Target-cell-dependent plasticity within the mossy fibre–CA3 circuit reveals compartmentalized regulation of presynaptic function at divergent release sites

Abstract

Individual axons of central neurons innervate a large number of distinct postsynaptic targets belonging to divergent functional categories such as glutamatergic principal cells and inhibitory interneurons. While each bouton along a common axon should experience the same activity pattern in response to action potential firing within the parent presynaptic neuron, accumulating evidence suggests that neighbouring boutons contacting functionally distinct postsynaptic targets regulate their release properties independently, despite being separated by only a few microns. This target-cell-specific autonomy of presynaptic function can greatly expand the computational prowess of central axons to allow for precise coordination of large neuronal ensembles within a given circuit. An excellent example of target-cell-specific presynaptic mechanisms occurs in the CA3 hippocampus where mossy fibre (MF) axons of dentate gyrus granule cells target both principal cells and local circuit inhibitory interneurons via both anatomically and functionally specialized terminals. Of particular interest, mechanisms of both short- and long-term plasticity remain autonomous at these divergent release sites due to an anatomical and biochemical segregation of discrete molecular signalling cascades. Here we review roughly a decades worth of research on the MF–CA3 pathway to showcase the target-cell dependence of presynaptically expressed NMDA receptor-independent synaptic plasticity.

Opening remarks

Synaptic contacts represent the primary functional unit for information transfer between central neurons, and plasticity of synaptic efficacy is considered an essential feature for information processing and storage within the CNS (Bliss & Collingridge, 1993; Bear & Malenka, 1994; Malenka & Nicoll, 1999; Malenka & Bear, 2004). Moreover, many neurological disorders such as epilepsy and chronic pain manifest as inappropriate alterations in synaptic efficacy (Kullmann et al. 2000; Woolf & Salter, 2000). As such, cellular mechanisms governing synaptic plasticity represent an area of intense investigation with both physiological and a pathological relevance. While NMDAR-dependent forms of bidirectional plasticity, typified by LTP and LTD observed at contacts between hippocampal pyramidal cells, remain the most intensely investigated and ubiquitously expressed prototypic models of synaptic plasticity, a more pluralistic view encompassing distinct forms of plasticity that proceed independent of NMDAR activation has evolved over the past decade (Malenka & Bear, 2004; Duguid & Sjostrom, 2006). In particular, it has become clear that functionally distinct postsynaptic targets within a given neuronal circuit contacted by the same afferent population may exhibit disparate forms of long-term plasticity with regards to induction protocols, polarity and molecular mediators. Here we provide a brief review of our work examining NMDAR-independent plasticity at divergent postsynaptic targets in the hippocampal mossy fibre–CA3 circuit that serves to illustrate this target cell-dependent nature of long-term plasticity (for a detailed review see McBain, 2008).

Target-cell-dependent plasticity within the MF–CA3 circuit

Mossy fibre axons (MFs) arising from dentate gyrus granule cells provide a major conduit for information flow from higher cortical areas into the intrinsic hippocampal circuitry by linking the entorhinal cortex with the CA3 hippocampus forming the second relay of the classic trisynaptic loop. Within the CA3 region individual MF axons provide excitatory glutamatergic input to both CA3 pyramidal cells (PYRs) and a diverse population of inhibitory interneurons with somata located within or on the edges of stratum lucidum (for simplicity, here termed stratum lucidum interneurons; SLINs). As SLINs subsequently provide GABAergic inhibitory input to PYRs, MF-mediated monosynaptic excitation of PYR targets is accompanied by disynaptically mediated inhibition via the simultaneous recruitment of the SLIN feedforward inhibitory network (Bragin et al. 1995; Penttonen et al. 1997; Henze et al. 2002; Mori et al. 2004; Bischofberger et al. 2006). In fact, at relatively low-frequency firing levels afferent throughput in the MF pathway is dominated by feedforward inhibition as MFs contact interneuron targets approximately 10 times more frequently than PYRs and are more effective at firing SLIN targets (Acsady et al. 1998; Lawrence & McBain, 2003). From this brief description it is clear that one must consider potential plasticity mechanisms at all divergent postsynaptic targets of MFs to predict and understand how a given plasticity-inducing protocol, such as high-frequency firing of a granule cell, will sculpt cortico-hippocampal information propogation through the MF–CA3 pathway.

Roughly 10 years ago our lab began to examine the effects of a well known plasticity-inducing protocol (high-frequency stimulation, HFS; 3 rounds of 100 Hz stimulation lasting 1 s each given at 10 s intervals) on divergent CA3 postsynaptic targets of MF afferents. At the time it was well established that MF–PYR synapses exhibit robust NMDAR-independent LTP in response to HFS (reviewed in Nicoll & Schmitz, 2005). Although the pre- and postsynaptic requirements for induction of this MF–PYR LTP remain controversial (Yeckel et al. 1999; Mellor & Nicoll, 2001; Contractor et al. 2002) it is widely accepted that expression resides presynaptically manifesting as an increase in initial release probability. This HFS-induced persistent increase in release relies in part on the activation of an adenylyl cyclase (AC)–cAMP–PKA cascade in the presynaptic terminal that ultimately leads to phosphorylation of the active zone protein RIM1alpha which somehow translates to increased efficiency of synaptic vesicle fusion independent of changes in presynaptic Ca²⁺ dynamics (Fig. 1A; Regehr & Tank, 1991; Castillo et al. 1994; Weisskopf et al. 1994; Weisskopf & Nicoll, 1995; Castillo et al. 1997, 2002; Kamiya et al. 2002). Consistent with this scheme, MF–PYR LTP can be mimicked by transient application of chemical reagents that directly activate the AC–cAMP–PKA cascade such as forskolin. Armed with this knowledge we naively hypothesized that such a form of plasticity should distribute evenly to all release sites of a common MF axon regardless of the identity of the postsynaptic target being innervated, and thus, MF–SLIN transmission should also exhibit HFS-induced presynaptic LTP. However, to our great surprise, MF–SLIN synapses yielded presynaptically expressed LTD (∼40–60% depression at 15–30 min post-induction) in response to the exact same non-associative HFS protocol that reliably induced presynaptic LTP at PYR targets (Maccaferri et al. 1998; see also Toth et al. 2000; Lei & McBain, 2002; Pelkey et al. 2005). Moreover, in marked contrast to MF–PYR transmission, synaptic efficacy at MF–SLIN contacts was completely insensitive to direct activation of AC with exogenously applied forskolin (Maccaferri et al. 1998).

Figure 1. Target-cell-specific plasticity at MF synapses.

Three schematics depicting the three most common forms of non-associative, HFS-induced activity-dependent plasticity observed at MF synapses in the CA3 hippocampus. A, a non-associative HFS-induction protocol typically produces a presynaptically expressed NMDAR-independent form of LTP at CA3 pyramidal cell synapses. This HFS-induced persistent increase in release relies in part on the activation of an adenylyl cyclase (AC)–cAMP–PKA cascade in the presynaptic terminal that ultimately leads to phosphorylation of the active zone protein RIM1alpha to enhance initial releases probability independent of changes in presynaptic Ca²⁺ dynamics. B and C, at filopodial or en passant synapses onto SLINs the same induction protocol induces two forms of LTD. B, LTD at CI-AMPAR-dominated synapses has a postsynaptic locus for both induction and expression. This form of LTD is NMDAR dependent, requires an elevation of postsynaptic Ca²⁺ and relies on internalization of surface AMPARs. C, LTD at CP-AMPAR-containing synapses also requires an elevation of postsynaptic Ca²⁺ for induction but expression is presynaptic. This NMDAR-independent form of plasticity requires presynaptic mGluR7 activation which triggers a PKC- and retrograde messenger-dependent persistent inhibition of P/Q-type VGCCs leading to a reduction in glutamate release. Please note that the schematics are for illustrative purposes only and the absence of mGluR7 in the middle panel is not meant to imply that CI-AMPAR-containing MF–SLIN synapses are devoid of mGluR7. Also for simplicity of highlighting intraterminal signalling cascades, the mGluR7 and P/Q-type VGCCs are mislocalized as both should concentrate within the presynaptic active zone. Modified from Maccaferri & McBain (2008).

Although puzzling at the time these results clearly established that the nature of long-term plasticity within a common afferent system was entirely dependent upon the postsynaptic target being innervated. Thus, neighbouring functionally divergent release sites along the same axon can serve as autonomous computational elements capable of regulating release properties independently of each other despite experiencing the same activity pattern. Additionally, it was immediately appreciated that the opposite polarities of plasticity observed at the disparate postsynaptic targets, strengthening at PYR targets combined with simultaneous weakening at SLIN targets, would create an uncoupling of the excitation–inhibition balance within the MF–CA3 system that could dramatically impact information throughput in the circuit.

Since the initial phenomenological observation of target cell-dependent long-term plasticity in the MF pathway we have endeavoured to determine the cell/molecular mechanisms responsible for this divergence in the regulation of synaptic transmission. The forskolin insensitivity in addition to the failure to induce LTP clearly indicated that MF release sites innervating SLINs either did not posses the same molecular machinery (AC–PKA–RIM1alpha) responsible for the robust potentiation at neighbouring PYR targeting MF terminals, or somehow render such machinery inactive. This implies an anatomical segregation of distinct biochemical signalling cascades to functionally divergent presynaptic terminals of the same afferent (see Pelkey & McBain, 2007 for further discussion). Fortuitously, within months of having published our initial findings, an elegant study revealed a potential anatomical substrate for such target cell-dependent biochemical compartmentalization within the MF pathway (Acsady et al. 1998). Upon examining morphologically distinct release sites of single granule cell axons Ascady and colleagues noted a strong segregation of presynaptic terminals contacting PYRs and SLINs: inputs onto PYRs are formed by the characteristic large complex multi-release site containing MF boutons (MFBs), whereas SLINs are principally innervated by a single release site containing small en passant varicosities or filipodial extensions (Fils) emanating from MFBs, but only rarely by large the MFBs themselves (Acsady et al. 1998; reviewed in Frotscher et al. 2006). With this anatomical division of labour one could envision distinct biochemical compartments along MFs coursing through stratum lucidum with PYR targeting MFBs housing elements of the cAMP cascade necessary for LTP generation and SLIN targeting en passant/Fil terminals endowed with an as yet unidentified signalling cascade capable of depressing release. Indeed, it was this model that fuelled much of our future research to determine the molecular nature of presynaptic MF–SLIN long-term plasticity.

Two forms of MF–SLIN LTD

Before discussing our findings concerning presynaptic mechanisms accounting for NMDAR-independent MF–SLIN LTD it is important to point out that postsynaptic differences between PYRs and SLINs also have the potential to influence long-term plasticity. Postsynaptically AMPARs at MF–SLIN synapses comprise a continuum ranging from GluR2-lacking, Ca²⁺-permeable channels (CP-AMPARs), to GluR2-containing Ca²⁺-impermeable channels (CI-AMPARs) (Toth & McBain, 1998; Toth et al. 2000; Lei & McBain, 2002). Interestingly, at extreme ends of the spectrum the differential expression of CP- and CI-AMPARs is mirrored by distinct NMDAR expression profiles: CI-AMPAR-dominated synapses have a large NMDAR-mediated component contributed by NR2B-lacking NMDARs, while CP-AMPAR-dominated synapses have small NMDAR-mediated components carried by NR2B-containing NMDARs (Lei & McBain, 2002). At the time of our initial report these postsynaptic variabilities in receptor complements at MF–SLIN synapses were not appreciated and all experiments were performed with NMDARs blocked (Maccaferri et al. 1998). However, after realizing that MF–SLIN synapses encompass distinct subpopulations with signature AMPA and NMDAR complements the HFS-induced MF–SLIN plasticity experiments were revisited and synapses at the two ends of the spectrum were segregated (Toth et al. 2000; Lei & McBain, 2002, 2004).

Consistent with the initial observations both CP- and CI-AMPAR-containing MF–SLIN synapses exhibited HFS-induced LTD. However, only CP-AMPAR-containing synapses yielded presynaptically expressed NMDAR-independent LTD (Lei & McBain, 2002, 2004). In contrast, CI-AMPAR-dominated synapses expressed a postsynaptic form of LTD triggered by Ca²⁺ influx through NMDARs and subsequent internalization of AMPARs similar to NMDAR-dependent LTD observed at excitatory synapses throughout the CNS (Fig. 1B; Bischofberger & Jonas, 2002; Lei & McBain, 2002, 2004). Hence, for a subset of MF–SLIN synapses the target-cell divergence of MF plasticity (LTP and LTD at PYR and SLIN synapses, respectively) can be explained by a purely postsynaptic difference. CI-AMPAR-containing MF–SLIN synapses probably represent the small population of recordings where no HFS-induced plasticity occurred under conditions of NMDAR blockade in our initial study (Maccaferri et al. 1998). Although no presynaptic specialization is required to explain this purely postsynaptic form of plasticity, it is likely that MF terminals supplying CI-AMPAR-rich SLIN postsynaptic sites remain biochemically distinct from PYR-targeting MFBs as no MF–SLIN inputs exhibited HFS-induced LTP under NMDAR blockade and all MF–SLIN recordings were forskolin insensitive (Maccaferri et al. 1998).

Despite the presynaptic expression locus of LTD at CP-AMPAR-containing MF–SLIN synapses, induction can be blocked by postsynaptic Ca²⁺ chelation (Lei & McBain, 2002; Pelkey et al. 2005). This implies that postsynaptic Ca²⁺ influx, presumably through the CP-AMPARs themselves, drives the generation of a retrograde trans-synaptic signal that participates in depressing presynaptic release. A similar requirement for trans-synaptic retrograde signalling is suggested for many forms of presynaptic long-term plasticity at various central synapses, including MF inputs to PYR and hilar interneuron targets (Alle et al. 2001; Contractor et al. 2002; Armstrong et al. 2006). In particular, recent years have witnessed a flood of publications describing a critical role for endocannabinoids (eCBs) as retrograde mediators of presynaptic LTD throughout the CNS, and some reports suggest a priviledged role for Ca²⁺ influx through CP-AMPARs in eCB generation (Gerdeman et al. 2002; Chevaleyre & Castillo, 2003, 2004; Sjostrom et al. 2003; Bender et al. 2006; Soler-Llavina & Sabatini, 2006; Chevaleyre et al. 2007; Kreitzer & Malenka, 2007; Oliet et al. 2007). Thus, although the identity of the retrograde messenger for presynaptic LTD at CP-AMPAR-containing MF–SLIN synapses remains enigmatic, eCBs would seem a logical candidate. However, in stark contrast to almost every other central neuron population examined, dentate granule cells appear completely devoid of CB1 receptors at both the message and protein levels suggesting that even if SLINs can liberate eCBs during HFS, MF terminals may not possess the machinery typically required to undergo eCB-mediated presynaptic depression (Katona et al. 2006). Clearly further work will be required to elucidate the molecular nature of any retrograde mediators of presynaptic MF–SLIN LTD.

The role of mGluR7 in presynaptic MF–SLIN LTD

In considering potential presynaptic mediators of LTD at CP-AMPAR-containing MF–SLIN synapses we came across an elegant study from Shigemoto and colleagues describing the distribution of metabotropic glutamate receptors (mGluRs) throughout the hippocampal formation (Shigemoto et al. 1997). Most interestingly, mGluR7 (specifically the mGluR7b splice variant) was found to localize very strongly to the mossy fibre pathway in stratum lucidum, and in particular, to aggregate in the active zone of release sites targeting interneurons without significant accumulation in PYR-targeting MFBs. Moreover, mGluR7 has a very low affinity for glutamate suggesting that the receptor will only be activated during periods of intense presynaptic activity such as during HFS. These properties combined with a well-defined role for mGluR7 in depressing release at various central synapses (e.g. O'Connor et al. 1999) prompted us to investigate whether mGluR7 participates in presynaptic LTD at CP-AMPAR-containing MF–SLIN synapses (Pelkey et al. 2005, 2006). Parenthetically, it was at this time that we switched our animal model from rat to mouse after noting that CP-AMPAR-containing MF–SLIN synapses were by far the more prevalent population in mouse providing an excellent system to study specifically NMDAR-independent presynaptic MF–SLIN LTD (K. A. Pelkey and C. J. McBain, unpublished observations). Thus, the expression and effects of mGluR7 at CI-AMPAR-containing MF–SLIN synapses remain unknown.

Consistent with a role for mGluR7, we found that the group III mGluR antagonist MSOP prevented HFS-induced LTD at CP-AMPAR-containing synapses without altering baseline transmission (Pelkey et al. 2005). Moreover, transient application of the group III mGluR agonist l-AP4 at concentrations sufficient to activate mGluR7 induced a long-lasting depression of MF–SLIN synapses that could not be reversed upon l-AP4 washout, even with washout medium containing MSOP (Pelkey et al. 2005). Intriguingly, this chemically induced LTD shared two hallmark features with HFS-induced LTD: (1) presynaptic expression as shown by increases in the paired-pulse ratio, CV, and failure rate; and (2) a requirement for postsynaptic Ca²⁺ influx during induction since only reversible depression occurred when BAPTA was included in the recording pipette or presynaptic stimulation was interrupted during l-AP4 treatment (Pelkey et al. 2005). Additionally, l-AP4-induced LTD was peculiar to MF–SLIN synapses since only a transient reversible depression of MF–PYR transmission was observed by l-AP4-mediated activation of high affinity group III mGluRs (probably mGluR4 or 8), further supporting efficient biochemical partitioning between functionally distinct MF terminals. Incidentally, MF–PYR LTD requires group II mGluR (mGluR2/3) activation (Tzounopoulos et al. 1998), indicating reciprocal roles for group II and III mGluRs in short- and long-term depression at functionally divergent MF release sites. Indeed, mGluR2/3 activation with agonist DCGIV, typically used to characterize afferent inputs as being of MF origin (Kamiya et al. 1996), yields fully reversible depression at MF–SLIN synapses (Maccaferri et al. 1998; Pelkey et al. 2005).

Together these findings indicate a critical role for mGluR7 activation in the induction of presynaptic MF–SLIN LTD. The discovery of this crucial trigger reveals that, despite being NMDAR independent, presynaptic MF–SLIN LTD is ‘Hebbian’ in nature requiring coincident pre- and postsynaptic activation at levels sufficient to ensure conjunctive mGluR7 and retrograde messenger signalling. While mGluR7 activation in isolation is clearly able to initiate presynaptic depression, the short lived nature of such inhibition upon agonist removal in the absence of postsynaptic Ca²⁺ elevation suggests that mGluR7 subserves an additional role in making the presynaptic terminal receptive to LTD consolidation by the retrograde messenger. Such ‘terminal priming’ by a presynaptic coincidence detector may be a common feature of retrograde messenger-dependent presynaptic LTD at diverse central synapses ensuring synapse specificity of plasticity, and could potentially explain the need for constant presynaptic stimulation during agonist treatment for some forms of presynaptic chemical LTD (Sjostrom et al. 2003; Ronesi et al. 2004; Bender et al. 2006; Duguid & Sjostrom, 2006; Singla et al. 2007).

In addition to requiring mGluR7 activation we also found that both l-AP4- and HFS-induced MF–SLIN depression rely on PKC signalling (Fig. 1C; Pelkey et al. 2005). Although heterologously expressed mGluR7 typically signals via Gi/o to inhibit AC, coupling through PKC has been observed in neurons and is consistent with the lack of AC–cAMP–PKA sensitivity of MF–SLIN transmission (Maccaferri et al. 1998; Pelkey et al. 2005). Indeed, mGluR7 directly interacts with the PKC scaffolding protein PICK1, and this interaction along with PKC activity is necessary for mGluR7-mediated depression of release from cerebellar granule cells (Boudin et al. 2000; Dev et al. 2000; El Far et al. 2000; Perroy et al. 2002a; Enz & Croci, 2003). Most interestingly, PKC acting downstream of mGluR7 activation has been reported to persistently depress P/Q-type voltage-gated calcium channels (VGCCs; Perroy et al. 2000, 2002b). Similarly, using combined electrophysiological recording and two-photon Ca²⁺ imaging techniques, we recently found that presynaptic MF–SLIN LTD is ultimately expressed as a persistent reduction in P/Q-type VGCC function (Fig. 1C; Pelkey et al. 2006). While monitoring stimulus-evoked presynaptic Ca²⁺ transients in SLIN targeting Fil release sites both HFS and l-AP4 treatment produced LTD of P/Q-type VGCC-mediated Ca²⁺ transients. Importantly, like presynaptic LTD of MF–SLIN transmission, Fil Ca²⁺ transient LTD was NMDAR independent but relied on both mGluR7 and CP-AMPAR activation. Incredibly, this Ca²⁺ transient LTD was highly specific to SLIN-targeting release sites as neighbouring parent MFBs showed no long-term effects of l-AP4 or HFS. This highly disparate and compartmentalized regulation of Ca²⁺ transients does not result from distinct complements of VGCCs at Fils and parent MFBs since both transmitter release, and the Ca²⁺ transients themselves, are mediated by similar mosaics of N- and P/Q-type VGCCs at the two divergent terminals. Rather the data further indicate a striking partitioning of biochemical signalling cascades between functionally distinct release sites of the same parent axon. Exactly how P/Q-type VGCCs remain depressed following the transient l-AP4 or HFS conditioning epochs remains to be determined; however, one attractive possibility is their removal from the Fil terminal surface membrane in conjunction with agonist-induced mGluR7 internalization (Altier et al. 2006; and see below).

mGluR7 as a metaplastic switch controlling bidirectional plasticity at MF–SLIN terminals

While attempting a simple occlusion experiment to further confirm whether chemical- and activity-induced MF–SLIN LTD share common cell signalling mechanisms, we made the surprising observation that the same conditioning stimulation that induces presynaptic LTD at naïve synapses completely reverses chemical LTD by persistently enhancing presynaptic release (Pelkey et al. 2005). In other words, if MF–SLIN synapses first undergo l-AP4-induced LTD subsequent HFS reverses this depression by inducing presynaptic LTP/dedepression, revealing that the polarity of plasticity observed in response to HFS critically depends upon the recent history of the synapse. This newly uncovered metaplasticity provided the first demonstration of bidirectional MF–SLIN plasticity and now forms a major focus of our ongoing research. Although the molecular details remain to be elucidated there are relatively clear consequences for the MF–CA3 circuit. The initial depression of feedforward inhibition by an appropriate plasticity-inducing level of activity in the MF pathway may be critical for excitation-spike generation in CA3 pyramid targets (Lawrence & McBain, 2003), and hence efficient dentate–CA3 information transfer. However, continued depression of MF–SLIN synapses in response to multiple barrages of dentate activity would create an inherently unstable circuit predisposed to epileptiform activity and excitotoxicity. The newly discovered ability of MF–SLIN synapses to reverse LTD is probably crucial for restoring balance to CA3 excitation/inhibition dynamics (see also (Froemke et al. 2007).

Interestingly, MF–SLIN LTP was not revealed by simply blocking mGluR7 with MSOP during HFS at naïve synapses, nor was it uncovered by other manipulations that depressed presynaptic function of naïve MF–SLIN inputs such as mGluR2/3 activation. These findings lead us to propose that following intense activation by exogenously applied agonist, mGluR7 is internalized from the surface of MF terminals contacting SLINs unmasking the ability of ‘naked’, mGluR7-lacking MF–SLIN releases sites, to persistently enhance release in response to the same HFS that typically induces LTD at naïve mGluR7 surface-expressing boutons. Consistent with this hypothesis, mGluR7 does indeed exhibit prominent agonist-induced internalization (Pelkey et al. 2005, Pelkey et al 2007). In this scheme, mGluR7 can be viewed as a metaplastic switch at CP-AMPAR-containing MF–SLIN synapses, whose activation and surface expression governs the direction of plasticity. Thus, in addition to enabling target-cell-specific presynaptic LTD in the MF pathway, selective accumulation of mGluR7 at MF–SLIN presynaptic terminals imparts a novel ‘state-dependent’ feature to long-term plasticity of MF–SLIN transmission. As agonist-induced internalization is a common property of diverse G-protein-coupled receptors (GPCRs; Gainetdinov et al. 2004), it is interesting to speculate that internalization of other presynaptic GPCRs may similarly impart state dependence to presynaptic regulation throughout the CNS.

Concluding remarks

The above discussion has highlighted that through methodical study of the individual components underlying release and plasticity at divergent MF targets we have gained a greater appreciation for target-cell-specific forms of plasticity arising from a common axon. At the network level we speculate that HFS-induced strengthening of mossy fibre transmission onto principal cells, in conjunction with a concomitant weakening of excitatory drive onto interneurons providing feedforward inhibition, will act to broaden the temporal window for downstream activation of the associative network between CA3 pyramidal cells. We predict that the subsequent internalization of mGluR7 and unmasking of an as yet unknown mechanism to strengthen excitatory drive onto interneurons should truncate or close this temporal window, effectively restoring feedforward inhibitory control of MF-driven CA3 network activity.

Although MF synapses are anatomically unique among central synapses, numerous examples of target-specific plasticity have emerged in recent years within a number of different central structures, indicating that anatomical specialization is not always a necessity. We suggest that systematic characterization of the distribution and function of presynaptic proteins and biochemical cascades will ultimately reveal that compartmentalization, and target-cell-specific mechanisms, are the rule and not the exception for presynaptic boutons along a common axon.