Activity and Synaptic Receptor Targeting

This review focuses on vertebrate synapses, and it is noteworthy that Drosophila glutamatergic neuromuscular synapses are like vertebrate glycinergic synapses in requiring activity for stabilization of postsynaptic receptors, but they also show homeostatic regulation at the level of bouton number and transmitter release.

Abstract

A novel mode of regulation of synaptic function has recently been described for central synapses: activity-dependent redistribution of receptors to or from synaptic sites. Despite the current interest in fast changes in receptor targeting that may underlie conversion of silent synapses to functional ones, a compelling case for activity-dependent modulation of receptor targeting has as yet been demonstrated only for regulation on the timescale of days. On this long timescale, many receptors exhibit activity-dependent targeting, including inhibitory glycine receptors (GlyRs) on spinal cord neurons, excitatory AMPA- and NMDA-type glutamate receptors on neocortical and hippocampal neurons, and acetylcholine receptors (AChRs) at the neuromuscular junction. Surprisingly, the direction of the effect is not conserved. Blockade of one receptor type induces its removal from the synapse, whereas blockade of another receptor type stabilizes it at the synapse (Table 1). On the one hand, these results have further dashed hopes for common principles underlying regulation of synapse composition. On the other hand, understanding why activity regulates the synaptic targeting of each receptor type in a different way may help us to understand fundamental operational principles of different synapse types and their impact on central nervous system development and plasticity. Interestingly, several of these phenomena, including synaptic scaling and activity-dependent stabilization, had been predicted previously on purely theoretical grounds.Table 1Effect of Long-Term Activity Blockade on Receptor Targeting at Vertebrate SynapsesSynapseReceptorEffect of BlockadeProposed FunctionReferenceNeuromuscularAChRGlobal: ↑ nonsynapticDenervation supersensitivityReviewed by 9Duclert A Changeux J.-P Physiol. Rev. 1995; 75: 339-368Crossref PubMed Scopus (185) Google Scholar)(Focal: focal ↓ synapticCompetitive synapse elimination2Balice-Gordon R.J Lichtman J.W Nature. 1994; 372: 519-524Crossref PubMed Scopus (236) Google ScholarSpinal glycinergicGlyR↓ SynapticActivity-dependent stabilization12Lévi S Vannier C Triller A J. Cell Sci. 1998; 111: 335-345PubMed Google Scholar, 10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google ScholarHippocampal or neocortical glutamatergicAMPAR↑ SynapticMultiplicative homeostatic scaling19Turrigiano G.G Leslie K.R Desai N.S Rutherford L.C Nelson S.B Nature. 1998; 391: 892-896Crossref PubMed Scopus (1529) Google Scholar, 13Lissin D.V Gomperts S.N Carroll R.C Christine C.W Kalman D Kitamura M Hardy S Nicoll R.A Malenka R.C Von Zastrow M Proc. Natl. Acad. Sci. USA. 1998; 95: 7097-7102Crossref PubMed Scopus (271) Google ScholarNMDAR↑ SynapticMetaplasticity17Rao A Craig A.M Neuron. 1997; 19: 801-812Abstract Full Text Full Text PDF PubMed Scopus (418) Google Scholar Open table in a new tab At the vertebrate neuromuscular junction, which has been the primary model system for studying regulation of the molecular composition of synapses, activity is one of the major signals regulating AChR distribution. However, at this synapse, activity selectively regulates the level of nonsynaptic receptor (Figure 1A), whereas molecular signals including agrin and ARIA/neuregulin regulate the level of synaptic receptor (reviewed by 9Duclert A Changeux J.-P Physiol. Rev. 1995; 75: 339-368Crossref PubMed Scopus (185) Google Scholar). Chronic blockade of evoked neural activity with tetrodotoxin (TTX) or chronic receptor blockade with α-bungarotoxin induces a large increase in extrasynaptic receptor, through increased transcription from nonsynaptic nuclei, but no change in the level of synaptic receptor for several days. This phenomenon may best be thought of in relation to denervation supersensitivity. In the event of muscle paralysis, a response that reinduces AChR over the entire muscle would be the best adaptation to attempt to rapidly reestablish functional synapses at the site of any nerve contact. Thus, from these classic studies, we can predict one kind of general response to activity or receptor blockade: upregulation of receptor over the entire cell surface. As we shall see, it appears that none of the central synapse types studied to date follow this simple prediction. Pioneering studies imaging AChRs in living mice over time (2Balice-Gordon R.J Lichtman J.W Nature. 1994; 372: 519-524Crossref PubMed Scopus (236) Google Scholar, 15Nguyen Q.T Lichtman J.W Curr. Opin. Neurobiol. 1996; 6: 104-112Crossref PubMed Scopus (97) Google Scholar, and references therein) have uncovered new principles at the neuromuscular junction that add further complexity to our models of activity-dependent receptor targeting. Focal blockade induces completely different effects from studies such as those described above utilizing global synaptic blockade. Focal blockade is also by its very nature focal activity; it is a means of generating a spatial or temporal disparity in synaptic activity. Although activity blockade of AChRs throughout a junction has no effect on their maintenance at the synapse, blockade of a portion of the junction induces a selective loss of the blockaded receptors and the overlying nerve terminal (Figure 1B). A simple determination of the fate of each AChR molecule by its own activity state is not consistent with the results; for example, active receptors that were newly inserted in the blocked regions were also eliminated, even though they themselves were not exposed to α-bungarotoxin. Redistribution of receptors and synapse reorganization was most effective when a small portion (5%–40%) of the junctional area was inactive, suggesting that the level of activity in the unblocked region was the critical determinant. The loss or maintenance of each receptor appears to depend on a balance of elimination and protection signals. According to the proposed model (15Nguyen Q.T Lichtman J.W Curr. Opin. Neurobiol. 1996; 6: 104-112Crossref PubMed Scopus (97) Google Scholar), in a region of high AChR activity, a local signal (perhaps Ca2+) induced by AChR activity spreads small distances and exerts a protective effect on AChRs in that region. At the same time, high AChR activity induces another signal (perhaps depolarization) that spreads throughout the junction and in the absence of the local protective signal induces elimination of AChRs and subsequently the overlying nerve terminal. Although the complete molecular cascade for AChR elimination is not known, the AChR anchoring protein rapsyn disappears at the same rate as AChRs themselves, suggesting a regulatory event further upstream, perhaps at the level of MuSK–rapsyn signaling. This redistribution of AChRs away from focally blockaded sites was proposed to share a common mechanism with synapse elimination during normal development, and indeed this idea is consistent with previous and subsequent studies. By recording from multiply innervated neonatal mouse muscle preparations6Colman H Nabekura J Lichtman J.W Science. 1997; 275: 356-361Crossref PubMed Scopus (217) Google Scholar found an increasing disparity in quantal content between two inputs to an individual muscle fiber during the first 2 postnatal weeks correlating with the change from multiple to single innervation. In addition, the input with the lower quantal content (number of quanta of neurotransmitter released) often exhibited lower quantal size (depolarization response to individual quanta). These studies indicate a scenario in which any local inactivity may lead to loss of local AChR and the overlying release site, leading to a decrease in quantal content for that nerve. The unaffected nerve would then gain a slight advantage in generating a stronger local protective signal and spreading elimination signal in subsequent cycles of competition. This phenomenon may, in general, underlie activity-dependent synapse elimination during development and may share common mechanisms with synaptic plasticity in the adult. Early this year, two groups reported that glycine receptor activation is required for stabilization of synaptic receptor in cultured spinal neurons (10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google Scholar, 12Lévi S Vannier C Triller A J. Cell Sci. 1998; 111: 335-345PubMed Google Scholar). Chronic strychnine-mediated GlyR blockade prevents GlyR from accumulating at synaptic sites (Figure 1C). In blockaded neurons, overall GlyR protein levels do not change, but the receptor distribution pattern changes such that GlyR is found in large intracellular aggregates. The effect of strychnine on GlyR distribution could be mimicked by L-type Ca2+ channel inhibitors, leading to the proposal that synaptic GlyR activation, which causes membrane depolarization early in development, may lead to opening of L-type Ca2+ channels, local Ca2+ influx, and, through additional unknown mechanisms, local stabilization of GlyRs (10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google Scholar). However, the reports from these two groups differ in important parameters that may lead to further insights into the molecular mechanism and physiological role of this activity-dependent redistribution. 10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google Scholar reported that gephyrin, the postsynaptic GlyR anchoring protein, redistributes with the receptor, whereas 12Lévi S Vannier C Triller A J. Cell Sci. 1998; 111: 335-345PubMed Google Scholar reported a redistribution of GlyR but no change in gephyrin distribution. The former would indicate regulation upstream of gephyrin localization, whereas the latter would point to regulation of the interaction between gephyrin and GlyR. Furthermore, it is not clear whether spontaneous quantal events are sufficient for stabilization of synaptic GlyRs (12Lévi S Vannier C Triller A J. Cell Sci. 1998; 111: 335-345PubMed Google Scholar) or whether evoked synaptic activity is required (10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google Scholar). Nor is it clear whether activity is required only for initial synaptic GlyR cluster formation (10Kirsch J Betz H Nature. 1998; 392: 717-720Crossref PubMed Scopus (227) Google Scholar) or also for the maintenance of such clusters (12Lévi S Vannier C Triller A J. Cell Sci. 1998; 111: 335-345PubMed Google Scholar). It may be that subtle differences in cell culture conditions modulate the effect of activity and account for some of these differences in the two studies. The significance of this activity-dependent GlyR distribution is superficially obvious, to stabilize active synaptic receptors and remove nonfunctional surface receptors. Activity-dependent stabilization is a fundamentally satisfying means of regulating the formation of postsynaptic sites. However, this form of regulation is seen only at vertebrate glycinergic synapses, not at neuromuscular, glutamatergic (see below), or even the most closely related synapses, GABAergic synapses. Synaptic GABAA receptors are also ligand-gated Cl− channels, but GABAA receptor clustering at synapses is not inhibited by activity blockade (8Craig A.M Blackstone C.D Huganir R.L Banker G Proc. Natl. Acad. Sci. USA. 1994; 91: 12373-12377Crossref PubMed Scopus (231) Google Scholar). Whereas for GlyRs, activity as well as molecular signals are required for synaptic localization, for the other vertebrate synapse types, receptor clustering appears to be induced purely by molecular signals. It remains puzzling exactly why glycinergic synapses would need such regulation; there is presumably some other reason for such dual regulation aside from cellular economy in removing nonfunctional surface receptors. Yet another fundamentally different kind of long-term activity-dependent regulation has been discovered in the past year for glutamate receptors, based on the work of three independent groups using cell culture models (17Rao A Craig A.M Neuron. 1997; 19: 801-812Abstract Full Text Full Text PDF PubMed Scopus (418) Google Scholar, 13Lissin D.V Gomperts S.N Carroll R.C Christine C.W Kalman D Kitamura M Hardy S Nicoll R.A Malenka R.C Von Zastrow M Proc. Natl. Acad. Sci. USA. 1998; 95: 7097-7102Crossref PubMed Scopus (271) Google Scholar, 19Turrigiano G.G Leslie K.R Desai N.S Rutherford L.C Nelson S.B Nature. 1998; 391: 892-896Crossref PubMed Scopus (1529) Google Scholar). While it is not yet clear how all of these studies mesh in detail, it is clear that activity blockade can increase synaptic levels of AMPA and/or NMDA receptor and, conversely, higher activity levels can decrease synaptic levels of AMPA and/or NMDA receptors (Figure 1D and Figure 1E). Furthermore, although both are driven in the same direction, AMPA receptor targeting is regulated independently of NMDA receptor targeting. 19Turrigiano G.G Leslie K.R Desai N.S Rutherford L.C Nelson S.B Nature. 1998; 391: 892-896Crossref PubMed Scopus (1529) Google Scholar found in visual cortical cultures that activity blockade using TTX or CNQX increases miniature excitatory postsynaptic current (mEPSC) amplitudes by multiplicative scaling. Furthermore, increased activity in the network by blockade of GABAergic transmission decreased mEPSC amplitudes in both neocortical and hippocampal cultures (13Lissin D.V Gomperts S.N Carroll R.C Christine C.W Kalman D Kitamura M Hardy S Nicoll R.A Malenka R.C Von Zastrow M Proc. Natl. Acad. Sci. USA. 1998; 95: 7097-7102Crossref PubMed Scopus (271) Google Scholar, 19Turrigiano G.G Leslie K.R Desai N.S Rutherford L.C Nelson S.B Nature. 1998; 391: 892-896Crossref PubMed Scopus (1529) Google Scholar). The changes in postsynaptic sensitivity that were observed to accompany these changes in mEPSC amplitude could be due to changes in the number of synaptic AMPA receptors or to posttranslational changes in their properties (19Turrigiano G.G Leslie K.R Desai N.S Rutherford L.C Nelson S.B Nature. 1998; 391: 892-896Crossref PubMed Scopus (1529) Google Scholar). A change in the number of synaptic AMPA receptors is suggested by the change in the number of synaptic cell surface clusters of virally expressed epitope-tagged GluR1 (13Lissin D.V Gomperts S.N Carroll R.C Christine C.W Kalman D Kitamura M Hardy S Nicoll R.A Malenka R.C Von Zastrow M Proc. Natl. Acad. Sci. USA. 1998; 95: 7097-7102Crossref PubMed Scopus (271) Google Scholar; Figure 1D). This result is in apparent contrast with previous reports of no change in AMPA receptor cluster number following TTX or CNQX treatment of spinal hippocampal cultures (16O'Brien R.J Mammen A.L Blackshaw S Ehlers M.D Rothstein J.D Huganir R.L J. Neurosci. 1997; 17: 7339-7350Crossref PubMed Google Scholar, 17Rao A Craig A.M Neuron. 1997; 19: 801-812Abstract Full Text Full Text PDF PubMed Scopus (418) Google Scholar). This discrepancy may be due to methodological differences such as measurement of exogenous versus endogenous receptor, or differences in definition of a cluster. Alternatively, there may be real differences in the behavior of AMPA receptors as a result of some difference in cell culture parameters: preparation of neurons from embryonic versus postnatal brain, cell density, or presence or absence of glial contact or serum. Sorting out these differences and the conditions under which AMPA receptor clustering is and is not modified will be an important next step. If modulation of AMPA receptor can be shown to underlie activity-dependent synaptic scaling in vivo, the next important challenge will be to determine the molecular basis, perhaps starting by analyzing the regulation and effects of AMPA receptor binding to glutamate receptor–interacting protein (GRIP) and N-ethylmaleimide–sensitive factor (NSF) (see 11Kornau H.C Seeburg P.H Kennedy M.B Curr. Opin. Neurobiol. 1997; 7: 368-373Crossref PubMed Scopus (305) Google Scholar and 20Ziff E.B Neuron. 1997; 19: 1163-1174Abstract Full Text Full Text PDF PubMed Scopus (325) Google Scholar, for a discussion of glutamate receptor binding proteins). Independent of the mechanism of this synaptic scaling, we may ask ourselves why the neuron would have evolved such a regulatory process. In fact, a similar type of process was previously predicted as a means of homeostasis, based on physiological studies in slices and in vivo (4Bienenstock E.L Cooper L.N Munro P.W J. Neurosci. 1982; 2: 32-48PubMed Google Scholar, 3Bear M.F Proc. Natl. Acad. Sci. USA. 1996; 93: 13453-13459Crossref PubMed Scopus (234) Google Scholar, 1Abraham W.C Tate W.P Prog. Neurobiol. 1997; 52: 303-323Crossref PubMed Scopus (289) Google Scholar). Positive feedback loops of Hebbian synaptic modification tend to drive LTP to saturation and LTD to a nonfunctional state. To prevent these runaway effects, a sliding synaptic modification threshold has been postulated that would reset the requirements for inducing LTP or LTD, for example, to favor LTD following potentiation. It is just such a resetting of thresholds that this synaptic scaling would accomplish, by a global decrease in synaptic AMPA receptor activity following a long-term overall increase in synaptic activity. Even more generally, synaptic scaling is a way of changing the gain of the system, such as in response to large changes in input during development. Determining the functional significance of synaptic scaling will clearly require a demonstration that such phenomena occur in vivo as well as in cell culture. NMDA receptors can also exhibit decreased synaptic targeting with activity and increased synaptic targeting with blockade (17Rao A Craig A.M Neuron. 1997; 19: 801-812Abstract Full Text Full Text PDF PubMed Scopus (418) Google Scholar; Figure 1E). However, in another study, activity had no effect on synaptic surface targeting of an expressed epitope-tagged NR1 subunit (13Lissin D.V Gomperts S.N Carroll R.C Christine C.W Kalman D Kitamura M Hardy S Nicoll R.A Malenka R.C Von Zastrow M Proc. Natl. Acad. Sci. USA. 1998; 95: 7097-7102Crossref PubMed Scopus (271) Google Scholar), again suggesting that some as yet undefined difference in parameters between the two hippocampal cell culture systems may determine whether glutamate receptors undergo this activity-dependent redistribution. In both systems, NMDA and AMPA receptors were independently targeted, with NMDA receptors exhibiting activity-dependent localization in one and AMPA receptors in the other. The NMDA receptor redistribution occurred with no change in overall protein levels of NR1 and no change in subunit ratio NR1:NR2A/B of synaptic versus nonsynaptic clusters (17Rao A Craig A.M Neuron. 1997; 19: 801-812Abstract Full Text Full Text PDF PubMed Scopus (418) Google Scholar). The putative NMDA receptor anchoring/scaffolding protein PSD-95 (for background, see 11Kornau H.C Seeburg P.H Kennedy M.B Curr. Opin. Neurobiol. 1997; 7: 368-373Crossref PubMed Scopus (305) Google Scholar, 20Ziff E.B Neuron. 1997; 19: 1163-1174Abstract Full Text Full Text PDF PubMed Scopus (325) Google Scholar) did not change localization but was clustered at synapses in the presence or absence of detectable synaptic NMDA receptor. These results point toward one possible mechanism: activity, perhaps through regulation of levels of local Ca2+ influx through the NMDA receptor, may modify and inhibit the interaction between NR2 and PSD-95, allowing PSD-95 to remain at synapses but uncoupling NMDA receptors. This regulation of NMDA receptor localization, if in fact it occurs in vivo, would also function to reset the system in a form of homeostasis and metaplasticity (4Bienenstock E.L Cooper L.N Munro P.W J. Neurosci. 1982; 2: 32-48PubMed Google Scholar, 3Bear M.F Proc. Natl. Acad. Sci. USA. 1996; 93: 13453-13459Crossref PubMed Scopus (234) Google Scholar, 1Abraham W.C Tate W.P Prog. Neurobiol. 1997; 52: 303-323Crossref PubMed Scopus (289) Google Scholar). While counterintuitive in terms of Hebbian or LTP-like processes, the activity-mediated redistribution of NMDA receptors away from synapses over a long time course would function to reset the synaptic modification threshold to stop runaway potentiation and regain a balance. In fact, modulation of NMDA receptor distribution may be the most effective way to reset plasticity thresholds with minimal effect on baseline transmission. The idea that the synaptic modification threshold, i.e., whether an event will induce potentiation or depression, is determined by the level of postsynaptic Ca2+ entry through the NMDA receptor fits with much experimental data from the hippocampus and neocortex (3Bear M.F Proc. Natl. Acad. Sci. USA. 1996; 93: 13453-13459Crossref PubMed Scopus (234) Google Scholar). The large magnitude of the redistribution in embryonic hippocampal neuron cultures also suggests a particular significance for this phenomenon during development, in a manner very similar to that postulated by 18Scheetz A.J Constantine-Paton M FASEB J. 1994; 8: 745-752Crossref PubMed Scopus (256) Google Scholar. As has been most elegantly shown in the visual system but is also true in many other systems, there is a critical period of development during which activity competitively induces a structural reorganization of synaptic connections, stabilizing some and eliminating others. This critical period for synaptic reorganization correlates with and may be due to the highest levels of synaptic NMDA receptor activation. Furthermore, activity blockade during this stage of development prolongs both the period of synaptic reorganization and the period of high NMDA receptor effectiveness. Thus, it is possible that the activity-dependent subcellular distribution of NMDA receptors observed in culture may contribute in vivo to determining the critical period. According to this model, early in development when connectivity is low, NMDA receptors would be in the synaptic distribution, allowing activity-dependent reorganization. As connectivity and thus overall activity levels increased, NMDA receptors would redistribute away from synapses, thus bringing to a close the critical period for synaptic reorganization. This scenario would also be homeostatic in the sense that the lowered synaptic NMDA receptor levels would reset the dynamic range of the system in reponse to the higher activity levels and may be important in preventing excitotoxicity. In each of these synapses, activity of a particular receptor type regulates its own subcellular distribution over a time course of days. Considering that these systems have evolved completely different modes and means of activity-dependent regulation of synaptic receptor targeting, it is particularly satisfying that each of these phenomena can not only be rationalized post hoc but were actually predicted on theoretical grounds. The general applicability of these models of receptor targeting underlying denervation supersensitivity, activity-dependent synapse stabilization, synaptic scaling and metaplasticity, and competitive synapse elimination needs to be assessed in other synapse types. For example, although this review focuses on vertebrate synapses, it is noteworthy that Drosophila glutamatergic neuromuscular synapses are like vertebrate glycinergic synapses in requiring activity for stabilization of postsynaptic receptors, but they also show homeostatic regulation at the level of bouton number and transmitter release (reviewed by 5Broadie K Curr. Opin. Neurobiol. 1998; 8: 128-138Crossref PubMed Scopus (19) Google Scholar). One of the outstanding questions regarding central synapse formation is how the appropriate pre- and postsynaptic components become matched and aligned for each chemical synapse type (8Craig A.M Blackstone C.D Huganir R.L Banker G Proc. Natl. Acad. Sci. USA. 1994; 91: 12373-12377Crossref PubMed Scopus (231) Google Scholar). Theoretically, stabilization via activation of synaptic receptors by their appropriate transmitters would be a satisfying means of matching presynaptic and postsynaptic components. According to this hypothesis, synaptic release of GABA could stabilize apposing GABA receptors but not glutamate receptors, and synaptic release of glutamate onto a neighboring site on the same dendrite could stablilize apposing glutamate but not GABA receptors. However, the prediction of this model, that activity is required for stabilization of synaptic receptors, was found only for GlyRs, but not for GABAA receptors or AMPA- or NMDA-type glutamate receptors, as reviewed above. Thus, receptor activation does not appear to be a general means for generating selectivity in receptor clustering among postsynaptic sites, and we are forced to contemplate more complex molecular models. The molecular mechanisms underlying the activity-dependent control of receptor localization are largely understood only for global activity manipulations at the neuromuscular junction (9Duclert A Changeux J.-P Physiol. Rev. 1995; 75: 339-368Crossref PubMed Scopus (185) Google Scholar). An immediate goal for the other model systems is to uncover the mechanistic cascades leading from depolarization, local Ca2+ influx, or changes in receptor conformation to changes in receptor targeting. Synaptic receptor density is determined by a large number of steps, collectively referred to here as receptor targeting. These steps include rates of receptor synthesis and degradation, sorting into transport vesicles, trafficking of these vesicles, insertion into or removal from the plasma membrane by exocytosis and endocytosis, and diffusion, trapping, or anchoring within the membrane (for a general discussion, see 7Craig A.M Banker G Annu. Rev. Neurosci. 1994; 17: 267-310Crossref PubMed Scopus (622) Google Scholar). It is likely that many of these steps are regulated, as is true for targeting of other receptors in polarized epithelial cells (14Mostov K.E Altschuler Y Chapin S.J Enrich C Low S.-H Luton F Richman-Eisenstat J Singer K.L Weimbs T Cold Spring Harbor Symp. Quant. Biol. 1995; 60: 775-781Crossref PubMed Scopus (31) Google Scholar). One of the most interesting sequelae of the recent cell culture studies of activity-dependent modulation of glycinergic and glutamatergic synapses is the number of important questions raised. Can these phenomena be observed in vivo, and are they developmentally restricted? In contrast to the large pharmacological manipulations, can more subtle long-term physiological changes in activity levels also modulate synaptic receptor targeting, and can we quantitatively predict outcomes? Does activity control receptor targeting at the level of an individual synapse? These questions offer a challenge for creative use of recent advances in imaging techniques.