Subsynaptic spatial organization as a regulator of synaptic strength and plasticity (original) (raw)

. Author manuscript; available in PMC: 2019 Aug 1.

Published in final edited form as: Curr Opin Neurobiol. 2018 Jun 11;51:147–153. doi: 10.1016/j.conb.2018.05.004

Abstract

Synapses differ markedly in their performance, even amongst those on a single neuron. The mechanisms that drive this functional diversification are of great interest because they enable adaptive behaviors and are targets of pathology. Considerable effort has focused on elucidating mechanisms of plasticity that involve changes to presynaptic release probability and the number of postsynaptic receptors. However, recent work is clarifying that nanoscale organization of the proteins within glutamatergic synapses impacts synapse function. Specifically, active zone scaffold proteins form nanoclusters that define sites of vesicle release, and these sites align transsynaptically with clustered postsynaptic receptors. These nanostructural characteristics raise numerous possibilities for how synaptic plasticity could be expressed.

Transsynaptic alignment can control synaptic function

The efficacy of synaptic transmission may rely in part on efficient organization of presynaptic release sites to postsynaptic receptors. While in some synapse types, ultrastructural landmarks such as ribbons or T-bars provide a natural hint for where release occurs, the organization of release sites in brain glutamatergic synapses has been more difficult to measure. New work on this includes increasingly sophisticated physiology and imaging. Release characteristics were examined through patch-clamp at the synapse between cerebellar granule cells and molecular layer interneurons, a prototypic “simple” synapse involving a single contact between one active zone (AZ) and one postsynaptic density (PSD) [1*]. Quantal analysis of release events counted with high temporal precision here revealed that action potentials drove release at multiple discrete and essentially independent sites in the AZ. Consistent with this, a novel imaging approach called pHuse (pHluorin uncovering sites of exocytosis), using vesicle-resident sensors to map sites of single-vesicle fusion events within individual synapses, demonstrated that action potentials evoked vesicle fusion within a smaller area of the presynaptic bouton than spontaneous release in hippocampal terminals [2**]. Critically, trains of stimuli prompt release from repeatedly used sites [3*]. Thus, release probability (Pr) in a single AZ is spatially heterogeneous.

The molecular basis for this map of release likelihood presumably arises from a small set of multidomain “scaffold” proteins that coordinate the positioning of the vesicle, Ca2+ channels, and other release machinery [4, [5]. Amongst these critical scaffold proteins are Rab-interacting molecules (RIMs), Munc13/Unc13, RIM-binding proteins (RBPs), and ELKs family proteins [5]. By super-resolution imaging, these scaffolds form nanoclusters typically of ~60 to 80 nm within the AZ [2**, [6, [7**]. Varied approaches have now been used to deduce that these nanoclusters dictate the spatial distribution of Pr across the AZ surface. Combined pHuse and live-cell, single-molecule imaging of expressed RIM1 found that AZ subregions with high RIM1 density were the preferred sites of vesicle fusion after an action potential [2**]. In the glutamatergic Drosophila neuromuscular junction, (M)Unc13 distribution measured by STED aligned with functional synaptic vesicle release sites as measured by postsynaptic Ca2+ sensor GCaMP [8**]. In particular, the isoform Unc13-A appeared critical for evoked release, as deletion of the isoform resulted in drastic reduction in evoked release stemming from a reduced number of vesicles docked close to Ca2+ channels [9]. In hippocampal cultures, a newly synthesized fluorescent glutamate sensor to measure quantal release characteristics and deduce the number of release sites in individual boutons (~3 to over a dozen), was combined with STORM to measure the distribution of Munc13 in the same terminals [7**]. Remarkably, the number of Munc13 clusters observed per bouton tightly correlated with the number of functionally defined release sites. Interestingly, while modulated by AZ scaffold proteins, Ca2+ channels themselves can be mobile [10] but are frequently clustered [11], and release probability scales not only with the number of channels but the number of channel nanoclusters [12*]. This suggests a model whereby release sites arising from protein clustering may occupy stable positions, but that release likelihood per site may vary as scaffolds and channels move among them.

These observations of AZ organization strikingly parallel nanostructure within the PSD. Both AMPA and NMDA type receptors are subsynaptically nanoclustered, in spatial correlation with PSD-95 [13, [14, [15], which itself is clustered [14, [16]. In fact, PSD nanoclusters are very similar in size and number to AZ protein nanoclusters [2]. Critically, protein clustering on both sides of the synapse creates the potential for alignment or mis-alignment of these functionally critical features. Indeed, in hippocampal neurons both in culture and in acute slices, RIM nanoclusters align transsynaptically with concentrated postsynaptic receptors and scaffolding proteins [2**]. This organization brings to mind a transcellular protein “nanocolumn” that guides release to occur near receptors (Fig. 1). Such alignment is important because the surprisingly small amount of glutamate released from a single vesicle is not likely to activate relatively distant receptors before dispersing by diffusion [5, [14, [17, [18]. Direct experimental support for the idea that spatially confined subsets of receptors are activated by single vesicles remains elusive, but there is substantial indirect evidence [19, [20, [21, [22]. Indeed, given maps of RIM abundance and receptor distribution, the relationship between release location and AMPAR opening probability has been used to model synaptic strength and predict the effect of release misalignment [2**], which can be up to even 50% change in signal amplitude [2**, [14], similar in magnitude to many forms of synaptic plasticity.

Figure 1.

Figure 1

Key synaptic proteins are enriched within nanocolumns. High density nanoclusters of RIM1/2 and (M)Unc13 dictate sites of vesicle fusion following action potentials, which are assembled in alignment with postsynaptic nanoclusters of receptors [2**,8**]. The coupling between the release and receptors through transsynaptic proteins together with the distributions of proteins in grey is hypothetical, while the distributions of color-coded proteins have been confirmed. Adapted from [2**].

Alignment-mediated plasticity arising from postsynaptic reorganization

Sensitivity of receptor activation to alignment with release site suggests many forms of synaptic plasticity (Fig. 2). Classic models of potentiation and depression have established that these canonical forms of plasticity involve alterations in the number of postsynaptic receptors. However, the functional effects of adding receptors, for instance, will depend on their alignment with release sites. In one influential model, synapses contain multiple “modules” spanning the two cells, and under basal conditions, some modules are opposed to release sites but are “silent” because they lack AMPARs [23]. In this model, LTP converts silent modules to become responsive through addition of AMPARs to these regions (Fig. 2). Support for such a modular nature of synaptic organization came from recent work using STED imaging to examine pre- and post-synaptic proteins simultaneously even in live synapses undergoing LTP [24]. Single spines were observed to contain several sometimes widely spaced, distinguishable assemblies of PSD-95, suggestive of modules. These were associated with colabeled presynaptic proteins, and glycine-induced LTP increased the number of these coordinated modules, as suggested in Figure 2h. further work will be required to reveal the nature of these modules and their relation with the nanocolumns seen with STORM. It will also be important to determine whether these modules of PSD-95 contain AMPARs. In fact, the location of receptors newly added during physiological LTP has not (somewhat shockingly) been deduced, and there is not concrete evidence for whether they intermix with pre-existing receptors or occupy distinct sites. An important recent test used an optically regulated protein dimerization system to recruit exogenous AMPARs to the PSD without LTP induction [25**]. Surprisingly, the newly added receptors had no apparent effect on quantal amplitude even though they responded to glutamate uncaging at the synapse. One interpretation of these intriguing results is that the receptors’ position within the PSD was not optimal to sense evoked glutamate release.

Figure 2.

Figure 2

Potential mechanisms of synaptic plasticity mediated by changes in release-receptor alignment. These changes may arise from repositioning or change in the number or position of release sites presynaptically, from altering the number, position, size, or internal density of postsynaptic receptor clusters, or from changing the nature of transsynaptic coupling. Panels with notations depict mechanisms that have support from direct or indirect evidence in the literature.

Posttranslational modification of PSD-resident receptor subsets may alter transmission through the local effects of kinases docked in PSD nanoregions (Fig. 2f), for instance CaMKII on specific NMDARs [26]. Alternatively, regulation of receptor-anchoring PSD proteins may be required to drive receptor exchange in local regions of the synapse. Indeed, multiple mechanisms control PSD-95 abundance, including phosphorylation [27], palmitoylation [13] and direct binding to the actin modulator α-actinin [28]. Because PSD-95 displays such limited intra-PSD mobility [29, [30], spatially restricted modification of the scaffold may be a mechanism for nanoscale remodeling of receptor distribution. Another exciting possibility is the potential for extracellular interactions to control receptors’ ability to be activated. In C elegans, presynaptic secretion of the protein NRAP-1 modifies gating of postsynaptic NMDARs and completely controls their ability to be opened by glutamate (Fig. 2f) [31*]. Whether a related mechanism is at play in vertebrates is unknown, but this could offer a potential mechanism to activate pre-clustered receptors.

Perhaps most intriguingly, reorganization of receptor distribution—regardless of changes in receptor number—could alter synaptic strength (Fig. 2). Simulations indicate that over a broad range of starting geometrical conditions, simply condensing a cluster of existing AMPARs strongly potentiates the AMPAR current [32], and redistribution of NMDARs similarly was predicted to strongly regulate their activation probability [31]. The receptor distribution is likely established and adjusted by diverse mechanisms. Critically, AMPARs can be mobile within the PSD, but the diffusion rate and fraction which is mobile are affected not only by binding of auxiliary subunits [33, [34, [35], but also by synaptic activity [36] and receptor desensitization [37]. Recent work makes clear that the extraordinarily high density of protein within the PSD has profound consequences for receptor mobility. Steric hindrance in the PSD obstructs receptor motion [38*, [39] and may even help lock receptors within dense subdomains [40]. In fact, immediately after chemical LTP induction, PSD-95 is more concentrated within nanoclusters, which may in turn increase the AMPAR density locally within nanocolumns (Fig. 2g) [2**]. Even more unexpectedly, PSD-95 in the synapse can exist in both condensed liquid and aqueous phases, with the transition between these states regulated by density and through binding of the highly abundant GTPase activating protein SynGAP [41**]. It is easy to imagine that mobility of receptors and other proteins may alter within PSD regions of differing phase states. These complex interactions that control receptor exchange and diffusion within the synapse are likely to be particularly important during synaptic plasticity, since reducing receptor mobility (via antibody-based cross-linking) not only magnifies paired-pulse depression [42], but impairs LTP and learning [43*].

Alignment-mediated plasticity arising from presynaptic reorganization

Just as net receptor activation is not completely predicted from only the number of postsynaptic receptors, Pr may not be sufficient per se to determine the functional impact of increasing or decreasing glutamate release. Release position matters, suggesting potential mechanisms of plasticity that arise from alterations to vesicle release sites.

Following the idea of synaptic protein “modules,” release sites may be reorganized by the addition of functional presynaptic units. One mechanism for this may be through “nascent zones,” synaptic areas adjacent to AZs that lack presynaptic vesicles (Fig. 2c) [44]. It has been proposed that with PSD expansion in LTP, nascent zones are converted to functional domains of the AZ through addition of synaptic vesicles [44], though the molecular requirements for such a transition have not been evaluated.

The AZ itself is also a dynamic structure. AZ proteins can undergo vigorous reorganization on a time scale of minutes [45], which may consequently affect release site organization. For instance, following prolonged synaptic silencing, Bassoon was found to be less clustered while overall quantity remained unchanged [46]. This was associated with increased recruitment of presynaptic vesicle fusion machinery including Ca2+ channels [46] and a decreased AZ-PSD distance [47], both presumably facilitating the known homeostatic increase of synaptic transmission. These activity-dependent reorganizations could dynamically modulate the properties of release sites, including not only Pr but also spatial distribution (Fig. 2a). Indeed, release mapping with pHuse showed that high-frequency stimulation reduced the frequency with which release sites were re-used, and shifted their location away from the AZ center [3*]. This redistribution of glutamate release sites would lead to reduced EPSC amplitude due to an increased distance to the receptor densities.

Another mechanism by which the location of release could be triggered to change during plasticity is through regulation of presynaptic Ca2+ channels. This may occur through regulation by AZ scaffold proteins that engage with Ca2+ channels. For example, RBP-null mutants have impaired presynaptic homeostatic plasticity resulting from lack of enhancement in presynaptic Ca2+ influx, looser coupling between Ca2+ channels and vesicles, and the number of releasable vesicles [48]. Alternatively, if mobile Ca2+ channels dynamically regulate effectiveness of individual release sites [10], their mobility may be independently modulated by activity, as suggested by decreased Ca2+ channel mobility after lowering intracellular Ca2+ with chelators [10].

Cleft reorganization

Given these considerations, the organization of the synaptic cleft appears to play a critical role in governing synapse function. Though molecular mechanisms underlying release-receptor alignment are still unknown, the most intriguing hypothesis is that trans-synaptic cell adhesion molecules (CAMs) might organize and modulate these functional modules (Fig. 2). CAMs include a large number of protein families, and some members directly interact with glutamate receptors and scaffolds [49]. Cadherin family members adopt a perisynaptic distribution, and recent EM and super-resolution imaging has revealed nanoclustered organizations of several CAMs including SynCAM [50], neuroligin and LRRTM2 [51], and Neurexin1β [52]. The colocalization of these with vesicle fusion sites or receptors remains to be tested, but CAMs appear poised to play a key role in establishing and maintaining release-receptor alignment [5] and thus controlling synaptic strength. CAMs undergo vigorous activity-dependent remodeling, and may indeed be a critical site of action. An NMDAR activation inducing long-term depression leads to a rapid disassembly of NRXβ1-NLG1 complexes (Fig. 2d) [51], and this may result from proteolytic activity of the extracellular metalloproteases MMP9 and/or ADAM 10 [53, [54] as well as CAMKII dependent phosphorylation [55] of neuroligin. A similar treatment also enlarges postsynaptic SynCAM1 puncta on the cell surface [50]. Further studies are necessary to test whether these reorganizations of CAMs result in dismantling or strengthening of release-receptor alignment, and if so, whether these alignment changes contribute to the alteration in synaptic transmission.

By acting as a signaling bridge, CAMs can also mediate the trans-synaptic coordination during maturation and plasticity of synapses [56]. In C. elegans neuromuscular junction, postsynaptic neurexin1 plays a key role in mediating a retrograde inhibition of presynaptic transmitter release [57**]. Emerging evidence suggests that this kind of retrograde coordination could happen structurally at the nanoscale level. Following chemical LTD induction with NMDA treatment, hippocampal synapses underwent a two-step change of synaptic nanoarchitecture: initially, PSD-95 nanoclusters were disrupted (Fig. 2e), but delayed effects after 30 min recovery included an increase in size of presynaptic RIM nanoclusters that remarkably appeared only in those aligned with postsynaptic PSD-95 and AMPARs (Fig. 2b). Meanwhile there was no change in RIM nanoclusters not aligned with PSD-95, suggesting an alignment-specific process involved in this plasticity [2**].

Conclusions

Together, these observations indicate that glutamatergic synapses are built with a delicate subsynaptic spatial organization of vesicle release sites relative to clustered receptors, and suggest that this architecture may help establish synaptic strength. Many forms of plasticity may thus arise from spatial reorganization of release sites, receptors, or cleft proteins—cooperating with or even instead of changes in release probability or receptor number. Exciting work lies ahead to link particular nanostructural changes to functional changes in synaptic strength, and to discover which of these many potential means of regulation are in fact taken advantage of by cells as they undergo plasticity. A deeper grasp of the diversity of mechanisms, including between synapses and amongst different cell types, will give greater insight to their unique contributions to neural circuits, ultimately enabling a better understanding of both healthy and diseased brain function.

Highlights.

Acknowledgments

This work was supported by the Kahlert Foundation and grants to HC (F30 MH105111), AT (NARSAD YIA), and TAB (R01 MH080046 and NS090644).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References