The amyloid cascade hypothesis
The first suggestion of an ‘amyloid hypothesis’ to explain the pathology of AD was that of Wong and colleagues [
11], who postulated that Aβ-derived cerebrovascular amyloid caused seepage of Aβ and other substances from plasma into the brain, leading to the formation of Aβ plaques and possibly neurodegeneration. This was revised into the more well-known amyloid cascade hypothesis that proposed that deposition of Aβ as neuritic plaques caused AD and led to neurofibrillary tangles, cell loss, vascular damage, and dementia [
12]. The amyloid hypothesis linking Aβ to AD catalyzed much of AD and Aβ research over the past two decades, and key studies during that period led to important revisions of the hypothesis that highlighted the central role of soluble Aβ oligomers in synaptic dysfunction and loss [
4,
13‐
19].
The current understanding of the Aβ cascade is derived primarily from
in vitro studies, the vast majority of which were conducted by using Aβ concentrations orders of magnitude greater than those found
in vivo. The assembly of Aβ peptides to form soluble oligomers, protofibrils, and fibrils is well documented to be affected by the isoform of the starting peptide, how the peptide is treated prior to assembly, the nature of the buffer, pH, peptide concentration, assembly temperature and time, agitation, and presence of other peptides or biological materials [
20‐
24]. Moreover, preparations of soluble Aβ species have been shown to change with time or upon dilution in different buffers, particularly in cell culture media [
19]. Thus, caution must be taken in extrapolating the results and conclusions of
in vitro studies to
in vivo reality. Although precise mechanistic details remain to be elucidated, a multitude of studies by numerous researchers support the conclusion that monomeric Aβ peptides assemble to form soluble Aβ oligomers, which further aggregate to form fibrillar Aβ [
17,
25].
Three distinct pools of Aβ species exist: Aβ monomers, soluble Aβ oligomers, and insoluble fibrillar Aβ. Each of these pools encompasses an array of individual species. Thus, monomeric Aβ peptides encompass various isoforms, including Aβ(1-40), Aβ(1-42), and Aβ(1-43), as well as numerous N-terminal truncated isoforms. (For example, see the introductory paragraphs of Tekirian and colleagues [
26].) Insoluble fibrillar Aβ aggregates are also known to be heterogeneous in structure and composed of various Aβ isoforms, both full-length as well as N-terminal and C-terminal truncated isoforms. (For example, see the introductory paragraphs of Roher and colleagues [
27] and Thal and colleagues [
28].)
Soluble Aβ oligomers are also heterogeneous and perhaps more ambiguous because of the different terminologies used by different researchers to describe them. (For an excellent review of soluble Aβ oligomers, see Benilova and colleagues [
9].) Thus, soluble Aβ oligomer species reported by various researchers have been termed sodium dodecyl sulfate (SDS)-stable Aβ oligomers [
29,
30], low-n-oligomers [
31‐
33], dimers [
33‐
35], trimers [
33,
36‐
38], tetramers [
37], paranuclei [
38,
39], dodecamers and Aβ*56 [
37,
40,
41], amyloid-derived diffusible ligands (ADDLs) [
42‐
44], Aβ oligomers [
45], prefibrillar oligomers [
46], Aβ globulomers [
40,
47‐
49], spherical oligomers [
50], amylospheroids [
51,
52], protofibrils [
20,
53,
54], and annular protofibrils [
55,
56]. Most of these terminologies refer to a mixture of metastable, soluble Aβ oligomer species in equilibrium rather than a discrete, stable species. In many cases, there is similarity in the species comprising the different preparations. In this review, we will use the terminology soluble Aβ oligomers to describe Aβ species composed of more than one Aβ peptide that remain in solution following centrifugation. Aβ fibrils or fibrillar Aβ will be used as a general description of insoluble Aβ plaques and vascular amyloid. (The term Aβ plaques rather than amyloid plaques will be used to more precisely describe amyloid plaques comprising primarily Aβ peptides versus those comprising primarily non-Aβ peptides.)
It has been established that the levels of Aβ monomer and deposited fibrillar Aβ in the AD brain are orders of magnitude greater than soluble Aβ oligomer levels [
57‐
65]. However, more than three decades of intense investigation has not provided a precise understanding of the extent of interconversion among the various Aβ species.
Aβ dimers have been demonstrated to form at physiologically relevant concentrations of Aβ monomer
in vivo[
66,
67]. Numerous
in vitro studies have shown that soluble Aβ oligomers and protofibrils form under similar conditions and that both species can proceed to form larger fibrillar species [
20,
43,
53,
68‐
74]. More recent studies provide further support for the addition of soluble Aβ oligomers to protofibrillar and fibrillar assemblies and also provide data indicating that oligomerization can occur via a secondary nucleation mechanism caused by fibrillar Aβ [
73,
74], possibly suggesting a mechanistic linkage between fibrillar Aβ and soluble Aβ oligomers. However, the precise mechanism of
in vivo fibril formation has not been fully established.
Aβ plaques are now generally considered to be a relatively benign species [
19,
72,
75]; however, whether Aβ plaques are a sink or a source for toxic soluble Aβ oligomers is a subject of debate. Several studies with γ-secretase inhibitors [
76‐
79] and two studies in transgenic mice that overexpress mutant amyloid precursor protein (APP) via a doxycycline-regulated promoter [
80,
81] show that subchronic or chronic suppression of Aβ production arrested Aβ plaque formation but did not result in observable clearance of existing plaques. In a related study [
82], Aβ(1-42) immunization of Tg2576 mice prior to significant Aβ deposition, at an age with modest deposition or at an age with significant deposition, showed that immunization prevented additional Aβ deposition but did not result in significant clearance of pre-existing Aβ. These studies indicate that mature, dense-core Aβ plaques are not in significant equilibrium with soluble Aβ pools. Studies in transgenic mouse models of AD reporting decreased soluble Aβ oligomer levels and reduced cognitive deficits with increasing Aβ plaque levels also provide support for the concept that Aβ plaques may be a sink for soluble Aβ species [
83,
84].
However, in an elegant study in APP transgenic mice using a novel microdialysis technique, the half-life of low-molecular-weight Aβ species in hippocampal interstitial fluid following inhibition of Aβ production by a potent secretase inhibitor was doubled in 12- to 15-month-old mice with Aβ deposits compared with 3-month-old mice without Aβ deposits [
85]. On the basis of these results, it was suggested that insoluble Aβ was in equilibrium with soluble Aβ in the interstitial fluid. In a related study using similar techniques, the temporal changes of low-molecular-weight Aβ species in interstitial fluid and Aβ levels in Tris-buffered saline (TBS), SDS, and formic acid extracts of brain tissues of 3-, 12-, and 24-month-old APP transgenic mice were reported [
86]. A significant age-dependent decrease in low-molecular-weight Aβ in interstitial fluid and significant increases in Aβ species in SDS and formic acid extracts of brain tissues were found. Although Aβ increased approximately seven-fold from baseline in the TBS extract of brain tissues, the level of Aβ in TBS extracts was less than 2% of the total Aβ in SDS and formic acid extracts. These results indicated an age-dependent sequestration of Aβ as non-diffusible cell matrix and membrane-bound Aβ and deposited Aβ plaques. Acute γ-secretase inhibition of Aβ production in plaque-free and plaque-rich mice suggested that Aβ(1-42) in the interstitial fluid of plaque-rich mice was derived primarily from Aβ(1-42) sequestered in brain parenchyma rather than from new biosynthesis. However, it was not possible to determine whether cell matrix and membrane-bound Aβ or Aβ plaques or both forms of Aβ were the source of Aβ(1-42) in the interstitial fluid of aged mice.
In a more recent study of the temporal changes of Aβ species in the interstitial fluid and brain tissues of APP transgenic mice, Takeda and colleagues [
87], using a 1,000-kDa-molecular-weight cutoff microdialysis probe, reported the temporal changes in soluble Aβ oligomer levels. In that study, consistent with previous studies [
85,
86], a significant, age-dependent increase in Aβ levels in TBS and formic acid extracts of brain tissues was found, and TBS extractable Aβ was less than 1% of formic acid extractable Aβ. However, unlike previous studies using a 35-kDa-molecular-weight cutoff microdialysis probe [
85,
86], a significant, age-dependent increase in interstitial fluid Aβ levels was found by using the larger-pore-sized microdialysis probe. The majority of the Aβ in interstitial fluid was determined to be higher-molecular-weight soluble Aβ oligomers, which showed an age-dependent increase relative to lower-molecular-weight Aβ species. Temporal changes in soluble Aβ oligomer levels in interstitial fluid and TBS brain extracts showed a significant positive correlation with formic acid Aβ extract levels. Comparison of high- and low-molecular-weight interstitial fluid Aβ levels at baseline and following acute treatment with a γ-secretase inhibitor showed slower clearance of higher-molecular-weight Aβ oligomers compared with low-molecular-weight Aβ species.
Narayan and colleagues [
88] have recently used single-molecule imaging techniques to investigate interactions between Aβ peptides and hippocampal cell membranes and reported results indicating that Aβ oligomers preferentially interact with membranes compared with Aβ monomer, thereby providing support for the results observed in the microdialysis studies. Thus, this study, coupled with those of the temporal changes of Aβ species in APP transgenic mice [
85‐
87], does not resolve the controversy regarding the sink/source relationship between fibrillar and soluble Aβ species.
The sink/source relationship between soluble and insoluble Aβ pools is complex, not fully understood, and subject to ongoing debate. Two different forms of Aβ plaques are present in the AD brain: vascular amyloid plaques that are primarily composed of Aβ(1-40) and Aβ plaques that are primarily composed of Aβ(1-42) [
89].
In vitro studies have shown that fibrillar Aβ(1-40) and Aβ(1-42) are in equilibrium with soluble Aβ [
90,
91] but that recycling of Aβ(1-40) fibrils is significantly faster that Aβ(1-42) fibrils [
91]. Moreover, it has been reported that plaque deposition proceeds in two distinct kinetic phases: an initial, reversible deposition phase followed by a time-dependent irreversible deposition phase [
92]. Furthermore, a recent study of the rates of formation of Aβ oligomers and fibrils provided evidence that the formation of soluble Aβ oligomers from monomeric Aβ is catalyzed by fibrillar Aβ [
73]. This study not only indicated a mechanistic linkage between soluble Aβ oligomers and fibrillar Aβ but also provided evidence showing that fibrillar Aβ can be a ‘source’ of soluble Aβ via catalyzed oligomerization of Aβ monomer rather than via a disaggregation process. This study also provided an alternative explanation for the study by Koffie and colleagues [
93], who reported a halo of soluble Aβ oligomers surrounding Aβ plaques and proposed that Aβ plaques were a possible source of soluble Aβ oligomers.
Although precise details remain to be fully elucidated, collectively, over two decades of studies on the mechanisms of formation of oligomeric and fibrillar Aβ, the temporal distribution of Aβ species in vitro and in vivo, the age-dependent effect of Aβ immunization, and the effects of subchronic and chronic suppression of Aβ production upon in vivo Aβ plaque levels show a significant, age-dependent increase in brain levels of soluble Aβ oligomers and deposited fibrillar Aβ. Collectively, the studies suggest that mature, dense-core Aβ plaques are not in equilibrium to any significant extent with soluble pools of Aβ, but that Aβ sequestered in the cell matrix and membranes and immature plaques is in equilibrium with soluble Aβ pools, and that fibrillar Aβ catalyzes Aβ monomer oligomerization, giving rise to soluble Aβ oligomers and growth of Aβ plaques.
Neuronal toxicity of amyloid-beta species
Monomeric Aβ, primarily the Aβ(1-40) and Aβ(1-42) peptides, is produced in various cell types throughout the body and reported to have trophic properties
in vitro[
94,
95]. There are no reports suggesting that monomeric Aβ possesses any direct cellular toxicity at physiologically relevant concentrations. Insoluble fibrillar aggregates of Aβ, vascular amyloid and Aβ plaques, exhibit relatively low
in vitro toxicity and have been proposed to be an
in vivo mechanism for removal of the more toxic soluble Aβ species [
83,
96,
97]. It was first suggested in 1995 that soluble Aβ species rather than fibrillar plaques could trigger neurotoxicity leading to AD [
98], and in the subsequent decades, many studies have shown soluble Aβ oligomers to be the most toxic Aβ form, causing both acute synaptotoxicity and inducing neurodegenerative processes [
5‐
10,
99‐
102].
Low-picomolar levels of soluble Aβ oligomers have been reported to have trophic properties
in vitro[
103‐
105], suggesting that therapeutic targeting of soluble Aβ oligomers may need to modulate oligomer levels versus completely sequestering or preventing formation of soluble Aβ oligomers. However, the concentration of Aβ42 at which enhancement of long-term potentiation (LTP) was observed in these studies was one to two orders of magnitude greater than levels of soluble Aβ oligomers reported in the cerebrospinal fluid (CSF) of human patients with AD, and the concentration above which inhibition of LTP was observed was an order of magnitude greater than the total levels of soluble Aβ species reported in the CSF of human patients with AD [
65]. Thus, the relevance of the reported
in vitro trophic properties of soluble Aβ oligomers to
in vivo conditions remains to be established.
Soluble Aβ oligomers bind with high affinity to synapses on a subset of hippocampal and cortical neurons [
19,
40,
106‐
108], indicative of specific binding to discrete cell surface receptors. In rodent hippocampal slice preparations, synaptic binding leads to rapid inhibition of LTP [
19,
40,
109], and injection of various soluble Aβ oligomer preparations directly into the rodent brain leads to reversible impairment of cognitive function [
31,
33,
110]. This aberrant signaling also causes accumulated biochemical damage within neurons [
100,
111,
112], such as hyperphosphorylation of tau [
100,
111‐
113], suggesting a linkage between Aβ and tau pathologies [
114‐
116]. Soluble Aβ oligomers have been isolated from extracts of postmortem AD brain tissue and from transgenic AD animal models [
37,
52,
117‐
119] and have been reported to be elevated in human AD brain relative to non-demented older patients [
59,
61,
62,
65,
119‐
124]. Importantly, a recent study suggests a correlation between CSF levels of soluble Aβ oligomers and cognitive deficits in human patients with AD [
61,
65]. These findings support the view that soluble Aβ oligomers interfere acutely with normal synaptic functions and contribute significantly to the memory loss and cognitive dysfunction characteristic of AD.
Structure and activity of soluble amyloid-beta oligomer species
There is substantial ongoing debate and research concerning the structure and activity of soluble Aβ species [
5,
7‐
9,
33,
41,
52,
72,
99,
125‐
127]. Various soluble Aβ oligomer species have been reported to display synaptic toxicity or induce cognitive deficits, including dimers [
33,
35,
128,
129], trimers [
32], dodecamers [
33,
37,
40], and larger soluble Aβ oligomers with molecular weights of 90 to 650 kDa (20 to 150 mers) [
19,
130,
131]. Unfortunately, the different methodologies for the preparation, characterization, and evaluation of soluble Aβ oligomer species by various research groups impede a direct comparison of the results reported, and few studies have directly compared the toxicities of different soluble Aβ species while using the same techniques.
Two studies have reported the comparative toxicities of different soluble Aβ oligomer species by using LDH (lactate dehydrogenase release) and MTT (oxidoreductase activity) cell viability assays. Deshpande and colleagues [
45] examined the relative toxicities of purified spherical Aβ(1-42) oligomers [
50], ADDLs [
132], and fibrils [
50]. However, because solutions of soluble Aβ oligomers in the neurobasal medium used in this study have been shown to change with time [
19], it is not possible to draw definitive structure-activity conclusions from the results of this study. In the second study, Ono and colleagues [
133] reported relative toxicities of purified, cross-linked Aβ(1-40) dimers, trimers, and tetramers, which were shown to be relatively stable under the assay conditions. In an MTT assay, half maximal effective concentration values were 67.3, 41.6, 24.5, 20.5, and 57.6 μM, respectively, for monomer, dimer, trimer, tetramer, and fibrils. Comparable toxicity was obtained in an LDH assay. The micromolar concentrations of Aβ species used in this study were approximately six orders of magnitude greater than
in vivo Aβ concentrations, and the relevance of cell culture MTT-type assays to the
in vivo synaptotoxicity of Aβ species has been questioned [
134]. Therefore, the results of the study by Ono and colleagues provide little understanding of the relative
in vivo toxicities of soluble Aβ species.
More recently, cognitive effects
in vivo were assessed in rats by using the alternating lever cyclic ratio assay following intracerebroventricular (ICV) injections of cell- and synthetically derived soluble Aβ oligomers [
33]. Monomer and low-n-mer soluble Aβ oligomers derived from 7PA2 cells [
29,
135‐
137], trimer and a dodecameric species (Aβ*56) extracted from Tg2576 mouse brain [
37], and synthetic soluble Aβ oligomers (ADDLs) [
43] were compared in this study. Injection of conditioned media from 7PA2 cells caused significant cognitive deficits. Evaluation of size exclusion chromatography (SEC)-enriched dimer or trimer fractions of 7PA2 cell-conditioned media showed significant cognitive deficits following injection of dimer-enriched fractions but a non-significant effect upon injection of trimer-enriched fractions. SEC-purified monomer had no effect. Aβ monomer was the predominant Aβ species in unfractionated 7PA2 conditioned media, and the amounts of dimer and trimer injected in the dimer- and trimer-enriched fractions were considerably greater than the amounts injected in unfractionated conditioned media. However, cognitive effects following injection of unfractionated conditioned media were comparable to or exceeded the observed effects following treatment with dimer- and trimer-enriched fractions. Thus, there is an inherent ambiguity regarding the results reported for 7PA2-derived Aβ species that is difficult to explain. One possible explanation is that a higher-order soluble Aβ oligomer species contributes to the cognitive deficits observed upon injection of unfractionated 7PA2 conditioned media and is a more potent inhibitor of cognitive function than Aβ dimer or trimer species or both. Another possible explanation is that combinations of different soluble Aβ oligomer species, perhaps interacting differently with different neuronal receptors, have an additive or synergistic toxicological effect. Trimers extracted from aged Tg2576 mouse brain also failed to elicit significant cognitive deficits. However, consistent with other
in vivo efficacy studies [
37], a dodecameric soluble Aβ oligomer extracted from aged Tg2576 mouse brain (Aβ*56) caused significant cognitive deficits. Synthetic soluble Aβ oligomers (ADDLs) also caused cognitive deficits following ICV injection. The results of this study show that ICV injection of soluble Aβ oligomers from different sources causes cognitive deficits in wild-type rats and that these deficits are reversible. The results of the study show that soluble Aβ oligomer containing conditioned media of 7PA2 cells is more potent than solutions of Aβ*56 obtained from aged Tg2576 mouse brains or solutions of synthetically prepared soluble Aβ oligomers (ADDLs). However, because it is not possible to quantitatively characterize the exact nature or distribution of soluble Aβ species in these different preparations, it is not possible to draw definitive conclusions from this study regarding the structure-activity relationships between individual soluble Aβ oligomer species.
In a more recent study, Moreth and colleagues [
19] prepared, characterized, and evaluated the hippocampal binding and effects on neurotransmission of spheroidal, protofibrillar, and fibrillar Aβ aggregates. Under the conditions used to test for hippocampal binding and neurotransmission, the different species where shown to be relatively unchanged. Monomeric and fibrillar Aβ did not bind to mature hippocampal neurons (DIV21) or effect neurotransmission at concentrations as high as 1 μM. In contrast, spheroidal and protofibrillar Aβ aggregates displayed punctate binding to mature hippocampal neurons and impaired neurotransmission with nanomolar potency. Significantly, the mode of impairment of neurotransmission was different for spheroidal Aβ aggregates, which impaired LTP at 30 nM, compared with protofibrillar aggregates, which impaired basal neurotransmission at 100 nM. Spheroidal Aβ aggregates had no effect on basal neurotransmission at concentrations as high as 100 nM. Although this study was unable to address the comparative neurotoxicity of discrete soluble Aβ oligomers, it did show that different forms of soluble Aβ oligomers can trigger distinct neuronal activities.
At this point, the exact structures of the toxicologically relevant soluble Aβ oligomer species have not been determined to the complete exclusion of other possible structures, and analytical tools do not exist to characterize the Aβ oligomers that form at concentrations in the AD brain [
44,
138]. The numerous studies reporting neuronal toxicity for different soluble Aβ oligomers support the conclusion that multiple soluble Aβ oligomer species exhibit neuronal toxicity, rather than a single, discrete toxic species. This suggestion of multiple toxic soluble Aβ oligomer species may also explain the plethora of reported neuronal receptors that mediate the effects of soluble Aβ oligomers (Table
1) [
139‐
165], a discussion of which is well beyond the scope of this review. (See the perspective of Benilova and De Strooper [
166] for a good introduction to this complex and controversial field of study.)
Table 1
Reported receptors that bind or mediate the toxicity of soluble amyloid-beta oligomers
| | |
P53-Bax cell death pathway [ 146] | cAMP/PKA/CREB-signaling pathway [ 147, 148] | Tau protein kinase 1/glycogen synthase kinase-3β [ 51] |
c-Jun N-terminal kinase [ 149] | | P/Q-type calcium currents [ 49] |
Cyclin-dependent kinase 5 [ 149] | | |
| | |
p38 mitogen-activated protein kinase [ 149] | | |
α 7-nicotinic receptors [ 105] | | Brain-derived neurotrophic factor [ 131] |
| | |
| | |
Despite recognition that soluble Aβ oligomers are key structures causing AD memory malfunction and cognitive deficits, drug discovery efforts targeting these species have been hampered by perceived technical difficulties of generating physiologically relevant preparations of synthetic soluble Aβ oligomers and by differing terminologies and methodologies used by various researchers [
167]. However, well-characterized and documented preparations of synthetic soluble Aβ oligomers have been reported by numerous researchers [
19,
40,
43,
95,
100,
107,
108,
111,
168‐
170] that generate soluble Aβ oligomers with little or no detectable fibrillar Aβ species. With the availability of a number of well-documented preparations of different soluble Aβ oligomer species and tools for comparative characterization, it is hoped that additional side-by-side testing of various soluble Aβ oligomer preparations in different toxicity paradigms such as reversal of basal neurotransmission, LTP inhibition, changes in AMPA receptor trafficking, tau phosphorylation, and loss of dendritic spines [
19,
108,
109,
170] will be conducted and reported.
Competing interests
All authors are employees or paid consultants of Acumen Pharmaceuticals, Inc. and own stock or stock options of this company.
Authors’ contributions
All authors read and approved the final manuscript.