Biochemical Society Transactions

Central Nervous System

Synaptic memory mechanisms: Alzheimer's disease amyloid β-peptide-induced dysfunction

M.J. Rowan, I. Klyubin, Q. Wang, N.W. Hu, R. Anwyl


There is growing evidence that mild cognitive impairment in early AD (Alzheimer's disease) may be due to synaptic dysfunction caused by the accumulation of non-fibrillar, oligomeric Aβ (amyloid β-peptide), long before widespread synaptic loss and neurodegeneration occurs. Soluble Aβ oligomers can rapidly disrupt synaptic memory mechanisms at extremely low concentrations via stress-activated kinases and oxidative/nitrosative stress mediators. Here, we summarize experiments that investigated whether certain putative receptors for Aβ, the αv integrin extracellular cell matrix-binding protein and the cytokine TNFα (tumour necrosis factor α) type-1 death receptor mediate Aβ oligomer-induced inhibition of LTP (long-term potentiation). Ligands that neutralize TNFα or genetic knockout of TNF-R1s (type-1 TNFα receptors) prevented Aβ-triggered inhibition of LTP in hippocampal slices. Similarly, antibodies to αv-containing integrins abrogated LTP block by Aβ. Protection against the synaptic plasticity-disruptive effects of soluble Aβ was also achieved using systemically administered small molecules targeting these mechanisms in vivo. Taken together, this research lends support to therapeutic trials of drugs antagonizing synaptic plasticity-disrupting actions of Aβ oligomers in preclinical AD.

  • Alzheimer's disease
  • amyloid β-peptide (Aβ)
  • glutamatergic transmission
  • hippocampus
  • integrin
  • long-term potentiation (LTP)
  • synaptic memory
  • tumour necrosis factor α (TNFα)


Clinically, patients with AD (Alzheimer's disease) suffer from a progressively worsening dementia that necessitates full time care and is poorly responsive to currently available drug therapy [1]. AD is an aging-dependent neurodegenerative disorder that is characterized neuropathologically by the deposition of insoluble fibrillar Aβ (amyloid β-peptide) in extracellular plaques and aggregated tau protein, which is found largely in the intracellular neurofibrillary tangles. There is now extensive evidence that abnormal processing of Aβ, as a result of altered production by β-secretase and γ-secretase cleavage of APP (amyloid precursor protein) or impaired Aβ clearance mechanisms, leading to the accumulation of toxic aggregates, is a causal factor in AD [2]. The mild cognitive impairment that presages the insidious onset of AD may be due to synaptic dysfunction caused by the accumulation of non-fibrillar, misfolded Aβ, long before large-scale synaptic loss and neurodegeneration occur [3]. LTP (long-term potentiation), a form of synaptic plasticity that has been widely used as a cellular model of learning and memory mechanisms, is inhibited by Aβ [4,5]. Such an inhibition of LTP by Aβ has been proposed to underlie the early synaptic dysfunction in AD.

Although several conformations of aggregated Aβ have been implicated in mediating the cognitive and synaptic deficits of AD [6,7], we and others have provided convincing experimental evidence that highly mobile soluble oligomers of Aβ rapidly and potently inhibit LTP of synaptic transmission in the hippocampus both in vivo and in vitro [3]. For example, subnanomolar concentrations of relatively stable, natural, low-n oligomers of Aβ in conditioned medium from cells that overexpress APP inhibited LTP in vivo, whereas monomers were inactive [8]. Indeed, recently, we found that certain human CSF (cerebrospinal fluid) samples that contain Aβ dimers at similar concentrations also inhibit LTP (I. Klyubin, V. Betts, K. Blennow, H. Zetterberg, A. Wallin, C.A. Lemere, W.K. Cullen, A. Welzel, Y. Peng, T. Wisniewski, D.J. Selkoe, R. Anwyl, D.M. Walsh and M.J. Rowan, unpublished work). In contrast, the same CSF samples in which Aβ had previously been removed by immunodepletion failed to affect LTP induction. Intriguingly, these samples of human CSF were from clinically normal older individuals as well as patients with AD. It seems likely that low-n oligomers of Aβ are actively disrupting synaptic plasticity mechanisms underlying memory processes in the early prodromal phase of AD prior to clinical or biochemical diagnosis. Soluble Aβ oligomers are assumed to be non-physiological assembly states generated by misprocessing and therefore would not be expected to have ‘physiological’ receptors. However, the high potency of these oligomers to rapidly inhibit LTP induction points to oligomeric Aβ having a conformation(s) that interact(s) selectively with a specific binding domain(s) of a receptor(s), directly or indirectly regulating synaptic plasticity. Many different receptors have been proposed to mediate the toxic, pathological effects of Aβ [9], but it is unclear which, if any of them, bind the Aβ oligomer species that causes the inhibition of LTP at subnanomolar concentrations.

In a previous review [10], we summarized evidence that Aβ oligomers cause activation of glial/neuronal stress kinases and downstream production and release of nitric oxide, superoxide and other mediators. This then compromises normal activity-dependent physiological activation of these cellular pathways by synaptically released glutamate acting through NMDARs [NMDA (N-methyl-D-aspartate) receptors] and mGluR5 (metabotropic glutamate receptors, subtype 5), resulting in inhibition of high-frequency stimulation-induced LTP. In the present review, we outline recent evidence for the key roles of TNFα (tumour necrosis factor α) and αv integrins in mediating the disruption of synaptic plasticity by Aβ. Both the type-1 (p55) TNFα receptor [TNF-R1 (type-1 TNFα receptor)] [11] and αv integrins [12] have been proposed to act as receptors for Aβ.

The pro-inflammatory cytokine TNFα mediates the disruption of LTP by Aβ

The brains of patients who die from AD harbour extensive evidence of inflammation, including activation and proliferation of glia and increased concentrations of inflammatory mediators. However, it is not clear how early in the disease process brain inflammation occurs or the relative contribution of potentially detrimental or positive effects of inflammatory mechanisms to clinical symptoms and disease progression [13].

TNFα is a key cytokine that has been implicated in the normal physiological function of the central nervous system as well as mediating pro-inflammatory processes in neurodegenerative diseases including AD [14]. One putative physiological role of glia-derived TNFα is the promotion of synaptic homoeostasis in developing hippocampal pyramidal cells by triggering a slow compensatory strengthening of synaptic inputs after prolonged periods of inactivity [15]. Another function may be the mediation of the induction of LTD (long-term depression) in vitro [16]. Indeed we found that activation of TNF-R1 appears to be essential for mGluR-dependent LTD at perforant pathway synapses with granule cells in the dentate gyrus of adult animals [17]. TNFα has been shown to enhance Ca currents [18] and Ca mobilization from intracellular ryanodine-sensitive Ca stores [19], either of which may be required for synaptic scaling and synaptic LTD.

In striking contrast, TNFα has a disruptive effect on the induction of LTP in the hippocampus ([2022] but see [15]). Intriguingly, the inhibition of LTP by TNFα is dependent on activation of mGluR5 and the stress kinase p38 MAPK (mitogen-activated protein kinase) [2123], both of which are critically implicated in LTD induction [17,24]. Thus the inhibition of LTP by TNFα may be due to a shift towards mechanisms promoting LTD.

In AD, the brain and plasma concentration of TNFα is increased [25,26], while brain TNF-R1 shows increased expression [11]. Indeed cognitively normal older people who have elevated spontaneous TNFα production from peripheral mononuclear cells are at significantly increased risk of developing AD compared with those with low production [27]. The Aβ-induced activation of cultured microglia is known to result in the production of TNFα [28]. Moreover, TNFα was identified as the principal neurotoxic agent resulting from Aβ-induced pro-inflammatory transcriptional changes [29]. Furthermore, pre-aggregated Aβ rapidly increases TNFα protein levels in the rat brain [30], and TNFα and the Aβ-containing C-terminus of APP synergistically interact to disrupt working memory after intrahippocamal infusion [31].

On the basis of this highly suggestive background information, we tested the idea that the Aβ42-mediated inhibition of LTP in the dentate gyrus in vitro occurs as a consequence of release of endogenous TNFα and activation of TNF-R1. We examined the role of endogenous TNFα using agents that bind and neutralize TNFα or that inhibit the production of TNFα. We also determined whether Aβ-mediated inhibition of LTP induction occurred in mutant mice deficient in TNF-R1.

In order to neutralize TNFα, we used two TNFα ligands, infliximab (a chimaeric IgG1κ monoclonal antibody containing a murine TNFα-binding region and a human IgG1 backbone) and a TNFα peptide antagonist (an analogue of the amino acid sequence 159–178 of the human TNF-R1 with specific and high-affinity binding to TNFα), at concentrations that did not alter LTP induction when applied alone to rat hippocampal slices. Pre-treatment with either of these agents prevented Aβ inhibition of LTP induction at medial perforant pathway synapses. We also studied the effect of an agent that inhibits TNFα production, thalidomide [32], which on its own did not affect LTP. Pre-treatment with thalidomide prevented the disruptive effect of Aβ. Significantly, Aβ had no effect on the induction of LTP in mutant mice deficient in TNF-R1, whereas it had the same inhibitory effect in wild-type mice as was found in rat hippocampus. Taken together, these findings are strong evidence that the inhibition of LTP by Aβ in vitro requires TNFα activation of TNF-R1.

Given the potential therapeutic implications of these findings, we recently examined the effects of targeting TNFα on the inhibition of LTP at CA1 synapses by soluble Aβ42 in vivo (N.W. Hu, R. Anwyl and M.J. Rowan, unpublished work). Consistent with the in vitro findings described above, intracerebroventricular injection of either infliximab or the TNFα antagonist peptide prevented the inhibition of LTP at CA1 synapses caused by intracerebroventricular injection of Aβ. As thalidomide is a drug that can be used in humans under controlled conditions, we examined the effectiveness of systemic treatment against the synaptic plasticity-disrupting effect of Aβ. The intracerebroventricular injection of Aβ failed to inhibit LTP in animals that received peripheral administration of thalidomide. Importantly, none of these interventions affected LTP when used on their own.

The involvement of TNFα in mediating the inhibition of LTP in the hippocampus by soluble Aβ complements the recent reports of Aβ-induced TNFα-dependent delayed impairment of spatial learning and reduction in hippocampal synaptophysin [33,34]. Similar to the inhibition of LTP, the delayed disruption of spatial learning and apparent loss of synapses were also mediated through TNF-R1 and were dependent on iNOS (inducible nitric oxide synthase) [34].

The cell adhesion molecule αv integrin mediates the disruption of LTP by Aβ

The integrin class of cell adhesion molecules are transmembrane glycoprotein heterodimers of non-covalently linked α- and β-subunits that mediate communication between cells and the extracellular framework, enabling focal cell adhesion, growth, migration and survival. Each subunit has a large ligand-binding extracellular domain and a short cytoplasmic tail that links to the actin cytoskeleton and orchestrates multiple signalling pathways, both outside-in and inside-out. Integrins are expressed widely at dendritic spines and synapses in the brain and are known to have a physiological role in the regulation of synaptic transmission and synaptic plasticity [3537]. Activity-dependent synaptic plasticity was strongly reduced in Drosophila mutants, which have deletion of a gene encoding synaptic integrins [38]. Peptides containing the RGD (Arg-Gly-Asp)-binding sequence are recognized by a large number of integrins present in neurons, and such peptides rapidly increase NMDAR-mediated synaptic transmission [39] and inhibit LTP, especially the expression/maintenance phase [40]. Disintegrins, which are potent antagonists of RGD-binding integrins, also inhibit the expression/maintenance of LTP [41].

In contrast, there is increasing evidence that integrins have a pathological role acting as potential neuronal and glia receptors mediating Aβ-induced cytotoxicity [42]. Soluble and fibrillar Aβ can substitute for extracellular matrix proteins and bind integrins, resulting in partial activation of focal adhesion signalling and, potentially, neurodegeneration [9,4345]. Following the evidence for an involvement of α1β1 integrins in Aβ-induced cell death in cultures of aging hippocampal neurons [46], Wright et al. [12] used selective integrin-blocking antibodies to demonstrate that both the binding/deposition and neurotoxicity of Aβ in human cortical primary neurons are mediated through αvβ1 and α2β1 integrins. However, some integrin-dependent effects of Aβ that have been characterized in vitro, such as microglial activation, may not normally be fully recapitulated in vivo. Thus a β1 integrin-dependent activation of microglia by Aβ leads to Aβ phagocytosis in vitro [47], whereas activated microglia may not usually phagocytose Aβ in vivo [48]. Based on these reports, we investigated whether selective αv integrin-blocking antibodies and a small-molecule inhibitor of αv integrins can prevent the Aβ inhibition of LTP both in hippocampal slices and in the intact hippocampus in vivo (Figure 1) [49]. We also examined the ability of other integrin ligands including echistatin and superfibronectin to block the Aβ inhibition of LTP.

Three different antibodies to αv-containing integrins, 18C7, 20A9 and 17E6, none of which affected LTP induction alone, prevented the Aβ-mediated inhibition of LTP at medial perforant pathway synapses with dentate granule cells in hippocampal slices. In contrast, two isotype control antibodies did not significantly affect the Aβ inhibition of LTP. Similarly in vivo, pre-treatment of anaesthetized rats with the αv integrin antibody 17E6, but not an isotype control antibody, prevented the inhibition of LTP at CA3 to CA1 synapses in the dorsal hippocampus caused by intracerebroventricular injection of soluble fibril-free Aβ. Consistent with a requirement for integrins, both superfibronectin, which is a ligand for αvβ1, and echistatin, which is a snake venom-derived disintegrin that can inhibit RGD-dependent integrins, prevented the Aβ-mediated block of LTP at a concentration that alone did not significantly affect LTP induction in vitro. Although actin stabilizers and destabilizers prevented the death of primary cortical neurons caused by Aβ [12], the actin stabilizer phalloidin failed to affect the block of LTP by Aβ at a concentration that did not affect control LTP.

Figure 1 Aβ-induced disruption of glutamatergic synaptic transmission in the hippocampus is mediated through αv integrins

(a) LTP of excitatory synaptic transmission in the CA1 area of anaesthetized rats was inhibited by intracerebroventricular (i.c.v.) injection of soluble Aβ42 (50 pmol) (left-hand panel). An antibody to αv integrin, 17E6, injected intracerebroventricularly, or a small-molecule non-peptide inhibitor of αv integrins, SM256, injected systemically, prevented the inhibition of LTP by Aβ42. Similarly, bath application of Aβ42 (500 nM) inhibited LTP at medial perforant pathway synapses with dentate granule cells in hippocampal slices (right-hand panel). Such an inhibition was abrogated by pre-treatment with an anti-αv antibody 17E6, SM256, the integrin ligand superfibronectin and the disintegrin echistatin. The actin stabilizer phalloidin had no significant effect. LTP was not inhibited when these agents were administered alone and isotype control antibodies were inactive. Values are the means±S.E.M. for % baseline EPSP (field excitatory postsynaptic potential) measured before the HFS (high-frequency conditioning stimulation). *P<0.05 compared with baseline. (b) Typical example of the inhibition of LTP in vivo by i.c.v. injection of soluble Aβ42 (asterisk) 10 min before HFS (arrow). Animals received a systemic injection of vehicle 40 min before the i.c.v. injection. (c) As in (b), except that animals received an injection of SM256 [40 mg, intraperitoneal (i.p.)] 40 min prior to Aβ42. Based on data from [49].

Since antibodies have relatively poor brain penetrance due to their large size, we also tested SM256 (3-[1-[3-(N-imidazol-2-ylamino)propyl]indazol-5-ylcarbonylamino]-2(S)-(2,4,6-trimethylbenzenesulfonylamino)propionic acid trifluoroacetate), a non-peptide small molecule which is a potent αv antagonist and which inhibits αv-mediated cell adhesion [50] and Aβ-induced neurodegeneration and signalling in cultured cortical neurons [12]. SM256 applied in vitro in the perfusion medium prevented the Aβ-mediated block of LTP without affecting control LTP. Moreover, in in vivo experiments, systemic (intraperitoneal) pre-administration of SM256 abrogated the inhibition of LTP caused by intracerebroventricular injection of soluble, fibril-free Aβ. This confirms the role of αv-containing integrins and demonstrates that they can be targeted in the brain by peripheral infusion of a small-molecule inhibitor.


The Aβ-induced inhibition of LTP may involve direct Aβ binding and activation of TNF-R1, as Aβ was found to bind to the intracellular death domain of TNF-R1 with high affinity [11]. Importantly, soluble oligomers of Aβ bound to TNF-R1 rather than insoluble fibrils [11]. Thus it is possible that in the presence of Aβ oligomers TNF-R1 is activated both by (i) endogenously released TNFα acting extracellularly and (ii) Aβ acting from inside the cell. An Aβ-induced release of endogenous TNFα is likely to be mediated by the binding of Aβ to a receptor complex that contains integrin proteins [47]. In the case of fibrillar Aβ, a microglial receptor complex has been found to contain α6β1 integrin, the integrin-associated protein CD47 and the B-class scavenger receptor CD36 [47], whereas a neuronal receptor is known to include αv/α2-containing integrins [12]. One possible scenario is that Aβ oligomers bind αv-containing integrins on glia and neurons and this triggers either TNFα release or Aβ internalization, both effects resulting in TNF-R1 activation. This in turn will activate stress kinases and downstream production of nitric oxide, superoxide and other mediators from intracellular sites including mitochondria, which causes inhibition of LTP and perhaps promotion of LTD. Many of the key factors mediating the rapid inhibition of LTP by Aβ, including involvement of oligomers and cellular stress and LTD-like mechanisms, have been implicated in Aβ-induced synaptic pruning and neurodegeneration.


This work was supported by the Science Foundation Ireland and Programme for Research in Third Level Institutions in Ireland. We thank Dr Liam Cullen, Dr Sarah Wright, Dr Irene Griswold-Prenner and Professor Dominic Walsh for extensive help.


  • Central Nervous System: A Focus Topic at Life Sciences 2007, held at SECC Glasgow, U.K., 9–12 July 2007. Edited by C. Dart (Liverpool, U.K.), M. Houslay (Glasgow, U.K.), M. Ludwig (Edinburgh, U.K.), R. Porter (Trinity College Dublin, Ireland) and J. Potts (Misouri-Columbia, U.S.A.).

Abbreviations: Aβ, amyloid β-peptide; AD, Alzheimer's disease; APP, amyloid precursor protein; CSF, cerebrospinal fluid; HFS, high-frequency conditioning stimulation; LTD, long-term depression; LTP, long-term potentiation; mGluR, metabotropic glutamate receptor; NMDA, N-methyl-D-aspartate; NMDAR, NMDA receptor; TNFα, tumour necrosis factor α; TNF-R1, type-1 TNFα receptor


View Abstract