Current Alzheimer Research (www.bentham.org/car), 2004, 1, 27-32
Bentham Science Publishers Ltd.(www.bentham.org)

---

Mechanisms of Synaptic Homeostasis in Alzheimer’s Disease

David H. Small^*

Laboratory of Molecular Neurobiology, Dept. of Biochemistry and Molecular Biology, Monash University, Victoria 3800, Australia

*Address correspondence to this author at the Laboratory of Molecular Neurobiology, Dept. of Biochemistry and Molecular Biology, Monash University, Victoria 3800, Australia; Tel: 61-3-9905-1563; Fax: 61-3-9905-3726; E-mail: david.small@med.monash.edu.au

Abstract: Despite considerable progress in defining the role of the b-amyloid protein (Ab) in the pathogenesis of Alzheimer’s disease (AD), the mechanism by which accumulation of Ab causes dementia remains elusive. Memory loss is probably caused by an Ab-induced change in synaptic plasticity. Computational neuroscience (neural network modelling) studies demonstrate that cell death (or synaptic loss as a consequence of cell death) per se cannot cause the specific pattern of gradual amnesia that occurs in AD. Amnesia typical of that seen in AD can only be produced when synaptic scaling occurs. Synaptic scaling is a compensatory homeostatic mechanism which maintains the excitatory response of individual neurons and prevents the catastrophic amnesia associated with synapse loss. In this review, several possible mechanisms of synaptic scaling are described.

Keywords: Amyloid, nicotinic receptor, neural network, computational neuroscience, amnesia, acetylcholinesterase, synaptic scaling.

introduction

In recent years, considerable progress has been made in elucidating the biochemical pathology of Alzheimer’s disease (AD). It is now generally accepted that the accumulation of the b-amyloid protein (Ab) in the brain, either as a soluble oligomeric protein or in the form of insoluble amyloid fibrils, is an early step in the pathogenesis of the disease [1]. However, one of the central unanswered questions in the field is the mechanism of Ab toxicity. In general, this question has been addressed using the “bottom-up” approach [2]. This approach involves analysis of the direct biochemical effects of Ab in vitro. Using this approach, it is hoped that understanding the biochemical effects of Ab (at the bottom) will elucidate the mechanism of cognitive deficit (at the top).

The idea that Ab is toxic was first suggested by Yankner et al. [3], who demonstrated that an amyloidogenic fragment of Ab (Ab25-35) was neurotoxic. A large number of subsequent studies have confirmed this result [1]. The mechanism of Ab neurotoxicity in cell culture remains unclear, although there is evidence implicating a variety of different biochemical pathways [2].

One of the major problems with the bottom-up approach is that it is difficult to ascertain which biochemical pathways are important for the onset of AD. Although it may be argued that many different biochemical pathways could contribute towards dementia, defining the most important pathways is not a trivial problem. If effective drugs are to be developed, based on the inhibition of Ab toxicity, then it will be very important to target only those pathways that primarily induce the neurodegenerative changes that lead to dementia.

Despite the large number of studies that have been undertaken, the suitability of cell culture models of neurotoxicity can be questioned. In many studies, neurotoxicity has been measured using assays of cell death or apoptosis [4]. However, the role of cell death in the manifestation of cognitive changes in AD is unclear [5]. Very high concentrations of Ab are required to produce neurotoxicity in culture and it is not known whether similar high concentrations occur in vivo. In addition, it may be argued that although a number of drugs (e.g. antioxidants) can attenuate Ab toxicity in vitro [6], these same drugs may not have clinical efficacy [7].

We have argued elsewhere that a “top-down” approach is needed to understand Ab toxicity [2]. Specifically, it is important to understand what sorts of cellular changes in the brain could lead to dementia before attempting to understand the biochemical mechanisms that may be involved. Evidence from two very different sources, brain pathology and computational neuroscience, suggests that neither cell death nor any non-specific toxic mechanism is likely to cause the memory loss typical of that seen in AD. Instead, a specific mechanism targeted at synaptic efficacy is likely to be the underlying cause of cognitive decline.

Evidence for a mechanism involving synaptic plasticity

Neuronal loss is readily apparent in the AD brain [8]. However, despite this, neuronal loss does not correlate as well with the cognitive decline as the loss of neocortical synapses [5]. Dystrophic neurites, which result from synaptic loss, are frequently found in the regions containing amyloid plaques. By contrast, cell death is not a common feature associated with amyloid deposits [5].

A key to understanding the mechanism of action of Ab is to determine how the protein affects neural networks in which memories are stored. Specifically, it is important to understand how cellular changes can impact upon memory. This is the “top-down” approach [2]. In AD, memory loss proceeds gradually with progressive loss, first of recent memories and the sparing of older more established memories [9]. As the disease progresses, older memories are slowly lost.

It is now known that interconnected networks of model neurons can form associative memories, provided that the strength of the synaptic connections between neurons changes as a result of prior activity [10, 11]. Memories are thereby encoded in the strength of synaptic potentials. Attractor neural networks can interpret a pattern of inputs on the basis of previous inputs and become “attracted” to these previous patterns by the current input [12]. In this way, the network can learn from previous experience, i.e. it can form memories.

Studies on attractor networks very clearly demonstrate that neuronal loss cannot cause the type of memory loss seen in AD [13, 14]. By contrast, loss of synapses can cause memory loss typical of AD, but only when synaptic scaling occurs. Synaptic scaling is a process which acts to keep the overall firing of each neuron constant over time (Fig. 1). This homeostatic plasticity may be an important mechanism by which the brain maintains the stability of neural circuits. Under conditions where synaptic scaling does not occur, a loss of synapses can lead to a catastrophic loss of memory once a critical number of synapses is reached. However, with synaptic scaling, memory loss is more gradual, with more recent, less established memories being the first to be lost [15]. This is very similar to the loss that occurs in the retrograde amnesia of AD [9].

The mechanism by which synaptic scaling occurs may involve alterations in both neuronal excitability and dendritic architecture. At the Drosophila neuromuscular junction, synaptic efficacy is regulated through the control of neurotransmitter release and through the number of postsynaptic receptors [16]. The probability of neurotransmitter release at hyper-innervated muscles is lower than in normal muscle tissue, while in hypo-innervated muscle, the number of post-synaptic receptors increases. In cortical pyramidal neurons, long-term potentiation may result in a compensatory decrease in the number of AMPA receptors, thereby reducing target activity [17]. Homeostatic regulation of pyramidal neurons may also be achieved by the local secretion of growth factors such as brain-derived neurotrophic factor (BDNF) which act to increase inhibitory signalling under conditions of high neuronal activity [18,19].

Molecular mechanisms controlling synaptic plasticity in AD

The formation of synapses can be controlled by the expression of both homophilic and heterophilic adhesion proteins including members of immunoglobulin superfamily such as N-CAM/Fasciclin II, L1, side-kicks and nectin, protocadherins, neurexins, neuroligins and proteoglycans [20].

Synapse disassembly and input elimination are less well understood. At the neuromuscular junction, loss of acetylcholine receptors, utrophin and rapsyn are early events [21]. However, it is unclear whether the loss of any one of these components triggers synapse disassembly. Loss of basal lamina components is a later event [21], but it may be a more important trigger. Input elimination (i.e. the removal of presynaptic inputs) may also be driven by activity-dependent competition mediated through the postsynaptic cell [22, 23]. Whether this is due, in turn, to competition for trophic support through the release of limiting amounts of growth factors secreted by the postsynaptic neuron [22] is still unclear.

In the case of cortical and hippocampal pyramidal neurons, the dendritic spine is a key site of synaptic plasticity [24]. The structure of dendritic spines is controlled by the level of calcium influx into the spine, which in turn is regulated by voltage-gated and ligand-gated ion channels. This pathway controls the Ras/mitogen activated protein kinase pathway which is involved in filopodia formation [25]. The balance of long-term potentiation (LTP) and long-term depression (LTD) (both of which are needed for memory formation) is also controlled by the entry of calcium during temporally associated dendritic action potentials and synaptic potentials. Induction of LTP is associated with spine enlargement and the growth of new spines [26]. The entry of calcium is to a large extent controlled by L-type calcium channels which are located on dendritic spines and which in turn are activated by receptors such as the AMPA receptor, the NMDA receptor and the a7 nicotinic acetylcholine receptor (a7 nAChR) [24, 27].

THE ROLE OF THE a7 NACHR IN SYNAPSE FORMATION

There is abundant evidence that a7 nAChRs play an important role in memory and synapse formation [28]. a7 nAChRs belong to the family of ligand-gated ion channel receptors. In the central nervous sytem, nAChRs contain various combinations of a and b subunits arranged in a basic pentameric structure which forms an ion pore [29] (Fig. 2). The two major classes of nAChRs in the brain are the heteromeric a4b2 receptor and the homomeric a7 receptor [29].

Because of its relatively high permeability to calcium, the a7 nAChR is thought to play an important role in development and synaptic plasticity [28]. In the developing rat somatosensory cortex, a7 nAChRs are transiently expressed at precisely the period when thalamocortical afferents are migrating into the cortex and forming synapses [30]. Thalamocortical afferents transiently express acetylcholinesterase (AChE), which is localised adjacent to cortical cells which express the a7 nAChR [31]. Synaptogenesis within the cortex is probably regulated by cholinergic activity, as altered cholinergic function is known to influence neurite outgrowth and synaptogenesis [32].

Similarly, cholinergic systems may regulate dendritic remodelling in the mature hippocampus and cortex, regions which are essential for memory formation. a7 nAChRs are expressed highly in both regions of the brain [29]. In the chick ciliary ganglion, a7 nAChRs are located on somatic spines adjacent to L-type calcium channels [33, 34], and the relatively greater permeability of these receptors for calcium, compared with other classes of nicotinic receptors suggests that they play an important role in the regulation of calcium levels in dendritic spines [27]. The a7 nAChRs can activate the mitogen-activated protein kinase pathway [2], which in turn regulates the phosphorylation of cytoskeletal proteins controlling dendritic morphology [25]. Further evidence for the view that the cholinergic system regulates memory and synaptic plasticity comes from the finding that cigarette smoking (i.e. nicotine) can temporarily improve memory performance [35].

Ab AND THE a7 NACHR

Ab peptides can bind to several different types of nAChRs. We were the first group to demonstrate that Ab25-35 can bind to and modulate the nAChR present on bovine adrenal chromaffin cells [36]. More recently, full-length Ab peptides have been shown to bind to nAChRs, most notably the a7 receptor [37-49]. For example, Wang et al. [37] showed that the a7 nAChR and Ab1-42 can be co-immunoprecipitated from human brain tissue and that neuronal cell lines overexpressing the a7 nAChR can bind Ab1-42. This binding was inhibited by the a7 nAChR specific antagonist a-bungarotoxin. Pettit et al. [42], Tozaki et al. [43] and Liu et al. [39] have all reported that Ab blocks the a7 nAChR receptor. However, Dineley et al. reported in two separate studies [40, 41] that Ab stimulates the a7 nAChR. Studies in Xenopus oocytes have found that Ab1-42 can stimulate a7 nAChRs [41, 44].

Our own studies support the view that Ab1-42 can stimulate the receptor [50]. We have found that in neuronal cultures, Ab1-42 (but not Ab1-40) can increase levels of a minor form of AChE that is also increased in the AD brain in regions rich in amyloid deposits. The levels of AChE are also selectively increased by a7 nAChR agonists and decreased in the presence of a7 antagonists [50]. These studies suggest that the increase in AChE levels seen around amyloid plaques in the AD brain is due to activation of a7 nAChRs close to the amyloid deposits.

Although the binding site for Ab on the a7 nAChR is unknown, it may be close to the acetylcholine binding site. In support of this idea, we have found that a peptide that is conformationally constrained to mimic a loop region in the a7 nAChR (loop C) is able to bind Ab peptides in vitro (Fig. 2B). Therefore, this binding domain represents a potential target for AD therapeutics.

Synaptic dysfunction in AD

Ab does not necessarily have to cause synaptic loss to influence memory. More subtle changes in dendritic architecture may be all that is needed. There is ample evidence that Ab can affect the growth of neurites in cell culture. Addition of Ab to neuronal cultures in both soluble and substrate-bound forms can influence the normal growth of neurites [51-53]. Amyloid plaques have been found to provide a poor adhesive substrate for neurite growth on tissue slices [54]. In transgenic mice, disturbances in the growth of neurites around amyloid plaques with the formation of abnormal ectopic synapses have been reported in vivo [55].

Computational studies show that memory is dependent upon the balance between LTP and LTD that is induced at different synapses on the same neuron by specific patterns of pre- and postsynaptic firing. These computational studies indicate that unless the ratio of LTP to LTD is within a narrow range, the number of memories that can be stored within a network of neurons is limited [56]. The ratio of LTP to LTD is controlled (among other things) by the timing of pre- and postsynaptic activity at individual synapses [57, 58]. Alterations in the dendritic length can upset this timing and thereby influence the ratio of LTP to LTD. In support of this mechanism operating in AD, Knowles et al. [59] estimated that the disturbances in dendritic architecture in the region of amyloid deposits in the AD brain are sufficient to account for changes in cognitive function.

If synaptic dysfunction is fairly easy to understand, then what homeostatic mechanisms could mediate the synaptic compensation necessary to produce the gradual amnesia that is typical of AD? As described previously, decreased synaptic activity can result in increased levels of postsynaptic receptors, which help to maintain the overall level of excitatory input (Fig. 1). However, in the case of AD, there is little evidence that the levels of either AMPA or NMDA receptors, or indeed most other types of receptors is increased. However, most interestingly, unlike other classes of nAChRs, the a7 nAChR is increased in APP (SW) transgenic Tg 2576 mice engineered to produce human Ab [40, 60] and the level of a7 nAChR mRNA is increased in the AD brain [61]. The increase in the transgenic mice occurs at a very early age, well before most other cellular changes are observed and well before the formation of amyloid plaques. At this early stage of development, the levels of human Ab are measurable, but are not sufficiently high to result in plaque formation. It is also interesting to note that levels of AChE are altered at a very early stage in Tg 2576 transgenic mice as well [62].

The effect of AChE inhibitors on synaptic scaling

It is generally thought that AChE inhibitors act to relieve the deficit in cholinergic neurotransmission that occurs when basal forebrain cholinergic neurons die [63]. However, this concept can be called into question [64-66], particularly as the loss of cholinergic neurons probably occurs at later stages of AD, well after the onset of cognitive decline [67]. Indeed, at the earliest stages of AD, there may even be an upregulation of cholinergic activity [67] similar to that seen in APP transgenic mice [40, 60, 62]. Furthermore, the possibility that AChE inhibitors may benefit a range of other neurodegenerative conditions such as vascular dementia [64], Lewy body dementia [68] and Huntington’s disease [69] also calls this hypothesis into question.

An alternative hypothesis is that AChE inhibitors act to enhance synaptic scaling, perhaps via an effect on an a7 nAChR-mediated mechanism. If this is the case, then this might explain some of the clinical effects of AChE inhibitors and the fact that these drugs do not benefit all AD patients [70]. One possibility is that an intact cholinergic system is needed for synaptic scaling and that AChE inhibitors act by boosting the normal mechanism of synaptic scaling that occurs in AD. A certain number of a7 nAChRs would be needed, or else the compensatory postsynaptic response would not be sufficient to increase the number of excitatory postsynaptic potentials (Fig. 3). This concept may explain why ultimately AChE inhibitors become less effective as the disease progresses [71]. The loss of a7 nAChRs and/or the loss of cholinergic innervation at later stages of the disease may prevent synaptic compensation from operating. Thus synaptic loss may finally reach a severe stage in which the normal mechanism of synaptic scaling is overwhelmed. The resulting decline would then be more rapid and patients would be unresponsive to AChE inhibitor drugs.

Summary

More attention should be paid to the mechanism by which Ab affects synaptic plasticity in the brain. There is now increasing evidence that Ab can bind to a7 nAChRs directly and modify their function. This interaction is a potential therapeutic target. In addition, a7 nAChRs may be important for synaptic scaling, a mechanism of homeostatic regulation in the brain. Some of the beneficial effects of AChE inhibitors may be due to an enhancement of synaptic scaling.

References

[1]             Small DH, McLean CA. Alzheimer's disease and the amyloid beta protein: What is the role of amyloid? J Neurochem 73: 443-449 (1999).

[2]             Small DH, Mok SS, Bornstein JC. Alzheimer's disease and Abeta toxicity: from top to bottom. Nat Rev Neurosci 2: 595-598 (2001).

[3]             Yankner BA, Dawes LR, Fisher S, Villa-Komaroff L, Oster-Granite ML, Neve RL. Neurotoxicity of a fragment of the amyloid precursor associated with Alzheimer's disease. Science 245: 417-420 (1989).

[4]             Forloni G, Chiesa R, Smiroldo S, Verga L, Salmona M, Tagliavini F, et al. Apoptosis mediated neurotoxicity induced by chronic application of beta amyloid fragment 25-35. Neuroreport 4: 523-526 (1993).

[5]             Terry RD. Cell death or synaptic loss in Alzheimer disease. J Neuropathol Exp Neurol 59: 1118-1119 (2000).

[6]             Munoz FJ, Opazo C, Gil-Gomez G, Tapia G, Fernandez V, Valverde MA, et al. Vitamin E but not 17beta-estradiol protects against vascular toxicity induced by beta-amyloid wild type and the Dutch amyloid variant. J Neurosci 22: 3081-3089 (2002).

[7]             Tabet N, Birks J, Grimley Evans J. Vitamin E for Alzheimer's disease. Cochrane Database Syst Rev CD002854 (2000).

[8]             Terry RD, Peck A, DeTeresa R, Schechter R, Horoupian DS. Some morphometric aspects of the brain in senile dementia of the Alzheimer type. Ann Neurol 10: 184-192 (1981).

[9]             Butters N, Delis DC, Lucas JA. Clinical assessment of memory disorders in amnesia and dementia. Annu Rev Psychol 46: 493-523 (1995).

[10]          Hopfield JJ. Neural networks and physical systems with emergent collective computational abilities. Proc Natl Acad Sci USA 79: 2554-2558 (1982).

[11]          Hopfield JJ. Neurons with graded response have collective computational properties like those of two-state neurons. Proc Natl Acad Sci USA 81: 3088-3092 (1984).

[12]          Amit DJ, Treves A. Associative memory neural network with low temporal spiking rates. Proc Natl Acad Sci USA 86: 7871-7875 (1989).

[13]          Ruppin E, Reggia JA. A neural model of memory impairment in diffuse cerebral atrophy. Br J Psychiatry 166: 19-28 (1995).

[14]          Horn D, Levy N, Ruppin E. Neuronal-based synaptic compensation: a computational study in Alzheimer's disease. Neural Comput 8: 1227-1243 (1996).

[15]          Turrigiano GG. Homeostatic plasticity in neuronal networks: the more things change, the more they stay the same. Trends Neurosci 22: 221-227 (1999).

[16]          Davis GW, Goodman CS. Synapse-specific control of synaptic efficacy at the terminals of a single neuron. Nature 392: 82-86 (1998).

[17]          Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB. Activity-dependent scaling of quantal amplitude in neocortical neurons. Nature 391: 892-896 (1998).

[18]          Rutherford LC, DeWan A, Lauer HM, Turrigiano GG. Brain-derived neurotrophic factor mediates the activity-dependent regulation of inhibition in neocortical cultures. J Neurosci 17: 4527-4535 (1997).

[19]          Rutherford LC, Nelson SB, Turrigiano GG. BDNF has opposite effects on the quantal amplitude of pyramidal neuron and interneuron excitatory synapses. Neuron 21: 521-530 (1998).

[20]          Goda Y, Davis GW. Mechanisms of synapse assembly and disassembly. Neuron 40: 243-264 (2003).

[21]          Culican SM, Nelson CC, Lichtman JW. Axon withdrawal during synapse elimination at the neuromuscular junction is accompanied by disassembly of the postsynaptic specialization and withdrawal of Schwann cell processes. J Neurosci 18: 4953-4965 (1998).

[22]          Sanes JR, Lichtman JW. Development of the vertebrate neuromuscular junction. Annu Rev Neurosci 22: 389-442 (1999).

[23]          Katz LC, Shatz CJ. Synaptic activity and the construction of cortical circuits. Science 274: 1133-1138 (1996).

[24]          Hering H, Sheng M. Dendritic spines: structure, dynamics and regulation. Nat Rev Neurosci 2: 880-888 (2001).

[25]          Wu GY, Deisseroth K, Tsien RW. Spaced stimuli stabilize MAPK pathway activation and its effects on dendritic morphology. Nat Neurosci 4: 151-158 (2001).

[26]          Toni N, Buchs PA, Nikonenko I, Bron CR, Muller D. LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature 402: 421-425 (1999).

[27]          Fabian-Fine R, Skehel P, Errington ML, Davies HA, Sher E, Stewart MG, et al. Ultrastructural distribution of the alpha7 nicotinic acetylcholine receptor subunit in rat hippocampus. J Neurosci 21: 7993-8003 (2001).

[28]          Broide RS, Leslie FM. The alpha7 nicotinic acetylcholine receptor in neuronal plasticity. Mol Neurobiol 20: 1-16 (1999).

[29]          Sargent PB. The diversity of neuronal nicotinic acetylcholine receptors. Annu Rev Neurosci 16: 403-443 (1993).

[30]          Broide RS, Robertson RT, Leslie FM. Regulation of alpha7 nicotinic acetylcholine receptors in the developing rat somatosensory cortex by thalamocortical afferents. J Neurosci 16: 2956-2971 (1996).

[31]          Robertson RT, Poon HK, Mirrafati SJ, Yu J. Transient patterns of acetylcholinesterase activity in developing thalamus: a comparative study in rodents. Dev Brain Res 48: 309-315 (1989).

[32]          Small DH, Michaelson S, Sberna G. Non-classical actions of cholinesterases: role in cellular differentiation, tumorigenesis and Alzheimer's disease. Neurochem Int 28: 453-483 (1996).

[33]          Shoop RD, Martone ME, Yamada N, Ellisman MH, Berg DK. Neuronal acetylcholine receptors with alpha7 subunits are concentrated on somatic spines for synaptic signaling in embryonic chick ciliary ganglia. J Neurosci 19: 692-704 (1999).

[34]          Shoop RD, Chang KT, Ellisman MH, Berg DK. Synaptically driven calcium transients via nicotinic receptors on somatic spines. J Neurosci 21: 771-781 (2001).

[35]          Peters R, McGee R. Cigarette smoking and state-dependent memory. Psychopharmacology (Berl) 76: 232-235 (1982).

[36]          Cheung NS, Small DH, Livett BG. An amyloid peptide, beta A4 25-35, mimics the function of substance P on modulation of nicotine-evoked secretion and desensitization in cultured bovine adrenal chromaffin cells. J Neurochem 60: 1163-1166 (1993).

[37]          Wang HY, Lee DH, D'Andrea MR, Peterson PA, Shank RP, Reitz AB. beta-Amyloid(1-42) binds to alpha7 nicotinic acetylcholine receptor with high affinity. Implications for Alzheimer's disease pathology. J Biol Chem 275: 5626-5632 (2000).

[38]          Wang HY, Lee DH, Davis CB, Shank RP. Amyloid peptide Abeta(1-42) binds selectively and with picomolar affinity to alpha7 nicotinic acetylcholine receptors. J Neurochem 75: 1155-1161 (2000).

[39]