Biochemical Society Transactions


The role of cell-derived oligomers of Aβ in Alzheimer's disease and avenues for therapeutic intervention

D.M. Walsh, I. Klyubin, G.M. Shankar, M. Townsend, J.V. Fadeeva, V. Betts, M.B. Podlisny, J.P. Cleary, K.H. Ashe, M.J. Rowan, D.J. Selkoe


Burgeoning evidence suggests that soluble oligomers of Aβ (amyloid β-protein) are the earliest effectors of synaptic compromise in Alzheimer's disease. Whereas most other investigators have employed synthetic Aβ peptides, we have taken advantage of a β-amyloid precursor protein-overexpressing cell line (referred to as 7PA2) that secretes sub-nanomolar levels of low-n oligomers of Aβ. These are composed of heterogeneous Aβ peptides that migrate on SDS/PAGE as dimers, trimers and tetramers. When injected into the lateral ventricle of rats in vivo, these soluble oligomers inhibit hippocampal long-term potentiation and alter the memory of a complex learned behaviour. Biochemical manipulation of 7PA2 medium including immunodepletion with Aβ-specific antibodies and fractionation by size-exclusion chromatography allowed us to unambiguously attribute these effects to low-n oligomers. Using this paradigm we have tested compounds directed at three prominent amyloid-based therapeutic targets: inhibition of the secretases responsible for Aβ production, inhibition of Aβ aggregation and immunization against Aβ. In each case, compounds capable of reducing oligomer production or antibodies that avidly bind Aβ oligomers also ameliorate the synaptotoxic effects of these natural, cell-derived oligomers.

  • alternating lever cyclic ratio
  • Alzheimer's disease
  • amyloid β-protein
  • long-term potentiation
  • oligomer
  • size-exclusion chromatography


Alzheimer's disease (AD) is the most common human dementia. In the year 2000, there were an estimated 4.5 million persons with AD in the U.S.A., and in the absence of effective therapy, this number is set to triple by 2050 [1]. The precise onset of clinical AD is very difficult to discern, with mild memory impairment the earliest symptom, but as the disease progresses other cognitive and behavioural changes accrue [2,3]. The end-stage AD brain is characterized by atrophy of the hippocampal formation and cerebral cortex and ventricular enlargement. Microscopically, amyloid plaques and NFTs (neurofibrillary tangles) are detected throughout the hippocampus and cerebral cortex and are often accompanied by variable numbers of amyloid-bearing meningeal and cortical microvessels. The principal component of NFTs is the microtubule-associated protein, tau. Plaques and vascular deposits are chiefly composed of Aβ (amyloid β-proteins) that are generated by sequential proteolysis of the APP (β-amyloid precursor protein) by enzymes known as β- and γ-secretase [46] (Figure 1). Although many studies have sought to correlate the severity of dementia with the number of amyloid plaques or NFTs, the best statistical correlations exist between measures of synaptic density and degree of dementia [79]. In fact, the decrease in synapse number and density is disproportionate to the loss of neuronal cell bodies [7,10,11] and suggests that pruning of processes occurs prior to neuronal death.

Figure 1 Aβ production and assembly

Aβ is generated by sequential proteolysis of APP first by β-secretase and then by γ-secretase. Monomer levels are controlled by both the rate of production and rate of degradation. Above a certain critical concentration Aβ monomer can self-associate forming dimers, trimers and other low-n oligomers. Such oligomers have been detected in the study of both cell-derived and synthetic Aβ. Fibril intermediates such as ADDLs and PF have been detected in studies employing synthetic peptides, but their existence in vivo has not been confirmed.

The emerging role of soluble Aβ

Diverse lines of evidence suggest that Aβ plays a central role in the pathogenesis of neuronal dysfunction in AD [1214], yet the Aβ hypothesis remains controversial, not least because the quantity and temporal progression of amyloid plaques do not show a simple relationship to the clinical progression of the disease [15]. However, recent studies suggest that the relatively weak correlation between plaque burden and severity of cognitive impairment may be explained by the activity of multiple different Aβ assembly forms and that early memory impairment may be mediated by soluble low-n oligomers. To model Aβ-mediated neurotoxicity, many investigators have used synthetic peptides (for a review, see [16]). At ambient or body temperature and at concentrations ≥10–20 μM, both synthetic Aβ1-40 and Aβ1-42 self-associate to form low-n oligomers, PFs (protofibrils) and fibrils (Figure 1). An important caveat when considering the cellular effects of different Aβ assemblies is the highly dynamic nature of Aβ aggregation. Because intermediates can further associate into higher-ordered aggregates, it is difficult to unambiguously ascribe cytopathological activity to a discrete species. Nonetheless, several groups have attempted to isolate pre-fibrillar synthetic Aβ assemblies and probe their toxic activity. In 1998, Lambert et al. [17] presented the first experimental evidence that certain soluble, non-fibrillar assemblies of synthetic Aβ [which they called ADDLs (Aβ-derived diffusible ligands)] could be neurotoxic. ADDLs are only formed under certain specific conditions, but once formed they are relatively stable [17,18]. By atomic force microscopy ADDLs appear as spheres with a diameter of approx. 5 nm, and migrate on SDS/PAGE at approx. 4, 8, 16 and 18 kDa. ADDLs have been shown to cause neuronal death in culture, block LTP (long-term potentiation) [17,19] and inhibit reduction of MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] by neural cell lines [17,18]. When incubated with organotypic mouse brain slices at 500 nM for 45–60 min, cell loss was not evident but a near-complete block of LTP was observed [17,19]. It is conceivable that during their incubation with neurons, ADDLs may form larger Aβ assemblies; however, the electrophysiological experiments were performed over a short-time course (1–2 h) and at concentrations (∼500 nM) well below the critical concentration for synthetic Aβ fibril formation in vitro, suggesting that ADDLs are themselves synaptotoxic.

Another non-fibrillar assembly, PF, can also rapidly alter neuronal function. PF range from spherical assemblies of approx. 5 nm diameter to short, flexible rods of up to 200 nm in length [20,21], but unlike, ADDLs, PF can be generated under a variety of biochemical conditions, and their rate of formation is dependent on Aβ concentration, pH and ionic strength [22]. PF appear to behave as true fibril intermediates in that they can both form fibrils and dissociate to lower-molecular-mass species [22,23]. Using whole-cell patch-clamp recordings, PF composed of Aβ1-40 induced an instantaneous increase in EPSCs (excitatory post-synaptic currents) in rat cortical neurons [24]. Fibril preparations also enhanced EPSCs, whereas monomeric Aβ had no effect. This excitability was entirely reversible and was concentration-dependent, with activity starting at low micromolar concentrations. Moreover, PF appear to have inherent electrophysiological activities distinct from fibrils [25] suggesting that PF and fibrils may act by separate mechanisms.

Cell-derived oligomers of Aβ disrupt both synaptic plasticity and learned behaviour

While there is no doubt that soluble pre-fibrillar assemblies of synthetic Aβ can alter synaptic function, there is as yet no confirmation that these species actually occur in nature. Thus instead of studying the activity of Aβ assemblies generated from synthetic peptides, we chose to study the activity of naturally produced, cell-derived Aβ oligomers. SDS-stable oligomers (∼8 and 12 kDa) of Aβ have been detected in the buffer-soluble fraction of human cerebral cortex [26], in human cerebrospinal fluid [27,28] and in a variety of cultured cells [2831]. 7PA2 cells [CHO (Chinese-hamster ovary) cells that express a mutant V717F (Val717→Phe) human APP] produce and secrete significant amounts of SDS-stable Aβ low-n oligomers [29] that migrate in denaturing gels as dimers, trimers and occasionally tetramers [32]. Importantly the species detected in 7PA2 CM (conditioned medium) has been confirmed as bona fide Aβ oligomer by both N-terminal radiosequencing and precipitation with Aβ40- and Aβ42-specific C-terminal antibodies [28,29]. Because of the easy maintenance and fast growth rate of 7PA2 cells, 7PA2 CM has been our medium of choice to investigate the biological activities of cell-derived Aβ oligomers. Microinjection of small volumes (∼1.5 μl) of such 7PA2 CM into the lateral ventricle of the brain of a live rat inhibited hippocampal LTP [33]. Evidence that the blockage of LTP was mediated by Aβ oligomers emerged from biochemical manipulation of the sample. Immunodepletion of the CM with Aβ-specific antibodies prevented the blockage of LTP, whereas immunodepletion of the abundant soluble APP-α derivative had no effect. Most importantly, preincubation of the CM with insulin degrading enzyme, a protease that efficiently degrades Aβ monomer but not oligomers, did not alter the LTP effect [33]. In addition, we employed SEC (size-exclusion chromatography) to fractionate 7PA2 and CHO CM (using non-denaturing, non-disaggregating buffers) and showed that the block of LTP was specifically mediated by low-n oligomers, not by Aβ monomers or any larger aggregates [34]. Taken together, these results demonstrate that a biochemically defined, oligomeric assembly of naturally secreted human Aβ alters hippocampal synaptic plasticity both in vivo and in vitro.

Whether LTP is a valid electrophysiological surrogate of learning and memory is still contentious [35]. Therefore, we proceeded to assess whether an impairment of short-term memory, the earliest symptom of AD, could actually be induced by soluble low-n oligomers of Aβ. For these experiments we utilized the alternating lever cyclic ratio test, a procedure proven to be highly sensitive for measuring drug effects on cognitive function [36,37]. In this procedure, rats learn a complex sequence of lever-pressing requirements. The animals must alternate between two levers, switching to the second lever after pressing the first lever enough times to get a food pellet. The number of presses required for each reward proceeds from 2 to 56, incorporating intermediate values based on the quadratic function, x2x. One cycle is an entire ascending and descending sequence of these response requirements (e.g. 2, 6, 12, 20, 30, 42, 56, 56, 42, 30, 20, 12, 6 and 2 presses per food reward). Six such full cycles are presented during each session. Errors are scored when the rat perseveres on a lever after reward (a ‘perseveration error’), or when an animal switches levers before completing the required number of presses on that lever (a ‘switching error’). Rats micro-injected with Aβ-containing CM showed a marked increase in both switching and perseveration errors when tested 2 h after injection, but recovered to baseline when retested 24 h later [38]. Evidence that this transient interruption of a learned behaviour was attributable to Aβ oligomers came from the findings that immunodepleting the CM of Aβ rendered the CM inactive, and that SEC fractions containing oligomers induced the deficits, whereas monomer-containing fractions had no effect [38]. Independent support for our finding that dimers and trimers of Aβ can interfere with the memory of a learned behaviour, comes from the report that the appearance of dimeric Aβ in cortical lipid raft fractions coincides with the first indicators of behavioural compromise in APP transgenic mice [39].

Avenues for therapeutic intervention

The data reviewed above clearly demonstrate that cell-derived oligomers of Aβ disrupt both synaptic plasticity and learned behaviour in vivo and recommend prevention of the formation of oligomers as an attractive therapeutic approach. We have shown that γ-secretase inhibitors can markedly decrease Aβ oligomer formation by cells at doses that still allow appreciable monomer production [33], and it seems likely that other agents that reduce intracellular monomer levels could have similar effects. Although no physiological function has been confirmed for the Aβ monomer, substantial or complete depletion of monomers in vivo could potentially result in adverse effects. In contrast, Aβ oligomers presumably arise solely as a pathological event, hence titrating monomer to levels that cannot support oligomerization or direct targeting of oligomers should relieve the toxic effects of oligomers and also minimize side-effects due to loss of monomer. We have characterized the effects on natural Aβ oligomerization of a number of compounds known to inhibit synthetic Aβ fibril formation or to prevent Aβ-mediated toxicity. Thus far we have identified only two compounds capable of inhibiting intracellular oligomer formation and of relieving the oligomer-mediated block of LTP [34]. Both compounds are hydroxyanaline derivatives, which were previously shown to prevent toxicity mediated by synthetic Aβ [40,41].

The most clinically advanced amyloid-directed therapy, Aβ immunization, has been shown to reduce cerebral Aβ levels, decrease amyloid-associated gliosis and neuritic dystrophy, and alleviate memory impairment in transgenic mice [4246]. Similarly, we have found that the intracerebro-ventricular injection of anti-Aβ monoclonal antibodies prevented the oligomer-mediated block of LTP, and that active immunization against Aβ was partially effective also [47]. Importantly, the degree of protection given by endogenous antibodies was directly related to their ability to recognize Aβ oligomers. These results suggest that anti-Aβ antibodies could bind and help clear soluble oligomers of Aβ, so that the latter are no longer present at sufficient concentrations to alter synaptic physiology. Such a mechanism could explain the rapid reversal of cognitive deficits in APP transgenic mice treated acutely with an anti-Aβ monoclonal antibody [48] and suggests the neutralization of Aβ oligomers as a potentially powerful mechanism for immunotherapy, distinct from microglial-mediated clearance and peripheral sink effects.


Adverse effects of Aβ on hippocampal synaptic plasticity and learned behaviour in vivo can be attributed to a specific, biochemically defined species of secreted Aβ, namely soluble low-n oligomers. The experimental system described herein should prove useful to further dissect the synaptotoxic effects of Aβ and to test new ways to neutralize them therapeutically. Specifically, agents that reduce Aβ production, or inhibit Aβ aggregation, and antibodies that avidly bind Aβ have already been shown to ameliorate the toxic effects of cell-derived Aβ oligomers, and recommend this paradigm as a useful pre-clinical screen for amyloid-directed therapies.


This work was supported by Wellcome Trust grant 067660 (to D.M.W.), National Institutes of Health (Bethesda, MD, U.S.A.) grant AG05134 (to D.J.S.) and by the Foundation for Neurological Diseases.


  • Proteins in Disease: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by B. Austen (St George's Hospital Medical School, London, U.K.), C. Connolly (Dundee, U.K.), B. Irvine (Belfast, U.K.), M. Sugden (Queen Mary, London, U.K.) and V. Zammit (Hannah Research Institute, Ayr, U.K.).

Abbreviations: Aβ, amyloid β-protein; AD, Alzheimer's disease; ADDL, Aβ-derived diffusible ligand; APP, β-amyloid precursor protein; CHO cells, Chinese-hamster ovary cells; CM, conditioned medium; EPSC, excitatory post-synaptic current; LTP, long-term potentiation; NFT, neurofibrillary tangle; PF, protofibril; SEC, size-exclusion chromatography


View Abstract