involved in mounting a crystal. The numbers refer to the
steps described in Protocol 7 in the text.

Protocol 1. Mounting crystals for room temperature data

collection

Equipment and reagents

• Glass slides and coverslips • Scalpel with a new and pointed blade
• Lindemann tubes (thin-walled glass or • Spirit burner, low temperature soldering
quartz capillary tubes, 1.0 or 0.7 mm iron, or Bunsen burner with which to melt
diameter) wax, wax, forceps, supply of appropriate
• A supply of pipettes sufficiently small to fit buffer
inside a 1.0 mm Lindemann tube: use • Filter paper strips or cotton thread for
smaller bore Lindemann tubes or Pasteur soaking up excess mother liquor inside the
pipettes drawn to narrow diameter in a Lindemann tubes
flame • Nylon fishing line, fine wire, or glass fibres
• Small glass vials ('pots' made by cutting a for manipulating crystals
vial to give a container perhaps 8 mm in « Diamond cutting tool for quartz capillaries
diameter by 6 mm deep are ideal) • Rubber or plastic tubing to connect
• Sealing grease (silicone grease, petroleum Lindemann capillary to syringe
jelly) • Disposable tuberculin (1 ml) syringe
• Modelling clay (Plasticine)
• Probe with a fine tip (like a sewing needle)

Method
1. Attach a 1.5 cm length of rubber tubing to the end of a disposable
tuberculin syringe (1 ml). With the diamond tool, score a Lindemann
tube near the closed end and break it neatly. Insert the wide end of the
406
 14: X-ray analysis
Lindemann tube into the rubber tubing and if necessary, roll back the
ends of the tubing to improve the seal. In this way, you have created a
narrow-bore pipette.
2. Draw the crystal with a small amount of mother liquor or handling
buffer into the Lindemann tube. The tuberculin syringe is small
enough that the apparatus can be held in one hand with the thumb
available for drawing up on the plunger. Remove the end of the tube
from the crystal droplet and continue drawing the crystal further up
the Lindemann tube. With the crystal at the desired height in the tube,
draw a final small plug of mother liquor into the end of the Lindemann
tube.
3. Seal the open end of the Lindemann tube with wax. Soak a small piece
of tissue in water and drape the wet paper over the tube at the height
of the crystal. This is to protect the crystal from heat conduction up the
tube while melted wax is being applied to the end. Seal the end of the
tube with wax. Applying a small piece of Plasticine to the wax makes
handling easier and allows the capillary to be stuck on a microscope
slide or the table top for subsequent manipulation.
4. Score and break with the diamond tool the Lindemann tube a second
time — this time 'above' the crystal. If breaking or cutting a glass tube
without scoring it, add a drop of wax just to the crystal side of where
the break is to be made, but well clear of the crystal, to prevent the
tube collapsing when being broken.
5. Dry the remaining buffer from around the crystal with the aid of a
shred of filter paper inserted through the open end of the tube. If
necessary, larger volumes can be removed with a finely drawn-out
Pasteur pipette or small bore Lindemann/syringe assembly as in step
1, before the drying stage. A dry mount is preferred for two reasons.
The faces are more easily visible when aligning the crystal and the
absence of solvent may reduce the effects of crystal slippage. It should
be kept in mind, however, that in this dry atmosphere, the crystal is
susceptible to solvent loss, thus the following steps should be
performed as quickly as possible.
6. Seal the other end with wax using the wet tissue draped once again
over the tube to protect the crystal.

Many variations are possible at the discretion of the mounter. For example,
it may be preferable to have two plugs of buffer in the Lindemann tube; one
on either side of the crystal. This can be accomplished by adding a small
amount of buffer to the top of the tube before the final wax seal is applied. If
it is necessary to reposition the crystal, opening up the wax plug is most easily
done with a heated needle.
407
 L. Sawyer and M. A. Turner

3,2.2 Mounting for low temperatures

It is often necessary to maintain the crystal at temperatures lower than
ambient. For example, crystal stability may require working at around 0°C.
Working at low (around 0°C) temperatures can be accomplished by housing
the X-ray equipment in a cold room but, since such a system is not always
available, the alternative is to pass a stream of cooled, dried air or nitrogen
over the crystal from a nozzle mounted on the X-ray instrument as close to
the crystal as possible (18), Usually the stream is co-axial with the capillary
tube and goniometer head (see Figure 9) and a plastic collar added to protect
the instrument. Below room temperature but above the free/ing point of the
solution, the crystals should be mounted in the cold room otherwise when the
crystal is cooled on the camera or diffractometer, the temperature gradient
produced by the cooler will lead to water distilling along the tube and
dissolving the crystal. Arrange the cooler to pass cold gas along the tube and
ensure that a plug of mother liquor is only at the end of the capillary closest to
the cooler so that condensation will occur preferentially at the drop rather
than at the crystal.
However, if cooling is required at all, it now makes sense to 'flash-freeze'
the crystal and carry out the data collection at around 100 K (near liquid

Figure 9. A typical goniometer head with the crystal sealed in a capillary fixed upon it.
The key shown is for adjusting the slides and arcs. It has a fine Alien key at the other end
for locking the arcs after adjustment. The threaded ring at the base will screw onto an
X-ray camera or ctiffractometer.

408
Figure 10. The cryo-loop method of crystal mounting. (a) A goniometer head with the
magnetic base and mounted loop. (b) A close-up of the tip of a typical mounted loop of
about 0.5 mm diameter. (c) A protein crystal flash-frozen in its cryo-solvent, (d) Typical
equipment for handling frozen crystals. Left to right: a goniometer head, a magnetic
base, a CrystalCap with mounted cryo-loop, an 18 mm cryo-vial, a cryo-vial in a plastic
handling tube, a mounted cryo-loop in a plastic pipette tip for mounting a crystal, a
plastic handling tube made from a Pasteur pipette and ideal for filling the cryo-vial with
liquid nitrogen.

409
 L. Sawyer and M. A. Turner
nitrogen temperature). As noted already, recent developments in the cryo-
crystallography of biological molecules have meant that in many laboratories
data collection at 100 K is now routine. A general overview of these develop-
ments is given by Rodgers (19). If it becomes apparent that very low tempera-
tures will be required (because conventionally mounted crystals have
unworkably short lifetimes in the X-ray beam) a different mounting pro-
cedure must be applied. Whilst there is some benefit in equilibrating the
protein crystals in cryo-protectant, it is not strictly necessary, though it is
essential to have a suitable cryo-protectant mother liquor. This can often be
obtained by mixing crystal mother liquor with increasing concentrations of
glycerol until a capillary containing the solution remains transparent when
plunged into liquid nitrogen.
In addition to the equipment mentioned in Section 3.2.1, some special
equipment is needed, both for mounting but also for X-ray work. A popular
and convenient device for maintaining the crystal at 100 K whilst in the X-ray
beam is the Cryostream made by Oxford Cryosystems but most X-ray gener-
ator manufacturers provide an equivalent. Much of the equipment for crystal
mounting can conveniently be obtained from Hampton Research but it can
also be hand-made in the laboratory. Protocol 2 and Figure 10 illustrates how
the crystal is mounted in a cryo-loop and also shows a convenient and cheap
way of handling the mounted loop once frozen in liquid nitrogen.

Protocol 2. Mounting crystals for cryo-crystallography

Equipment and reagents
• Mounted cryo-loops • A goniometer head with arcs: modified
• A magnetic base heads (20, or bought from Charles Supper)
• A CrystalCap: these are convenient both for simplify the placing and removing the
handling and for long-term storage crystal on the goniometer head, but are not
strictly necessary for horizontal goniometer
• Handlers for the CrystalCap: these can spindles (e.g. the Mar Image Plate), once
conveniently be made out of disposable the method below has been mastered
Pasteur pipettes or disposable pipette tips,
but tweezers can also be used • A Dewar of liquid nitrogen

Method
1. Attach a holder (a 1 ml plastic pipette tip is suitable) to the magnetic
base end of a mounted cryo-loop. Insert the vial part of the
CrystalCap into a small (5 x 5 x 1 cm) expanded polystyrene float
with a hole to fit the vial firmly, the open end being up.
2. Select the crystal to be frozen and place on the same slide a drop of
cryo-protectant buffer solution.
3. Submerge the float and vial in the liquid nitrogen until it is cold
(boiling ceases) and the vial is full of liquid nitrogen.
4. Carefully scoop up the crystal with the cryo-loop in which it will be
held by surface tension, and immediately immerse it in the cryo-

410
 14: X-ray analysis
protectant. The time required for cryo-protection varies depending on
buffer system and cryo-protectant used. This will need to be deter-
mined by trial and error and it is sensible to begin with some less
good crystals.
5. Carefully scoop up the crystal from the cryo-protectant and plunge it
into the liquid nitrogen in the vial. Allow the magnetic base to cool
down (boiling ceases) before screwing the cap onto the vial.
6. The crystal is now mounted and frozen—it must now be maintained
at approximately this temperature for as long as it is required.
7. Either transfer the vial and cap to a suitable storage Dewar containing
liquid nitrogen if the X-ray work is not to proceed immediately or take
the crystal in the Dewar to the X-ray laboratory. To transfer the crystal
to the X-ray diffractometer, the method will depend upon the exact
arrangement of the goniometer spindle. If vertical (e.g. R-Axis Image
Plate), a goniometer with an extension to permit the frozen crystal to
be positioned so that it points downwards, is essential. If the spindle
is horizontal (e.g. Mar Image Plate) proceed as follows.
8. Place the goniometer head with magnetic base attached on the
spindle and adjust the z-translation so that the cryo-loop when in
position will be in the centre of the cold gas stream. This adjustment
is conveniently done with the mounted loop when finding a suitable
cryo-protectant. It may be helpful to withdraw the nozzle of the cold
stream slightly to allow some extra room for the next stage. Also
arrange for the arc with the largest angular displacement to be
vertical and at the upper extremity of its travel. The crystal when
mounted will then be pointing downwards by some 20-30°
depending on the goniometer head.
9. Attach the mounted loop to the base, unscrew the cap and withdraw
the vial of nitrogen, holding the vial (re-)filled with liquid nitrogen in a
holder so that the metal part of the cap is free to be located on the
magnetic base. The crystal should now be in the stream of cold dry
nitrogen. Care should be taken not to disturb the X-ray back-stop.
10. Reposition the nozzle, if necessary, to be as close to the crystal as
possible without interfering with the X-ray beam. This is conveniently
done by a second person as soon as the vial is removed. Ensure that
there are no drafts in the laboratory which might deflect the flow from
the cryo-cooler and similarly do not breathe at the crystal whilst
mounting it.
11. Begin the X-ray measurements.

4. X-ray data
The fundamental data about a crystal which must be known before the
structure solution can be attempted are the unit cell dimensions and the space
411
 L. Sawyer and M. A. Turner
group. Until relatively recently, these data were always determined first, in
order that the strategy for data collection could be optimized, a necessary
prerequisite for crystals with limited X-ray lifetime. Nowadays however, most
data are recorded automatically by the oscillation/rotation method, often
before the space group and cell dimensions are known, and the main purpose
of examining the first images is to determine that the crystal is single, un-
cracked, and diffracts X-rays. In addition, some clues about the space group
can be obtained from the symmetry of the pattern near the principal zones,
but this is not really necessary. The data images are stored on tape or disc and
the images further processed (fairly) automatically by computer. The unit cell
dimensions are calculated and the space group is determined, once the data
have been processed, by plotting out layers as 'mock' precession photos to
observe the systematic absences more easily and hence determine the space
group. Thus the strategy has now become one of shoot first and ask questions
later.

4.1 Oscillation methods for data collection

Most data collection nowadays is done by the oscillation method (21) with
some sort of area detector mostly, imaging plate systems. Recently, the
development of the charge coupled device (CCD) has provided the sensitivity
and dynamic range of the imaging plate but has eliminated the time-
consuming scanning step typical of phosphor-imaging plate technology.
Whatever device is used for data capture, however, the oscillation method
remains the technique of choice. Unlike the precession method, it is not
necessary and indeed is undesirable, that the crystal be perfectly aligned
before data collection starts. The crystal is rotated through a small angle
(0.1-1.5°) about an axis perpendicular to the X-ray beam. As the crystal
rotates (about an axis through X perpendicular to the page in Figure 5), the
successive reciprocal lattice planes (rotating about O in Figure 5) cut the
Ewald sphere producing extended circles or lunes as shown in Figures 6 and 7.
Several passes or oscillations through the rotation range minimize the effects
of fluctuations of X-ray intensity. Provided the rotation angle is not too large,
adjacent levels will not overlap and data from many layers can be collected on
each image. The size of the oscillation range is chosen depending on the
detector, crystal cell dimensions, Bravais lattice type, properties of the
incident X-ray beam, and crystal mosaic spread. An estimate of the maximum
permissible rotation angle can be obtained from:

where dmax is the maximum resolution for which data are required, q is the
spacing of planes perpendicular to the X-ray beam (e.g. a when the a axis is
parallel to the X-ray beam), and A, which is typically 0.1-0.3°, is the reflecting
range of the crystal, or mosaic spread. The strategy adopted in current prac-
tice is to use a relatively large rotation range, or as large a range as possible.
412
 14: X-ray analysis
However, the high degree of automation available with area detector and
image plate software and the cheapness of disc storage allow oscillation
ranges less than or comparable to the actual diffraction spot size to be used
which in turn allows integration of the spot as it traverses the reflecting
position. Oscillation ranges larger than a typical spot result in the collection of
intensity not just of the spot itself but of background 'in front of and 'behind'
the spot as well, thus reducing the effective signal-to-noise ratio for that spot.
Larger oscillation ranges, however, are used to minimize the time of overall
data collection with image plates because of the relatively large time require-
ment for scanning each image before the plate can be used for collection of
the next frame.

4.2 Optical alignment

Adjust the 'height' (the distance of the crystal from the base of the
goniometer head) of the crystal using the adjustment on the instrument and
the z-translation on the goniometer head, as necessary. Centring of the crystal
is then carried out to ensure that it remains in the X-ray beam during rotation
about the spindle. The two bottom sledges on the goniometer head are used
to do this (NB: NOT the arcs). First, rotate the crystal through 360°, noting its
position in the microscope cross-hairs at 0, 90, 180, and 270°. To centre the
crystal, put one sledge perpendicular to the direction of view — this will
normally correspond to either the 0/180° or the 90/270° positions. Move the
sledge to place the centre of the crystal at the midpoint of the 0/180° (or
90/270°) readings and repeat for the other sledge. The process is repeated
until the crystal (not the tube or the cryo-loop) is stationary through a full
rotation. Note that the cross-hairs on the telescope may not define the centre
of the rotation.

4.3 Crystal characterization with an area detector/image

plate
It is assumed that assistance is available to the user getting started on data
collection. Two pre-collection files are required for area detector data collec-
tion, the flood field and brass plate images, necessary to calibrate respectively
intensity and spatial fluctuations on the surface of the detector. Ideally, these
calibrations should be done every time the detector is moved to a new
distance and they must be done at the distance at which data collection will
occur. The calibrations are done using an 57Fe source and cannot be done
when the crystal is in place on the goniometer. Thus, initially, the calibrations
must be carried out before any knowledge of crystal cell dimensions is
available.
Calibration is also required for image plate data collection in order to
determine the position of the beam as the centre of the diffraction pattern.
This is often carried out by capturing the concentric circle diffraction pattern
413
 L. Sawyer and M. A. Turner
of wax but may also be recorded on each image if the beamstop has a tiny
hole in it allowing a 'centre' spot to be exposed. A typical strategy for data
collection is described in Protocol 3.

Protocol 3. Data collection

Equipment
• X-ray generator or synchrotron producing • Imaging plate diffractometer
monochromatic radiation around 0.1 nm • Appropriate graphics workstation and
wavelength software

Method
1. With the detector at the desired distance, centre the crystal in the X-ray
beam as described above. The shorter the crystal-to-detector distance,
the higher is the resolution to which data can be measured but the
greater the likelihood of spot overlap. If nothing is known beforehand,
it saves time to use the setting already in use.
2. Check the diffraction pattern by exposing a frame for an arbitrary
length of time and oscillation range, for example, 120 sec and 0.25°
with an area detector, 10 min and 1° with an image plate. If spots are
visible to the edges of the image, it may be desirable to swing the
detector out to a non-zero 20 angle, or decrease the crystal-to-detector
distance. This will help determine, according to Bragg's equation, the
resolution to which the crystal diffracts.
3. Having decided the length of time to be spent exposing each frame
and the oscillation range desired to achieve spot separation even at
the edges of the detector, begin data collection. Depending on the
system and programs used, it is recommended that data processing
be started as soon as possible. It may become obvious while
attempting to process the data that problems exist with the crystal. If
this is the case, the decision can be made to end the measurement and
try another crystal without wasting detector time.

Data processing packages are as varied as the hardware used to measure

intensities. In general, after the calibrations of the detector face are made, the
orientation of the crystal with respect to the laboratory system must be
determined. A series of frames is read and a peak search procedure records
the positions of strong, well-defined spots to be used in autoindexing. Figure
11 shows a typical workstation screen with the observed pattern displayed and
the predicted pattern superimposed. The unit cell dimensions, the crystal
orientation, and the crystal-to-detector distance are modified to obtain the
best fit of predicted to observed pattern. The procedure now to be described
414
Figure 11. A picture of a typical workstation during the processing of oscillation data from
an image plate system, (a) The image recorded is shown on the left wit ha magnified part of
the image shown inset on the right from which it can be seen that the spots are single and
not overlapping, (b) The same image as in (a) but with the predicted diffraction pattern
superimposed. The inset on the right shows how well the prediction fits the image.

415
 L. Sawyer and M, A. Turner
is that used in the program XDS developed by Kabsch (22). The autoindexing
routine begins by assigning a reciprocal-space vector to each spot. Low
resolution differences between these reciprocal lattice points are accumulated
in clusters and are sorted by decreasing population. The first two which are at
an angular separation > 45° are chosen and indices are assigned to them.
These are used as a basis set from which the remaining difference-vector
clusters can be indexed. Originally, it was expected that space group and cell
dimensions of the crystal were known prior to running the autoindexing
routine however contemporary algorithms allow both orientation and un-
known cell dimensions to be determined. Alternative choices of cell dimen-
sions are given with associated agreement factor allowing statistical
consideration of all the possibilities.
The autoindexing routine employed by DENZO (23) uses a different
algorithm coined 'real space indexing' whereby a complete search of all
possible indices of a reflection is carried out using a Fast Fourier transform.
Once the three best linearly independent vectors with minimal unit cell
volume are found, the cell is 'reduced' to describe a standard basis for the
description of the unit cell. In DENZO, a basis set for each of the 14 Bravais
lattices is found and a distortion index is calculated for the peaks in the peak
search list. The user must then, on the basis of the magnitude of the deviation
from ideal Bravais lattice symmetry, decide upon most likely cell dimensions
and the space group.

4.4 Determination of space group

As noted above, most software packages which are used to control area
detectors provide a means of estimating resolution limits and unit cell dimen-
sions. Thus there is no real need to obtain cell dimensions by an independent
method. The software (see Protocol 4) also usually gives the Bravais lattice
and all that remains to be done is to determine the space group which is best
done by displaying the various principal zones (e.g. hk0, h0l, 0kl) on the
workstation used for data processing.

Protocol 4. Space group determination

Equipment and reagents

• Appropriate graphics workstation, software and associated printer/plotter

Method
1. Observe the symmetry of the diffraction pattern of the zero level zones
('mock' precession photographs) which must be consistent with the
unit cell parameters and lattice type already determined. This gives the
diffraction (Laue) symmetry given in Table 2 for the 11 relevant classes
(and helps to ensure that principal zones have indeed been identified).
416
 14: X-ray analysis
Diffraction symmetry always has an inversion centre. For example, a
triclinic cell, P1, has -1 diffraction symmetry. It is the appearance of
extra symmetry which allows labelling of crystal class. At this point
the axes can be assigned as a, b, or c such that a, p, and -y are close or
equal to 90° (unless a trigonal or hexagonal cell is suspected) and the
cell is primitive (see Table 2). A zero layer photograph by definition,
arises from either the hk0, h0I, or 0kl sets of planes. These can be
assigned arbitrarily in the case of certain space groups. The Inter-
national tables for X-ray crystallography (24) will help with the task of
assigning axes according to crystallographic convention. In general,
the unique axis is b for monoclinic cells and c for cells of higher
symmetry. The upper layer images, which contain no reciprocal axes,
must also be assigned as hkn, hnl, or nkl where n > 1.
2. Index the spots, h, k, I on each image. Be aware that systematically
absent reflections also require indexing.
3. Analyse the systematically absent reflections in the diffraction pattern.
This pin-points the space group often, but not always, uniquely. Use
the axial absences to identify any screw axes.
4. Check that assignments of systematic absences are consistent with
upper level images as well. The upper layers also allow, for example,
distinction between a sixfold and a threefold axis. (These look the
same on a zero level photograph.)
5. Identify, with use of the International tables, a list of space groups
compatible with the observed diffraction patterns. In some cases there
is no ambiguity: e.g. P212121, whilst in others no distinction is possible
until the structure solution is under way, e.g. 1222 and 1212121 have
identical systematic absences as have the enantiomorphs P3-|21 and
P3221 where only the hand of the screw axis differs.
6. Try to find as high a symmetry space group which is consistent with
your observations and work to lower symmetries as need be.
7. Determine the approximate number of molecules in the unit cell from
the unit cell dimensions, the molecular weight of the molecule, and
Vm. Knowing the crystal system helps in this, e.g. if the crystals are
orthorhombic, there must be a multiple of four molecules in the unit
cell. (Note that a 'molecule' may also be some identically repeated
portion of the protein or polynucleotide.)

4.5 Other techniques for diffraction data collection

Precession photography gives an undistorted image of a reciprocal lattice
plane. It does, however, require care and experience to align the crystal and it
is fairly slow and as has been said, it is seldom used now. Two other methods
which may be used are mentioned here only for completeness. Protein crystal-
417
 L. Sawyer and M. A. Turner
lographers do occasionally use the Laue technique but this is exclusively
carried out at synchrotron sources since white radiation is required. Speed is
the main benefit of this technique which can record a full diffraction pattern in
a few seconds. It is not a technique used for the initial characterization of a
newly crystallized protein, however. The other method which can be used is
diffractometry. Modern four-circle diffractometers are the backbone of small
molecule crystal structure laboratories and have sophisticated control pro-
grams which allow cell dimensions to be obtained with little if any user inter-
vention. However, most instruments use Mo radiation and the crystal-to-
detector distance is often not large enough to allow easy resolution of spots
with spacings typical of proteins. The instrument consists of a series of
concentric circles, three forming an Eulerian cradle capable of rotating the
crystal to (nearly) any angle relative to the X-ray beam, the fourth moving the
detector in a horizontal plane. Considering the Ewald construction again (see
Figure 5), using the three circles of the goniometer, it is possible to orient the
normal to any desired set of crystal planes in the horizontal plane in such a
way that the normal bisects the angle between the incident X-ray beam and
the detector. This satisfies Bragg's Law and the reflection is observed by
stepping the crystal from one side of the exact bisecting position to the other,
thus moving the crystal through the reflecting position. A plot of detector
counts versus angle will then show a peak as the reflection passes through the
Ewald sphere. If such an instrument is available, preferably with a Cu tube
and an extension on the detector arm fitted with a helium path, it may be
worth trying to determine the cell dimensions. Whilst this approach may
appear to be the simplest, and with good crystals it is convenient and very
accurate, it is very time-consuming since reflections are measured one at a
time.

5. Concluding remarks
The object of this brief excursion into X-ray crystallography has been to intro-
duce the ideas and methods required to collect the information necessary for
the first publication on a new crystalline material. Such papers should include
not only the purification and crystallizing conditions, which should be repro-
ducible, but also the techniques employed to obtain the X-ray diffraction data
and the crystal lifetime in the X-ray beam and on the shelf. The unit cell
dimensions and space group together with the resolution obtainable from a
crystal have been the main concern of this chapter. Vm, the number of
molecules in the asymmetric unit, the solvent content, and any comments about
the subunit structure are also generally mentioned. Increasingly, molecular
replacement techniques will reveal similarities to known structures and so the
'crystallization note' is often superseded by the preliminary structure,
obtained rapidly from the complete data set which is collected from the first
crystals. Finally, protein crystallographers always enjoy talking about their
418
 14: X-ray analysis
subject and the number of groups around the world has risen considerably
since the first edition of this book. You will have discovered that the
technique requires a modicum of dedication and therefore do seek guidance
in getting your project under way.

References
1. Glusker, J. P., Lewis, M., and Rossi, M. (1994). Crystal structure analysis for
chemists and biologists. VCH, New York.
2. Stout, G. H. and Jensen, L. H. (1989). X-ray structure determination, 2nd edn.
John Wiley, New York.
3. McRee, D. E. (1993). Practical protein crystallography. Academic Press Inc., New
York.
4. Drenth, J. (1994). Principles of protein X-ray crystallography. Springer-Verlag,
New York.
5. Blundell, T. L. and Johnson, L. N. (1976). Protein crystallography. Academic
Press, London.
6. Bragg, W. L. (1968). Sci. Am., 219, 58.
7. Glusker, J. P. (1994). Methods Biochem. Anal, 37, 1.
8. Johnson, L. N. and Hajdu, J. (1990). Eur. J. Biochem., 29, 1669.
9. Rossmann, M. G. (1990). Acta Cryst., A46, 73.
10. Hope, H. (1990). Annu. Rev. Biophys. Biophys.Chem.,26, 107.
11. Acharya, R., Fry, E., Stuart, D., Fox, G., Rowlands, D., and Brown, F. (1989).
Nature, 337, 709.
12. Quiocho, F. A. and Richards, F. M. (1964). Proc. Natl. Acad. Sci. USA, 52, 833.
13. Matthews, B. W. (1968). J. Mol. Biol, 33, 491.
14. Hodgkin, D. C. and Riley, D. P. (1968). In Structural molecular biology (ed. A.
Rich and N. Davidson), pp. 15-28. Freeman, San Francisco.
15. Henderson, R. (1990). Proc. Roy. Soc. Land., B241,6.
16. Teng, T. Y. (1990). J. Appl. Cryst., 23, 387.
17. Garman, E. A. and Mitchell, E. P. (1996). J. Appl. Cryst., 29, 584.
18. Hajdu, J., McLaughlin, P. J., Helliwell, J. R., Sheldon, J., and Thompson, A. W.
(1985). J. Appl. Cryst., 18, 528.
19. Rodgers, D. W. (1994). Structure, 2, 1135.
20. Engel, C., Wierenga, R., and Tucker, P. A. (1996). J. Appl. Cryst., 29, 208.
21. Arndt, U. W. and Wonacott, A. (ed.) (1978). The rotation method in crystallo-
graphy. North-Holland Publishers, Amsterdam.
22. Kabsch, W. (1988). J. Appl. Cryst., 21, 67.
23. Otwinowski, Z. and Minor, W. (1997). Methods in enzymology (eds C. W. Carter
and R. M. Sweet), Academic Press, London. Vol. 276, pp. 307.
24. Hahn, T. (ed.) (1987). International tables for X-ray crystallography. D. Reidel
Publishing Co., Dordrecht, Netherlands.

419
This page intentionally left blank
 Al
List of suppliers
Aldrich-Chemical Co., Inc., 1001 W. St Paul Avenue, PO Box 355, Milwaukee,
WI 53201, USA. (chemicals)
Alpha Laboratories Ltd., Eastleigh, Hampshire, UK. (multiple liquid dispenser)
American Can Company, Greenwich, CT 06830, USA. (Parafilm® 'M',
laboratory film)
Amersham
Amersham International pic., Lincoln Place, Green End, Aylesbury,
Buckinghamshire HP20 2TP, UK.
Amersham Corporation, 2636 South Clearbrook Drive, Arlington Heights, IL
60005, USA.
Amicon Division, W. R. Grace and Co., 72 Cherry Hill Drive, Beverly, MA
01915, USA. (filters, membranes)
Anderman
Anderman and Co. Ltd., 145 London Road, Kingston-Upon-Thames, Surrey
KT17 7NH, UK.
Applied Biosystems, Inc., 850 Lincoln Center Dr., Foster City, CA 94404,
USA and Birch wood Science Park North, Warrington, Cheshire WA3 7PB,
England, (biochemical instrumentation, chemicals)
Appligene, route du Rhin, BP 72, 67402 Illkirch Cedex, France, (biochemicals)
Bachem, Hauptstrasse 144, CH-4416 Bubendorf, Switzerland, (detergents)
BDH Limited, Broom Road, Poole, BH12 4NN, UK. (electrophoresis
products)
Beckman, 4550 Noris Canyon Road, PO Box 5101, San Ramon, CA 94583,
USA. (centrifugation, pipetting station)
Beckman Instruments
Beckman Instruments UK Ltd., Progress Road, Sands Industrial Estate,
High Wycombe, Buckinghamshire, HP12 4JL, UK.
Beckman Instruments Inc., PO Box 3100, 2500 Harbor Boulevard, Fullerton,
CA 92634, USA.
Becton-Dickinson and Co., Clay Adams Div., 299 Webro Road, Parsippany,
NJ 07054, USA. (Falcon plasticware)
Becton Dickinson
Becton Dickinson and Co., Between Towns Road, Cowley, Oxford OX4 3LY,
UK.
 List of suppliers
Becton Dickinson and Co., 2 Bridgewater Lane, Lincoln Park, NJ 07035, USA.
Bender and Hobein GmbH, D-8000 Munchen 2, Lindwurmstrasse 71,
Germany, (free flow electrophoresis)
Bijhoelt and Heuvelen SV, The Netherlands, (transparent and adhesive
plastic foils)
BioBlock Scientific, BP 111, F-67403 Illkirch Cedex, France, (scientific
equipments)
BioRad, 1414 Harbour Way South, Richmond CA 94804, USA. (HPLC, IEF
equipments, biochemicals)
Boehringer Mannheim, PO Box 310120, D-6800 Mannheim 31, Germany.
(biochemicals)
Bio
Bio 101 Inc., c/o Statech Scientific Ltd, 61-63 Dudley Street, Luton, Bedford-
shire LU2 OHP, UK.
Bio 101 Inc., PO Box 2284, La Jolla, CA 92038-2284, USA.
Bio-Rad Laboratories
Bio-Rad Laboratories Ltd., Bio-Rad House, Maylands Avenue, Hemel
Hempstead HP2 7TD, UK.
Bio-Rad Laboratories, Division Headquarters, 3300 Regatta Boulevard,
Richmond, CA 94804, USA.
BioWhittaker, Inc., 8830 Biggs Ford Road, Walkersville, MD 21793, USA.
Boehringer Mannheim
Boehringer Mannheim UK (Diagnostics and Biochemicals) Ltd, Bell Lane,
Lewes, East Sussex BN17 1LG, UK.
Boehringer Mannheim Corporation, Biochemical Products, 9115 Hague Road,
P.O. Box 504 Indianapolis, IN 46250-0414, USA.
Boehringer Mannheim Biochemica, GmbH, Sandhofer Str. 116, Postfach
310120 D-6800 Ma 31, Germany.
British Drug Houses (BDH) Ltd, Poole, Dorset, UK.
Brookhaven Instrument Corp., 750 Blue Point Road, Holtsville, NY 11743,
USA. (light scattering instrumentation)
Bunton Instrument Co., Inc., 615 South Stonestreet Avenue, Rockville, MD
20850, USA. (microgrippers)
Calbiochem Behring Diagnostics, 10933 N. Torrey Pines Road, La Jolla, CA
923037, USA. (biochemicals, detergents)
Cambridge Repetition Engineers Ltd., Green's Road, Cambridge, CB4 3EQ,
UK. (dialysis buttons for crystallization)
CEA verken AB, S-152 01 Strangnas, Sweden. (X-ray films)
Charles Supper Company Inc., 15 Tech Circle, Natick, MA 07160, USA.
(crystallographic equipment)
CJB Developments Limited, Airport Service Road, Portsmouth, Hampshire
PO35PG, UK. (large-scale preparative electrophoretic apparatus)
Cole and Palmer Instrument Co., 7425 N. Oak Park Avenue, Chicago, IL
60648, USA. (scientific equipments)
422
 List of suppliers
Coming, Inc., Science Products, MP-21-5-8, Corning, NY 14831, USA.
(glassware, pipettes)
Costar Nucleopore®, One Alewife Center, Cambridge, MA 02140, USA and
Costar Europe, Ltd., PO Box 94, 1170 AB Badhoevedrop Sloterweg 305a,
1171 VC Vadhoevedrop, The Netherlands, (titration and crystallization
plates, pipettors)
Cruachem Ltd., West of Scotland Science Park, Acre Road, Glasgow G20
0UA.
Difco Laboratories
Difco Laboratories Ltd., P.O. Box 14B, Central Avenue, West Molesey, Surrey
KT8 2SE, UK.
Difco Laboratories, P.O. Box 331058, Detroit, MI 48232-7058, USA.
Douglas Instruments Ltd., 255 Thames House, 140 Battersea Park Road,
London SW11 4NB, UK. (automatic batch crystallization system)
Dow Corning Corp., Dow Corning Center, Box 0994, Midland, MI 48686-
0994, USA. (silicone oil, grease)
Dupont de Nemours and Co., Concord Plaza, Wilmington, DE 19898, USA.
(chemicals, instrumentation)
Du Pont
Dupont (UK) Ltd., Industrial Products Division, Wedgwood Way, Stevenage,
Herts, SGI 4Q, UK.
Du Pont Ltd., NEN Life Science Products, PO Box 66, Hounslow TW5 9RT,
UK.
Du Pont Co. (Biotechnology Systems Division), P.O. Box 80024, Wilmington,
DE 19880-002, USA.
Dynatech Laboratories, Inc., 14340 Sullyfield Circle, Chantilly, VA 22021,
USA. (titration plates for crystallization robots)
Eastman-Kodak Co., 343 State St., Rochester, NY 14650, USA and Kodak
House, Station Road, Hemel Hempstead, Herts HP1 1JU, UK. (chemicals,
films)
Enraf Nonius Delft, PO Box 483, 2600 AL Delft, The Netherlands.
(diffractometry)
Euromedex, Produits de Recherche, 29 rue Herder, F-67000 Strasbourg,
France, (chemicals, protease inhibitors)
European Collection of Animal Cell Culture, Division of Biologies, PHLS
Centre for Applied Microbiology and Research, Porton Down, Salisbury,
Wilts SP4 OJG, UK.
Everett's Co., Parkgate, Nr, Southampton, UK. (vacuum wax, seals, and
lubrifiants)
Falcon (Falcon is a registered trademark of Becton Dickinson and Co.).
Fisher Scientific Company, 711 Forbes Avenue, Pittsburgh, PA 15219^1785,
USA. (biochemicals, scientific equipments)
Flow Laboratories, Woodcock Hill, Harefield Road, Rickmansworth, Herts.
WD3 1PQ, UK.
423
 List of suppliers
Flow Laboratories International SA, via Lambro 23/25, I-20090 Opera
(MI), Italy, (biochemical equipments, Linbro plate, CrystalPlate and
coverslips)
Fluka
Fluka-Chemie AG, CH-9470, Buchs, Switzerland.
Fluka Chemicals Ltd., The Old Brickyard, New Road, Gillingham, Dorset
SP8 4JL, UK.
Fluka Chemie AG, Industriestrasse 25, CH-9470 Buchs, Switzerland, (bio-
chemicals, detergents)
Genset SA, 1, rue Robert et Sonia Delaunay, 75011 Paris, France.
Genzyme Corporation, 75 Kneeland Street, Boston, MA 02111, USA.
(protease-free deglycosylation enzymes)
Gibco BRL, Bethesda Research Laboratories, Life Technologies, Inc. PO
Box 6009, Gaithersburg, MD 20877, USA. (biochemicals, growth media)
Gibco BRL (Life Technologies Inc.), 3175 Staler Road, Grand Island, NY
14072-0068, USA.
Gibco BRL (Life Technologies Ltd.), Trident House, Renfrew Road, Paisley,
Scotland, PAS 4EF, UK.
Gilson Medical Electronics, Inc., 72 rue Gambetta, BP 45, F-95400 Villers-le-
Bel, France and 3000 W. Beltine Hwy., PO Box 27, Middleton, WI 53562,
USA. (sample changers)
Gow-Mac Inc., PO Box 32, Bound Brook, NJ 08805-0032, USA. (thermal
conductivity detectors)
Hamilton Co., PO Box 10030, Reno, NV 89520-0012, USA. (syringes)
Hampton Research, 27632 El Lazo Road, Suite 100, Laguna Beach, CA
92677-3913, USA.
Heraeus Feinchemikalien und Forchungsbedarf GmbH, Alter Weinberg,
D-7500 Karlsruhe 41-Ho., Germany, (chemicals, reagents for silaniz-
ation)
Hewlett-Packard Co., Analytical Group, Mailstop 20B AE, Palo Alto, CA
94403, USA. (robotics)
Hilgenberg Glass Company, D-3509 Malsfeld, Germany. (X-ray glass/quartz
capillaries)
Arnold R. Horwell, 73 Maygrove Road, West Hampstead, London NW6 2BP,
UK.
Huber Diffraktionstechnik GmbH, D-8219 Rimsting, Germany, (diffractometry)
Hybaid
Hybaid Ltd., 111-113 Waldegrave Road, Teddington, Middlesex TW11 8LL,
UK.
Hybaid, National Labnet Corporation, P.O. Box 841, Woodbridge, NJ. 07095,
USA.
HyClone Laboratories 1725 South HyClone Road, Logan, UT 84321, USA.
IBF Biotechnics, 35 avenue Jean-Jaures, 92290 Villeneuve-la-Garenne,
France. (chromatographic matrices, biochemicals)
424
 List of suppliers
ICI Cambridge Research Chemicals, Gadbrook Park, Northwich, Cheshire
CW9 7RA, UK. (chemicals)
ICN Biomedicals, Inc., Micromedica Systems Diagnostic Division, 102 Witmer
Road, Horsham, PA 19044-2281, USA. (robotic protein crystallization
system II, pipetting stations)
ICN Flow, 330 Hyland Avenue, Costa Mesa, CA 92626, USA. (biochemical
equipments, Linbro plate, CrystalPlate® and coverslips)
Imaging Technology Inc., 600 West Cummings Park, Woburn, MA 01801,
USA. (digitizers)
Intermec Corp., 4405 Russell Road, PO Box 360602, Lynnwood, WA 98046-
9702, USA. (barcode printer)
International Biotechnologies Inc., 25 Science Park, New Haven, Connecticut
06535, USA.
Invitrogen Corporation
Invitrogen Corporation 3985 B Sorrenton Valley Building, San Diego, CA.
92121, USA.
Invitrogen Corporation do British Biotechnology Products Ltd., 4-10 The
Quadrant, Barton Lane, Abingdon, Oxon OX14 SYS, UK.
Jouan SA, rue Bobby Sands, F-44800 Saint Herblain, France, (laboratory
equipments)
Keithley Data Acquisition and Control, 28775 Aurora Road, Cleveland, OH
44139, USA. (instrument interfaces)
Kodak: Eastman Fine Chemicals 343 State Street, Rochester, NY, USA.
Kohyo Trading Company, Kyodo Bldg 4-1,2 Chome, Iwando-cho, Chiyoda-ky,
Tokyo, Japan, (detergents)
Leica SARL, see Wild-Leitz.
Leitz/Leica. Ill Deer Lake Road, Deerfield, IL 60015, USA.
Life Technologies Inc., 8451 Helgerman Court, Gaithersburg, MN 20877,
USA.
Marresearch, Grosse Theaterstrasse 42, Postfach 303670, 2000 Hamburg 36,
Germany, (image plate)
Memmert GmbH and Co., Aeussere Ritterbacherstrasse 38, D-8540
Schwabach, Germany, (laboratory equipments, thermostated cabinets)
Merck
Merck Industries Inc., 5 Skyline Drive, Nawthorne, NY 10532, USA.
Merck, Frankfurter Strasse, 250, Postfach 4119, D-64293, Germany.
Merck, Frankfurter Strasse 250, D-6100 Darmstadt, Germany, (chemicals and
biochemicals)
Microflex Technology, Inc., The Millennium Centre, PO Box 31, Triadelphia,
WV 26059, USA. (microgrippers)
Micromedica System, Inc., (see ICN Biomedicals). (pipetting station)
Millipore
Millipore (UK) Ltd., The Boulevard, Blackmoor Lane, Watford, Herts WD1
8YW, UK.
425
 List of suppliers
Millipore Corp./Biosearch, P.O. Box 255, 80 Ashby Road, Bedford, MA
01730, USA.
Millipore Waters, PO Box 255, Bedford MA 01730, USA and Zone
Industrielle, F-67120 Molsheim, France. (filtration, membranes, HPLC
equipments)
NAPS Gottingen GmbH, Nucleic Acids Products Supply, Rudolf-Wissel Str.
28, 37070 Gottingen, Germany.
National Institute of Standards and Technology, (Standard Reference Data)
Bldg. 221/A323, Gaithersburg, MD 20899, USA. (Software with crystalliz-
ation data bank)
National Instruments Corp., 12109 Technology Blvd, Austin, TX 78727-6204,
USA. (instrument interfaces, laboratory software)
Neosystem Laboratories, Technopole du Rhin, 21 rue du la Rochelle, F-67100
Strasbourg, France, (peptides)
New England Biolabs (NBL)
New England Biolabs (NBL), 32 Tozer Road, Beverley, MA 01915-5510, USA.
New England Biolabs (NBL), c/o CP Labs Ltd., P.O. Box 22, Bishops Stortford,
Herts CM23 3DH, UK.
Nikon Corporation Instrument Div., Fuji Bldg 2-3, 3-Chome, Maranouchi,
Chiyoda ku, Tokyo 100, Japan, (stereo microscopes)
Nikon Europe BV, Shipholm weg 321, 1171 AE Badhoevedorp, The
Netherlands. (stereo microscopes)
Nunc Inc., 2000, North Aurora Road, Naperville, IL 60566, USA. (plastic
tubes and plates)
Ominifit Ltd., 51 Norfolk Street, Cambridge CB1 2LE, UK and 2005 Park
Street, Box 56, Atlantic Beach, NY 11509, USA. (valves)
Omnilabo Holland BV, Breda, The Netherlands, (multi-well plates)
Oxyl, Peter Henlein Strasse 11, D-8903 Bobingen, Germany, (detergents)
Panasonic Inc., One Panasonic Way, Secaucus, NJ 07094, USA. (optical disc
recorder)
Pentapharm Ltd., Engelgasse 109, CH-4002 Basel, Switzerland, (protease
inhibitors)
Peptide Institute, 476 Ina Miush-shi, Osaka 562, Japan, (protease inhibitors)
Perkin-Elmer
Perkin-Elmer Ltd., Maxwell Road, Beaconsfield, Bucks. HP9 1QA, UK.
Perkin-Elmer Ltd., Post Office Lane, Beaconsfield, Bucks, HP9 1QA, UK.
Perkin-Elmer-Cetus (The Perkin-Elmer Corporation), 761 Main Avenue,
Norwalk, CT 0689, USA.
Perpetual Systems Corporation, 2283 Lewis Avenue, Rockville, Maryland
20851, USA. (sitting-drop rods for cystallization)
PerSeptive Biosystems, City of Dover, Kent County, DL 19901, USA.
Pfanstiel Laboratory, Inc., 1219 Glen Rock Avenue, Wavkega, IL 60085 0439,
USA. (detergents)

426
 List of suppliers
Pharmacia Biosystems
Pharmacia Biotech Europe Procordia EuroCentre, Rue de la Fuse-e 62,
B-1130 Brussels, Belgium.
Pharmacia Biosystems Ltd. (Biotechnology Division), Davy Avenue,
Knowlhill, Milton Keynes MK5 8PH, UK.
Pharmacia LKB Biotechnology AB, Bjorngatan 30, S-75182 Uppsala, Sweden.
Phenomenex, 6100 Palos Verdes Drive S., Rancho Palos Verdes, CA 90274,
USA. (HPLC columns for tRNA)
Pierce, PO Box 1512, 3260 BA Oud-Beijerland, The Netherlands, (laboratory
supplies)
Polycrystal book service, PO Box 3439, Dayton, Ohio 45401, USA.
(crystallography books)
PolyLabo Paul Block et Cie, BP 36, F-67023 Strasbourg Cedex, France.
(scientific equipments)
The Product Integrity Company, Enfield, CT 06082, USA. (programs for
factorial analysis)
Prolabo, 12 rue Pelee, F-7511 Paris, France, (chemicals, equipments)
Promega
Promega Ltd., Delta House, Enterprise Road, Chilworth Research Centre,
Southampton, UK.
Promega Corporation, 2800 Woods Hollow Road, Madison, WI 53711-5399,
USA.
Protein Solutions Incorporated, 2300 Commenwealth Drive, Suite 102,
Charlottesville, VA 22901, USA.
Pye Unicam Ltd, York Street, Cambridge CB1 2PX, UK. (Philips X-ray
generator)
Qiagen
Qiagen Inc., do Hybaid, 111-113 Waldegrave Road, Teddington, Middlesex,
TW11 8LL, UK.
Qiagen Inc., 9259 Eton Avenue, Chatsworth, CA 91311, USA.
Radiometer, A/S 49 Krogshojvej, DK 2880 Dagsvaerd, Denmark. (pH-meter,
conductimeter)
Rainin Instrument Co. Inc., Mack Road, Woburn, MA 01801, USA. (filters)
Resolution Technology, 26000 Avenida Aeropuerto 22, San Juan Capistrano,
CA 92675, USA. (time-lapse VCR)
Rigaku, Monschauer Strasse 7, D-4000 Diisseldorf-Heerdt, Germany & 3
Electronics Avenue, Danvers, MA 01923, USA. (X-ray generators, image
plate)
Roucaire, BP 65, F-78143 Velizy-Villacoublay Cedex, France, (scientific
equipments)
Schleicher and Schuell
Schleicher and Schuell Inc., Keene, NH 03431 A, USA.
Schleicher and Schuell Inc., D-3354 Dassel, Germany.
Schleicher and Schuell Inc., c/o Andermann and Company Ltd.

427
 List of suppliers
Seikagaku Kogyo Co. Ltd., 1-5, Nihonbashi-Honcho 2-Chome Chuo-ku,
Tokyo, 103, Japan, (biochemicals, glycosylases)
Serva Feinbiochemica GmbH and Co., PO Box 105260, D-6900 Heidelberg,
Germany, (biochemicals)
Setaram, 7 rue de 1'Oratoire, BP. 34, F-69641 Caluire Cedex, France.
(instrumentation, calorimeters)
Shandon Scientific Ltd., Chadwick Road, Astmoor, Runcorn, Cheshire WA7
1PR, UK.
Siemens AG, Mess., Pruf. und Prozesstechnik, Ostl. Rheinbriickenstrasse 50,
D-7500 Karlsruje 21, Germany, (diffractometry)
Sigma Chemical Company
Sigma Chemical Company (UK), Fancy Road, Poole, Dorset BH17 7NH, UK.
Sigma Chemical Company, 3050 Spruce Street, P.O. Box 14508, St. Louis,
MO 63178-9916, USA.
Societe 3412, 65 avenue de Stalingrad, F-95104 Argenteuil, France.
(crystallization boxes)
Sofranel, 59 rue Parmentier, 78500 Sartouville, France. (X-ray glass/quartz
capillaries)
Sorvall DuPont Company, Biotechnology Division, P.O. Box 80022,
Wilmington, DE 19880-0022, USA.
Speciality Chemicals, PO Box 1466, Gainesville, FL 32602, USA. (Prosil®-28
reagent for silanization)
Spectrum Medical Industries, Inc., 8430 Santa Monica Blvd, Los Angeles, CA
90069, USA. (dialysis membranes-Spectrapore®)
Stratagene
Stratagene Ltd., Unit 140, Cambridge Innovation Centre, Milton Road,
Cambridge CB4 4FG, UK.
Strategene Inc., 11011 North Torrey Pines Road, La Jolla, CA 92037, USA.
Tosohaas, 6th and Market Streets, Philadelphia, PA 19105, USA. (HPLC
columns)
Transformation Research Inc., PO Box 241, Framington, MA 01701, USA.
(protease inhibitors)
United States Biochemical, P.O. Box 22400, Cleveland, OH 44122, USA.
Vegatec S.A.R.L., 7 place des Onze Arpents, F-94800 Villejuif, France.
(detergents)
Velmex, Inc., PO Box 38, E. Bloomfield, NY 14443, USA. (stepper motors,
motorized slides)
Wellcome Reagents, Langley Court, Beckenham, Kent BR3 3BS, UK.
Whatman Laboratory Sales Ltd, Unit 1, Colred Road, Parkwood, Maidstone,
Kent, ME15 9XN, UK. (chromatography supports)
Wild-Leitz (Leica SARL), 86 avenue du 18 juin 1940, F-92563 Rueil-
Malmaison Cedex, France and CH-9435 Heerbrugg, Switzerland, (stereo
microscopes)
Wolfgang Miiller, Reierallee 12, D-1000 Berlin 27, Germany. (X-ray glass/
quartz capillaries)
 Index
ACA CrystalPlates® 134-6 stoichiometry 226
activation free energy 316-21 streak seeding 204
acupuncture method 166-70 column method of solubility measurement 273
additives 335-8, 369 complexes, see co-crystallizations
for co-crystallizations 227 computer software
divalent cations 222, 223 experimental design 86-8
for membrane protein crystallization 261 net charge estimation 284
monovalent ions 222 statistics 105
for nucleic acid crystallization 221-3 X-ray data processing 414-16
polyamines 221 concentration 124-5
spermine 221, 223 estimation 25-6
agarosegel 150 measurement 125
preparation 159, 161 consolution boundary 246
see also gel crystallization Costar plates 182, 184
ageing 23 countercurrent fractionation 215
ammonium sulfate 9 counter-diffusion techniques 164-6
amorphous precipitate 279 coverslip preparation 131
animal whisker probes 185 critical micellar concentration 248, 249-50
cross-linking 38, 377
cross-seeding 177, 180, 197-200
bacterial expression systems 46; see also cryo-crystallography 379, 381, 385-8
Escherichia coli expression system cryo-protectants 277-8, 369, 379, 381
baculoviral expression system 46-7, 51 mounting crystals for 410-11
batch crystallization 138-41, 142-3 Cryschem plates 136
in gels 161-4 crystallization
biochemical analysis databases 288-9
of crystals 34-6 historical aspects 4-7
of samples 24-6 kinetics 281-3, 334-5, 337-8
Bragg'slaw 395 methods 121, 126-41, 145; see also specific
buffers 122, 220-1, 294 methods
parameters affecting 7-9
practising 143-5
calorimetry for solubility measurement 274 strategy choice 12-13, 306-8
capillary crystals 137 see also crystals
soaking 374-6 crystallography grade purity 8-9, 27
carbon films 349-54 CrystalPlates® 134-6
centrifugation experiments 172-3 crystals
chromatography analysis 34-6
detergent exchange 258 classes 394, 395
of detergents 249 face types 329-30
hydrophobic interaction 216 growth, see growth of crystals
oflipids 256 properties 2-4
co-crystallizations 202-4, 224-34, 367 368 structure 392-5
additives 227 symmetries 393,395
agents 227 systems 393-5
analysis techniques 204-6
DNA:drug 219, 224
DNA:protein 227-31 databases 288-9
homogeneity problems 202-3, 226 density measurement 36-9
protocols 228-30, 232-3 by cross-linking 38
purification 226-7 Ficoll™ method 37-8
RNA:protein 231-4 using molar absorption coefficient 38-9
stability 226 using organic solvents 37
 Index
detergent 245, 246-51, 254-5 orthogonal arrays 84-5
choice 260 sampling 82-8
concentration 260-1 see also mathematical models
critical micellar concentration 248, 249-50 expression systems 18-19, 46-55
exchange 257-8 bacterial 46; see also Escherichia coll
purification 248-9, 251, 255 baculoviral 46-7, 51
thin-layer chromatography 249 Escherichia coll 46, 48-50, 51
dialysis techniques methionine auxotroph strains 65
crystallization 126-30, 141-2, 145 strain DL41: 65
detergent exchange 257-8 mammalian cells 47, 51
double 129-30 yeast 47, 51
Hofmeister series testing 300-1
membrane proteins 263-4
microcap 127-8 Fab-peptide complex cross-seeding 197-9
nucleation zone location 306, 307-8 face types 329-30
salt removal 123-4, 286, 288 factorial experimental design 77-82, 103
for screening 306, 307 fractional 84
tubing preparation 124 incomplete 82-4, 103
diffractometry 418 Ficoll™ method 37-8
divalent cation additives 222, 223 floating drops 140-1
DMA fusion tags 50-1, 52
co-crystallization
with drugs 219, 224
with proteins 227-31 gel crystallization 149-70
crystallization protocols 218-19, 228-30 acupuncture method 166-70
purification 22 agarosegel 150
synthesis preparation 159, 161
chemical 211-12 batch method 161-4
fragment design 210 cavity formation 153
see also nucleic acid crystallization and counter-diffusion 164-6
double dialysis 129-30 crystal characteristics 170
crystal preparation 170
diffusion properties 152
gel incorporation into crystals 153
electron microscopy gel preparation 158-61
grid preparation 349-54 and impurities 157
negative staining 355 inside gel 154-7, 161-4
specimen preparation 355 of membrane proteins 263
transfer of films onto grids 354 nucleation 154-7
of two-dimensional crystals 355-60 outside gel 166-70
electrophoresis 24 silica gel 150-2
nucleic acids 24-5 preparation 158-9
energy, activation free 316-21 gel electrophoresis 24
epitaxial nucleation 180, 200-2 nucleic acids 24-5
Escherichia coll expression system 46, 48-50, glass coverslip preparation 131
51 gold compounds 381
methionine auxotroph strains 65 COSSET 86-8
strain DL41: 65 gravity manipulation 170-3
evaporation kinetics 137-8 growth of crystals
Ewald sphere 398 cessation 10, 338
experimental design control ll,330-4
computer-generated 86-8 and face type, 329-30
experimental matrix preparation 90-1 kinetics 281-3, 334-5, 337-8
factorial 77-82, 103 rate 334-5
fractional 84 impurities 337-8
incomplete 82-4, 103 spiral 331-4
Hardin-Sloane 85-6, 87, 88 and temperature 334-5
minimum-prediction variance 85 by two-dimensional nucleation 330

430
 andex
handling samples 26-7 lipid layer crystallization 341-63
hanging drops 130 advantages 344
withACACrystalPlates® 134-6 electron microscopy 349-60
with Cryschem plates 136-7 grid preparation 349-54
in Linbro boxes 132-4 negative staining 355
recrystallization 134 specimen preparation 355
soaking 373 transfer of films onto grids 354
Hardin-Sloane designs 85-6, 87, 88 helical crystallization 362
heavy-atom derivatives 224, 366, 377-9 lipid solution preparation, 344-5
soaking compounds 379, 380-4 optical diffraction 360
suppliers 385 protein solution preparation 345-6
heterogeneity, see homogeneity of samples reproducibility 362
HIVintegrase 56 setting up 344-8
Hofmeister series 94, 298-301 Teflon supports 346-7
testing by dialysis 300-1 lipids
homogeneity of samples 28-31 solution preparation 344-5
improving 33—4 thin-layer chromatography 256
probing 31-2 lysozyme crystallization
hosts, see expression systems by gel acupuncture 166-7
hydrophobic interaction chromatography in Linbro boxes 144
216 polymorphism 308-10
hydrophobic ligands, soaking 367-8
hydrostatic pressure 140
hypergravity 172-3 macromolecular crystals, see crystals
macromolecular samples, see samples
macroseeding 178, 180, 191, 193-6
impurities 27-8, 335-8 of needles 196
and gel crystallization 157 MAD 64, 366
see also purification/purity mammalian cell expression systems 47, 51
inclusion bodies 48-50 mathematical models 75-120
INFAC 86, 87 analysis 100-12
insect cell expression systems 46-7, 51 contrast analysis 101—2
interface diffusion 141 crystal property scoring 96-100
interferometry 274 multiple regression analysis 102-12
internet/web sites for optimization 112-16
crystallization databases 288-9 polymorph resolution 116-18
E. coll Genetic Stock Center 65 for screening 82-5, 89-96
experimental design 86-7 stationary point identification 108-12
heavy-atom compound suppliers 385 Matthew's coefficient 403
protein net charge estimates 286 membrane protein crystallization 245-68
ionic strength 295-8, 304 additives 261
iridium compounds 382 agents 261-3
isomorphous replacement 366, 377-9 concentration 258-9
detergent 245, 246-51, 254-5
choice 260
kinetics concentration 260-1
of crystallization and growth 281-3, 334-5, critical micellar concentration 248,
337-8 249-50
of evaporation 137-8 exchange 257-8
krypton 379 purification 248-9, 251, 255
thin-layer chromatography 249
gel techniques 263
lanthanide compounds 384 homogeneity 257
Laue technique 418 lipid analysis and elimination 256
lead compounds 382 microdialysis 263-4
light microscopy 121, 404-5 optimization 263
light scattering techniques 274, 322-5, 326-7 PEG use 261-2
Linbro boxes 132-4, 137 protocols 252-3, 259-64

431
 Index
membrane protein crystallization (continued) DNAiprotein 227-31
purification 255-7 homogeneity problems 226
'salting-out' 262-3 protocols 228-30, 232-3
solubility problems 245 purification 226-7
vapour diffusion 263-4 RNA:protein 231-4
without detergent 265 stability 226
mercury compounds 380 stoichiometry 226
methionine auxotroph strains 65-7 concentration 216-18
cell growth 65-7 engineering 224
fermentation medium 66 experimental design 223
starter medium 66 gel electrophoresis 24-5
methionine pathway inhibition 67-8 pH 220
micellar solutions 248 preparation 210-18
critical micellar concentration 248, 249-50 protocols 218-24
Michelson interferometry 274 purification 22, 215-16
microcap dialysis 127-8 co-crystallizations 226-7
microdialysis 126-S screening kits 223
of membrane proteins 263-4 storage of samples 23
microgravity 170-2 synthetic fragment synthesis 210-14
microheterogeneity of samples 28-31 and temperature 220
improving 33-4 see also RNA
probing 31-2
microscopy 121, 404-5; see also electron
microscopy optimization 308
microseeding 178, 180, 188-91, 192 membrane protein crystallization 263
mixed-bed resins 287-8 modelling 112-16
modelling, see mathematical models orthogonal arrays 84-5
monovalent ion additives 222 osmium compounds 383
mother liquor 123 osmotic pressure techniques 324, 325-6
mounting solutions and residual protein overlap extension method 59-61
concentration 277
multiple anomalous dispersion 64, 366
multiple regression 102-12 packing 10
mutagenesis 59-64 PEG 122,261-2
random 61-2, 63-4 purification 122-3
site-directed 59-61 periodic bond chain theory 329
pH 220, 293-5, 304
phase diagrams 141-3, 278-81
net charge 283-6, 288-90 phenol extraction of RNA 214-15
websites 286 platinum compounds 381
neutron scattering 324, 325 polarized light 404-5
niobium compounds 384 polyamine additives 221
nuclease inhibitors 30, 34 polyethylene glycol (PEG) 122, 261-2
nucleation 10, 279-81 purification 122-3
epitaxial 180,200-2 polymorphism 336-7
heterogeneous 280, 315, 318-19 lysozyme 308-10
homogeneous 279, 315, 316-18 resolution by modelling 116-18
primary versus secondary 315 precipitation 279
rate 280-1,315-16 pre-nucleation 321-8
two-dimensional 330 pre-seeding 179-84
nucleic acid crystallization 209-24 pressure manipulation 140
additives 221-3 probes, making and cleaning 185
agents 220 proteases 29, 58
buffers 220-1 inhibitors 30,33
co-crystallizations 224-34 preparation 33-4
additives 227 proteins
agents 227 aggregation 55, 56
DNArdrug 219, 224 co-crystallizations 203, 224-34

432
 Index
additives 227 phenol extraction 214—15
agents 227 purification 22,215-16
analysis techniques, 204-6 storage 23
DNA:protein 227-31 synthesis
homogeneity problems 226 chemical 213-14
protocols 228-30, 232-3 fragment design 211
purification 226-7 in vitro 212-13
RNA:protein 231-4 tRNAs
stability 226 co-crystallization with proteins 231-4
stoichiometry 226 modifications 30
streak seeding 204 purification 22, 215-16
concentration estimation 25 sources 214—15
modifications 29-30 see also nucleic acid crystallization
oxidation 56
protein-protein interactions 293
purification 21-2, 283-4
residual, concentration measurement 274-8 salt 298-301
sequence modification 55-64 crystal recognition 405
shortening 57-9 removal
storage 22-3 by dialysis 123-4, 286,288
streak seeding 204 by mixed-bed resins 287-8
tagged 50-1, 52 and solubility 292, 295-301
see also membrane protein crystallization 'salting-in' 296
proteolysis 29 'salting-out' 9, 296, 298
limited 58-9 membrane proteins 262-3
purification/purity 7-9 samples
co-crystallizations 226-7 ageing 23
crystallography grade 8-9, 27 biochemical analysis 24-6
of detergents 248-9, 251, 255 concentration 124-5
improving 32-3 estimation 25-6
of membrane proteins 255—7 measurement 125
of nucleic acids 22, 215-16 handling 26-7
of polyethylene glycol (PEG) 122-3 homogeneity 28-31
probing 31-2 improving 33-4
of proteins 21-2, 283-4 probing 31-2
of RNA 22, 215-16 preparation 121-5
of selenomethionyl proteins 68-9 purification, see purification/purity
techniques 19-22,28 solid particle removal 125
see also impurities sources 18-19
Pyrex plates 136 storage 22-3
sampling 82-8
sandwich drops 130
random mutagenesis 61-2, 63—4 with ACA CrystalPlates® 134-6
refrigerators 121 Schaeffer method 354
residual protein concentration 274—8 scintillation and solubility measurement 274
resins, mixed-bed 287-8 screening 306, 307
response surface 75,77 kits 84, 223
ribosome crystals 234-5 models 82-5, 89-96
RNA SDS-PAGE analysis 204-5
co-crystallization with proteins 231-4 seeding 177-208
concentration 216-18 analytical 185-8
estimation 26 complexes 204
crystallization 210 cross-seeding 177, 180, 197-200
protocols 219 heterogeneous 180, 196-200
handling 216-18 homogeneous 180
natural macroseeding 178, 180, 191, 193-6
purification 215-16 of needles 196
sources 214-15 microseeding 178, 180, 188-91, 192

433
 Index
seeding (continued) microscopy 273-4
pre-seeding 179-84 scintillation 274
by vapour diffusion 180-4 temperature controlled light scattering
probes 185 274
residual protein concentration measurement and pH 293-5, 304
276-7 and salts 292, 295-301
streak seeding 177 and temperature 121, 302-3, 304
analytical 185-8 see also supersaturation
complexes 204 solution preparation 121-3
microseeding 190-1 solvents 369
selenolsubtilisin 199-200 space crystallization 170-2
selenomethionine substitution 64-70 space group determination 416-17
crystallization 69 sparse matrix kits 84
crystal storage 69-70 spermine 221, 223
eukaryotes 68 spiral growth 331-4
health and safety measures 70 stationary point identification 108-12
prokaryotes 65-8 statistics packages 105
purification 68-9 stepwise replacement 304-5
silica gel 150-2 storing samples 22-3
preparation 158-9 streak seeding 177
see also gel crystallization analytical 185-8
site-directed mutagenesis 59-61 complexes 204
sitting drops 130 microseeding 190-1
with ACA CrystalPlates® 134-6 subcloning 53
in capillaries 137 supersaturation 9, 178-9, 270, 271, 279,
with Cryschem plates 136 314-15
in Linbro boxes 137 rate 279
making plates for 184
soaking 372-3
small angle neutron scattering 324, 325
small angle X-ray scattering 324, 325, 327-8 tantalum compounds 384
soaking 365-90 Tag DNA polymerase 61
applications 366, 377-88 Teflon supports 346-7
capillaries 374-6 temperature
for cryo-crystallography 379, 381, 385-8 and crystal growth 334-5
in dilute ligand solutions 376-7 and nucleic acid crystallization 220
hanging drops 373 regulation 121, 181
for heavy-atom derivatives 366, 377-9 and solubility 121, 302-3, 304
soaking compounds 379, 380-5 thallium compounds 382
hydrophobic ligands 367-8 thaumatin crystallization 144-5
for isomorphous replacement 366, 377-9 thin-layer chromatography
sitting drops 372-3 of detergents 249
techniques 368-77 oflipids 256
software packages thymol 137
experimental design 86-8 tRNAs
net charge estimation 284 co-crystallization with proteins 231—4
statistics 105 modifications 30
X-ray data processing 414-16 purification 22, 215-16
solubility 9, 269-83 sources 214-15
denned 270 tungsten compounds 383
diagrams 141-3,278-81 two-dimensional crystallization 341-63
factors affecting 291-305 advantages 344
and ionic strength 295-8, 304 electron microscopy 349-60
measurement 271-4 grid preparation 349-54
calorimetry 274 negative staining 355
column method 273 specimen preparation 355
crystallization 271—3 transfer of films onto grids 354
Michelson interferometry 274 lipid solution preparation, 344-5

434
 Index
xenon 379
X-ray crystallography 391-419
reproducibility 362 crystal choice 405
setting up 344-8 data 411-18
Teflon supports 346-7 collection 411-14, 417-18
two-dimensional nucleation 330 processing 414-17
resolution 400, 402
diffraction patterns 395-9
diffractometry 418
uranium compounds 383 Laue technique 418
mounting crystals 404-11
for cryo-crystallography 410-11
at low temperatures 408-11
vapour diffusion 130-8, 142, 145 at room temperature 405-7
of membrane proteins 263-4 optical alignment 413
and pre-seeding 180-4 oscillation data 412-16
vectors, see expression systems preliminary investigations 402-4
virus crystals 235—8 pre-nucleation investigations, 324, 325,
327-8
space group determination 416-17
X-ray sources 392
web/internet sites
crystallization databases 288-9
E. coli Genetic Stock Center 65 yeast expression systems 47, 51
experimental design 86-7
heavy-atom compound suppliers 385
protein net charge estimates 286 Zeppenzauer cells 126

435