In the CNS (central nervous system), nerve cells communicate by transmitting signals from one to the next across chemical synapses. Electrical signals trigger controlled secretion of neurotransmitter by exocytosis of SV (synaptic vesicles) at the presynaptic site. Neurotransmitters diffuse across the synaptic cleft, activate receptor channels in the receiving neuron at the postsynaptic site, and thereby elicit a new electrical signal. Repetitive stimulation should result in fast depletion of fusion-competent SVs, given their limited number in the presynaptic bouton. Therefore, to support repeated rounds of release, a fast trafficking cycle is required that couples exocytosis and compensatory endocytosis. During this exo-endocytic cycle, a defined stoichiometry of SV proteins has to be preserved, that is, membrane proteins have to be sorted precisely. However, how this sorting is accomplished on a molecular level is poorly understood. In the present chapter we review recent findings regarding the molecular composition of SVs and the mechanisms that sort SV proteins during compensatory endocytosis. We identify self-assembly of SV components and individual cargo recognition by sorting adaptors as major mechanisms for maintenance of the SV protein complement.
- adaptor proteins
- compensatory endocytosis
- molecular sorting
- readily retrievable pool
- release site clearance
- synaptic vesicle
Early EM studies revealed clusters of granular-like structures in presynaptic terminals that were designated as SVs (synaptic vesicles) . At the same time, the quantal hypothesis of neurotransmitter release was proposed  and several years later, SVs were unambiguously identified as subcellular compartments releasing discrete packages of neurotransmitter (‘quanta’) upon fusion with the PM (plasma membrane) [3,4]. These electrophysiological recordings of neuromuscular synapses, however, showed that neurons can sustain high rates of SV release without exhausting their SV pools, although the cell body with its precursor-supplying organelles may be as far as 1 m away. Thus, synapses must have developed efficient endocytic mechanisms to recapture and reuse fused SVs in a compensatory way. Further EM studies using extracellular endocytic tracers indeed ultimately established the model of local SV recycling via exocytosis and subsequent compensatory endocytosis [5,6]. What is particularly striking about this compensatory endocytosis triggered by exocytosis is the remarkable precision with which the surface area is reset to its original value after a secretory event. This is best seen in electrophysiological recordings of the electrical capacitance (e.g. ), which scales linearly with the membrane area, but is also observed in fluorescence recordings of live neurons expressing a pH-sensitive green fluorescent protein (pHluorin) fused to luminal domains of SV proteins (e.g. [8,9]) or using antibodies against luminal domains of SV proteins coupled to pH-sensitive organic dyes .
Many aspects of the spatio-temporal dynamics of SV cycling have been studied in detail, including molecular mechanisms of exocytic membrane fusion [11,12] and endocytic retrieval of SVs [8,13]. For a tight spatio-temporal coupling between these two processes, presynaptic scaffolding proteins have been suggested to be essential [14,15]. However, SV function relies on a distinct set of proteins present in a defined stoichiometry, and the molecular sorting mechanisms for individual SV components during exo-endocytosis still remain largely unresolved.
Three principal scenarios for maintenance of SV integrity can be put forward (Figure 1). In the simplest case, SV integrity would be preserved by a so-called kiss-and-run mechanism of exo-endocytosis: SVs at the AZ (active zone) would only transiently fuse, forming a narrow fusion pore, just wide enough for transmitter release. Then, they would rapidly (in <1 s) pinch off . This way, no proteins would be lost during exo-endocytosis (Figure 1A, upper panel). However, the significance and contribution of this mechanism to compensatory endocytosis in the CNS remains questionable: the most recent results obtained by high-resolution cryo-EM after rapid high pressure freezing indeed revealed an ultrafast endocytosis mechanism starting 50–100 ms after stimulation , but SVs appeared to fully collapse and flatten within the PM.
EM data are in favour of CME (clathrin-mediated endocytosis) being the major mechanism for SV recycling [5,18]. Here SV exo- and endo-cytosis, although temporally tightly coupled, are spatially segregated: SVs fully collapse and flatten into the PM, and SV proteins and lipids have to be retrieved elsewhere outside the AZ (Figures 1A, lower panel, and 1B). As clathrin is neither able to bind directly to the membrane nor to cargo proteins of SVs, specific adaptor proteins are needed to link the clathrin coat formation to the concomitant selection and concentration of cargo proteins . Two modes, fundamentally distinct with respect to the molecular identity, are possible. First, SV constituents could remain clustered on fusion, and diffuse as a raft-like patch along the membrane into the peri-AZ (peri-active zone), where CME takes place  (Figure 1A, lower panel). In this case, no sorting of SV proteins and only one specific adaptor protein would be needed for clathrin-coat formation. Second, SV constituents could disperse and freely diffuse in the presynaptic PM (Figure 1B). Thus, the SV identity would be fully lost, and SV constituents would have to be re-clustered and re-sorted [9,21]. This process entails either a multitude of different endocytic adaptor proteins, that is, one type of adaptor for each SV protein species, or involves a combination of self-assembly interactions among the cargo along with a distinct set of adaptors targeting some key SV proteins, pivotal in the self-assembly process.
Starting from the SV proteomics and lipidomics we will review the known interactions of identified adaptor proteins with SV proteins as well as the known interactions among them. We develop an argument as to why the most complex of all possible three scenarios (Figure 1B), that is, spatial segregation of exo- and endocytosis along with a combination of self-assembly and adaptor protein-induced sorting (Figure 2), might have evolved for maintaining SV integrity throughout the exo-endocytic cycle.
Lipid and protein composition of SVs
SVs of 40 nm diameter can be purified at large scale and high purity, giving access to analysis of protein and lipid content [22,23]. The lipid composition of purified SVs is characterized by a low phospholipid proportion of phosphatidylinositol and negligible amounts of glycolipids or lyso-phosphatidylcholine [23–25]. Sphingomyelins make up about 10% of all phospholipids, and the phosphatidylethanolamine fraction contains a high proportion of plasmalogen ether-phospholipids [23,25]. The distribution of phospholipids between the two SV leaflets is asymmetric, with most of the phosphatidylethanolamine and phosphatidylinositol located in the outer leaflet , largely reflecting the phospholipid distribution of the presynaptic PM. The cholesterol content of isolated SVs is higher than that reported for synaptosomal preparations, that is, preparations of complete presynaptic terminals (∼40 mol% and ∼30 mol%, respectively) [24–26].
Proteomic data revealed a surprisingly large number of SV proteins [25,27]. Up to 80 different integral membrane proteins and an additional large number of peripheral proteins were identified, indicating that SVs are dominated by proteins. Within the SV proteome, the majority of identified proteins is involved in trafficking, for example SNARE proteins, Rab proteins and endocytic adaptor proteins. Additionally, SVs contain the molecular machinery required for neurotransmitter uptake, including transporters, ion channels and the v-ATPase (v-type ATPase). Quantitative proteomic analyses partially resulted in different absolute copy numbers per SV and a high intervesicle variability for proteins such as the vesicular SNARE Syb2 (synaptobrevin 2; 10–70 copies per SV) or the tetraspan membrane protein Syp1 (synaptophysin 1; 13–30 copies per SV) [25,28]. However, for some proteins such as SV2 (synaptic vesicle 2-related protein) or v-ATPase, little intervesicle variation in copy number has been reported, with absolute numbers of eventually only one or two molecules per SV [25,28], indicating that they are sorted with high precision.
Recognition of synaptic vesicle cargo by sorting adaptors
As discussed above, CME is thought to be the main mechanism of SV retrieval at the peri-AZ (Figure 2). This cargo-specific mechanism of endocytosis has been studied intensively in non-neuronal and neuronal cells, revealing that CME can be separated into several spatially distinct steps: the recruitment of clathrin from the cytosol to the PM, subsequent invagination and fission of clathrin-coated vesicles from the PM and finally disassembly of the clathrin coat in the cytosol [14,29]. Two sets of proteins are involved: those forming the clathrin coat and a number of accessory proteins. The basic module of the coat is the three-legged clathrin triskelion, which forms a lattice attaching to the PM via adaptor proteins such as the tetrameric adaptor protein complex AP-2. Among the accessory proteins that associate with the coat are dynamin (a GTPase that has a prime function in the fission reaction of clathrin-coated vesicles) and amphiphysin (which targets dynamin and other proteins to the coat) .
Endocytic adaptor proteins
The term ‘endocytic adaptor’ was developed in 1981. Originally, it described a family of molecular units that would sort out those molecules that were removed from the membrane in clathrin cages from those that remained behind . These units were predicted to (1) interact with clathrin on the cytoplasmic side, (2) recognize specifically an ‘address ticket’ carried by a receptor and (3) include some signal indicating to which organelle they should go. Nowadays, a number of adaptors have been identified that bind to clathrin, which itself can neither interact directly with cargo nor membranes, and to sorting signals within the cytoplasmic tails of transmembrane cargo, but also to PI(4,5)P2 (phosphatidylinositol 4,5-bisphosphate). PI(4,5)P2 is a minor phospholipid of the cytoplasmic leaflet of the PM that is enriched at sites of endocytosis. Therefore endocytic adaptors have the ability to directly interact with all components important for endocytosis and are ideal candidates to mediate efficient sorting of cargo into clathrin-coated vesicles.
Characterization and classification of endocytic adaptor proteins
Although endocytic adaptor proteins vary greatly in size [∼300–3000 aa (amino acids)] and structure, all of them consist of discretely folded domains, which bind to PM components, and unstructured domains, which contain multiple peptide-binding motifs important for recruitment of clathrin and other accessory proteins [32,33]. This modular design enables the interaction with numerous partners in an apparently synchronized fashion  and allows efficient internalization of cargo during endocytosis.
Endocytic adaptors can be categorized in two major classes: the classic multimeric adaptor proteins and the non-classic CLASPs (clathrin-associated sorting proteins) . AP-2 is so far the best known and characterized endocytic adaptor of the multimeric adaptor protein class. It was identified originally using Triton X-100 extractions of clathrin-coated vesicles  and consists of two ∼100 kDa subunits named α and β2, the ∼50 kDa subunit μ2 and a small ∼17 kDa subunit δ2 . Each subunit interacts with numerous other proteins and membrane lipids. Owing to its direct interaction with clathrin, it provides a straight mechanism to assemble proteins into clathrin-coated pits. The μ2 subunit is thought to play the major role in cargo recognition , but using quantitative endocytosis assays it was shown that no single aspect of μ2 function is critical, implying that a more distributed network of interactions supports AP-2 function in SV endocytosis . In contrast, CLASPs are monomeric or dimeric, and vary in structure and binding properties.
Adaptor proteins involved in SV endocytosis
The minimal protein complement of SVs absolutely required for efficient neurotransmission (Figures 3 and 4) comprises a specific transporter for neurotransmitter uptake (e.g. vGlut1), a driving force for neurotransmitter uptake (v-ATPase), the vesicular part of the SNARE fusion complex (Syb2), and finally, a calcium sensor for triggering exocytosis [Syt1 (synaptotagmin 1)]. In general, CME of the majority of transmembrane receptors is either directly or indirectly dependent on AP-2 . Indeed, short cytosolic sorting motifs recognized by AP-2 have also been identified and validated for AP-2 binding in the SV proteins Syt1  and SV2 , and in a number of transporters, including vGlut1 .
However, knockdown or knockout of AP-2 subunits in neurons resulted in only minor inhibition of SV retrieval [42,43]. These results suggest that alternative adaptors may also be employed (Figures 3 and 4).
Stonin 2 was identified as a sorting adaptor for Syt1: the first and so far only endocytic protein specifically dedicated to SV recycling [19,44]. The interaction between the adaptor and its cargo takes place at the μHD domain of stonin 2 . Whether other SV proteins are co-sorted in a stonin 2-dependent manner is so far unclear. Analysis from Drosophila melanogaster stonedB (stonin 2) mutant flies indicate that stonedB/stonin 2 is also involved in the retrieval of vGlut1 . But Syt1 has also been suggested to interact with other adaptors such as AP-2  and Eps15 or intersectin . The SH3 domain-containing accessory protein intersectin in addition scaffolds the endocytic process by directly associating with AP-2 . This indicates that multiple adaptor proteins with overlapping functions may co-operate in parallel to completely retrieve the entire set of SV cargo.
So far, little is known about the endocytic sorting of the vesicular SNARE protein Syb2. Two suggested candidate proteins are the CLASPs AP-180 and its ubiquitous homologue CALM, which were shown to recognize specific regions in the SNARE motif of Syb2. Deletion of AP-180/CALM caused selective accumulation of Syb2 at the neuronal surface .
Given the lack of identified adaptor proteins for some SV proteins, their low affinity to their cargo, and the fact that even AP-2 [42,43] and clathrin  are not absolutely required for compensatory endocytosis, other molecular mechanisms must coexist. One attractive candidate is self-assembly.
SV component retrieval by self-assembly
Proteomic data indicate that most SV proteins are present in surprisingly low copy numbers per vesicle – in some cases, such as SV2 or the v-ATPase, even just one or two copies are present per vesicle . How such extremely precise sorting of SV proteins can be accomplished molecularly is enigmatic. Even with several layers of sorting adaptor and endocytic accessory protein interactions, it is expected on physical grounds that individual numbers of molecules per SV follow Poisson statistics: an average of one or two copies thus would imply that 10–30% of SVs must lack this particular protein. In this light, the astonishingly high copy number of some key proteins such as the SNARE Syb2 (∼70 copies per SV) becomes understandable as a safety margin to ensure the insertion of at least two copies, necessary and sufficient for fast synchronous fusion . However, in the case of the v-ATPase (∼1.4 copies), up to 30% of SVs should not contain transmitter, which appears not to be the case. Thus, a scenario in which SV components simply remain clustered after exocytosis (Figure 1B), as suggested by STED (stimulated emission depletion) confocal microscopy , is very attractive. Such strong self-assembly forces would also explain why rather weak adaptor protein cargo interactions  suffice for successful retrieval of all SV components. However, previous experiments  and more recent combined dual-colour three-dimensional STED and physiological analysis  suggested that SV proteins at least partially disperse in the PM and are re-sorted and re-assembled in an RRetP (readily retrievable pool) of SV protein patches on the presynaptic membrane (Figures 1B and 2).
The finding that SV proteins disperse after fusion and recluster implies that self-assembly forces are too weak to alone account for clustering. Only few interactions between SV proteins have been documented and biochemically confirmed (Figure 4). These data show that oligomerization forces are indeed weak and in the micromolar range. Interestingly, the two most abundant proteins, the vesicular SNARE Syb2 and the tetraspan protein Syp1, were shown to oligomerize. Syp1 was found to form homohexamers in artificial membranes [52,53], whereas Syb2 can dimerize . Moreover, both are able to form heterodimers with each other [54–56]. This interaction was shown to control and mediate targeting of Syb2 to SVs , suggesting that Syp1 simply functions as an SV-integral adaptor for Syb2 (reviewed in ). However, Syp1-knockout mice show only a mild defect in retrieval efficacy for Syb2 [59–61]. Interestingly, the Syp1–Syb2 interaction also critically depends on the membrane cholesterol content . Cholesterol appears to be enriched at sites of endocytosis, as SV membranes display a high cholesterol content . Thus, cholesterol might indirectly contribute to self-assembly of exocytosed SV by enforcing interactions during endocytosis.
Furthermore Syb2 has been shown to directly interact via its membrane-proximal domain with a cytosolic loop of the c-subunit of the V0 sector of the v-ATPase . Acute perturbation of this interaction did not affect proton pump activity, but surprisingly caused a decrease in neurotransmitter release. How could one or two v-ATPases efficiently re-sort tens of Syb2? The v-ATPase is a huge multidomain protein comprising at least 13 subunits and thus provides a substrate for many binding partners, including actin and AP-2 [64,65]. Therefore, even a single copy of this multimeric enzyme might serve as a major hub and nucleation site for SV protein self-assembly. SV2 is another large protein that is present only in one or two copies per SV. Yet it appears to be required for trafficking of Syt1 to SVs . Neurons lacking SV2 contained less Syt1 and had a higher proportion of Syt1 at the PM. Besides binding to Syt1, SV2 in vitro also binds to the clathrin adaptors AP-2, Eps15, and amphiphysin 2 . Again, with only one or two copies per SV, it remains unclear how it can efficiently contribute to retrieval of ∼15 Syt1 molecules . Maybe both v-ATPase and SV2 provide multiple interaction sites for nucleation of self-assembly and subsequent adaptor protein recruitment, and serve as a sort of ‘Swiss army knife’ for endocytosis.
Active zone integrity: clearance and scaffolds
The tight temporal coupling between SV exo- and endocytosis suggests that SV cargo protein sorting and recycling probably involves precise control of the localization and dynamics of proteins or protein complexes as they partition between sites of fusion, that is, AZ, and sites of endocytosis, that is, peri-AZ (Figure 2). It has been proposed that a co-ordinated translocation of SV components from the AZ to the peri-AZ , which has been termed ‘clearance’ [15,67], is not only essential for proper retrieval of SV components, but also necessary for maintaining the integrity of the release sites at the AZ. Interestingly, previous studies [7,68] could indeed demonstrate that impairing endocytosis leads to a novel form of fast short-term depression, which is not related to insufficient SV supply at the presynaptic bouton, but is a result of slow clearance of vesicular components from the release site. This finding implies an important role of endocytic proteins in conjunction with scaffold proteins at and outside the AZ for sustained synaptic transmission at high rates.
Giant scaffold proteins such as RIMs, piccolo, bassoon and BRP [14,15,69] may assemble into a cytosolic matrix that provides attachment sites for the movement of SV membrane constituents from the AZ. Maintenance of the integrity of release sites might be the key for understanding exo-endocytic coupling: exocytosed SV components must be swiftly cleared away to give way for a new SV to dock at the release site (Figure 2). Thus, exo- and endocytosis need to be spatially segregated, a condition not provided by the kiss-and-run mechanism (Figure 1A, upper panel). Although remaining raft-like patches could meet this demand (Figure 1A, lower panel), the required strong self-assembly forces in turn would result in a bulky (∼80 nm diameter) and rather rigid protein cluster that could not easily diffuse away through the tight protein scaffold meshwork of the AZ and thus would obstruct fusion sites and perturb their distances to Ca2+ channels, that is, disintegrate release site function. Because Syb2 engages with plasma membrane SNAREs for SNARE complex formation during fusion, these become an integral part of the raft-like patches and thus newly formed SVs. Hence, the seemingly most inefficient and complex of all coupling sequences (Figures 1B and 2), namely full collapse, quick dispersion in all directions, and re-assembly at the periphery, is the only mechanism that allows both fast clearance and spatial segregation of AZ membrane SNAREs and vesicular Syb2. In this light, co-operative coupling of weak self-assembly forces with weak cargo–adaptor protein interactions appears to be the best compromise to enable integrity of both the AZ and SVs during their exo-endocytic itinerary. In addition, such a complex network (Figure 4) provides compensation even for a lack of AP-2 or clathrin.
But why then are these re-assembled and re-sorted patches not completely removed during compensatory endocytosis, but in part maintained as RRetP at steady-state in the absence of stimulation (Figure 2)? Maybe the major advantage of the RRetP is to provide a sink and buffer for SV proteins diffusing away from the release site. This way, clearance is facilitated by counteracting the thermodynamically favoured reformation of SNARE complexes in the AZ membrane.
Clearly, the analysis of endocytic and AZ/peri-AZ components and their functional relationship to exo-endocytic coupling at different types of synapses is far from being complete. As described above, several steps in the SV cycle necessitate a tight spatio-temporal co-ordination in defined membrane domains: precise timing of transmitter release relative to the action potential requires proximity between Ca2+ channels and release-ready vesicles; high release rates of SVs call for an effective clearance mechanism in order to give way for newly arriving SVs and also to preserve the AZ integrity; and finally resorting of SV components from those of the PM is needed prior to endocytic retrieval within the RRetP at the peri-AZ. All of these tasks entail a complex set of transient, that is rather weak, protein–protein interactions among SV proteins, with adaptor proteins, and with structural scaffolds for the spatiotemporal co-ordination. The future task will not only be to define the complete network of interactions, but also visualize their dynamics in live cells.
• For maintaining synaptic transmission, SV components have to be cleared away swiftly from the AZ after fusion. Otherwise crowding of exocytosed SV proteins at the AZ perturbs its integrity, resulting in short-term depression.
• For efficient clearance, sites of exocytosis and endocytosis must be spatially segregated, that is, SVs need to fully collapse and flatten in the membrane, and SV proteins need to disperse at least partially and diffuse to sites of endocytosis outside the AZ.
• For retrieval of SV proteins at the peri-AZ (sites of endocytosis), they have to be efficiently re-sorted and re-clustered in the PM prior to endocytosis. This process involves protein–protein interactions between SV proteins (self-assembly) as well as with cytosolic adaptor proteins (cargo sorting).
• Neither self-assembly forces nor adaptor interactions alone are strong enough for correct clustering. This way, efficient clearance after fusion in addition to proper SV protein function (i.e. no steric obstruction) within SVs in the absence of adaptors is warranted.
• Productive SV cargo clustering at endocytic sites requires co-operativity between self-assembly and several layers of adaptor protein interactions, and probably scaffolding by large cytomatrix proteins. These multiple layers of interactions, however, also provide a backup, such that even a lack of AP-2 or clathrin can be tolerated.
We thank all laboratory members for intensive and fruitful discussions, and especially Nataliya Glyvuk and Yaroslav Tsytsyura for providing the EM picture. J.K. was supported by grants from the Deutsche Forschungsgemeinschaft [ESF EuroMembrane, grant numbers SFB 629, SFB 944, and DFG EXC 1003, Cells in Motion – Cluster of Excellence, Muenster, Germany].
- © The Authors Journal compilation © 2015 Biochemical Society