REVIEWS

MECHANISMS OF DISEASE

Soluble protein oligomers in

neurodegeneration: lessons from
the Alzheimer’s amyloid β-peptide
Christian Haass* and Dennis J. Selkoe‡
Abstract | The distinct protein aggregates that are found in Alzheimer’s, Parkinson’s,
Huntington’s and prion diseases seem to cause these disorders. Small intermediates —
soluble oligomers — in the aggregation process can confer synaptic dysfunction,
whereas large, insoluble deposits might function as reservoirs of the bioactive oligomers.
These emerging concepts are exemplified by Alzheimer’s disease, in which amyloid
β-protein oligomers adversely affect synaptic structure and plasticity. Findings in other
neurodegenerative diseases indicate that a broadly similar process of neuronal dysfunction
is induced by diffusible oligomers of misfolded proteins.

Amyloid
Degenerative diseases of the human brain have long been notably with the seminal studies in 1984 of Glenner on
Tissue deposits of insoluble, viewed as among the most enigmatic and intractable the cerebrovascular amyloid deposits of Alzheimer’s
proteinaceous fibrils that are problems in biomedicine. As research on human neuro- disease (AD). As research on the protein deposits of AD
rich in β-pleated sheet degeneration has moved from descriptive phenomenology grew increasingly intensive, partly because of the com-
structure and therefore bind to
to mechanistic analysis, it has become increasingly monness of the disorder, similar concepts and methods
the histochemical dye Congo
red in a polarized manner. apparent that the morphological lesions long used by began to be applied to other human neurodegenerative
neuropathologists to confirm a clinical diagnosis after disorders.
death might provide an experimentally tractable handle to In this review, we will illustrate the emerging trend
understand causative pathways. This concept came as no to define several human neurodegenerative syndromes
surprise to those pursuing a small and arcane field of medi- as disorders of protein folding and oligomerization
cine, the study of amyloidosis. Amyloids were defined some through the example of AD. In AD, a small peptide,
150 years ago as tissue deposits of extracellular filaments amyloid β-protein (Aβ), forms long, insoluble amy-
(usually called fibrils) that were microscopically — and in loid fibrils, which accumulate in spherical micro-
severe cases, macroscopically — visible in various organs scopic deposits known as senile plaques. However,
in several seemingly unrelated human diseases. the relevance of these plaques to AD pathogenesis was
In the twentieth century, amyloidologists found that unclear and even questioned by many investigators, a
amyloid fibrils were proteinaceous in origin, and by the contentious issue that might now be coming to a rather
1960s some amyloid subunits (such as the fragments of simple and surprising solution. Therefore, an emphasis
*Adolf Butenandt Institute, immunoglobulins that accumulate in certain immune- of this review will be on an important recent develop-
Department of Biochemistry,
Laboratory for Alzheimer’s
related amyloidoses) were identified as specific proteins. ment from studies that attempted to identify the toxic
and Parkinson’s Disease Biochemical and ultrastructural experiments established moiety responsible for synaptic dysfunction and neuro-
Research, Ludwig the remarkable insolubility of amyloid fibrils in aque- nal cell loss in AD. The production of the apparent
Maximilians University, ous buffers. X-ray-diffraction analyses on purified fibrils toxic species — soluble Aβ oligomers — and their
80336 Munich, Germany.
‡
indicated that the constituent proteins were unusually subsequent ability to cause neuronal injury depends
Center for Neurologic
Diseases, Harvard Medical rich in highly ordered, β-pleated sheet structure, pro- on the precision of an intramembranous proteolytic
School and Brigham and viding an apparent explanation for their long-known cleavage, which is unprecedented and will therefore
Women’s Hospital, Boston, property of binding to the histochemical dye Congo red be described in detail. We will also discuss how the
Massachusetts 02115, USA. in a polarized manner. production of Aβ and its capacity to form oligomers
e-mails: chaass@med.uni-
muenchen.de; dselkoe@rics.
These advances in elucidating amyloid deposits in is affected by mutations that cause aggressive forms of
bwh.harvard.edu peripheral tissues encouraged amyloidologists to apply familial AD. Some oligomeric species of Aβ are small
doi:10.1038/nrm2101 their methods to neurodegenerative disorders, most and soluble enough to diffuse readily through the

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 101

© 2007 Nature Publishing Group
REVIEWS

‘Reverse genetics’ enables AD gene discovery

Box 1 | Pathological inclusions in neurodegenerative disorders
Spectacular advances in human genetics and the avail-
Many neurodegenerative disorders are characterized by the formation of insoluble ability of the DNA sequence of the entire human genome
inclusions. However, recent evidence indicates that these deposits might not be have greatly facilitated the identification of genes
responsible directly for the primary neurotoxic effects and memory loss of these involved in diverse hereditary disorders, including sev-
disorders. They might function as a reservoir for soluble oligomers, which are able to eral prominent neurodegenerative disorders (for neuro-
diffuse through the brain parenchyma and into synaptic clefts. Insoluble deposits and
degenerative disease histopathology, see BOX 1). A recent
diffusible oligomers from different diseases are composed of individual amyloidogenic
proteins as different as amyloid β-protein (Aβ), tau, prion protein (PrP), α-synuclein and
and particularly compelling example is PD1. Although
huntingtin. They might share common structural epitopes and common mechanisms of the cytopathological hallmarks of PD, known as Lewy
neurotoxicity and memory impairment79. bodies (LBs) and Lewy neurites (LNs) (BOX 1), have long
Panel a shows Alzheimer’s disease with Aβ-positive senile plaques, and panel b shows been known, the amyloidogenic neuronal protein that is
tau-positive neurofibrillary tangles, neurophil threads and dystrophic neurites. Panel c deposited in LBs and LNs remained unidentified. Until
shows PrP deposits in the cerebellum of a Creutzfeldt–Jakob case. Panel d shows substantia a decade ago, PD was considered by most authorities to
nigra of a Parkinson’s disease case with an α-synuclein-positive Lewy body (right) and Lewy be essentially a sporadic (that is, non-familial) disorder.
neurites (left). Panel e shows a ubiquitin-positive huntingtin inclusion in a case of Yet it was through genetics, not biochemistry, that the
Huntington’s disease. The scale bar represents 100 µm. protein component of LBs and LNs was identified.
a b Linkage analysis and positional cloning of the first
PD-associated gene in a family with autosomal dominant
parkinsonism identified a missense mutation in the
gene that encodes the neuronal protein α-synuclein2.
α-synuclein is a naturally unfolded, soluble cytoplasmic
protein that is highly expressed in neurons and that can
bind to the membranes of synaptic, and other, vesicles3.
After the discovery that α-synuclein was genetically
implicated in familial PD, immunohistochemistry com-
bined with biochemical purification quickly showed that
α-synuclein was the principal protein component of LBs
c d and LNs4,5. In this example, geneticists directly enabled
biochemists to isolate the needle from the haystack of
candidate proteins. In a similar manner, breakthrough
discoveries of causative genes allowed the subsequent
identification of aggregation-prone proteins in disorders
such as amyotrophic lateral sclerosis (ALS)6,7 and HD8.

The discovery of the first AD causative gene. By contrast,

a biochemical hypothesis preceded and guided the dis-
covery of the first causative gene in AD. The AD causative
gene encodes β-amyloid precursor protein (APP), which
e is a single-transmembrane, receptor-like protein that is
expressed ubiquitously in neural and non-neural cells. After
the Bavarian psychiatrist Alois Alzheimer presented his
first clinicopathological case in Tübingen on November 3,
1906 (REF. 9), it became clear that amyloid plaques in
the cerebral cortex (BOX 1) were invariably associated with the
disease that now bears his name.
In the 1980s, biochemists focused on the isolation of
the amyloid to identify its principal component. Glenner
and Wong purified microvascular amyloid deposits
from the meninges of AD brains and provided a partial
sequence of an ~4-kDa subunit protein that they named
amyloid β-protein (Aβ)10. Shortly thereafter, Masters,
brain parenchyma and affect synaptic structure and Beyreuther and co-workers identified the same protein
function and, ultimately, neuronal survival. We will as the subunit of amyloid plaque cores that were isolated
also describe the emergence of highly specific treat- from post-mortem AD cortices11. Around the same time,
ment strategies that are based on the new molecular the microtubule-associated protein tau was identified
understanding of AD. Throughout the review we as the main constituent of the hallmark neurofibrillary tangles
Lewy body will indicate the timeliness of the topic — rapidly (BOX 1) that accumulate inside many neurons and their
(LB). A deposit of α-synuclein expanding research on other brain disorders such as processes in AD brains12–14. Tau is a highly soluble cyto-
typically found in neuronal cell
bodies of patients with
Parkinson’s disease (PD) and Huntington’s disease plasmic protein that binds to tubulin during its poly-
Parkinson’s disease or related (HD) has suggested common features that resemble merization into microtubules in neurons and thereby
disorders. the fundamental pathogenic process in AD. stabilizes these important cytoskeletal organelles.

102 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group
 REVIEWS

of this small hydrophobic peptide initiates a slow but

Box 2 | The amyloid (or Aβ) cascade hypothesis
deadly cascade that leads to synaptic alterations, micro-
Gradual changes in the steady-state levels Changes in Aβ metabolism glial and astrocytic activation, the modification of the
of amyloid β-protein (Aβ) in the brain are • Increase in total Aβ production normally soluble tau protein into oligomers and then
thought to initiate the amyloid cascade22–24. • Increase in the Aβ42/Aβ40 ratio
• Reduced Aβ degradation/clearance into insoluble paired helical filaments, and progressive
Aβ levels can be elevated by enhanced
neuronal loss associated with multiple neurotransmitter
production and/or reduced clearance. Oligomerization of Aβ42 and initial deficiencies and cognitive failure (BOX 2).
In particular, the Aβ42/Aβ40 ratio can be (diffuse) Aβ42 deposits
augmented by mutations in three different
genes (β-amyloid precursor protein (APP), Subtle effects of soluble Aβ42 oligomers Regulated intramembrane proteolysis of APP
on synaptic function Initially, the mechanism by which the partially intra-
presenilin-1 (PS1) and PS2) that cause
familial forms of Alzheimer’s disease. Inflammatory responses (microglial and membrane Aβ region (FIG. 1a) could be liberated as a free
This relative increase of Aβ42 enhances astrocytic activation) and amyloid peptide from its precursor was enigmatic and was consid-
oligomer formation, which causes subtle plaque formation ered to require some pre-existing membrane injury. It was
and then increasingly severe and assumed that the hydrophobic interior of the membrane
Progressive synaptic/neuronal injury
permanent changes of synaptic function. bilayer would need to be damaged to allow access of a
In parallel, Aβ42 forms microscopically Altered neuronal ionic homeostasis protease and water to effect cleavage. But this concept was
visible deposits in the brain parenchyma, & oxidative injury disproven when Aβ was unexpectedly discovered to be
first as relatively benign diffuse (non-
fibrillar) plaques. As the diffuse plaques Aberrant oligomerization and produced normally by the intramembranous proteolysis
begin to acquire fibrils of Aβ, local hyperphosphorylation of tau of APP throughout life and APP was found to circulate in
inflammatory responses (microgliosis and extracellular fluids, including cerebrospinal fluid (CSF)
Widespread neuronal dysfunction and
astrocytosis) are observed. Synaptic spine cell death associated with and plasma25–28. This discovery opened up the dynamic
loss and neuritic dystrophy also occur. neurotransmitter deficits study of Aβ, which heretofore had only been obtained
Over time, these events result in oxidative through painstaking isolation from post-mortem
Dementia with plaque and tangle
stress, altered ionic (for example, calcium) pathology human brain. As predicted by the amyloid hypothesis,
homeostasis and a host of additional all AD-causing APP mutations that have been identified
biochemical changes. Neurofibrillary
so far occur either within or flanking the Aβ region of
tangles are induced by altered kinase and
this large polypeptide. Accordingly, the mutations that
phosphatase activities and contribute to
additional defects, including some in flank the Aβ region increase the production of the highly
axonal transport. The cascade culminates in amyloidogenic Aβ42 isoform, whereas the mutations
widespread synaptic/neuronal dysfunction within the Aβ region enhance the oligomerization of the
and cell death, leading to progressive peptide29 (BOX 3).
dementia associated with extensive Aβ
and tau pathology. Intramembrane proteolysis liberates Aβ. Because the
generation of Aβ species with variably hydrophobic
C termini — and therefore different propensities to
Glenner also showed that the amyloid deposits that oligomerize — is directly relevant to the development
occur in the brain vessels of young adults with Down of AD, we will review the current understanding of the
syndrome were composed of Aβ15. Since 1969, it had biochemical mechanisms that underlie Aβ production,
been known that middle-aged patients with Down syn- with a focus on intramembrane processing.
drome develop the amyloid plaques and neurofibrillary Aβ generation has turned out to be just one example
tangles that are typical of AD16. On this basis, Glenner of a general physiological mechanism now known as
assumed that the gene ultimately found to encode Aβ regulated intramembrane proteolysis (RIP)29–31. In a prin-
might be causative of AD cases. The subsequent cloning cipal version of this process, membrane proteins first
Lewy neurite of APP, which encodes a large, type 1 membrane glyco- undergo a regulated shedding of their ectodomains by
(LN). A deposit of α-synuclein
protein, by Beyreuther and co-workers in 1987, and its membrane-anchored proteases known as secretases or
typically found in swollen
neuronal processes of patients localization to the long arm of chromosome 21, was sheddases, releasing the large luminal domains into
with Parkinson’s disease or consistent with this hypothesis17. extracellular fluids. The membrane-retained stubs can
related disorders. These biochemical findings pointed strongly to the then be cleaved within their transmembrane domains
APP gene as a site which geneticists should search for (TMDs) to release small hydrophobic peptides (for
Amyloid plaques
Spherical extracellular deposits
AD-causing mutations. The first such mutation was example, Aβ in the case of APP) into the extracellular
predominantly composed of discovered in a family with hereditary cerebrovascular space and intracellular domains (ICDs) into the cytoplasm
the amyloid β-peptide and amyloidosis with multiple haemorrhages18. Shortly (FIG. 1a). The free ICDs might have specific functions,
found in the brains of all cases thereafter, a distinct APP missense mutation was found including the activation of nuclear signalling pathways,
of Alzheimer’s disease.
in a family with early-onset AD19, and then additional as in the case of the Notch ICD, which is liberated by
Neurofibrillary tangle mutations in other families were detected20,21. γ-secretase from membrane-bound Notch32.
An intraneuronal filamentous These and other findings led to the formal proposal of In the case of APP and certain other RIP substrates,
aggregate composed of a hypothesis of disease in which excessive Aβ accumula- ectodomain shedding can be mediated by either of two
abnormally phosphorylated tion and deposition could trigger a complex downstream distinct membrane-anchored proteases, α-secretase
tau protein found in several
human neurodegenerative
cascade that resulted in the symptoms of AD22,23. In its (believed to consist of one or more members of the
disorders, including Alzheimer’s most recent iteration24, the amyloid (or Aβ) hypothesis ADAM family of metalloproteases) or β-secretase
disease. states that the gradual accumulation and aggregation (also called β-site APP-cleaving enzyme (BACE))29.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 103

© 2007 Nature Publishing Group
REVIEWS

a
Extracelluar space/lumen Aβ

NCT
Shedding by PS1
BACE PS2 PEN2
Aβ Aβ Aβ
1 2 3 4 5 6 7 8 9
D D

APP CTFβ
Cytosol NTF CTF APH1a
APH1b
b

AICD
Aβ

6
9
0
8

ζ4
ε4
γ4
γ4
γ3
γ (38,40,42)
ζ (46) ...GGVVIATVIVITLVMLKKK...
ε (49)
CTFβ

Figure 1 | Amyloid β-protein generation by normal proteolytic processing of β-amyloid precursor protein.
a | Amyloidogenic processing of β-amyloid precursor protein (APP) by β-site APP-cleaving enzyme (BACE) and the
γ-secretase complex. In this pathway, full-length APP (left) is first processed by BACE, and the large ectodomain is
secreted. The remaining membrane retained stub (CTFβ) binds to a docking site on the surface of the γ-secretase
complex and is then transferred to the active site that includes transmembrane domains 6 and 7 of presenilin-1 (PS1)
or PS2. PS1 and PS2 are both activated by presumed autoproteolytic cleavages, which create their N-, and C-terminal
fragments (NTF and CTF). These bind to each other and also to 3 other essential γ-secretase components, APH1a (or
APH1b), PEN2 and nicastrin (NCT). All four proteins form the core complex required for γ-secretase activity (shown in
dashed box). The two intramembrane aspartate residues in the NTF and CTF of presenilins (marked with a D) are a
crucial part of the unusual catalytic site of the protease. The γ-secretase cleavage occurs in the middle of the
membrane and liberates amyloid β-protein (Aβ) and the APP intracellular domain (AICD). The function of the AICD is
unclear. b | Various proposed sites of intramembrane proteolysis by γ-secretase. The amino-acid sequence around the
cleavage sites of APP is shown (numbers refer to the sequence of Aβ; shaded amino acids are in the transmembrane
domain). γ-secretase cuts its substrates several times. The cleavage sites are referred to as ε, ζ and γ (from the C- to
N-terminal). The γ-site is variable and can occur at least after amino acids 38, 40 and 42. This cleavage is highly relevant
for the subsequent aggregation propensity of Aβ. Some γ-secretase-modifying drugs shift the cleavage at Aβ42 to
amino acid 38, and the resultant peptide aggregates much less readily.

The membrane-associated stub, which is created by after Aβ amino acids 38, 40 or 42. The precise sites of
BACE cleavage, can then undergo an intramembrane these γ-cleavages have an important influence on the
scission that is mediated by the γ-secretase complex, a self-aggregating potential and resulting pathogenicity
special type of aspartyl protease with a unique active of Aβ, as only the Aβ42 peptide has a strong propensity
site and cleavage mechanism33,34 (FIG. 1a). This complex is to oligomerize in vivo.
composed of presenilin-1 (PS1) or PS2, nicastrin, APH1
Regulated intramembrane and PEN2 (FIG. 1a)29,35–37. All four proteins are both neces- Aββ oligomers induce synaptic dysfunction
proteolysis sary and sufficient to reconstitute γ-secretase activity in Research on AD seeks to answer a central question: what
(RIP). Regulated cleavage of yeast, which lacks this enzyme37. causes the onset of a subtle, intermittent impairment of
the luminal domain of certain
membrane proteins is followed
At least in the case of its two most well studied hippocampal neuronal function and, therefore, episodic
by a constitutive cleavage in substrates, APP and Notch, γ-secretase can carry out memory? Substantial evidence indicates that Aβ might
the transmembrane domain. multiple intramembrane cleavages. Current evidence be the central player in AD pathology. However, per-
At least in some cases this indicates that the PS–γ-secretase complex can cleave at haps the most persistent argument against the amyloid
process is involved in signalling
different sites (referred to as γ, ε and ζ) in the TMD hypothesis as summarized earlier is that many appar-
pathways.
(FIG. 1b). The ε-cleavage close to the cytoplasmic border ently healthy older humans have substantial amounts of
Intracellular domain of the TMD releases the free ICD into the cytosol38–40. amyloid in their limbic and association cortices upon
(ICD). One of the cleavage It seems that the remaining membrane-anchored frag- post-mortem examination. These Aβ deposits are over-
products of intramembrane ment undergoes an intermediate scission ~3 residues whelmingly of the diffuse type — they are not composed
proteolysis in the RIP pathway,
it is liberated into the cytosol
N-terminal to the ε-cut at the so-called ζ-site 41,42. of amyloid fibrils and they have little or none of the
and in some cases targeted to Thereafter, Aβ is released into biological fluids by the surrounding neuritic and glial cytopathology found in
the nucleus. final cuts at the γ-site. The γ-cut is variable and occurs mature neuritic plaques43. Furthermore, reports of weak

104 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group
 REVIEWS

Box 3 | Familial Alzheimer’s disease and the role of amyloid β-protein oligomerization
A strong clue to the importance of amyloid β-protein (Aβ) oligomerization in Alzheimer’s disease (AD) pathogenesis came
from the observation that some AD-causing mutations in β-amyloid precursor protein (APP) and all AD-causing mutations
in presenilin-1 (PS1) or PS2 enhance the production of the Aβ42 peptide117,118. The enhanced production of the Aβ42
peptide results in an increase in absolute Aβ42 levels, or at least in an increase in the Aβ42/Aβ40 ratio (that is, Aβ42 can
increase at the expense of Aβ40). The more hydrophobic Aβ42 peptide, with its C-terminal alanine and isoleucine residues,
aggregates more rapidly, therefore forming stable Aβ oligomers at an earlier time point58,119–121. Moreover, Aβ42 tends to
form stable trimeric and/or tetrameric oligomers, whereas Aβ40 does not121. Inherited missense mutations directly in the
Aβ region of APP increase the propensity of the peptide to aggregate; for example, the E693G (Arctic) APP mutation
strongly enhances oligomerization122. The aggregation process is reflected in the initial microscopic deposition of Aβ42
in the form of early (diffuse) plaques in AD brains123.
In three genetic conditions, the increased production of all forms of Aβ precipitates AD. These conditions are Down
syndrome, the K595M and N596L(Swedish) double APP mutation at the β-site APP-cleaving enzyme (BACE) cleavage
site124,125, and duplication of a small region of chromosome 21q containing the APP gene126. Importantly, the recent
discovery of the families with a duplication at chromosome 21q has revealed an exciting analogy to inherited Parkinson’s
disease, where rare families carry a duplication or even triplication of the α-synuclein gene127.
On the basis of these and other findings, one might conclude that a relative increase in Aβ42 versus Aβ40 levels seems to
be sufficient to trigger the AD phenotype. The above results in familial AD fit nicely with the recent discovery in transgenic
mice that a marked elevation of only Aβ40 does not lead to plaque formation but can actually serve to retard the
deposition of Aβ42 in the brain128. In accordance, expression of various AD-causing presenilin mutations in cell lines
derived from PS1 PS2 double-knockout mice indicates that at least some mutations do not result in increased levels of
Aβ42 but rather result in decreased levels of Aβ40 (REF. 129), thereby reducing the levels of this putative ‘anti-aggregation
factor’. Furthermore, the deletion of the PS1 loop in vivo results in reduced Aβ40, but not Aβ42, generation and thereby
triggers amyloid plaque pathology in mice130.

quantitative correlations between manual microscopic However, it must also be pointed out that large
counts of amyloid plaques in post-mortem brain sec- plaques of fibrillar Aβ in AD brains typically show sur-
tions and the extent of cognitive symptoms measured rounding dystrophic neurites, indicating that insoluble
pre-mortem are fraught with methodological challenges. aggregates might contribute to neuronal injury. Indeed,
Counting spherical plaques in two-dimensional cross fibrillar Aβ deposits have been associated with local
sections provides an imprecise measure of Aβ amounts synaptic abnormalities and even with the breakage of
and misses small and heterogeneous Aβ-assembly forms. neuronal processes51. The problem is that large, insoluble
Last, the cognitive testing done before the patient’s death protein aggregates are likely to be intimately surrounded
has often been done with simple, insensitive mental by a number of smaller, more diffusible, assemblies (for
status screens. example, oligomers). So, it becomes difficult to ascertain
The advent of specific Aβ enzyme-linked-immuno- whether the large aggregates are directly inducing local
sorbent assays (ELISAs) coupled with western blotting neuronal injury and dysfunction. At the current stage
and mass spectrometry has now enabled a more pre- of research, one should not conclude that either large,
cise and comprehensive assessment of Aβ quality insoluble deposits or small, soluble oligomers represent
and quantity. Such studies indicate that biochemically- the sole neurotoxic entity; indeed, a continuous dynamic
measured levels of soluble Aβ, including soluble oligo- exchange between these forms might well be detrimental.
mers, correlate much better with the presence and degree Nevertheless, we hypothesize that diffusible oligomers
of cognitive deficits than do simple plaque counts44–47. This have the principal role, particularly during the earliest,
evidence, coupled with the fact that large (~20–120-µm even pre-symptomatic, stages of the AD process.
diameter) fibrillar plaques present much less Aβ surface
area to neuronal membranes than do a multitude of small Different types of synthetic and natural Aβ oligomers.
oligomers that can diffuse into synaptic clefts, indicates A large and confusing body of literature describes many
that such soluble assembly forms are better candidates types of assembly forms of synthetic Aβ, including proto-
for inducing neuronal and/or synaptic dysfunction than fibrils (PFs), annular structures, paranuclei, Aβ-derived
plaques, per se. diffusible ligands (ADDLs), globulomers and amyloid
Importantly, the idea that large aggregates of a disease- fibrils (for reviews, see REFS 52,53; the different types of
causing protein can actually be inert or even protective to recognized Aβ oligomers are summarized in TABLE 1). In
neurons has been supported by work on other protein- general, soluble oligomers are defined as Aβ assemblies
folding disorders. For example, in cell-culture studies of that are not pelleted from physiological fluids by high-
HD, less cell death has been observed when large aggre- speed centrifugation, and not all of the aforementioned
gates of polyglutamine-rich huntingtin protein are present synthetic assembly forms fulfil this definition. Moreover,
in the cells than when only soluble huntingtin is present soluble oligomers can bind to other macromolecules or
without these inclusions48,49. Analogous findings have to cell membranes and can therefore become insoluble.
been reported in a mouse model of spinocerebellar ataxia PFs are intermediates that were observed in the course
in which the polyglutamine-rich forms of the ataxin-1 of studying the fibrillization of synthetic Aβ54–57. They
protein are expressed50 . are flexible structures that can continue to polymerize

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 105

© 2007 Nature Publishing Group
REVIEWS

Table 1 | Oligomeric assemblies of Aβ

Oligomeric assembly Characteristics References
Protofibril (PF) Intermediates of synthetic Aβ fibrillization; up to 150 nm in length and 54–57
~5 nm in width; β-sheet structure: bind Congo red and Thioflavin T
Annular assemblies Doughnut-like structures of synthetic Aβ; outer diameter of ~8–12 nm; 58,59
inner diameter of ~2.0–2.5 nm
Aβ-derived diffusible Synthetic Aβ oligomers smaller than annuli; might affect neural signal- 60,61
ligands (ADDLs) transduction pathways
Aβ*56 Apparent dodecamer of endogenous brain Aβ; detected in the brains of an 62
APP transgenic mouse line and might correlate with memory loss
Secreted soluble Aβ Produced by cultured cells; resistant to SDS; resistant to the Aβ-degrading 63,64,69
dimers and trimers protease IDE; alter synaptic structure and function
Aβ, amyloid β-protein; APP, β-amyloid precursor protein; IDE, insulin-degrading enzyme.

in vitro to form amyloid fibrils or can depolymerize to amnestic activity. Like the Aβ oligomers produced from
lower-order species. PFs are narrower than bona fide cultured cells69, Aβ*56 might disrupt synaptic function
amyloid fibrils (~5 nm versus ~10 nm). Ultrastructural and therefore affect memory 62. Whether Aβ*56 and
analyses of synthetic PF preparations by electron micro- species that are similar to it are stable assemblies of only
scopy and atomic force microscopy have revealed both Aβ under native conditions, or whether smaller oligo-
straight and curved assemblies of up to 150 nm in length. meric assemblies can associate with another protein is
Synthetic Aβ PFs have been shown to contain substan- currently unknown. However, Aβ*56 and Aβ trimers
tial β-sheet structure, as they can bind to Congo red or secreted by cultured cells could turn out to share common
Thioflavin T in an ordered manner. synaptotoxic properties.
Annular assemblies of synthetic Aβ are doughnut-like
structures, with an outer diameter of 8–12 nm and an How does Aβ induce memory loss?
inner diameter of 2.0–2.5 nm, that can be distinguished An intensively studied electrophysiological correlate of
from PFs by atomic force microscopy and electron learning and memory is long-term potentiation (LTP).
microscopy58,59. Some laboratories have observed smaller Repetitive, high frequency electrical stimulation of certain
oligomeric species of synthetic Aβ than annuli and have synaptic circuits, for example the CA3–Schaefer collateral–
designated these ADDLs60. Apparent ADDL-like oligo- CA1 pathway in the mammalian hippocampus, can
meric assemblies have been isolated from post-mortem induce a prolonged potentiation of synapse firing (LTP)
AD brains, and their presence correlated with memory that is referred to as inducing synaptic plasticity.
loss61. Chemical stabilization of synthetic Aβ42 assembly There is now considerable evidence that ADDLs of
intermediates has revealed an apparent hexamer perio- synthetic human Aβ60 and soluble, low-number (low-n)
dicity, with hexamer, dodecamer and octadecamer struc- oligomers of naturally secreted human Aβ69,70 (FIG. 2) can
tures observed58. Whether the recently described Aβ*56, all inhibit the maintenance of hippocampal LTP. In the
an apparent dodecamer of natural Aβ detected in the case of the cell-derived oligomers, this inhibition occurs
brains of an APP transgenic mouse line62, might represent at low- to sub-nanomolar concentrations that are similar
an in vivo analogue of synthetic ADDLs remains unclear, to those that can be found in human CSF. This effect
as direct structural comparisons have not been possible. has been shown by both in vivo microinjection in living
Whereas most of the Aβ assembly intermediates rats69 and by treatment of hippocampal slices71. The effects
described above have only been observed upon in vitro of the natural oligomers on LTP are specifically neutral-
incubation of synthetic Aβ, small oligomeric Aβ forms ized by anti-Aβ antibodies in vivo, either through active
do occur in vivo and might therefore be relevant to vaccination or passive infusion72 (see below). The same
disease pathogenesis. Intracellular and secreted soluble oligomers have been shown to interfere rapidly and
dimeric and trimeric oligomers have been described in reversibly with the memory of a learned behaviour in
cultured cells63,64 (FIG. 2a), and SDS-stable oligomers of wake, behaving rats73. Taken together, these various
varying sizes have also been detected by western blotting results provide compelling evidence that decreased
in APP transgenic mouse brain and human brain62,65–68. hippocampal LTP and altered memory function can be
Such natural (that is, non-synthetic) Aβ oligomers can directly attributed to an isolated, biochemically defined,
be resistant not only to SDS but also to the Aβ-degrading assembly form of human Aβ (with low-n soluble oligo-
protease insulin-degrading enzyme (IDE), which can mers probably ranging from dimers to dodecamers), in
only digest monomeric Aβ69. Naturally secreted mono- the absence of amyloid fibrils or PFs. Whether Aβ*56
meric and oligomeric Aβ species are being character- (REF. 62), an apparent dodecameric Aβ assembly, can also
ized in experiments in vivo to decipher their effects on inhibit hippocampal LTP has not yet been investigated.
synaptic structure and function 69,70 (see below). Aβ It might turn out that a range of low-n oligomeric assem-
oligomers produced by cultured cells could be related blies of Aβ can affect synaptic structure and function.
to the aforementioned Aβ*56 (REF. 62), which seems to Therefore, the identification of a single cytopathological
represent a brain-derived soluble dodecamer that has Aβ species is unlikely.

106 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group
 REVIEWS

a b 180

160

EPSP amplitude (% of baseline)

Untransfected

APP (V717F)
140

Trimers 120

Dimers
100
Untransfected
Monomers APP (V717F)
80

0 40 80 120 160 200

HFS Time (min)
Figure 2 | Naturally secreted SDS-stable amyloid β-protein oligomers block hippocampal long-term
potentiation in vivo. a | Small oligomeric assemblies of amyloid β-protein (Aβ) are secreted from CHO cells that express
the familial Alzheimer’s disease causing β-amyloid precursor protein (APP) V717F mutation but are not present in
untransfected CHO cells. Monomers and SDS-stable dimers and trimers are indicated. All Aβ species were
immunoprecipitated from conditioned media with an Aβ polyclonal antibody, separated by SDS–PAGE and western
blotted with an Aβ monoclonal antibody. b | The SDS-stable oligomers of human Aβ shown in part a block hippocampal
long-term potentiation (LTP) in rats in vivo. Just 1.5 µl of conditioned medium from CHO cells secreting Aβ oligomers (blue)
or from untransfected CHO cells (orange) were injected intracerebroventricularly into wild-type adult rats. LTP of
excitatory synaptic transmission in the CA1 area of the hippocampus was induced by high frequency stimulation (HFS).
The conditioned medium that contained SDS-stable oligomers of human Aβ abrogated LTP, whereas the medium from
control cells had no effect. EPSP, excitatory postsynaptic potential.

Binding of oligomers to synaptic plasma membranes. cytoplasmic ubiquitin system. Again, selective receptors
The biochemical mechanism by which soluble oligomers might be involved.
bind to synaptic plasma membranes and interfere with
the complex system of receptor and/or channel proteins Future challenges. It is now important to confirm and
and signalling pathways that are required for synaptic extend the above findings using other experimental
plasticity is under intensive study. Intriguingly, Kamenetz systems in which natural Aβ oligomers are harvested
and colleagues70 showed that neuronal electrical activity and purified — preferably from APP transgenic mouse
stimulated BACE and therefore increased Aβ generation, brains or even AD brain tissue — and then systematically
and the resulting increased levels of Aβ then depressed characterized on hippocampal slices or transgenic ani-
synaptic transmission. Moreover, Cirrito and colleagues mals. The recently described experimental paradigm of
used in vivo microdialysis probes to demonstrate that exogenous induction of cerebral β-amyloidogenesis by
interstitial fluid Aβ concentrations correlate with the injection of amyloid-plaque-containing brain extracts77
synaptic activity in APP transgenic mice74. might provide an in vivo animal model to identify neuro-
It is possible that soluble Aβ oligomers interfere toxic properties of naturally produced and purified
with signalling pathways downstream of certain oligomeric assemblies of Aβ.
NMDA (N-methyl-d-aspartate) or AMPA (α-amino- Unfortunately, we are just at the beginning of deci-
3-hydroxy-5-methyl-4-isoxazole) receptors at synaptic phering the cellular and molecular mechanisms that
plasma membranes in a manner that allows an initial underlie neurotoxicity caused by soluble oligomers. It is
LTP response but not its persistence. In this regard, it is clear that human neurodegeneration in general and AD
interesting to note that the application of Aβ to cortical in particular are strongly age dependent, so that a host
slices has been reported to promote the endocytosis of age-related biochemical alterations can make neurons
of some NMDA receptors through a mechanism that and their processes more vulnerable to the effects of
involves initial binding of the Aβ to α7 nicotinic recep- soluble oligomers. We do not yet know whether soluble
tors75. Consistent with this model, Aβ treatment lowered oligomers bind to specific receptors, causing selective
NMDA-evoked currents. It has also been shown that malfunctions in a subset of neurons, or whether oligo-
oligomeric Aβ can interfere indirectly with LTP through mers adsorb nonspecifically to various receptors and
an inhibition of a ubiquitin C-terminal hydrolase channel proteins and cause a range of adverse signalling
(UCH). This enzyme enhances recycling of ubiquitin, effects. In addition, there is evidence that amyloidogenic
which is required to label unfolded proteins destined for proteins (at least in their synthetic form) can physically
proteasome degradation. Aβ treatment inhibited UCH, intercollate into and penetrate membranes, leading
which in turn blocked LTP76. However, it is currently to permeabilization78. In this scenario, pore-forming
unclear how the extracellular Aβ can reach and affect the assemblies might arise from low-n oligomers.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 107

© 2007 Nature Publishing Group
REVIEWS

Such diverse mechanisms of oligomer-induced neuro- Yet another example comes from the study of tau,
nal, and also glial, membrane alteration can presumably the amyloidogenic protein of familial frontotemporal
lead to numerous downstream biochemical changes, dementia and AD (BOX 1), which can form soluble oligo-
including altered signalling pathways, perturbed calcium mers. Dimeric intermediates of recombinant tau can
homeostasis, production of reactive oxygen species, trig- further assemble into 8–14-unit aggregates in vitro86.
gering of inflammatory cascades and mitochondrial dys- Cytotoxicity has been associated with such oligomeric
function. However, many of our current concepts about structures that seem to arise prior to the formation of
the molecular mechanisms involved in Aβ-mediated classical paired helical filaments87.
neuronal injury have arisen from in vitro experiments More recently, TDP43 was found to be deposited in tau-
and tissue-culture studies with synthetic Aβ fibrils at and synuclein-negative inclusions in a disorder designated
supraphysiological concentrations and should therefore as frontotemporal lobar degeneration with ubiquitin-
be considered with care until they are validated in vivo. positive inclusions (FTLD-U) as well as in ALS88. TDP43
Recent in vivo experiments such as those mentioned is a ubiquitously expressed, highly conserved nuclear pro-
above have supported the new concept that memory tein that can bind to nucleic acids. Pathological TDP43
loss can be caused by bioactive soluble oligomers that in FTLD-U becomes hyperphosphorylated, ubiquitylated
directly disrupt synaptic integrity in the hippocampus. and cleaved to generate an insoluble, disease-associated
25-kDa C-terminal fragment. Currently, it is not known
Soluble oligomers in other disorders whether the 25-kDa fragment forms any type of oligo-
Importantly, small oligomeric assemblies of misfolded mer. However, its apparent lack of affinity for Congo red
proteins have been identified in other neurodegenerative might indicate that at least large β-sheet-rich fibres are
disorders. This has been facilitated by the generation of not formed.
an antibody to synthetic Aβ oligomers that specifically Clearly, more work is needed to connect in vitro find-
identifies a common structure that is present in several ings that used high concentrations of recombinant and
different amyloid-prone synthetic proteins79. The smallest synthetic proteins to what occurs endogenously in bio-
soluble Aβ oligomers that are recognized by this antibody logical tissues. Structural data on different amyloidogenic
are probably octamers. Strikingly, this conformation- proteins that were affinity purified with, for example,
specific antibody not only detected oligomeric assemblies the antibody described above79 are required to better
of Aβ but also detected soluble oligomers derived from understand potentially shared mechanisms of neuronal
the following recombinant disease proteins: α-synuclein dysfunction in distinct neurodegenerative disorders.
(involved in PD), islet amyloid polypeptide (IAPP;
involved in type II diabetes), huntingtin with extended Targeting Aβ oligomers
polyglutamine stretches (involved in HD) and the prion In the case of AD, compounds that are currently in pre-
protein (PrP; involved in transmissible and inherited clinical and early clinical development that lower the
spongiform encephalopathies) (see BOX 1 for neuro- production of Aβ monomers, such as β- or γ-secretase
degenerative disease histopathologies). It seems that inhibitors (FIG. 3), also decrease the formation of soluble,
different aggregation-prone proteins that characterize bioactive oligomers69. But the findings on the specific
neurodegenerative diseases such as AD, PD, HD and pathogenicity of small, soluble oligomers (see above) has
prion disorders have some common structural features. raised the question of attempting to target the oligomeric
Therefore, soluble proteins of entirely different sequences forms directly without perturbing the enzymatic activi-
can fold into β-sheet-rich structures that contain one or ties of the secretases, which process many important
more shared conformational epitopes. This realization substrates32,89 (FIG. 3).
not only indicates that assemblies produced by different One way this might be accomplished is through
disease-causing amyloid proteins might initiate similar the new and promising approach of immunotherapy,
cytotoxic mechanisms, but also raises the possibility through either active Aβ-peptide vaccination 90 or
of targeting their common structures for therapeutic passive infusion of anti-Aβ monoclonal antibodies91
treatment (see below). (FIG. 3). A Phase 2 trial of an Aβ1–42 vaccine in patients
From studies on PD, there is evidence for pathogenic with AD was associated with the development of a
activity of oligomeric assemblies of α-synuclein. Like T-cell-mediated, autoimmune meningoencephalitis in
synthetic Aβ, recombinant α-synuclein can form pore- 6% of patients, leading to cessation of dosing92. Despite
like annular structures in vitro80. Moreover, in vitro PF this self-limited reaction, all patients were followed,
formation is accelerated by α-synuclein mutations that and a portion of the patients, who subsequently devel-
cause early onset PD; the A30P α-synuclein mutation oped anti-Aβ antibodies, seemed to have some slowing
promotes the formation of spherical PFs but actually of their cognitive decline, as measured by a general
slows the conversion to insoluble amyloid fibrils59,81–83. cognitive-status test93. A later report, which included all
Similar oligomeric assemblies have also been associated of the vaccine-trial subjects, could not confirm an overall
with other neurodegenerative diseases. These include benefit on such general cognitive tests94. However, when
annular structures that are composed of polyglutamine responders were grouped on the basis of their antibody
(for example, the aggregation-causing repetitive sequence titre, a titre-dependent stabilization of certain tests of
in huntingtin84) and ring-like protofibrillar structures in declarative memory was found to be statistically signifi-
the rare neurodegenerative disorder known as familial cant94. Clearly, additional studies of Aβ immunotherapy
British dementia85. with higher numbers of patients and without interruption

108 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group
 REVIEWS

Aβ monomers Soluble Aβ neurons, at least in a mouse model of AD97. This finding

oligomers supports the idea that Aβ oligomers are at the top of the
Anti-aggregation,
degradation, amyloid cascade (BOX 2) and can induce tau aggregation.
vaccination Proposed mechanisms of action for anti-Aβ immuno-
LTP therapy include local microglia-mediated plaque phago-
γ-secretase
inhibition cytosis91, systemic sequestration of Aβ in plasma that
β-secretase results in its efflux from the brain98 and/or the inhibi-
Aβ inhibition γ38 γ42 tion of Aβ oligomerization and cytotoxicity99,100. We will
Aβ
γ-secretase
modifier focus here on the neutralization of Aβ oligomers; refer to
...GGVVIATVIVITLVMLKKK... recent reviews101,102 for the other mechanisms.

Neutralization of oligomeric Aβ. An early hint that Aβ

γ-secretase immunotherapy might neutralize a subset of soluble
inhibition AICD Aβ assembly forms (FIG. 3) came from the finding that
Figure 3 | Therapeutic approaches targeting amyloid β-protein production and antibody-mediated reversal of cognitive dysfunction
oligomerization. Several therapeutic strategies, some of them already in early clinical in APP transgenic mice did not necessarily require a
trials, have emerged from the growing knowledge of amyloid β-protein (Aβ) generation decrease in total brain Aβ levels99,100. This result is con-
and the effects of soluble Aβ oligomers on synaptic function. Inhibition of either β- or sistent with in vitro studies showing that Aβ antibodies
γ-secretase should lower Aβ generation. However, this might also block the biological can inhibit synthetic Aβ fibrillogenesis and can even
function of these enzymes in cell-fate decisions (γ-secretase32) or in myelination disrupt pre-existing fibrils103,104. The disassembly of
(β-secretase89), potentially causing unwanted side effects. γ-secretase modulators such
fibrils is accompanied by strongly reduced neurotoxicity
as certain non-steroidal anti-inflammatory drugs (NSAIDs) and their derivatives, on the
other hand, will not block the γ-secretase cleavage but rather shift its cleavage site from in culture103. However, such in vitro cytotoxicity assays
the rapidly aggregating 42-residue variant to the far less amyloidogenic 38-residue form are rather crude, as ill-defined mixtures of synthetic
(shaded amino acids are in the transmembrane domain). This shift will not affect potential monomers, oligomers, PFs and fibrils are added to the
signalling functions of γ-secretase substrates (Fig. 1a). Other therapeutic strategies have media of cultured cells at supraphysiological (1 µM to
emerged from the concept of selectively lowering the levels of soluble Aβ oligomers. 1 mM) concentrations. Therefore, in vivo analysis of
All secretase inhibitors will also decrease oligomer formation; however, this could be the synaptic effects of natural oligomers was required to
accompanied by the side effects mentioned above. Anti-aggregation drugs, which determine whether active or passive Aβ immunotherapy
prevent or even disrupt oligomers, will shift the pool of oligomeric Aβ to the benign can specifically prevent oligomer-mediated synaptic
Aβ monomers (dashed arrow). Aβ immunotherapy might also allow the selective dysfunction.
neutralization of Aβ oligomers. Antibodies directed selectively against these oligomers
As reviewed earlier, naturally secreted soluble oligo-
might prevent their formation and/or disrupt pre-existing oligomers (dashed arrow).
Any Aβ-related treatment strategy will lower the amyloid plaque load as well. AICD, mers of human Aβ injected intraventricularly into wild-
β-amyloid precursor protein intracellular domain; LTP, long-term potentiation. type rats potently inhibited hippocampal LTP69 (FIG. 2),
providing one model system for studying potential
therapies (FIG. 3). In this regard, co-injection of an Aβ
monoclonal antibody completely prevented the inhibi-
by adverse events are required to fully confirm this tion of LTP72. Importantly, endogenous antibodies that
apparent slowing of decline in memory. An alternative were raised by active Aβ vaccination also protected rats
immunotherapeutic approach to circumvent brain from oligomer-mediated LTP inhibition; the degree of
inflammation from active vaccination uses passive infu- protection correlated roughly with the level of circulating
sions of a monoclonal antibody to the Aβ N terminus, antibodies that recognized oligomers72.
and this is currently in Phase 2 clinical testing.
Brain examination of a few patients who died of natural Selective degradation of oligomers and fibrils. Another
causes after receiving this active vaccine has revealed areas therapeutic approach could be to stimulate the selective
of the cerebral cortex with apparent decreases in amyloid degradation of oligomers and fibrils (FIG. 3). However,
plaque burden, accompanied by local Aβ-containing most Aβ fibrils can be largely resistant to proteolytic
microglia95,96. As a note of caution, one needs to consider degradation, as reflected by their progressive accumu-
recently reported neuropathological findings in two lation in AD brains. It is unlikely that the catalytic centres
patients from the interrupted vaccine trial; although of most enzymes that have been shown to degrade Aβ
cored amyloid plaque levels decreased, soluble Aβ levels monomers can accommodate larger oligomers and
increased sharply96. This might be due to a lack of full fibrils. In fact, one of the main Aβ proteases, IDE, cata-
clearance of all Aβ species under the conditions of the bolizes natural Aβ monomers but not soluble dimers
interrupted trial. Whether the increased levels of soluble and trimers105, and incubation of these secreted Aβ
Aβ also result in enhanced amounts of soluble Aβ oligo- species with IDE does not rescue LTP inhibition69.
mers in these cases is currently unknown. Ongoing trials Among the proteases tested so far, only plasmin
of passive immunotherapy use antibodies that can bind (a protease with a central role in fibrinolysis) and cathep-
not only to amyloid plaques but also to soluble monomers sin B are capable of directly degrading Aβ oligomers and
and oligomers. higher aggregates106,107. But even if only Aβ monomers
Strikingly, immunotherapy does not only affect Aβ could be degraded efficiently, oligomerization and con-
oligomerization and plaque formation, but it also seems sequent neurotoxicity should be reduced. This concept is
to lead to a clearance of hyperphosphorylated tau in supported by the beneficial effects of overexpressing either

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 109

© 2007 Nature Publishing Group
REVIEWS

IDE or neprilysin (another important Aβ-degrading enzyme) 115,116. An orally bioavailable compound,
protease108) in APP transgenic mice109. Moreover, Aβ R-flurbiprofen (Flurizan), which is derived from an
oligomers accumulated early in life in an APP transgenic NSAID but that lacks cyclo-oxygenase inhibitory activity,
mouse line that lacked neprilysin, and these led to altered is currently in a Phase 3 human trial (see Alzforum in
hippocampal synaptic plasticity110. Further information).

Destabilization of Aβ oligomers. The search is on for Conclusions and perspectives

small molecules that can specifically inhibit the form- The identification of soluble oligomers of Aβ as dif-
ation of Aβ oligomers and/or prevent their binding fusible assemblies that are capable of interfering with
to and stabilization on neuronal membranes (FIG. 3). synaptic function and integrity provides an important
Proteoglycans and their constituent glycosaminoglycans opening for understanding the basis of memory loss
are associated with amyloid plaques in AD brain tissue in AD. Importantly, similar findings in other disorders
and might stabilize the aggregates and make them more might indicate convergent mechanisms of synaptic
resistant to proteolysis. Studies of this interaction have failure in several neurodegenerative diseases. These
indicated that they specifically enhance the nucleation and other findings that are not discussed owing to
phase of Aβ aggregation111. On the basis of these findings, space limitations indicate that amyloid-based thera-
one compound (Alzhemed) that is designed to prevent peutic strategies in AD are beginning to specifically
Aβ from interacting with glycosaminoglycans and target the generation of soluble Aβ oligomers and
proteoglycans is now in a Phase 3 AD trial (see Alzforum their downstream consequences. The current hope is
in Further information). that one or more of the approaches described above
A related approach is derived from evidence that the can slow the progression of AD and its harbinger, mild
initiation and propagation of synthetic Aβ aggregates cognitive impairment, even as numerous questions
can be mediated in part by membrane glycolipids. about the precise roles of oligomers of misfolded pro-
Certain glycolipid-derived sugars, such as scyllo-inositol teins such as Aβ remain to be answered. For AD, these
(AZD-103), have been shown to stabilize synthetic Aβ unresolved questions include: is there a specific neu-
assemblies at the low-n oligomer stage112. Oral adminis- ronal (or glial) receptor for Aβ oligomers, or do they
tration of scyllo-inositol to APP transgenic mice resulted indiscriminately perturb diverse receptor or channel
in significant decreases in plaque burden, surrounding proteins on synaptic membranes? Is the induction of
cytopathology and associated deficits in water-maze tau hyperphosphorylation and its aggregation into
performance113. paired helical filaments in AD a direct consequence of
Last, certain non-steroidal anti-inflammatory drugs the accumulation of intra- or extraneuronal Aβ oligo-
(NSAIDs) indirectly reduce oligomer formation by the mers or are these much later downstream events? Can
modulation of γ-secretase activity114. Such drugs decrease soluble Aβ oligomers trigger a pro-inflammatory cas-
the formation of the aggregation-prone Aβ42 species by cade by directly or indirectly activating local microglia
shifting the γ-secretase cut by 4 amino acids N-terminal and astrocytes, and is this an important contributor
to this site to produce the smaller, non-aggregating Aβ38 to synaptic failure? Is there a common structure and
(FIGS 1b,3). Because these drugs do not block γ-secretase cytotoxic mechanism of oligomeric assemblies derived
activity per se, they do not show deleterious side effects from different neurodegenerative diseases? Let us hope
on the cell-fate decisions that require proper cleavage of that the development of safe and efficacious therapeu-
Notch by the PS–γ-secretase complex (these deleterious tics for AD will not require the full resolution of these
effects are observed with classical inhibition of this and other burning questions.

1. Gasser, T. Genetics of Parkinson’s disease. Curr. Opin. 9. Alzheimer, A. Über eine eigenartige Erkrankung der 16. Olson, M. I. & Shaw, C. M. Presenile dementia and
Neurol. 18, 363–369 (2005). Hirnrinde. Allg. Z. Psychiatrie Psychisch-Gerichtl. Alzheimer’s disease in mongolism. Brain 92,
2. Polymeropoulos, M. H. et al. Mutation in the Med. 64, 146–148 (1907). 147–156 (1969).
α-synuclein gene identified in families with Parkinson's 10. Glenner, G. G. & Wong, C. W. Alzheimer’s disease: 17. Kang, J. et al. The precursor of Alzheimer’s disease
disease. Science 276, 2045–2047 (1997). initial report of the purification and characterization of amyloid A4 protein resembles a cell-surface receptor.
3. Kahle, P. J., Haass, C., Kretzschmar, H. A. & a novel cerebrovascular amyloid protein. Biochem. Nature 325, 733–736 (1987).
Neumann, M. Structure/function of α-synuclein in Biophys. Res. Commun. 120, 885–890 (1984). 18. Levy, E. et al. Mutation of the Alzheimer’s disease
health and disease: rational development of animal 11. Masters, C. L. et al. Amyloid plaque core protein in amyloid gene in hereditary cerebral hemorrhage,
models for Parkinson's and related diseases. Alzheimer disease and Down syndrome. Proc. Natl Dutch type. Science 248, 1124–1126 (1990).
J. Neurochem. 82, 449–457 (2002). Acad. Sci. USA 82, 4245–4249 (1985). 19. Goate, A. et al. Segregation of a missense mutation
4. Spillantini, M. G. et al. α-Synuclein in Lewy bodies. 12. Grundke-Iqbal, I. et al. Abnormal phosphorylation of in the amyloid precursor protein gene with familial
Nature 388, 839–840 (1997). the microtubule-associated protein τ (tau) in Alzheimer Alzheimer’s disease. Nature 349, 704–706
5. Baba, M. et al. Aggregation of α-synuclein in Lewy cytoskeletal pathology. Proc. Natl Acad. Sci. USA 83, (1991).
bodies of sporadic Parkinson's disease and dementia 4913–4917 (1986). 20. Chartier-Harlin, M. C. et al. Early-onset Alzheimer’s
with Lewy bodies. Am. J. Pathol. 152, 879–884 13. Kosik, K. S., Joachim, C. L. & Selkoe, D. J. disease caused by mutations at codon 717 of the
(1998). Microtubule-associated protein τ (tau) is a major β-amyloid precursor protein gene. Nature 353,
6. Rosen, D. R. Mutations in Cu/Zn superoxide dismutase antigenic component of paired helical filaments in 844–846 (1991).
gene are associated with familial amyotrophic lateral Alzheimer disease. Proc. Natl Acad. Sci. USA 83, 21. Mullan, M. et al. A pathogenic mutation for probable
sclerosis. Nature 364, 362 (1993). 4044–4048 (1986). Alzheimer’s disease in the APP gene at the
7. Deng, H. X. et al. Amyotrophic lateral sclerosis and 14. Nukina, N. & Ihara, Y. One of the antigenic N-terminus of β-amyloid. Nature Genet. 1, 345–347
structural defects in Cu, Zn superoxide dismutase. determinants of paired helical filaments is related to (1992).
Science 261, 1047–1051 (1993). τ protein. J. Biochem. 99, 1541–1544 (1986). 22. Selkoe, D. J. The molecular pathology of Alzheimer’s
8. A novel gene containing a trinucleotide repeat that is 15. Glenner, G. G. & Wong, C. W. Alzheimer’s disease and disease. Neuron 6, 487–498 (1991).
expanded and unstable on Huntington’s disease Down’s syndrome: sharing of a unique cerebrovascular 23. Hardy, J. A. & Higgins, G. A. Alzheimer’s disease: the
chromosomes. The Huntington’s Disease Collaborative amyloid fibril protein. Biochem. Biophys. Res. amyloid cascade hypothesis. Science 256, 184–185
Research Group. Cell 72, 971–983 (1993). Commun. 122, 1131–1135 (1984). (1992).

110 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group
 REVIEWS

References 22 and 23 set out the amyloid cascade Evidence that small, diffusible aggregates of Demonstrates the effects of Aβ on synaptic
hypothesis, for which strong experimental evidence intracellular huntingtin can confer neurotoxicity, function upon the stimulation of neuronal activity.
is now accumulating. perhaps analogously to soluble Aβ oligomers. 71. Townsend, M., Shankar, G. M., Mehta, T., Walsh, D. M.
24. Hardy, J. & Selkoe, D. J. The amyloid hypothesis of 49. Schaffar, G. et al. Cellular toxicity of polyglutamine & Selkoe, D. J. Effects of secreted oligomers of
Alzheimer’s disease: progress and problems on the expansion proteins: mechanism of transcription factor amyloid β-protein on hippocampal synaptic plasticity:
road to therapeutics. Science 297, 353–356 (2002). deactivation. Mol. Cell 15, 95–105 (2004). a potent role for trimers. J. Physiol. 572, 477–492
25. Haass, C. et al. Amyloid β-peptide is produced by 50. Cummings, C. J. et al. Mutation of the E6-AP ubiquitin (2006).
cultured cells during normal metabolism. Nature 359, ligase reduces nuclear inclusion frequency while 72. Klyubin, I. et al. Amyloid β protein immunotherapy
322–325 (1992). accelerating polyglutamine-induced pathology in SCA1 neutralizes Aβ oligomers that disrupt synaptic
26. Seubert, P. et al. Isolation and quantification of soluble mice. Neuron 24, 879–892 (1999). plasticity in vivo. Nature Med. 11, 556–561 (2005).
Alzheimer’s β-peptide from biological fluids. Nature 51. Tsai, J., Grutzendler, J., Duff, K. & Gan, W. B. Fibrillar LTP inhibition by soluble oligomers of human Aβ is
359, 325–327 (1992). amyloid deposition leads to local synaptic prevented by active and passive Aβ immunotherapy.
27. Shoji, M. et al. Production of the Alzheimer amyloid abnormalities and breakage of neuronal branches. 73. Cleary, J. P. et al. Natural oligomers of the amyloid-β
β-protein by normal proteolytic processing. Science Nature Neurosci. 7, 1181–1183 (2004). protein specifically disrupt cognitive function. Nature
258, 126–129 (1992). 52. Caughey, B. & Lansbury, P. T. Protofibrils, pores, fibrils, Neurosci. 8, 79–84 (2005).
28. Busciglio, J., Gabuzda, D. H., Matsudaira, P. & and neurodegeneration: separating the responsible 74. Cirrito, J. R. et al. Synaptic activity regulates
Yankner, B. A. Generation of β-amyloid in the secretory protein aggregates from the innocent bystanders. interstitial fluid amyloid-β levels in vivo. Neuron 48,
pathway in neuronal and nonneuronal cells. Proc. Natl Annu. Rev. Neurosci. 26, 267–298 (2003). 913–922 (2005).
Acad. Sci. USA 90, 2092–2096 (1993). 53. Teplow, D. B. Structural and kinetic features of amyloid An in vivo demonstration of the effects of synaptic
29. Haass, C. Take five-BACE and the γ-secretase quartet β-protein fibrillogenesis. Amyloid 5, 121–142 (1998). activity on Aβ levels.
conduct Alzheimer's amyloid β-peptide generation. 54. Harper, J. D., Wong, S. S., Lieber, C. M. & 75. Snyder, E. M. et al. Regulation of NMDA receptor
EMBO J. 23, 483–488 (2004). Lansbury, P. T. Observation of metastable Aβ amyloid trafficking by amyloid-β. Nature Neurosci. 8,
A review on APP processing by β- and γ-secretase. protofibrils by atomic force microscopy. Chem. Biol. 4, 1051–1058 (2005).
30. Brown, M. S., Ye, J., Rawson, R. B. & Goldstein, J. L. 119–125 (1997). A cellular mechanism that describes how Aβ lowers
Regulated intramembrane proteolysis: a control 55. Hartley, D. M. et al. Protofibrillar intermediates of NMDA-evoked currents.
mechanism conserved from bacteria to humans. Cell amyloid β-protein induce acute electrophysiological 76. Gong, B. et al. Ubiquitin hydrolase Uch-L1 rescues
100, 391–398 (2000). changes and progressive neurotoxicity in cortical β-amyloid-induced decreases in synaptic function
31. Weihofen, A. & Martoglio, B. Intramembrane-cleaving neurons. J. Neurosci. 19, 8876–8884 (1999). and contextual memory. Cell 126, 775–788
proteases: controlled liberation of functional proteins 56. Walsh, D. M., Lomakin, A., Benedek, G. B., (2006).
and peptides from membranes. Trends Cell Biol. 13, Condron, M. M. & Teplow, D. B. Amyloid β-protein 77. Meyer-Luehmann, M. et al. Exogenous induction of
71–78 (2003). fibrillogenesis. Detection of a protofibrillar cerebral β-amyloidogenesis is governed by agent and
32. Mumm, J. S. & Kopan, R. Notch signaling: from the intermediate. J. Biol. Chem. 272, 22364–22372 host. Science 313, 1781–1784 (2006).
outside in. Dev. Biol. 228, 151–165 (2000). (1997). 78. Glabe, C. G. & Kayed, R. Common structure and toxic
33. Wolfe, M. S. et al. Two transmembrane aspartates in 57. Walsh, D. M. et al. Amyloid β-protein fibrillogenesis. function of amyloid oligomers implies a common
presenilin-1 required for presenilin endoproteolysis Structure and biological activity of protofibrillar mechanism of pathogenesis. Neurology 66, S74–S78
and γ-secretase activity. Nature 398, 513–517 intermediates. J. Biol. Chem. 274, 25945–25952 (2006).
(1999). (1999). 79. Kayed, R. et al. Common structure of soluble amyloid
34. Steiner, H. et al. Glycine 384 is required for presenilin-1 58. Bitan, G. et al. Amyloid β-protein (Aβ) assembly: Aβ40 oligomers implies common mechanism of
function and is conserved in polytopic bacterial and Aβ42 oligomerize through distinct pathways. pathogenesis. Science 300, 486–489 (2003).
aspartyl proteases. Nature Cell Biol. 2, 848–851 Proc. Natl Acad. Sci. USA 100, 330–335 (2003). Describes common conformational epitopes on
(2000). 59. Lashuel, H. A., Hartley, D., Petre, B. M., Walz, T. & oligomers of completely distinct amyloidogenic
35. Kimberly, W. T. et al. γ-Secretase is a membrane Lansbury, P. T. Jr. Neurodegenerative disease: amyloid proteins.
protein complex comprised of presenilin, nicastrin, pores from pathogenic mutations. Nature 418, 291 80. Lashuel, H. A. et al. α-Synuclein, especially the
Aph-1, and Pen-2. Proc. Natl Acad. Sci. USA 100, (2002). Parkinson's disease-associated mutants, forms pore-
6382–6387 (2003). 60. Lambert, M. P. et al. Diffusible, nonfibrillar ligands like annular and tubular protofibrils. J. Mol. Biol. 322,
36. Takasugi, N. et al. The role of presenilin cofactors in derived from Aβ1–42 are potent central nervous 1089–1102 (2002).
the γ-secretase complex. Nature 422, 438–441 system neurotoxins. Proc. Natl Acad. Sci. USA 95, 81. Conway, K. A., Harper, J. D. & Lansbury, P. T. Jr. Fibrils
(2003). 6448–6453 (1998). formed in vitro from α-synuclein and two mutant forms
37. Edbauer, D. et al. Reconstitution of γ-secretase 61. Gong, Y. et al. Alzheimer’s disease-affected brain: linked to Parkinson's disease are typical amyloid.
activity. Nature Cell Biol. 5, 486–488 (2003). presence of oligomeric Aβ ligands (ADDLs) suggests a Biochemistry 39, 2552–2563 (2000).
38. Sastre, M. et al. Presenilin-dependent γ-secretase molecular basis for reversible memory loss. Proc. Natl 82. Conway, K. A., Harper, J. D. & Lansbury, P. T.
processing of β-amyloid precursor protein at a site Acad. Sci. USA 100, 10417–10422 (2003). Accelerated in vitro fibril formation by a mutant
corresponding to the S3 cleavage of Notch. EMBO 62. Lesne, S. et al. A specific amyloid-β protein assembly α-synuclein linked to early-onset Parkinson disease.
Rep. 2, 835–841 (2001). in the brain impairs memory. Nature 440, 352–357 Nature Med. 4, 1318–1320 (1998).
39. Weidemann, A. et al. A novel ε-cleavage within the (2006). 83. Conway, K. A. et al. Acceleration of oligomerization,
transmembrane domain of the Alzheimer amyloid Identification of a brain-derived Aβ oligomeric not fibrillization, is a shared property of both
precursor protein demonstrates homology with Notch assembly, which impairs memory. α-synuclein mutations linked to early-onset
processing. Biochemistry 41, 2825–2835 (2002). 63. Podlisny, M. B. et al. Aggregation of secreted amyloid Parkinson's disease: implications for pathogenesis and
40. Gu, Y. et al. Distinct intramembrane cleavage of the β-protein into sodium dodecyl sulfate-stable oligomers therapy. Proc. Natl Acad. Sci. USA 97, 571–576
β-amyloid precursor protein family resembling in cell culture. J. Biol. Chem. 270, 9564–9570 (2000).
γ-secretase-like cleavage of Notch. J. Biol. Chem. 276, (1995). 84. Marchut, A. J. & Hall, C. K. Spontaneous formation of
35235–35238 (2001). 64. Walsh, D. M., Tseng, B. P., Rydel, R. E., Podlisny, M. B. annular structures observed in molecular dynamics
41. Qi-Takahara, Y. et al. Longer forms of amyloid β & Selkoe, D. J. The oligomerization of amyloid simulations of polyglutamine peptides. Comput. Biol.
protein: implications for the mechanism of β-protein begins intracellularly in cells derived from Chem. 30, 215–218 (2006).
intramembrane cleavage by γ-secretase. J. Neurosci. human brain. Biochemistry 39, 10831–10839 85. Srinivasan, R., Marchant, R. E. & Zagorski, M. G.
25, 436–445 (2005). (2000). ABri peptide associated with familial British dementia
42. Zhao, G. et al. γ-cleavage is dependent on ζ-cleavage 65. Funato, H., Enya, M., Yoshimura, M., Morishima- forms annular and ring-like protofibrillar structures.
during the proteolytic processing of amyloid precursor Kawashima, M. & Ihara, Y. Presence of sodium dodecyl Amyloid 11, 10–13 (2004).
protein within its transmembrane domain. J. Biol. sulfate-stable amyloid β-protein dimers in the 86. Wille, H., Drewes, G., Biernat, J., Mandelkow, E. M. &
Chem. 280, 37689–37697(2005). hippocampus CA1 not exhibiting neurofibrillary tangle Mandelkow, E. Alzheimer-like paired helical filaments
43. Dickson, D. W. The pathogenesis of senile plaques. formation. Am. J. Pathol. 155, 23–28 (1999). and antiparallel dimers formed from microtubule-
J. Neuropathol. Exp. Neurol. 56, 321–339 (1997). 66. Enya, M. et al. Appearance of sodium dodecyl sulfate- associated protein τ in vitro. J. Cell. Biol. 118,
44. Naslund, J. et al. Correlation between elevated levels stable amyloid β-protein (Aβ) dimer in the cortex 573–584 (1992).
of amyloid β-peptide in the brain and cognitive decline. during aging. Am. J. Pathol. 154, 271–279 (1999). 87. Friedhoff, P., von Bergen, M., Mandelkow, E. M.,
JAMA 283, 1571–1577 (2000). 67. Kawarabayashi, T. et al. Dimeric amyloid β protein Davies, P. & Mandelkow, E. A nucleated assembly
45. Lue, L. F. et al. Soluble amyloid β peptide rapidly accumulates in lipid rafts followed by mechanism of Alzheimer paired helical filaments.
concentration as a predictor of synaptic change in apolipoprotein E and phosphorylated τ accumulation Proc. Natl Acad. Sci. USA 95, 15712–15717
Alzheimer's disease. Am. J. Pathol. 155, 853–862 in the Tg2576 mouse model of Alzheimer's disease. (1998).
(1999). J. Neurosci. 24, 3801–3809 (2004). 88. Neumann, M. et al. Ubiquitinated TDP-43 in
46. McLean, C. A. et al. Soluble pool of Aβ amyloid as a 68. Roher, A. E. et al. Morphology and toxicity of frontotemporal lobar degeneration and
determinant of severity of neurodegeneration in Aβ-(1–42) dimer derived from neuritic and vascular amyotrophic lateral sclerosis. Science 314,
Alzheimer's disease. Ann. Neurol. 46, 860–866 amyloid deposits of Alzheimer’s disease. J. Biol. Chem. 130–133 (2006).
(1999). 271, 20631–20635 (1996). 89. Willem, M. et al. Control of peripheral nerve
47. Wang, J., Dickson, D. W., Trojanowski, J. Q. & Lee, V. M. 69. Walsh, D. M. et al. Naturally secreted oligomers of myelination by the β-secretase BACE1. Science 314,
The levels of soluble versus insoluble brain Aβ amyloid β protein potently inhibit hippocampal long- 664–666 (2006).
distinguish Alzheimer's disease from normal and term potentiation in vivo. Nature 416, 535–539 90. Schenk, D. et al. Immunization with amyloid-β
pathologic aging. Exp. Neurol. 158, 328–337 (1999). (2002). attenuates Alzheimer-disease-like pathology in the
48. Arrasate, M., Mitra, S., Schweitzer, E. S., Segal, M. R. Defines a synaptotoxic function for small, soluble PDAPP mouse. Nature 400, 173–177 (1999).
& Finkbeiner, S. Inclusion body formation reduces oligomers of secreted Aβ in vivo. Initial report of the beneficial effects of Aβ
levels of mutant huntingtin and the risk of neuronal 70. Kamenetz, F. et al. APP processing and synaptic immunotherapy in a transgenic mouse model
death. Nature 431, 805–810 (2004). function. Neuron 37, 925–937 (2003). of AD.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 8 | FEBRUARY 2007 | 111

© 2007 Nature Publishing Group
REVIEWS

91. Bard, F. et al. Peripherally administered antibodies 106. Tucker, H. M. et al. The plasmin system is induced by 121. Chen, Y. R. & Glabe, C. G. Distinct early folding and
against amyloid β-peptide enter the central nervous and degrades amyloid-β aggregates. J. Neurosci. 20, aggregation properties of Alzheimer amyloid-β
system and reduce pathology in a mouse model of 3937–3946 (2000). peptides Aβ40 and Aβ42: stable trimer or tetramer
Alzheimer disease. Nature Med. 6, 916–919 (2000). 107. Mueller-Steiner, S. et al. Antiamyloidogenic and formation by Aβ42. J. Biol. Chem. 281,
92. Orgogozo, J. M. et al. Subacute meningoencephalitis neuroprotective functions of cathepsin B: 24414–24422 (2006).
in a subset of patients with AD after Aβ42 implications for Alzheimer’s disease. Neuron 51, 122. Nilsberth, C. et al. The ‘Arctic’ APP mutation (E693G)
immunization. Neurology 61, 46–54 (2003). 703–714 (2006). causes Alzheimer’s disease by enhanced Aβ protofibril
93. Hock, C. et al. Antibodies against β-amyloid slow 108. Iwata, N. et al. Identification of the major Aβ1–42- formation. Nature Neurosci. 4, 887–893 (2001).
cognitive decline in Alzheimer's disease. Neuron 38, degrading catabolic pathway in brain parenchyma: A genetic explanation for the development of rare
547–554 (2003). suppression leads to biochemical and pathological forms of AD that is strongly supportive of the
First report of the beneficial effects of Aβ deposition. Nature Med. 6, 143–150 (2000). amyloid hypothesis.
immunotherapy in a small cohort of vaccinated 109. Leissring, M. A. et al. Enhanced proteolysis of β- 123. Iwatsubo, T. et al. Visualization of Aβ42(43) and Aβ40
patients with AD. amyloid in APP transgenic mice prevents plaque in senile plaques with end-specific A β monoclonals:
94. Gilman, S. et al. Clinical effects of Aβ immunization formation, secondary pathology, and premature evidence that an initially deposited species is
(AN1792) in patients with AD in an interrupted trial. death. Neuron 40, 1087–1093 (2003). Aβ42(43). Neuron 13, 45–53 (1994).
Neurology 64, 1553–1562 (2005). 110. Huang, S. M. et al. Neprilysin-sensitive synapse- 124. Cai, X. D., Golde, T. E. & Younkin, S. G. Release of
95. Nicoll, J. A. et al. Neuropathology of human Alzheimer associated Aβ oligomers impair neuronal plasticity excess amyloid β protein from a mutant amyloid β
disease after immunization with amyloid-β peptide: and cognitive function. J. Biol. Chem. 281, protein precursor. Science 259, 514–516 (1993).
a case report. Nature Med. 9, 448–452 (2003). 17941–17951 (2006). 125. Citron, M. et al. Mutation of the β-amyloid precursor
First report of the apparent removal of Aβ deposits 111. McLaurin, J., Franklin, T., Zhang, X., Deng, J. & protein in familial Alzheimer's disease increases
in humans by a therapeutic agent. Fraser, P. E. Interactions of Alzheimer amyloid-β β-protein production. Nature 360, 672–674
96. Patton, R. L. et al. Amyloid-β peptide remnants in peptides with glycosaminoglycans effects on fibril (1992).
AN-1792-immunized Alzheimer's disease patients: a nucleation and growth. Eur. J. Biochem. 266, 126. Rovelet-Lecrux, A. et al. APP locus duplication causes
biochemical analysis. Am. J. Pathol. 169, 1048–1063 1101–1110 (1999). autosomal dominant early-onset Alzheimer disease
(2006). 112. McLaurin, J., Golomb, R., Jurewicz, A., Antel, J. P. & with cerebral amyloid angiopathy. Nature Genet. 38,
97. Oddo, S., Billings, L., Kesslak, J. P., Cribbs, D. H. & Fraser, P. E. Inositol stereoisomers stabilize an 24–26 (2006).
LaFerla, F. M. Aβ immunotherapy leads to clearance of oligomeric aggregate of Alzheimer amyloid β peptide 127. Singleton, A. B. et al. α-Synuclein locus triplication
early, but not late, hyperphosphorylated τ aggregates and inhibit aβ-induced toxicity. J. Biol. Chem. 275, causes Parkinson's disease. Science 302, 841
via the proteasome. Neuron 43, 321–332 (2004). 18495–18502 (2000). (2003).
Further evidence of a linear connection between Aβ 113. McLaurin, J. et al. Cyclohexanehexol inhibitors of Aβ 128. McGowan, E. et al. Aβ42 is essential for parenchymal
deposition and tau hyperphosphorylation in an aggregation prevent and reverse Alzheimer phenotype and vascular amyloid deposition in mice. Neuron 47,
animal model. in a mouse model. Nature Med. 12, 801–808 (2006). 191–199 (2005).
98. DeMattos, R. B., Bales, K. R., Cummins, D. J., New Aβ-aggregation inhibitors show beneficial 129. Bentahir, M. et al. Presenilin clinical mutations can
Paul, S. M. & Holtzman, D. M. Brain to plasma effects on plaque burden and behaviour in mice. affect γ-secretase activity by different mechanisms.
amyloid-β efflux: a measure of brain amyloid burden in 114. Weggen, S. et al. A subset of NSAIDs lower J. Neurochem. 96, 732–742 (2006).
a mouse model of Alzheimer's disease. Science 295, amyloidogenic Aβ42 independently of cyclooxygenase 130. Deng, Y. et al. Deletion of presenilin 1 hydrophilic loop
2264–2267 (2002). activity. Nature 414, 212–216 (2001). sequence leads to impaired γ-secretase activity and
99. Dodart, J. C. et al. Immunization reverses memory Discovery of certain NSAIDs as γ-secretase exacerbated amyloid pathology. J. Neurosci. 26,
deficits without reducing brain Aβ burden in modulators: they lead to shorter, less 3845–3854 (2006).
Alzheimer's disease model. Nature Neurosci. 5, amyloidogenic Aβ species.
452–457 (2002). 115. Doerfler, P., Shearman, M. S. & Perlmutter, R. M. Acknowledgements
100. Janus, C. et al. A β peptide immunization reduces Presenilin-dependent γ-secretase activity modulates We thank D. Teplow, G. Multhaup, E. Mandelkow and D.
behavioural impairment and plaques in a model of thymocyte development. Proc. Natl Acad. Sci. USA Walsh as well as the members of our laboratories for helpful
Alzheimer's disease. Nature 408, 979–982 (2000). 98, 9312–9317 (2001). discussions. We also thank M. Neumann for providing the
101. Gelinas, D. S., DaSilva, K., Fenili, D., St George- 116. Geling, A., Steiner, H., Willem, M., Bally-Cuif, L. & figures that are included in Box 1.
Hyslop, P. & McLaurin, J. Immunotherapy for Haass, C. A γ-secretase inhibitor blocks Notch
Alzheimer’s disease. Proc. Natl Acad. Sci. USA 101, signaling in vivo and causes a severe neurogenic Competing interests statement
14657–14662 (2004). phenotype in zebrafish. EMBO Rep. 3, 688–694 The authors declare competing financial interests: see Web
102. Selkoe, D. J. & Schenk, D. Alzheimer’s disease: (2002). version for details.
molecular understanding predicts amyloid-based 117. Scheuner, D. et al. Secreted amyloid β-protein similar
therapeutics. Annu. Rev. Pharmacol. Toxicol. 43, to that in the senile plaques of Alzheimer's disease is DATABASES
545–584 (2003). increased in vivo by the presenilin 1 and 2 and APP The following terms in this article are linked online to:
103. McLaurin, J. et al. Therapeutically effective antibodies mutations linked to familial Alzheimer’s disease. OMIM: http://www.ncbi.nlm.nih.gov/entrez/query.
against amyloid-β peptide target amyloid-β residues Nature Med. 2, 864–870 (1996). fcgi?db=OMIM
4–10 and inhibit cytotoxicity and fibrillogenesis. 118. Suzuki, N. et al. An increased percentage of long Alzheimer’s disease | Down syndrome | Huntington’s disease |
Nature Med. 8, 1263–1269 (2002). amyloid β-protein secreted by familial amyloid Parkinson’s disease
Important mechanistic insights about how Aβ β-protein precursor (βAPP717) mutants. Science 264, UniProtKB: http://ca.expasy.org/sprot
immunotherapy can prevent oligomerization and 1336–1340 (1994). APH1a | APH1b | IDE | huntingtin | nicastrin | PEN2 | PrP | PS1 |
cytotoxicity. 119. Burdick, D. et al. Assembly and aggregation
PS2
104. Solomon, B., Koppel, R., Hanan, E. & Katzav, T. properties of synthetic Alzheimer’s A4/β amyloid
Monoclonal antibodies inhibit in vitro fibrillar peptide analogs. J. Biol. Chem. 267, 546–554 FURTHER INFORMATION
aggregation of the Alzheimer β-amyloid peptide. Proc. (1992). Christian Haass’s homepage:
Natl Acad. Sci. USA 93, 452–455 (1996). 120. Jarrett, J. T., Berger, E. P. & Lansbury, P. T. Jr. The http://haass.web.med.uni-muenchen.de
105. Qiu, W. Q. et al. Insulin-degrading enzyme regulates carboxy terminus of the β-amyloid protein is critical Dennis J. Selkoe’s homepage: http://brain.bwh.harvard.edu
extracellular levels of amyloid β- protein by for the seeding of amyloid formation: implications for Alzforum: http://www.alzforum.org
degradation. J. Biol. Chem. 273, 32730–32738 the pathogenesis of Alzheimer's disease. Biochemistry Access to this links box is available online.
(1998). 32, 4693–4697 (1993).

112 | FEBRUARY 2007 | VOLUME 8 www.nature.com/reviews/molcellbio

© 2007 Nature Publishing Group