Soluble Aβ oligomer production and toxicity

Abstract

For nearly 100 years following the first description of this neurological disorder by Dr. Alois Alzheimer, amyloid plaques and neurofibrillary tangles have been hypothesized to cause neuronal loss. With evidence that the extent of insoluble, deposited amyloid poorly correlated with cognitive impairment, research efforts focused on soluble forms of Aβ, also referred as Aβ oligomers. Following a decade of studies, soluble oligomeric forms of Aβ are now believed to induce the deleterious cascade(s) involved in the pathophysiology of Alzheimer’s disease. In this review, we will discuss our current understanding about endogenous oligomeric Aβ production, their relative toxicity in vivo and in vitro, and explore the potential future directions needed for the field.

INTRODUCTION

With the vast majority of baby boomers now entering the dreaded 65 years of age mark associated with increased prevalence for Alzheimer’s disease (AD), numerous occidental countries are facing a possible dantesque scenario for their respective health plans. More than 100 years following the initial publication of the description of its symptoms and neuropathology, and despite an accumulation of knowledge about the pathophysiology of AD, drugs capable of preventing, limiting and halting the disease are desperately needed. The major hurdle to the identification and development of new compounds targeting AD relies partly on the paradigm shift that occurred in the last 10–12 years: the emergence of amyloid-β (Aβ) oligomers as potential culprits of this brain disorder.

I. PRODUCTION OF OLIGOMERIC Aβ ASSEMBLIES

In AD, soluble Aβ monomers can form higher-order assemblies ranging from low-molecular weight oligomers (e.g. dimers and trimers) to mid range-molecular weight oligomers (e.g. Aβ*56) to protofibrils and fibrils (Figure 1). However, the exact relationship of these species to each other remains speculative, mostly because of their intrinsic nature as soluble globular intermediates. Despite this hurdle, multiple independent groups have observed apparent endogenous Aβ oligomers in protein expression studies using human postmortem brain tissue and using animal and cell lines genetically engineered to produce and secrete Aβ species.

Figure 1. Proposed model for endogenous oligomeric Aβ production.

Based on our current knowledge, Aβ monomers and trimers are produced and secreted by neurons. Of note, the N-terminus of Aβ is color-coded in blue, the central domain of Aβ in white and the C-terminus of Aβ is shown in red. The assembling of Aβ molecules within oligomers was based on our observations of epitope availability using a large panel of Aβ antibodies (82E1, 6E10, BAM10, 4G8, 40- and 42-end specific Aβ antibodies, A11 and OC). Once in the interstitial fluid, Aβ oligomers follow a trimer-based expansion in size under physiological conditions/normal aging. These species evade fibrillar formation. Upon the presence of an unidentified factor, monomeric Aβ becomes misfolded allowing the peptide to enter the fibrillar oligomer pathway, whose first step corresponds to the formation of Aβ dimers. Dimers can rapidly expand to create dimer-based protofibrils, ultimately contributing to the formation of fibrillar Aβ as amyloid plaques.

A. Detection of soluble Aβ oligomers in early studies

In the first half of the 1990s, two studies revealed that fibrillar amyloid load poorly correlated with cognitive impairment (Terry et al. 1991, Dickson et al. 1995). In 1999, two additional reports indicated that soluble non-fibrillar Aβ levels correlated strongly with the severity of the disease (McLean et al. 1999) and synaptic loss (Lue et al. 1999), suggesting the possibility that soluble species of Aβ might play a larger role in AD pathophysiology than deposited amyloid. Particularly, McLean and coworkers found the presence of SDS-stable Aβ oligomers with an apparent molecular weight of ~8 and ~12 kDa in both soluble and insoluble fractions of AD brain tissue. Of note, apparent levels of Aβ dimers were more abundant in the insoluble fraction than those in the soluble fraction. In parallel to these in vivo studies, evidence emerged that cell lines overexpressing mutant forms of APP linked to familial AD were able to produce Aβ assemblies intracellularly and secrete soluble Aβ dimers and trimers (Podlisny et al. 1995, Podlisny et al. 1998, Walsh et al. 2000). In the latter study, the Chinese hamster ovary cell line 7PA2, stably overexpressing APP-V717F, was compared to human primary cortical neurons. Following immunoprecipitation with either 40- or 42-end specific antibodies (2G3 and 21F12), only monomeric Aβ was detected in the conditioned medium of human cells compared to monomers, dimers and trimers identified in 7PA2 (Walsh et al. 2000). Similar findings were found for cellular lysates even though the results need to be interpreted with caution, as carboxyl-terminus fragments of APP were not immunodepleted out of the protein extracts used.

Not only did these observations suggest that neurons might produce a different profile of Aβ oligomers than the genetically engineered line 7PA2, but they also demonstrated that Aβ oligomerization might be a natural process in situ.

B. Current understanding of oligomeric Aβ production

Following these original studies, soluble Aβ molecules became the specific focus of subsequent studies in attempt to shed light on their possible contribution to AD pathophysiology. In order to try to clarify the fast accumulating data on this topic, we classified these reports based on the putative Aβ oligomers directly involved. These endogenous assemblies included dimers, trimers, Aβ*56, species immunoreactive to ADDLs/globulomers antibodies, and annular protofibrils.

1. Dimers

Aβ dimers are probably the most studied oligomeric species at the present time. Multiple reasons account for the latter: (1) Aβ dimers are SDS-stable (McLean et al. 1999), (2) Aβ dimers brain levels are elevated in AD mouse models and in subjects with AD (Kawarabayashi et al. 2004, Shankar et al. 2008, Shankar et al. 2009, Mc Donald et al. 2010), (3) size-exclusion chromatography segregated Aβ dimers from AD brain extracts impair long-term potentiation (Shankar et al. 2008). While the potency of Aβ dimers appears to be remarkable (Reed et al. 2009), the question relative to the origin of dimers remains nevertheless speculative.

In animal models of AD such as Tg2576 (Kawarabayashi et al. 2004) and J20 (Shankar et al. 2009, Meilandt et al. 2009), Aβ dimers can be found in soluble protein extracts at 10–12 and 10–14 months of age, respectively. Interestingly, these ages correspond to amyloid plaques depositing in cortical areas (Mucke et al. 2000, Kawarabayashi et al. 2001), suggesting that Aβ dimers and fibrillar Aβ are related to each other. Figure 2 illustrates that the relative abundance of Aβ dimers and monomers detected in the formic-acid soluble fraction (containing fibrillar proteins) sharply increased starting at 12 months of age in Tg2576. Specifically, FA-soluble Aβ monomers followed a sigmoidal profile similar to the one previously described (Kawarabayashi et al. 2001). In contrast, dimeric Aβ levels showed an exponential profile in Tg2576 (Figure 2B). Since Aβ monomers reached a plateau prior than Aβ dimers in this model (Figure 2B, C), it is tempting to hypothesize that, dimeric Aβ levels appear to follow plaque accumulation in vivo in this model (Kawarabayashi et al. 2001). In plaque-free brains of young transgenic mice, Aβ dimers have yet to be found. Earlier work (Kawarabayashi et al. 2004) reported the presence of Aβ dimers in lipid raft preparations from 6-month-old Tg2576, but later studies revealed that specific lipid enrichment could trigger Aβ oligomerization artificially (Yu et al. 2005, Kim et al. 2006), questioning the intrinsic existence of these species in very young Tg2576 mice.

Figure 2. Temporal accumulation of endogenous Aβ dimers in brain tissue of Tg2576 mice.

(A) Western blots showing the relative levels of Aβ dimers detected following immunoprecipitation using formic acid extracts. Dimeric Aβ levels increase in parallel to plaque accumulation in Tg2576, suggesting that dimers are closely related to deposited amyloid. (B) Densitometry analysis of formic acid-soluble Aβ monomers and dimers in Tg2576. Empty circles represent monomeric Aβ while filled black circles correspond to Aβ dimers (n = 3/group). (C) Relative ratio between Aβ monomers and Aβ dimers in Tg2576.*, non-specific band due to the immunoprecipitation.

Similarly in human studies, apparent endogenous Aβ dimers were only found in brains of subjects with AD (Shankar et al. 2008, Mc Donald et al. 2010) when comparing individuals with extensive amyloid and tau pathology and subjects devoid of amyloid accumulation. In the former study (Shankar et al. 2008), proteins were extracted from subjects without AD (mean age = 65.6 ± 12.1 years; n = 3) or with severe AD (mean age = 78.1 ± 5.8 years; n = 7). Aβ dimers were identified only in the AD group, even though these groups were not normalized for age or amyloid burden. Thus, Aβ dimers appear to be closely linked to fibrillar amyloid in humans as well.

While these in vivo studies suggest that dimeric Aβ and plaques may play a deleterious and mysterious tango in the brain, the origin of dimers still remains elusive. Previous research revealed that primary cortical neurons from Tg2576 produce monomers and trimers; Aβ dimers were not detected under the experimental conditions used (Lesne et al. 2006) indicating that neurons preferentially produce and secrete monomers and trimers but not dimers. More recently, our own group observed that Aβ dimers are in fact produced and secreted by primary neurons but at very low amounts (1:20 and 1:15 ratio to monomers and trimers respectively; Figure 3A). These results are in sharp contrast to the relative abundance of dimers in the conditioned medium of 7PA2 cells (Podlisny et al. 1998, Walsh et al. 2002). The reason why dimers are produced and secreted so prominently in this cell line remains unknown, but the profile observed by biochemical analyses performed by our own group clearly differs from those from primary neurons (Figure 3B). Lastly, dimers are found as what appears to be the primary oligomeric species formed in mice overexpressing the E693Δ APP mutant (Tomiyama et al. 2010). Identified in 2008, this deletion within the mid-region of Aβ engages Aβ peptides to form preferentially oligomeric species instead of forming fibrils (Tomiyama et al. 2008, Tomiyama et al. 2010). Of interest, these Aβ dimers are accumulating intracellularly and co-segregated within the insoluble protein fraction extracted with formic acid but not within soluble protein fractions (Tomiyama et al. 2010). These independent findings further support the hypothesis that endogenous Aβ dimers are the molecular brick for fibrillar Aβ in vivo. Identifying what might favor dimeric Aβ formation in absence of mutation may shed light on the origin of these assemblies.

Figure 3. Comparison between endogenous Aβ species secreted by mature Tg2576 primary cortical neurons and 7PA2 cells.

(A, B) Western blots showing the relative levels of Aβ oligomers detected with 6E10 following immunoprecipitation using 1 mL of conditioned medium. While dimeric Aβ levels is preferentially secreted by 7PA2 cells (B), Aβ trimers are produced and secreted by primary neurons derived from Tg2576 mice (A). The observed profile in neurons (A) suggests a trimer-based oligomerization process under physiological conditions.

2. Trimers

Several lines of evidence indicate that Aβ trimers might constitute the molecular brick for non-fibrillar assemblies of Aβ. First, as mentioned in the above section, trimers are the preferential, most abundant species produced and secreted by primary neurons in vitro (Lesne et al. 2006), even though the origin of trimeric Aβ remains elusive. It possible that this preference might be due to lower energetic requirement to form trimers or due to the absence of a chaperone protein X allowing the conversion to dimers. Second, trimers are present in brain tissue of Tg2576 as early as embryonic day 14 and its expression persists throughout life (Lesne et al. 2006). Analyzing extracellular- and intracellular-enriched protein fractions in these mice indicated that brain levels of Aβ trimers steadily but very slowly rise with aging (Lesne et al. 2006). These findings were mirrored in human studies where trimers are found as early as 1 year of age, contrasting with brain Aβ dimers which were not detected in subjects until 50–60 years of age in TBS extracts (Lesne et al. unpublished data).

Importantly, when comparing Tg5469 and Tg2576 lines, which express equal amounts of human APP but vastly different levels of Aβ due to the presence of the Swedish mutation in the latter line, Aβ trimers could not be readily observed in Tg5469 brains (Ma et al. 2007). This suggests that the formation of trimers in vivo appears to be dependent on the levels of Aβ production. Cumulatively, these findings suggest that trimers are the earliest endogenous oligomeric Aβ species produced and secreted by neurons.

3. Aβ*56

Based on the finding that neurons preferentially secrete Aβ trimers, it was likely that these assemblies could form larger oligomeric species based on multiples of 3, i.e. hexamers (6-), nonamers (9-) and dodecamers (12-), in absence of deposited Aβ. Examining young Tg2576 mice prior to plaque formation, apparent trimers (13 kDa), hexamers (27 kDa), nonamers (40 kDa) and dodecamers (56 kDa) were detected using antibodies targeting the N-terminus and central domain of Aβ, such as 6E10, and using the A11 antibody detecting prefibrillar oligomeric proteins (Lesne et al. 2006, Kayed et al. 2003). The discovery of Aβ*56, a putative dodecameric Aβ assembly, derived from the observation that cognitive decline in Tg2576 starts at 6 months of age, when Aβ*56 first appears, and remains relatively stable for months (up to 15-months of age) when numerous endogenous Aβ species (soluble monomers, insoluble monomers, dimers, trimers) accumulate in brain tissues many fold (Lesne et al. 2006). Aβ*56 can be detected with multiple antibodies, which contribute to establishing the antibody reactivity profile of this Aβ species: A11+, OC−, 6E10+, 4G8+, Aβ_x-40+, Aβ_x-42+, 82E1−. This epitope specificity allows to categorize Aβ*56 as a non-fibrillar Aβ oligomer (Figure 4).

Figure 4. Epitope mapping for Aβ*56 in Tg2576 soluble extracts.

Immunoprecipitation/Western blots showing the detection profile of Aβ*56 present in Tg2576 mouse brains. A comparison between 6E10, A11 and OC antibodies (kind gifts from Rakez Kayed and Charles Glabe) is presented. The 6E10+/A11+/OC− profile characterizes Aβ*56 as a non-fibrillar Aβ oligomer. * indicate non-specific bands generated by the amplification process.

Importantly, the presence of Aβ*56 was also reported in other APP transgenic mouse models of AD, e.g. J20 (Cheng et al. 2007, Meilandt et al. 2009), Arc6/48 (Cheng et al. 2007), 3xTgAD (Oddo et al. 2006) and APP23 (unpublished data), clearly arguing against the possibility that it is an artifact intrinsic of one isolated mouse model.

Contrary to dimers and trimers, which can be detected in vitro, Aβ*56 is not detected in cellular protein extracts of cultured neurons or in conditioned media indicating that it is not produced and secreted by primary neurons (Lesne et al. 2006). This finding suggests that this assembly requires a co-factor X, present in brain tissue, to promote its formation. Furthermore, aging must regulate the expression of this unidentified co-factor X in order to explain the sudden accumulation of Aβ*56 in 6-month-old Tg2576 brains.

In mouse models, Aβ*56 is present in two soluble protein fractions, extracellular-enriched (Lesne et al. 2006, Lesne et al. 2008) and membrane-enriched extracts (Cheng et al. 2007), indicating that this assembly is formed extracellularly and may interact with surface receptors at neuronal membranes. Similar segregation was observed in studies with a large human cohort (n = 135; (Lesne et al. unpublished data)). When comparing across clinical diagnosis groups, Aβ*56 brain levels seem to peak during the preclinical stage of AD and decrease in MCI and AD subjects (Lesne et al. unpublished data). The data are in agreement with evidence gathered in Tg2576 suggesting that Aβ*56 may form in the aging brain prior to the formation of amyloid plaques.

Of interest, the relative levels of soluble Aβ*56 in the extracellular protein fraction of human brain tissue showed no relationship with Aβ dimers (Figure 5A). In contrast, brain levels of human Aβ*56 were positively correlated to trimeric soluble Aβ levels in the same fraction (Figure 5B). This observation further supports our hypothesis that trimeric Aβ might constitute the building block for larger oligomers in the non-fibrillar pathway.

Figure 5. Relationship between endogenous low molecular weight Aβ oligomers and Aβ*56 in human brain tissue.

(A, B) Linear regression analyses between the relative levels of soluble Aβ dimers (A) or trimers (B) and Aβ*56 detected in extracellular-enriched protein fractions of human brain tissue from the Religious Order Study (n = 88). Dimeric Aβ levels showed no association with Aβ*56 whereas the abundance of Aβ trimers was positively correlated with Aβ*56 brain levels. (Rho is indicated next to the P-value for Spearman Rank Correlation. Correlations were performed using the whole cohort, n = 88.)

4. ADDLs and Globulomers

Even though this review aims at discussing accumulated findings about endogenous oligomeric Aβ species, tools initially developed to target synthetic soluble Aβ molecules (e.g. Aβ derived diffusible ligands or ADDLs and globulomers) appear to detect endogenous Aβ entities, even though the exact nature of these molecules needs further studying (Gong et al. 2003, Georganopoulou et al. 2005, Barghorn et al. 2005, Hillen et al. 2010). The antibodies termed NU-x and anti-globulomer antibodies (5598, 8F5, A-887755), developed by Klein’s group at Northwestern and Abbott respectively, seem able to identify 35–60kDa oligomeric Aβ species recognized by these groups as dodecamers (Barghorn et al. 2005, Lambert et al. 2007, Hillen et al. 2010).

In Tg2576, ADDLs are increased ~5- to ~100-fold (Chang et al. 2003) at ages during which spatial reference memory remains unchanged. In addition, measurements of ADDLs in brain tissue and in CSF revealed 70-fold increases in AD compared to age-matched controls (Gong et al. 2003, Georganopoulou et al. 2005). These findings suggest that ADDLs levels rise sharply in AD, reminiscent of the marked elevation in Aβ dimers (Shankar et al. 2008).

Similarly to ADDLs, the use of mid-range molecular weight globular synthetic Aβ oligomers migrating at 38–48 kDa, termed globulomers, generated antibodies targeting alleged endogenous Aβ species. Due the apparent size of the synthetic assemblies detected, Barghorn and coworkers indicated that globulomers corresponded to Aβ dodecamers. In animal studies performed with Tg2576 mice, globulomers were mostly unchanged between 2.5 and 10 months of age prior to plaque formation (26%), but brain levels rose 496% at 12 months of age when amyloid plaques start forming (Kawarabayashi et al. 2001). Furthermore, immunohistochemical analyses using the anti-globulomer 8F5 antibody labeled amyloid plaques in brains of Tg2576 animals and subjects with AD (Barghorn et al. 2005). These findings indicate that, regardless of the exact endogenous Aβ species detected by anti-globulomer antibodies, these molecules are highly associated with amyloid plaques and are elevated in AD.

In both cases of ADDLs and globulomers, there has been some confusion that these “dodecamers” correspond to the endogenous 56 kDa Aβ assembly, Aβ*56 (Wilcox et al. 2011). However, several lines of evidence indicate that Aβ*56 and ADDLs/globulomers are different entities. First, Aβ*56 brain levels are remarkably stable between 6 and 15 months of age in extracellular-enriched fractions of Tg2576 (Lesne et al. 2006) while ADDLs and globulomers are vastly increased during that same time interval. Second, Aβ*56 levels in both extracellular and membrane-enriched protein fractions are lowered in AD compared to non-impaired age-matched controls (Lesne et al. unpublished data) while ADDLs/globulomers increase in AD (Gong et al. 2003, Georganopoulou et al. 2005, Barghorn et al. 2005, Nimmrich et al. 2008). Based on these observations, ADDLs antibodies likely detect in vivo species others than trimers and Aβ*56, possibly fibrillar oligomers or annular protofibrils (Glabe 2008). This subclass of oligomers is characterized by being immunopositive to the antibody OC (OC+) and immunonegative to A11 (A11−). Even though it is not known whether ADDLs and globulomers are OC+/A11−, the fact that ADDLs and globulomer antibodies detect amyloid plaques suggests that ADDLs correspond to fibrillar oligomers.

In addition to the point that ADDLs/globulomers and Aβ*56 represent different entities, the origin(s) of the endogenous Aβ species detected by ADDLs/globulomer antibodies is completely unknown. Further studies are needed to determine how these possible Aβ assemblies are produced and when they do form in the continuum aging-AD.

5. Annular protofibrils (~100–150kDa)

Annular protofibrils (APFs) are pore-like structures that are believed to derive from the circularization of non-fibrillar Aβ assemblies (Kayed et al. 2009). With a molecular size over 90 kDa, they resemble bacterial pore-forming toxins. APFs have been proposed to permeabilize lipidic membranes and induce cell death by altering neuronal homeostasis (Glabe & Kayed 2006, Lashuel & Lansbury 2006). Recent data revealed that APFs might constitute a distinct Aβ assembly from non-fibrillar oligomers (Kayed et al. 2009). Based on this work, APFs originate from pre-existing non-fibrillar oligomers, which upon some conformational change allows the recruitment of additional oligomers to form APFs. In 17–20 month-old APP23 mice, ~2% of examined synapses showed an accumulation of APFs using a specific αAPF antibody (Kokubo et al. 2009). On a follow-up study, these same groups recently demonstrated that APFs are associated with diffuse plaques and intracellular punctate deposits in AD brain sections (Lasagna-Reeves et al. 2011). Despite very small specimen numbers (n = 2/group; N = 4), these findings indicate that APF exist in vivo.

Even though these new studies and conformation-dependent antibodies are allowing us to discriminate among the multiple oligomeric forms of Aβ in vitro and in vivo, we know nothing about the age at which APFs first form in animal models and in human brain tissue and CSF.

C. Future directions

With the emergence of endogenous oligomeric forms of Aβ, perhaps more than anticipated, our understanding of the conditions necessary to prepare synthetic Aβ oligomers and novel detecting tools (e.g. conformation-dependent antibodies such as A11 and OC) still needs improvement. Great confusion has risen in our field when comparing specific oligomers, particularly regarding in vitro and in vivo Aβ assemblies. If the working hypothesis is that Aβ oligomers are the real culprits in AD, it is therefore imperative that we collectively determine the origins of endogenous soluble Aβ species, when their respective levels are modified in transgenic models and human brain tissue across ages and finally in longitudinal studies using CSF. A longitudinal model extrapolated from our work in human cohorts (Lesne et al. unpublished data) is shown in Figure 6. In the model proposed, lower energy requirements for Aβ to form high molecular weight oligomers compared to lower molecular weight species would support the concept that Aβ*56 peaks earlier than Aβ trimers and dimers. If correct, this phenomenon would lead these high-n molecules to reach a plateau first, then allowing the accumulation of the low-n Aβ oligomers, otherwise used for oligomerization under non-saturating conditions. Such scenario would therefore lead to the accumulation of larger species first, followed by a second accumulation of lower assemblies. Further studies will be needed to determine whether this proposed model is correct.

Figure 6. Proposed model for the expression profile of endogenous Aβ oligomers in human brain tissue.

We analyzed human brain tissue from 158 subjects with ages of 1 to 98 years at the time of death. Two cohorts were used for this study (Lesne et al., under review): 49 brain samples were from healthy subjects (from 1 to 62 years) and 89 from subjects with preclinical AD, MCI or AD. Briefly our findings indicate that in healthy subjects Aβ*56 was first detected in the 5^th decade of life preceding elevations in Aβ trimers and dimers observed in the 6^th and 7^th decades. In preclinical AD, Aβ*56 appeared to reach a plateau. In MCI brains, Aβ*56 levels decreased significantly while Aβ trimers levels rose sharply. At this stage, brain levels of Aβ dimers slowly elevated. One possible explanation for this change in the relative expression of these species might be due to a destabilization of the non-fibrillar oligomerization pathway in favor of the fibrillar oligomerization pathway. Finally in AD brain tissue, Aβ dimers were increased while levels of Aβ*56 and Aβ trimers fell to discrete amounts. Such proposed scenario must need to be tested in longitudinal studies.

Finally, the generation of novel, specific detection and diagnostic means is therefore a crucial avenue. With the aging of the population becoming a central concern for our respective health care systems, together, the field must find common grounds to foster collaborative projects and fasten our relatively speculative understanding of how endogenous Aβ oligomers are produced, when they appear and what/when they are modulated across preclinical subjects, patients with mild cognitive impairment or with AD.

II. TOXICITY OF OLIGOMERIC Aβ ASSEMBLIES

As described above, multiple forms of Aβ oligomers appear to exist and co-exist in brain tissue. While the field is still searching for avenues to identify the nature of the relationships between these various assemblies, we paradoxically know more about the deleterious effects of oligomeric Aβ than how they are produced in vivo.

A. Current knowledge about soluble Aβ oligomers-induced toxicity

1. Low molecular weight Aβ oligomers: dimers and trimers

Following the identification of SDS-stable Aβ oligomers in 7PA2 cells, it was important to determine whether these Aβ species were biologically active in order to set the stage for Aβ oligomers as potential toxins in AD. Conditioned medium of 7PA2 and CHO control cells were injected in rats and hippocampal long-term potentiation (LTP) was assessed. In vivo LTP was blocked upon injection of the medium containing Aβ oligomers (Walsh et al. 2002). After removal of Aβ species, the same medium had no effects on LTP indicating that soluble Aβ oligomers formed in 7PA2 cells inhibit neuronal function. To demonstrate that oligomers and not monomeric Aβ triggered this effect, 7PA2-derived species were injected into size-exclusion chromatography (SEC) columns. Fractions enriched for dimers/trimers inhibited LTP and cognition but not monomers (Walsh et al. 2005, Cleary et al. 2005).

To test whether 7PA2-derived Aβ dimers and trimers were directly related to synaptic loss, 7PA2 CM was applied to hippocampal organotypic cultures. Dendritic spine loss was observed after 3–5 days of exposure with these low molecular weight oligomers using a mechanism dependent on N-Methyl-D-Aspartate (NMDA) receptors (NMDAR) and calcineurin (Shankar et al. 2007). Despite these convincing findings, the nature of the soluble pathogenic Aβ species in human brain tissue or CSF was still speculative. To this end, soluble human brain tissue extracts from individuals with AD were obtained and segregated by SEC. Application of the total extract from AD brains blocked LTP (Shankar et al. 2008), reproducing the findings obtained with 7PA2 CM (Walsh et al. 2002, Walsh et al. 2005). In addition, the dimer-containing fraction inhibited LTP (Shankar et al. 2008) and facilitated LTD (Li et al. 2009) indicating that Aβ dimers alone can alter neuronal plasticity.

While attempting to elucidate the mechanism by which dimeric Aβ inhibits LTP, Li and colleagues demonstrated that NMDAR subunit 2B (NR2B or GluN2B)-containing receptors were indirectly necessary for LTP inhibition triggered by dimers. Briefly, dimers alter glutamate reuptake at synapses leading the accumulation of extracellular glutamate and excessive stimulation of extrasynaptic receptors (Li et al. 2009). As a result of activation of these receptors in vitro, the calpain/p38 axis is activated and the survival promoting ERK1/2 pathway is inhibited within 6 hours of exposure with 7PA2-derived Aβ dimers/trimers (Li et al. 2011).

2. Dimers

SEC-segregated Aβ dimers from AD cerebral cortex also induced tau-dependent cytoskeletal microtubule abnormalities in primary neurons (Jin et al. 2011). Treatment with dimers triggered tau hyperphosphorylation at disease-relevant epitopes (Jin et al. 2011). Altogether, these results demonstrate that endogenous Aβ dimers are synaptotoxic. Kawarabayashi and colleagues reported that Aβ dimers could be detected in lipid raft isolations of 6-month-old Tg2576 mice. These authors suggested that the appearance of dimers coincided with the onset of memory decline observed in this APP transgenic line (Kawarabayashi et al. 2004). However, while brain levels of Aβ dimers increased extraordinarily (~500 fold), memory function did not further worsen arguing against the implied relationship between dimers and the onset of cognitive decline in this mouse model. Ultimately, the mid-size molecular weight Aβ oligomer Aβ*56 was related to initiating this memory impairment (Lesne et al. 2006). Even though Aβ dimers may not be responsible for the initial decline observed in 6-month-old Tg2576 animals, it is tempting to hypothesize that the second decline occurring between 13 and 15 months of age is due to the accumulation of Aβ dimers, considering the association between amyloid plaques and dimers. In support of this postulate, a recent study proposed that Aβ dimers rapidly form stable protofibrils to mediate synaptotoxicity (O’Nuallain et al. 2010).

3. Trimers

Contrary to Aβ dimers, very few studies have been conducted with naturally formed Aβ trimers (Townsend et al. 2006, Reed et al. 2009). Applying SEC-separated Aβ trimers from 7PA2 cells onto hippocampal slices led to greater inhibition of LTP than any low-n oligomers derived from 7PA2 CM, including dimers (Townsend et al. 2006).

Reed and coworkers (2009) compared the effects of various Aβ assemblies on cognitive function in the alternating lever cyclic ratio (ALCR) behavioral paradigm (Cleary et al. 2005). Among the species tested were SEC-fractionated dimers and trimers from 7PA2, along with trimers and Aβ*56 from Tg2576 brain tissue. Both trimers altered cognitive function regardless of the source (cell line or mouse brain), even though their relative potency was quite weak compared to that of dimers or Aβ*56. We do not know yet whether the molecular mechanism(s) mediating these effects is the same identified for dimers but these results indicate that trimers are capable of impairing neuronal function. Additional indirect evidence supports a deleterious role of Aβ trimers in vivo. While young Tg2576 do not appear to be impaired in the Morris water maze compared to non-transgenic animals at ages where Aβ trimers are the only species detected (Lesne et al. 2006), this observation does not take into consideration the potential beneficial effect of soluble APP-α (sAPPα) resulting from human APP overexpression. When compared to Tg5469 mice, which overexpress APP at similar levels (Ma et al. 2007, Westerman et al. 2002), Tg2576 mice performance in the Morris water maze was much lower than Tg5469 transgenic animals, suggesting that trimers are already impairing cognition in Tg2576 prior to the formation of Aβ*56 occurring at 6 months of age. Also, using APP transgenic rats, intraneuronal trimers were found to be the only oligomeric Aβ species present at 3 months of age, when a cognitive deficit was detected (Leon et al. 2010).

4. Aβ*56

Since Aβ*56 is only formed at low levels in the aging brain, studying its effects following purification represents a real technical challenge. Nevertheless, following affinity purification combined with size-exclusion chromatography, purified Aβ*56 caused transient memory impairment in young healthy rats demonstrating that this Aβ assembly impairs brain function in vivo (Lesne et al. 2006). First identified in Tg2576 mice, the brain expression of Aβ*56 was also found associated with memory decline in other APP transgenic mice modeling AD, namely J20, Arc48 and 3xTgAD (Oddo et al. 2006, Cheng et al. 2007). In human studies (n = 88), we found that Aβ*56 was positively correlated with soluble hyperphosphorylated tau at disease-relevant epitopes in AD brains (Figure 7). Additional analyses linking Aβ*56 and tau in the aging human brain are currently awaiting peer-review (Lesne et al. unpublished data) and the exact molecular mechanism by which Aβ*56 induces tau changes is under investigation.

Figure 7. Correlation between extracellular levels of the oligomeric Aβ assembly, Aβ*56, and disease-relevant tau changes in AD brains.

Using a very well cohort of the Religious Order Study, we measured Aβ*56 levels in the extracellular-enriched protein fraction as previously reported (Lesne et al., 2006) as well as soluble total and disease-relevant tau species (using Tau5 and Alz50 antibodies). All protein levels were normalized to the neuron-specific nuclear protein, NeuN, to compensate for neuronal cell death across groups. (A) Representative quantitative western blot for tau5/Alz50 in 30 subjects of our cohort (10 preclinical, 10 MCI and 10 AD) using a Li-Cor Odyssey system. (B) Quantification of soluble Alz50-tau levels revealed disease-relevant tau changes in MCI and AD brains. (C) Linear regression analysis between soluble Alz50-tau and Aβ*56 levels indicated a strong positive correlation in AD brains. Abbreviations: PC, preclinical AD; MCI, mild cognitive impairment; AD, Alzheimer’s disease.

Aβ*56 is not toxic per se as there is no neuronal cell loss in Tg2576 at the age of Aβ*56 formation. We hypothesize that Aβ*56 is able to perturb neuronal physiology without triggering cell death. This postulate could explain why this Aβ assembly impairs cognition in middle-aged Tg2576 and in young rats as well as the reversibility of its effect following acute injection (Lesne et al. 2006).

5. ADDLs/globulomer immunoreactive species

The toxicity of synthetic Aβ oligomers referred as ADDLs have been reviewed recently (Wilcox et al. 2011) and we will therefore not extend this discussion much further than the points addressed there. Specifically, we want to include recent observations that application of ADDLs on primary neurons led to the abnormal localization of tau to dendrites (Zempel et al. 2010). At the same time, Ittner and coworkers elegantly demonstrated that tau could be missorted to dendrites under pathological conditions (Ittner et al. 2010, Ittner & Gotz 2011). This effect is mediated by the tyrosine Src kinase Fyn, which has been associated with soluble Aβ toxicity since the late 1990s (Lambert et al. 1998). Importantly, when APP23 mice were crossed with mice expressing a tau construct preventing Fyn-mediated tau relocalization to dendrites, neuronal/cognitive deficits induced by Aβ were blocked (Ittner et al. 2010). The activation of Fyn induced by ADDLs appears to be physiologically relevant as several in vivo studies demonstrated a crucial involvement of Fyn in Aβ-induced neuronal/cognitive dysfunction (Chin et al. 2004, Chin et al. 2005, Roberson et al. 2011). Moreover, Fyn can hyperphosphorylate tau at tyrosine 18 (Y18) which accumulates in brains of AD subjects and of tau-overexpressing mice (Lee et al. 1998, Lee et al. 2004, Bhaskar et al. 2005, Bhaskar et al. 2010). Based on the latter, genetic ablation of tau prevented cognitive impairment triggered by Aβ/Fyn (Roberson et al. 2011). Thus the triad Aβ/Fyn/tau seems to play an important role in Aβ-induced deficits related to AD (Haass & Mandelkow 2010, Ittner & Gotz 2011), even though the exact Aβ species and the molecular mechanism linking Aβ oligomers to Fyn is unknown. The status of pY18-tau also needs to be examined in the transgenic mouse lines used by Roberson and coworkers in order to validate whether Aβ modulates tau via Fyn signaling.

In contrast to Aβ molecules detected by ADDL antibodies, Aβ species immunoreactive for globulomer antibodies impair mitochondrial function (Eckert et al. 2008) and inhibit presynaptic P/Q-type calcium channels (Nimmrich et al. 2008). Based on the latter work, these assemblies appear to attack the presynaptic end of the synapse as opposed to the postsynaptic terminal for Aβ dimers and ADDL-positive entities.

6. Annular protofibrils

The existence of annular protofibrils (APFs) in vivo has only been recently confirmed using a novel antibody, αAPF, specifically designed to target this Aβ assembly (Kayed et al. 2009, Kokubo et al. 2009, Lasagna-Reeves et al. 2011). It is however unclear whether purified endogenous APFs display the same neurotoxic properties as their synthetic counterparts (Kayed et al. 2009). This has not been the case for dimers for instance, with Aβ S26C dimers and endogenous Aβ dimers displaying remarkable potency differences (Jin et al. 2011).

B. Proposed surface mediators of endogenous Aβ oligomers

Whereas endogenous oligomeric Aβ species have been linked to cognitive and neuronal dysfunction, findings related to the molecular mechanisms by which these Aβ assemblies display their effects are only starting to emerge (Figure 8).

Figure 8. Receptors and signaling pathways potentially modulated by endogenous Aβ oligomers.

Among the various receptors proposed to act as oligomeric Aβ receptors, only a handful of postsynaptic receptors appear to be relevant to endogenously produced soluble Aβ entities: PrP^c, EphB2, RAGE, α7nAChR, Glutamate transporters, and possibly NMDAR. Since no endogenous Aβ oligomer has been directly shown to bind to NMDARs, and since most signaling pathways altered by naturally produced Aβ oligomers disturb physiological NMDAR transducers or NMDAR expression, Aβ oligomers might act as indirect modulators of NMDAR activity.

Based on the observations that Aβ impairs synaptic plasticity and that NMDAR are crucial regulators of LTP and LTD, soluble Aβ exposure was shown to reduce surface NMDAR expression (Snyder et al. 2005). This internalization of NMDAR resulted from the activation of α7-nicotinic acetylcholine receptors (α7-nAChR) and downstream effectors calcineurin (formely known as protein phosphatase 2B) and striatal-enriched phosphatase STEP. The observed cascade of events briefly reported here suggested that certain oligomeric Aβ species present in the conditioned medium of APP-overexpressing N2A cells are capable of binding to the α7-nAChR, ultimately leading to synaptic plasticity impairments. Unfortunately, the nature of the Aβ molecules present in the material used was not disclosed (Snyder et al. 2005) making it difficult to place these results in a larger context when comparing the effects of specific oligomers.

Although some studies have shown Aβ colocalized with NMDA receptors (NMDAR) at synapses (Lacor et al. 2007, Dewachter et al. 2009), it remains unclear whether this is a direct target of the peptide. Several molecular targets for synthetic oligomeric Aβ at the neuronal surface have been proposed including the receptor for advanced glycation end products (RAGE) (Sturchler et al. 2008), metabotropic glutamate receptors mGluR5 (Renner et al. 2010), PrP^c (Lauren et al. 2009), EphB2 (Cisse et al. 2011a). Among them, two receptors particularly stimulate our interest: PrP^c and EphB2.

Aβ was first shown to interact with RAGE in the mid 1990s (Yan et al. 1996) and was subsequently proposed to regulate Aβ accumulation in the brain (Deane et al. 2003). While RAGE was identified as potential binding receptor for oligomeric Aβ species prior to the other protein receptors, its discovery as mediator of oAβ was obtained using high concentrations (10 μM) of synthetic Aβ_1–40 (Sturchler et al. 2008). Using 10 μM mixtures of oAβ with unknown stoichiometry or relative abundance, RAGE was found to bind oAβ when compared to A11-immunonegative Aβ fibrils or aggregates (Sturchler et al. 2008). It is unfortunate that monomeric Aβ preparations were not tested in parallel to synthetic oAβ. In addition, deletion of RAGE rescued LTP inhibition triggered by 200 nM of synthetic oAβ (Origlia et al. 2008). The protein kinase p38 was proposed to serve as a downstream effector of the RAGE pathway activated by oAβ in this paradigm (Origlia et al. 2008). It is however still unknown whether endogenous soluble Aβ oligomers do co-immunoprecipitate with RAGE in brains of AD patients or of transgenic mouse models of AD. Similarly, we do not know whether more physiological concentrations of oligomeric Aβ, i.e. low nM based on previous reports (Walsh et al. 2002, Shankar et al. 2008) would activate signaling pathways downstream of RAGE. In animal models, while overexpression of RAGE or a dominant negative form or RAGE in mice overexpressing the Swedish/Indiana mutant of APP enhanced or decreased behavioral deficits respectively (Arancio et al. 2004), gene ablation of RAGE in transgenic mice overexpressing the Artic mutant of APP did not rescue cognitive impairment (Vodopivec et al. 2009). Altogether, more evidence is needed to rigorously claim that endogenous soluble Aβ oligomers bind and activate RAGE.

Over the past few years, the prion protein has probably been the most controversial receptor for Aβ oligomers (Balducci et al. 2010, Kessels et al. 2010, Calella et al. 2010, Benilova & De Strooper 2010, Barry et al. 2011, Freir et al. 2011, Cisse et al. 2011b). The cellular form of the prion protein PrP^c has been identified in a high-throughput screen using non-neuronal cell lines and synthetic preparations of Aβ (Lauren et al. 2009). Based on this screen, very large synthetic Aβ oligomers (~500 kDa with an estimated number of Aβ molecules ranging between 50 < n < 100) were found to bind to the 95–105 region of PrP^c. When applied to primary neurons, these Aβ assemblies interacted with plasma membranes in a fashion reminiscent from the binding seen for ADDLs (Lacor et al. 2004). Genetic deletion of PrP^c partly reduced this apparent binding to cells indicating that PrP^c might act as one of multiple receptors for synthetic Aβ oligomers (Lauren et al. 2009). Controversy has existed with regards to the alleged requirement of PrP^c expression for synthetic Aβ to inhibit LTP (Lauren et al. 2009, Kessels et al. 2010, Calella et al. 2010, Barry et al. 2011, Freir et al. 2011). One possible explanation for the lack of reproducibility of these findings between in vitro and in vivo models may relate to the heterogeneity of synthetic oligomeric Aβ mixtures used and to a different relative abundance of the Aβ assemblies across systems. Another might relate to the fact that the effect of PrP was only tested in young APP transgenic mouse models (Calella et al., 2010; Cisse et al., 2011). Considering that endogenous brain Aβ oligomers are produced at different ages, it is therefore conceivable that PrP^c could mediate oAβ-induced neuronal dysfunction in older animals. To resolve part of this issue, Dr. Walsh’s and Dr. Rowan’s groups took advantage of an epitope masking PrP^c antibody (D13, targeting the 96–104 region of PrP) to prevent Aβ from interacting with PrP^c (Barry et al. 2011). When soluble Aβ-containing extracts of AD brain were injected intracerebroventricularly, LTP was inhibited as previously reported (Shankar et al. 2008). Pre-injection of D13 completely abolished, oligomeric Aβ-induced inhibition of LTP, indicating that endogenous Aβ oligomers can interact with PrP^c to alter LTP. Finally, these experiments were repeated in PrP-null mice in which LTP deficits induced by Aβ-containing extracts of AD brain were abolished indicating that PrP mediates the neurotoxic effects of certain Aβ species (Freir et al. 2011). More studies are however needed to directly show which endogenous Aβ oligomer mediates this effect, where they colocalize and what is the cellular pathway linking Aβ, PrP^c and neuronal dysfunction. The Ephrin B2 receptor EphB2 is enriched at synapses where it clusters with NMDA receptors (Dalva et al. 2000) and it has also been proposed to act as a receptor for endogenous Aβ oligomers (Cisse et al. 2011a). EphB2 expression was found decreased in human AD hippocampal extracts (Simon et al. 2009) and ADDL application to mature hippocampal neurons lowered EphB2 expression by 60% (Lacor et al. 2007). When using biotinylated synthetic Aβ oligomers, trimers co-immunoprecipitated with EphB2 in immature cortical and hippocampal neurons (DIV7); when 7PA2 CM was applied to neurons, dimers and trimers were pulled down with anti-EphB2 antibodies (Cisse et al. 2011a). This apparent discrepancy further warrants caution when designing experiments with synthetic intermediate assemblies. Even though these data indicate that EphB2 may bind oligomers in in vitro paradigms and that EphB2 is an important player in Aβ-induced neuronal dysfunction, it could have been more medically relevant to examine whether similar co-immunoprecipitations can be done using human and APP transgenic brain tissues, and specifically identify the endogenous Aβ species binding to EphB2 in situ.

C. Toxicity and relative concentration of oligomeric Aβ molecules

While intensive efforts are invested in trying to identify the potential mediators of Aβ oligomers at the neuronal plasma membrane, another “ghost from the past” parameter impedes the progress of our understanding of oAβ-induced toxicity: the relative concentration of exogenously applied Aβ oligomers. Already in the 1990s, the question of how much Aβ to use in experimental settings was questioned. Nowadays, this conundrum still holds true for Aβ oligomers, as recently illustrated (Puzzo et al. 2008). In a seminal article, low nanomolar (nM) concentrations of 7PA2-derived Aβ oligomers were only needed to disrupt LTP (Walsh et al. 2002). Of note, the relative concentration of Aβ oligomers was determined by comparison with monomeric Aβ integrating monomeric Aβ enzyme-linked immunosorbent assay (ELISA) with SDS-PAGE/densitometry analyses. Using relative concentrations within the low nM range (0.5–10 nM), exogenously applied Aβ oligomers were found capable of inhibiting cognition (Cleary et al. 2005, Reed et al. 2009), triggering dendritic spine loss (Shankar et al. 2007, Shankar et al. 2008), and tau hyperphosphorylation (Jin et al. 2011). In contrast, synthetic oligomeric Aβ preparations are traditionally used at much higher concentrations ranging from 0.1–20 μM (Lacor et al. 2004, Deshpande et al. 2006, Lacor et al. 2007, Shankar et al. 2008, Renner et al. 2010, O’Nuallain et al. 2010). The apparent difference in relative concentration of Aβ could therefore explain the discrepancies in the potential mechanisms mediating oligomeric Aβ toxicity.

D. Future directions

Our understanding of the deleterious events triggered by oligomeric Aβ species has greatly improved in the past 3–4 years but much more is needed before we can establish an integrated view on this problem. In particular, specific efforts must be committed to support studies that attempt to determine the molecular mechanism(s) of endogenous Aβ assemblies isolated by immuno-affinity combined with size-exclusion chromatography. In addition, there is an important need to standardize the relative concentrations of oligomeric Aβ used to assess their toxic properties. Only then, will we, as a field, reduce the amount of contradictory findings that plague our progression towards identifying specific signaling cascades and subsequent reasonable diagnostic and therapeutic interventions.

DISCUSSION

With growing evidence that at least 4 endogenous Aβ oligomeric assemblies, i.e. dimers, trimers, Aβ*56 and APFs, are produced in human and transgenic mouse brain tissue and are able to alter neuronal/cognitive function, it is now crucial that we establish the longitudinal profile of these molecules in human CSF and examine the relative relationship between the levels of these soluble Aβ assemblies and cognitive impairment. We must also focus our combined efforts on identifying the respective origins of these four molecules. While Aβ trimers are not necessarily the most abundant species when amyloid burden is prominent in vivo (Lesne et al. 2006, Shankar et al. 2008, Shankar et al. 2009), they are preferentially produced and secreted by mature neurons in vitro (Lesne et al. 2006). In addition, trimers are detected at very early ages in APP transgenic mice (Lesne et al. 2006) and humans (Lesne et al. unpublished data) suggesting that Aβ trimers are the molecular brick for oligomers evading fibrillization. Supporting this hypothesis, larger Aβ oligomers detected in mice, such as potential hexamers (~27 kDa) and Aβ*56, are formed at ages following the formation of trimers and prior to plaque formation in Tg2576 (Lesne et al. 2006). Inspired by the terminology advanced by Dr. Glabe (Glabe 2008), we propose that trimers and trimer-based oligomers (hexamers and Aβ*56) should be referred as non-fibrillar Aβ oligomers. APFs would also be part of this subgroup following the recent demonstration that APFs do not give rise to fibrils (Lasagna-Reeves et al. 2011). In contrast, Aβ dimers start accumulating extraordinarily paralleling or slightly preceding the earliest deposition of amyloid at 8–10 months of age in Tg2576 ((Kawarabayashi et al. 2004); Figure 5B). Similarly in human brain tissue, dimers are only detected in subjects with elevated amyloid loads (Shankar et al. 2008, Lesne et al. unpublished data). It is therefore likely that dimers might either constitute the seed of amyloid plaques or might be generated within plaques. We therefore propose that Aβ dimers are oligomers capable of converting to protofibrils and fibrils. Further supporting this hypothesis, studies using the artificially generated Aβ S26C dimers revealed that toxic protofibrils form rapidly from Aβ dimers. Accordingly, we would advance that Aβ dimers should be termed as prefibrillar oligomers. This new classification of Aβ oligomers is illustrated in Figure 1 modeling the possible relationships between Aβ entities in vivo.

Another major source of potential discrepancy between studies relies on the use of protein mixtures containing multiple Aβ species. To minimize this effect, the specific mechanism(s) of action of purified Aβ assemblies should also be examined individually. Using SEC to separate various endogenous Aβ assemblies has become widely accepted and has been useful for the characterization and isolation of Aβ oligomeric species (Townsend et al. 2006, Walsh et al. 2005, Lesne et al. 2006, Cheng et al. 2007, Shankar et al. 2008, Reed et al. 2009). It would then be extremely powerful to demonstrate that studies performed with synthetic preparations can be replicated with endogenously produced Aβ oligomers of similar molecular weight as determined by SDS-PAGE and SEC for instance. Moreover, these same techniques will allow us to create new tools to detect Aβ oligomers with the hope that they will also being able to block Aβ from interacting with its potential receptors.

Fundamentally, the amyloid cascade proposed nearly 20 years ago and revised in 2002 (Hardy & Higgins 1992, Hardy & Selkoe 2002) has reached another level of complexity with the emergence of multiple deleterious Aβ oligomers. Based on the findings integrated here, it is possible that Aβ dimers, trimers, Aβ*56 and APFs may activate specific signaling pathways, with differential consequences on neuronal synaptotoxicity and survival. Complicating this view further is the possibility that the Aβ species might act at different times of AD pathogenesis. Drawing a parallel with Tg2576, trimers, Aβ*56 and dimers appeared to form in brain tissues in this sequence (the onset of formation for APFs is not known in Tg2576). Such experimental design might offer insights in better improving our understanding of oligomeric Aβ-induced toxicity.

To conclude, the hypothesis that Aβ oligomers constitute the initiator culprits of AD is holding up fairly well to our scientific scrutiny. In a larger context, the upcoming identification of novel antibodies specific to each Aβ oligomer or preventing the interaction of oligomeric Aβ with its receptors offers hope to individuals diagnosed with AD and to their families.